From 3bcc60c479e52c00ff42f9afc9d601d24b60830a Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Wed, 29 Apr 2020 10:42:57 +0200
Subject: [PATCH 01/17] Fix a bug in increasing orders in MGE.

---
 gklearn/ged/median/median_graph_estimator.py | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/gklearn/ged/median/median_graph_estimator.py b/gklearn/ged/median/median_graph_estimator.py
index b5a829c..114c312 100644
--- a/gklearn/ged/median/median_graph_estimator.py
+++ b/gklearn/ged/median/median_graph_estimator.py
@@ -1115,7 +1115,12 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 						continue
 					for label in median_labels:
 						weights[label_id] = min(weights[label_id], self.__ged_env.get_node_rel_cost(dict(label), dict(node_labels[label_id])))
-				selected_label_id = urng.choice(range(0, len(weights)), size=1, p=np.array(weights) / np.sum(weights))[0] # for c++ test: xxx[iii] 
+				sum_weight = np.sum(weights)
+				if sum_weight == 0:
+					p = np.array([1 / len(weights)] * len(weights))
+				else:
+					p = np.array(weights) / np.sum(weights)
+				selected_label_id = urng.choice(range(0, len(weights)), size=1, p=p)[0] # for c++ test: xxx[iii] 
 # 				iii += 1 for c++ test
 				median_labels.append(node_labels[selected_label_id])
 				already_selected[selected_label_id] = True

From 9a365fac98a05e88343a9f1d7bf86a5b102da1da Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Thu, 30 Apr 2020 10:31:08 +0200
Subject: [PATCH 02/17] Fix bugs for enum and add test for median preimage
 generator.

---
 gklearn/ged/env/common_types.py                 | 11 ++--
 gklearn/ged/median/median_graph_estimator.py    |  2 +-
 gklearn/tests/test_median_preimage_generator.py | 71 +++++++++++++++++++++++++
 gklearn/utils/utils.py                          | 11 ++--
 4 files changed, 84 insertions(+), 11 deletions(-)
 create mode 100644 gklearn/tests/test_median_preimage_generator.py

diff --git a/gklearn/ged/env/common_types.py b/gklearn/ged/env/common_types.py
index 2face25..d195b11 100644
--- a/gklearn/ged/env/common_types.py
+++ b/gklearn/ged/env/common_types.py
@@ -6,12 +6,13 @@ Created on Thu Mar 19 18:17:38 2020
 @author: ljia
 """
 
-from enum import Enum, auto
+from enum import Enum, unique
 
+@unique
 class AlgorithmState(Enum):
     """can be used to specify the state of an algorithm.
     """
-    CALLED = auto # The algorithm has been called.
-    INITIALIZED = auto # The algorithm has been initialized.
-    CONVERGED = auto # The algorithm has converged.
-    TERMINATED = auto # The algorithm has terminated.
\ No newline at end of file
+    CALLED = 1 # The algorithm has been called.
+    INITIALIZED = 2 # The algorithm has been initialized.
+    CONVERGED = 3 # The algorithm has converged.
+    TERMINATED = 4 # The algorithm has terminated.
\ No newline at end of file
diff --git a/gklearn/ged/median/median_graph_estimator.py b/gklearn/ged/median/median_graph_estimator.py
index 114c312..df21601 100644
--- a/gklearn/ged/median/median_graph_estimator.py
+++ b/gklearn/ged/median/median_graph_estimator.py
@@ -302,7 +302,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		self.__median_id = gen_median_id
 		self.__state = AlgorithmState.TERMINATED
 		
-		# Get ExchangeGraph representations of the input graphs.
+		# Get NetworkX graph representations of the input graphs.
 		graphs = {}
 		for graph_id in graph_ids:
 			# @todo: get_nx_graph() function may need to be modified according to the coming code.
diff --git a/gklearn/tests/test_median_preimage_generator.py b/gklearn/tests/test_median_preimage_generator.py
new file mode 100644
index 0000000..c81bb7c
--- /dev/null
+++ b/gklearn/tests/test_median_preimage_generator.py
@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Jan 14 15:39:29 2020
+
+@author: ljia
+"""
+import multiprocessing
+import functools
+from gklearn.preimage.utils import generate_median_preimages_by_class
+
+		
+def test_median_preimage_generator():
+	"""MAO, Treelet, using CONSTANT, symbolic only.
+	"""
+	from gklearn.utils.kernels import polynomialkernel
+	# set parameters.
+	ds_name = 'MAO' #
+	mpg_options = {'fit_method': 'k-graphs',
+				   'init_ecc': [4, 4, 2, 1, 1, 1], #
+				   'ds_name': ds_name,
+				   'parallel': True, # False
+				   'time_limit_in_sec': 0,
+				   'max_itrs': 3, # 
+				   'max_itrs_without_update': 3,
+				   'epsilon_residual': 0.01,
+				   'epsilon_ec': 0.1,
+				   'verbose': 2}
+	pkernel = functools.partial(polynomialkernel, d=4, c=1e+7)
+	kernel_options = {'name': 'Treelet', #
+				      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+					  'n_jobs': multiprocessing.cpu_count(),
+					  'normalize': True,
+					  'verbose': 2}
+	ged_options = {'method': 'IPFP',
+				   'initialization_method': 'RANDOM', # 'NODE'
+				   'initial_solutions': 1, # 1
+				   'edit_cost': 'CONSTANT', # 
+				   'attr_distance': 'euclidean',
+				   'ratio_runs_from_initial_solutions': 1,
+				   'threads': multiprocessing.cpu_count(),
+				   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+	mge_options = {'init_type': 'MEDOID',
+				   'random_inits': 10,
+				   'time_limit': 600,
+				   'verbose': 2,
+				   'refine': False}
+	save_results = True
+	dir_save = ds_name + '.' + kernel_options['name'] + '.symb.pytest/'
+	irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+	edge_required = False #
+	
+	# print settings.
+	print('parameters:')
+	print('dataset name:', ds_name)
+	print('mpg_options:', mpg_options)
+	print('kernel_options:', kernel_options)
+	print('ged_options:', ged_options)
+	print('mge_options:', mge_options)
+	print('save_results:', save_results)
+	print('irrelevant_labels:', irrelevant_labels)
+	print()
+	
+	# generate preimages.
+	for fit_method in ['k-graphs', 'expert', 'random']:
+		print('\n-------------------------------------')
+		print('fit method:', fit_method, '\n')
+		mpg_options['fit_method'] = fit_method
+		generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=range(0, 4))
\ No newline at end of file
diff --git a/gklearn/utils/utils.py b/gklearn/utils/utils.py
index 9223b7a..14dc7c1 100644
--- a/gklearn/utils/utils.py
+++ b/gklearn/utils/utils.py
@@ -1,7 +1,7 @@
 import networkx as nx
 import numpy as np
 from copy import deepcopy
-from enum import Enum, auto
+from enum import Enum, unique
 #from itertools import product
 
 # from tqdm import tqdm
@@ -467,8 +467,9 @@ def get_mlti_dim_edge_attrs(G, attr_names):
 		attributes.append(tuple(attrs[aname] for aname in attr_names))
 	return attributes
 
-
+@unique
 class SpecialLabel(Enum):
-    """can be used to define special labels.
-    """
-    DUMMY = auto # The dummy label.
\ No newline at end of file
+	"""can be used to define special labels.
+	"""
+	DUMMY = 1 # The dummy label.
+	# DUMMY = auto # enum.auto does not exist in Python 3.5.
\ No newline at end of file

From c2e88d927429934d6065941662852b9080879194 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Thu, 30 Apr 2020 10:38:44 +0200
Subject: [PATCH 03/17] Update requirement.txt for pre-image.

---
 requirements.txt | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/requirements.txt b/requirements.txt
index 85aabf8..56401d4 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -5,6 +5,6 @@ networkx>=2.2
 scikit-learn>=0.20.0
 tabulate>=0.8.2
 tqdm>=4.26.0
-# cvxpy # for preimage.
-# cvxopt # for preimage.
-# mosek # for preimage.
+cvxpy # for preimage.
+cvxopt # for preimage.
+mosek # for preimage.

From 1191009381a92137d89c65dc217f44d30f714b92 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Sat, 2 May 2020 16:39:28 +0200
Subject: [PATCH 04/17] Update requirements.

---
 README.md        | 2 +-
 requirements.txt | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 3958704..7cfdecc 100644
--- a/README.md
+++ b/README.md
@@ -10,7 +10,7 @@ A Python package for graph kernels, graph edit distances and graph pre-image pro
 ## Requirements
 
 * python>=3.5
-* numpy>=1.15.2
+* numpy>=1.16.2
 * scipy>=1.1.0
 * matplotlib>=3.0.0
 * networkx>=2.2
diff --git a/requirements.txt b/requirements.txt
index 56401d4..853d485 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,4 +1,4 @@
-numpy>=1.15.2
+numpy>=1.16.2
 scipy>=1.1.0
 matplotlib>=3.0.0
 networkx>=2.2

From f5439ee8a1fd181907ec7142af9e789044773d4f Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Sat, 2 May 2020 16:47:38 +0200
Subject: [PATCH 05/17] Update requirements.

---
 requirements.txt | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/requirements.txt b/requirements.txt
index 853d485..4dcd662 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -5,6 +5,6 @@ networkx>=2.2
 scikit-learn>=0.20.0
 tabulate>=0.8.2
 tqdm>=4.26.0
-cvxpy # for preimage.
-cvxopt # for preimage.
-mosek # for preimage.
+cvxpy>=1.0.31 # for preimage.
+cvxopt>=1.2.5 # for preimage.
+mosek>=9.2.5 # for preimage.

From 1772f64ccba3d8cc2432ee8afbf54bc2835bdc51 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Sat, 2 May 2020 17:15:08 +0200
Subject: [PATCH 06/17] Update requirements.

---
 requirements.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/requirements.txt b/requirements.txt
index 4dcd662..47a91a1 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -7,4 +7,4 @@ tabulate>=0.8.2
 tqdm>=4.26.0
 cvxpy>=1.0.31 # for preimage.
 cvxopt>=1.2.5 # for preimage.
-mosek>=9.2.5 # for preimage.
+mosek>=9.2.5; python_version >= '3.6' # for preimage.

From 34e82bfce8e5860342cb4f7b3420be3773de1c12 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Thu, 7 May 2020 17:05:27 +0200
Subject: [PATCH 07/17] Add parallel functions to MGE.

---
 gklearn/ged/median/median_graph_estimator.py       |  282 +-
 gklearn/ged/median/test_median_graph_estimator.py  |    8 +-
 gklearn/ged/median/utils.py                        |    2 +
 gklearn/ged/util/util.py                           |    2 +
 gklearn/gedlib/gedlibpy.cpp                        | 2934 ++++++++++----------
 .../gedlibpy.cpython-36m-x86_64-linux-gnu.so       |  Bin 33390352 -> 33390352 bytes
 gklearn/gedlib/gedlibpy.pyx                        |   10 +-
 gklearn/preimage/median_preimage_generator.py      |    8 +-
 8 files changed, 1708 insertions(+), 1538 deletions(-)

diff --git a/gklearn/ged/median/median_graph_estimator.py b/gklearn/ged/median/median_graph_estimator.py
index df21601..9e0db71 100644
--- a/gklearn/ged/median/median_graph_estimator.py
+++ b/gklearn/ged/median/median_graph_estimator.py
@@ -13,6 +13,9 @@ import time
 from tqdm import tqdm
 import sys
 import networkx as nx
+import multiprocessing
+from multiprocessing import Pool
+from functools import partial
 
 
 class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined node?
@@ -47,6 +50,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		self.__desired_num_random_inits = 10
 		self.__use_real_randomness = True
 		self.__seed = 0
+		self.__parallel = True
 		self.__update_order = True
 		self.__refine = True
 		self.__time_limit_in_sec = 0
@@ -125,6 +129,16 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 	
 				else:
 					raise Exception('Invalid argument "' + opt_val  + '" for option stdout. Usage: options = "[--stdout 0|1|2] [...]"')
+
+			elif opt_name == 'parallel':
+				if opt_val == 'TRUE':
+					self.__parallel = True
+	
+				elif opt_val == 'FALSE':
+					self.__parallel = False
+	
+				else:
+					raise Exception('Invalid argument "' + opt_val  + '" for option parallel. Usage: options = "[--parallel TRUE|FALSE] [...]"')
 	
 			elif opt_name == 'update-order':
 				if opt_val == 'TRUE':
@@ -312,7 +326,6 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 #		print(graphs[0].nodes(data=True))
 #		print(graphs[0].edges(data=True))
 #		print(nx.adjacency_matrix(graphs[0]))
-
 			
 		# Construct initial medians.
 		medians = []
@@ -356,30 +369,12 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			self.__ged_env.load_nx_graph(median, gen_median_id)
 			self.__ged_env.init(self.__ged_env.get_init_type())
 			
-			# Print information about current iteration.
-			if self.__print_to_stdout == 2:
-				progress = tqdm(desc='Computing initial node maps', total=len(graph_ids), file=sys.stdout)
-				
 			# Compute node maps and sum of distances for initial median.
-			self.__sum_of_distances = 0
-			self.__node_maps_from_median.clear()
-			for graph_id in graph_ids:
-				self.__ged_env.run_method(gen_median_id, graph_id)
-				self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(gen_median_id, graph_id)
-# 				print(self.__node_maps_from_median[graph_id])
-				self.__sum_of_distances += self.__node_maps_from_median[graph_id].induced_cost()
-# 				print(self.__sum_of_distances)
-				# Print information about current iteration.
-				if self.__print_to_stdout == 2:
-					progress.update(1)
-					
+			self.__compute_init_node_maps(graph_ids, gen_median_id)
+			
 			self.__best_init_sum_of_distances = min(self.__best_init_sum_of_distances, self.__sum_of_distances)
 			self.__ged_env.load_nx_graph(median, set_median_id)
 # 			print(self.__best_init_sum_of_distances)
-			
-			# Print information about current iteration.
-			if self.__print_to_stdout == 2:
-				print('\n')
 				
 			# Run block gradient descent from initial median.
 			converged = False
@@ -434,7 +429,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 # 					print(self.__node_maps_from_median[graph_id].induced_cost())
 # 					xxx = self.__node_maps_from_median[graph_id]					   
 					self.__ged_env.compute_induced_cost(gen_median_id, graph_id, self.__node_maps_from_median[graph_id])
-#					print('---------------------------------------')
+# 					print('---------------------------------------')
 # 					print(self.__node_maps_from_median[graph_id].induced_cost())
 					# @todo:!!!!!!!!!!!!!!!!!!!!!!!!!!!!This value is a slight different from the c++ program, which might be a bug! Use it very carefully!
 					
@@ -637,6 +632,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		self.__desired_num_random_inits = 10
 		self.__use_real_randomness = True
 		self.__seed = 0
+		self.__parallel = True
 		self.__update_order = True
 		self.__refine = True
 		self.__time_limit_in_sec = 0
@@ -682,35 +678,123 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 	def __compute_medoid(self, graph_ids, timer, initial_medians):
 		# Use method selected for initialization phase.
 		self.__ged_env.set_method(self.__init_method, self.__init_options)
-		
-		# Print information about current iteration.
-		if self.__print_to_stdout == 2:
-			progress = tqdm(desc='Computing medoid', total=len(graph_ids), file=sys.stdout)
-			
+					
 		# Compute the medoid.
-		medoid_id = graph_ids[0]
-		best_sum_of_distances = np.inf
-		for g_id in graph_ids:
-			if timer.expired():
-				self.__state = AlgorithmState.CALLED
-				break
-			sum_of_distances = 0
-			for h_id in graph_ids:
-				self.__ged_env.run_method(g_id, h_id)
-				sum_of_distances += self.__ged_env.get_upper_bound(g_id, h_id)
-			if sum_of_distances < best_sum_of_distances:
-				best_sum_of_distances = sum_of_distances
-				medoid_id = g_id
-				
+		if self.__parallel:
+			# @todo: notice when parallel self.__ged_env is not modified.
+			sum_of_distances_list = [np.inf] * len(graph_ids)
+			len_itr = len(graph_ids)
+			itr = zip(graph_ids, range(0, len(graph_ids)))
+			n_jobs = multiprocessing.cpu_count()
+			if len_itr < 100 * n_jobs:
+				chunksize = int(len_itr / n_jobs) + 1
+			else:
+				chunksize = 100
+			def init_worker(ged_env_toshare):
+				global G_ged_env
+				G_ged_env = ged_env_toshare
+			do_fun = partial(_compute_medoid_parallel, graph_ids)
+			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
+			if self.__print_to_stdout == 2:
+				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
+							desc='Computing medoid', file=sys.stdout)
+			else:
+				iterator = pool.imap_unordered(do_fun, itr, chunksize)
+			for i, dis in iterator:
+				sum_of_distances_list[i] = dis
+			pool.close()
+			pool.join()
+			
+			medoid_id = np.argmin(sum_of_distances_list)
+			best_sum_of_distances = sum_of_distances_list[medoid_id]
+			
+			initial_medians.append(self.__ged_env.get_nx_graph(medoid_id, True, True, False)) # @todo
+
+		else:
 			# Print information about current iteration.
 			if self.__print_to_stdout == 2:
-				progress.update(1)
-		initial_medians.append(self.__ged_env.get_nx_graph(medoid_id, True, True, False)) # @todo
+				progress = tqdm(desc='Computing medoid', total=len(graph_ids), file=sys.stdout)
 		
-		# Print information about current iteration.
-		if self.__print_to_stdout == 2:
-			print('\n')
+			medoid_id = graph_ids[0]
+			best_sum_of_distances = np.inf
+			for g_id in graph_ids:
+				if timer.expired():
+					self.__state = AlgorithmState.CALLED
+					break
+				sum_of_distances = 0
+				for h_id in graph_ids:
+					self.__ged_env.run_method(g_id, h_id)
+					sum_of_distances += self.__ged_env.get_upper_bound(g_id, h_id)
+				if sum_of_distances < best_sum_of_distances:
+					best_sum_of_distances = sum_of_distances
+					medoid_id = g_id
+					
+				# Print information about current iteration.
+				if self.__print_to_stdout == 2:
+					progress.update(1)
+					
+			initial_medians.append(self.__ged_env.get_nx_graph(medoid_id, True, True, False)) # @todo
+			
+			# Print information about current iteration.
+			if self.__print_to_stdout == 2:
+				print('\n')
+			
+			
+	def __compute_init_node_maps(self, graph_ids, gen_median_id):
+		# Compute node maps and sum of distances for initial median.
+		if self.__parallel:
+			# @todo: notice when parallel self.__ged_env is not modified.
+			self.__sum_of_distances = 0
+			self.__node_maps_from_median.clear()
+			sum_of_distances_list = [0] * len(graph_ids)
+			
+			len_itr = len(graph_ids)
+			itr = graph_ids
+			n_jobs = multiprocessing.cpu_count()
+			if len_itr < 100 * n_jobs:
+				chunksize = int(len_itr / n_jobs) + 1
+			else:
+				chunksize = 100
+			def init_worker(ged_env_toshare):
+				global G_ged_env
+				G_ged_env = ged_env_toshare
+			do_fun = partial(_compute_init_node_maps_parallel, gen_median_id)
+			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
+			if self.__print_to_stdout == 2:
+				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
+							desc='Computing initial node maps', file=sys.stdout)
+			else:
+				iterator = pool.imap_unordered(do_fun, itr, chunksize)
+			for g_id, sod, node_maps in iterator:
+				sum_of_distances_list[g_id] = sod
+				self.__node_maps_from_median[g_id] = node_maps
+			pool.close()
+			pool.join()
 			
+			self.__sum_of_distances = np.sum(sum_of_distances_list)
+# 			xxx = self.__node_maps_from_median
+			
+		else:
+			# Print information about current iteration.
+			if self.__print_to_stdout == 2:
+				progress = tqdm(desc='Computing initial node maps', total=len(graph_ids), file=sys.stdout)
+				
+			self.__sum_of_distances = 0
+			self.__node_maps_from_median.clear()
+			for graph_id in graph_ids:
+				self.__ged_env.run_method(gen_median_id, graph_id)
+				self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(gen_median_id, graph_id)
+	# 				print(self.__node_maps_from_median[graph_id])
+				self.__sum_of_distances += self.__node_maps_from_median[graph_id].induced_cost()
+	# 				print(self.__sum_of_distances)
+				# Print information about current iteration.
+				if self.__print_to_stdout == 2:
+					progress.update(1)
+			
+			# Print information about current iteration.
+			if self.__print_to_stdout == 2:
+				print('\n')
+
 		
 	def __termination_criterion_met(self, converged, timer, itr, itrs_without_update):
 		if timer.expired() or (itr >= self.__max_itrs if self.__max_itrs >= 0 else False):
@@ -816,26 +900,57 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 
 
 	def __update_node_maps(self):
-		# Print information about current iteration.
-		if self.__print_to_stdout == 2:
-			progress = tqdm(desc='Updating node maps', total=len(self.__node_maps_from_median), file=sys.stdout)
-			
 		# Update the node maps.
-		node_maps_were_modified = False
-		for graph_id, node_map in self.__node_maps_from_median.items():
-			self.__ged_env.run_method(self.__median_id, graph_id)
-			if self.__ged_env.get_upper_bound(self.__median_id, graph_id) < node_map.induced_cost() - self.__epsilon:
-# 				xxx = self.__node_maps_from_median[graph_id]
-				self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__median_id, graph_id)
-# 				yyy = self.__node_maps_from_median[graph_id]
-				node_maps_were_modified = True
+		if self.__parallel:
+			# @todo: notice when parallel self.__ged_env is not modified.
+			node_maps_were_modified = False
+# 			xxx = self.__node_maps_from_median.copy()
+			
+			len_itr = len(self.__node_maps_from_median)
+			itr = [item for item in self.__node_maps_from_median.items()]
+			n_jobs = multiprocessing.cpu_count()
+			if len_itr < 100 * n_jobs:
+				chunksize = int(len_itr / n_jobs) + 1
+			else:
+				chunksize = 100
+			def init_worker(ged_env_toshare):
+				global G_ged_env
+				G_ged_env = ged_env_toshare
+			do_fun = partial(_update_node_maps_parallel, self.__median_id, self.__epsilon)
+			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
+			if self.__print_to_stdout == 2:
+				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
+							desc='Updating node maps', file=sys.stdout)
+			else:
+				iterator = pool.imap_unordered(do_fun, itr, chunksize)
+			for g_id, node_map, nm_modified in iterator:
+				self.__node_maps_from_median[g_id] = node_map
+				if nm_modified:
+					node_maps_were_modified = True
+			pool.close()
+			pool.join()
+# 			yyy = self.__node_maps_from_median.copy()
+
+		else:
 			# Print information about current iteration.
 			if self.__print_to_stdout == 2:
-				progress.update(1)
-			
-		# Print information about current iteration.
-		if self.__print_to_stdout == 2:
-			print('\n')
+				progress = tqdm(desc='Updating node maps', total=len(self.__node_maps_from_median), file=sys.stdout)
+				
+			node_maps_were_modified = False
+			for graph_id, node_map in self.__node_maps_from_median.items():
+				self.__ged_env.run_method(self.__median_id, graph_id)
+				if self.__ged_env.get_upper_bound(self.__median_id, graph_id) < node_map.induced_cost() - self.__epsilon:
+	# 				xxx = self.__node_maps_from_median[graph_id]
+					self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__median_id, graph_id)
+	# 				yyy = self.__node_maps_from_median[graph_id]
+					node_maps_were_modified = True
+				# Print information about current iteration.
+				if self.__print_to_stdout == 2:
+					progress.update(1)
+				
+			# Print information about current iteration.
+			if self.__print_to_stdout == 2:
+				print('\n')
 			
 		# Return true if the node maps were modified.
 		return node_maps_were_modified
@@ -1230,7 +1345,8 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 	
 	def __add_node_to_median(self, best_config, best_label, median):
 		# Update the median.
-		median.add_node(nx.number_of_nodes(median), **best_label)
+		nb_nodes_median = nx.number_of_nodes(median)
+		median.add_node(nb_nodes_median, **best_label)
 		
 		# Update the node maps.
 		for graph_id, node_map in self.__node_maps_from_median.items():
@@ -1494,4 +1610,40 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 #			median_label = {}
 #			for key, val in median.items():
 #				median_label[key] = str(val)
-#			return median_label
\ No newline at end of file
+#			return median_label
+
+
+def _compute_medoid_parallel(graph_ids, itr):
+	g_id = itr[0]
+	i = itr[1]
+	# @todo: timer not considered here.
+# 			if timer.expired():
+# 				self.__state = AlgorithmState.CALLED
+# 				break
+	sum_of_distances = 0
+	for h_id in graph_ids:
+		G_ged_env.run_method(g_id, h_id)
+		sum_of_distances += G_ged_env.get_upper_bound(g_id, h_id)
+	return i, sum_of_distances
+
+
+def _compute_init_node_maps_parallel(gen_median_id, itr):
+	graph_id = itr
+	G_ged_env.run_method(gen_median_id, graph_id)
+	node_maps_from_median = G_ged_env.get_node_map(gen_median_id, graph_id)
+# 				print(self.__node_maps_from_median[graph_id])
+	sum_of_distance = node_maps_from_median.induced_cost()
+# 				print(self.__sum_of_distances)
+	return graph_id, sum_of_distance, node_maps_from_median
+
+
+def _update_node_maps_parallel(median_id, epsilon, itr):
+	graph_id = itr[0]
+	node_map = itr[1]
+
+	node_maps_were_modified = False
+	G_ged_env.run_method(median_id, graph_id)
+	if G_ged_env.get_upper_bound(median_id, graph_id) < node_map.induced_cost() - epsilon:
+		node_map = G_ged_env.get_node_map(median_id, graph_id)
+		node_maps_were_modified = True
+	return graph_id, node_map, node_maps_were_modified
\ No newline at end of file
diff --git a/gklearn/ged/median/test_median_graph_estimator.py b/gklearn/ged/median/test_median_graph_estimator.py
index 7497bab..60bce83 100644
--- a/gklearn/ged/median/test_median_graph_estimator.py
+++ b/gklearn/ged/median/test_median_graph_estimator.py
@@ -53,7 +53,7 @@ def test_median_graph_estimator():
 	mge.set_refine_method(algo, '--threads ' + str(threads) + ' --initial-solutions ' + str(initial_solutions) + ' --ratio-runs-from-initial-solutions 1')
 	
 	mge_options = '--time-limit ' + str(time_limit) + ' --stdout 2 --init-type ' + init_type
-	mge_options += ' --random-inits ' + str(num_inits) + ' --seed ' + '1'  + ' --update-order TRUE --refine FALSE --randomness PSEUDO '# @todo: std::to_string(rng())
+	mge_options += ' --random-inits ' + str(num_inits) + ' --seed ' + '1'  + ' --update-order TRUE --refine FALSE --randomness PSEUDO  --parallel TRUE '# @todo: std::to_string(rng())
 	
 	# Select the GED algorithm.
 	algo_options = '--threads ' + str(threads) + algo_options_suffix
@@ -127,7 +127,7 @@ def test_median_graph_estimator_symb():
 	mge.set_refine_method(algo, '--threads ' + str(threads) + ' --initial-solutions ' + str(initial_solutions) + ' --ratio-runs-from-initial-solutions 1')
 	
 	mge_options = '--time-limit ' + str(time_limit) + ' --stdout 2 --init-type ' + init_type
-	mge_options += ' --random-inits ' + str(num_inits) + ' --seed ' + '1'  + ' --update-order TRUE --refine FALSE'# @todo: std::to_string(rng())
+	mge_options += ' --random-inits ' + str(num_inits) + ' --seed ' + '1'  + ' --update-order TRUE --refine FALSE --randomness PSEUDO --parallel TRUE '# @todo: std::to_string(rng())
 	
 	# Select the GED algorithm.
 	algo_options = '--threads ' + str(threads) + algo_options_suffix
@@ -155,5 +155,5 @@ def test_median_graph_estimator_symb():
 
 
 if __name__ == '__main__':
- 	set_median, gen_median = test_median_graph_estimator()
- 	# set_median, gen_median = test_median_graph_estimator_symb()
\ No newline at end of file
+ 	# set_median, gen_median = test_median_graph_estimator()
+ 	set_median, gen_median = test_median_graph_estimator_symb()
\ No newline at end of file
diff --git a/gklearn/ged/median/utils.py b/gklearn/ged/median/utils.py
index 908cb11..5c4c52f 100644
--- a/gklearn/ged/median/utils.py
+++ b/gklearn/ged/median/utils.py
@@ -30,6 +30,8 @@ def mge_options_to_string(options):
 			opt_str += '--randomness ' + str(val) + ' '
 		elif key == 'verbose':
 			opt_str += '--stdout ' + str(val) + ' '
+		elif key == 'parallel':
+			opt_str += '--parallel ' + ('TRUE' if val else 'FALSE') + ' '
 		elif key == 'update_order':
 			opt_str += '--update-order ' + ('TRUE' if val else 'FALSE') + ' '
 		elif key == 'refine':
diff --git a/gklearn/ged/util/util.py b/gklearn/ged/util/util.py
index a18b0cb..d72c2e6 100644
--- a/gklearn/ged/util/util.py
+++ b/gklearn/ged/util/util.py
@@ -54,6 +54,8 @@ def compute_geds(graphs, options={}, parallel=False):
 		ged_env.add_nx_graph(g, '')
 	listID = ged_env.get_all_graph_ids()	
 	ged_env.init()
+	if parallel:
+		options['threads'] = 1
 	ged_env.set_method(options['method'], ged_options_to_string(options))
 	ged_env.init_method()
 
diff --git a/gklearn/gedlib/gedlibpy.cpp b/gklearn/gedlib/gedlibpy.cpp
index 58aa1fd..18e7cd8 100644
--- a/gklearn/gedlib/gedlibpy.cpp
+++ b/gklearn/gedlib/gedlibpy.cpp
@@ -1173,9 +1173,9 @@ typedef npy_clongdouble __pyx_t_5numpy_clongdouble_t;
  */
 typedef npy_cdouble __pyx_t_5numpy_complex_t;
 
-/* "gedlibpy.pyx":180
- * 
+/* "gedlibpy.pyx":182
  * 
+ * # @cython.auto_pickle(True)
  * cdef class GEDEnv:             # <<<<<<<<<<<<<<
  * 	"""Cython wrapper class for C++ class PyGEDEnv
  * 	"""
@@ -2155,6 +2155,8 @@ static CYTHON_INLINE char *__pyx_f_5numpy__util_dtypestring(PyArray_Descr *, cha
 static PyTypeObject *__pyx_ptype_7cpython_5array_array = 0;
 static CYTHON_INLINE int __pyx_f_7cpython_5array_extend_buffer(arrayobject *, char *, Py_ssize_t); /*proto*/
 
+/* Module declarations from 'cython' */
+
 /* Module declarations from 'gedlibpy' */
 static PyTypeObject *__pyx_ptype_8gedlibpy_GEDEnv = 0;
 static CYTHON_INLINE PyObject *__pyx_convert_PyObject_string_to_py_std__in_string(std::string const &); /*proto*/
@@ -2685,7 +2687,7 @@ static PyObject *__pyx_codeobj__37;
 static PyObject *__pyx_codeobj__39;
 /* Late includes */
 
-/* "gedlibpy.pyx":128
+/* "gedlibpy.pyx":129
  * 
  * 
  * def get_edit_cost_options() :             # <<<<<<<<<<<<<<
@@ -2720,7 +2722,7 @@ static PyObject *__pyx_pf_8gedlibpy_get_edit_cost_options(CYTHON_UNUSED PyObject
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_edit_cost_options", 0);
 
-  /* "gedlibpy.pyx":139
+  /* "gedlibpy.pyx":140
  * 	"""
  * 
  * 	return [option.decode('utf-8') for option in getEditCostStringOptions()]             # <<<<<<<<<<<<<<
@@ -2729,13 +2731,13 @@ static PyObject *__pyx_pf_8gedlibpy_get_edit_cost_options(CYTHON_UNUSED PyObject
  */
   __Pyx_XDECREF(__pyx_r);
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 139, __pyx_L1_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 140, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     try {
       __pyx_t_2 = pyged::getEditCostStringOptions();
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 139, __pyx_L1_error)
+      __PYX_ERR(0, 140, __pyx_L1_error)
     }
     __pyx_t_4 = &__pyx_t_2;
     __pyx_t_3 = __pyx_t_4->begin();
@@ -2744,9 +2746,9 @@ static PyObject *__pyx_pf_8gedlibpy_get_edit_cost_options(CYTHON_UNUSED PyObject
       __pyx_t_5 = *__pyx_t_3;
       ++__pyx_t_3;
       __pyx_7genexpr__pyx_v_option = __pyx_t_5;
-      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_7genexpr__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 139, __pyx_L1_error)
+      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_7genexpr__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 140, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 139, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 140, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     }
   } /* exit inner scope */
@@ -2754,7 +2756,7 @@ static PyObject *__pyx_pf_8gedlibpy_get_edit_cost_options(CYTHON_UNUSED PyObject
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":128
+  /* "gedlibpy.pyx":129
  * 
  * 
  * def get_edit_cost_options() :             # <<<<<<<<<<<<<<
@@ -2774,7 +2776,7 @@ static PyObject *__pyx_pf_8gedlibpy_get_edit_cost_options(CYTHON_UNUSED PyObject
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":142
+/* "gedlibpy.pyx":143
  * 
  * 
  * def get_method_options() :             # <<<<<<<<<<<<<<
@@ -2809,7 +2811,7 @@ static PyObject *__pyx_pf_8gedlibpy_2get_method_options(CYTHON_UNUSED PyObject *
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_method_options", 0);
 
-  /* "gedlibpy.pyx":152
+  /* "gedlibpy.pyx":153
  * 		.. note:: Prefer the list_of_method_options attribute of this module.
  * 	"""
  * 	return [option.decode('utf-8') for option in getMethodStringOptions()]             # <<<<<<<<<<<<<<
@@ -2818,13 +2820,13 @@ static PyObject *__pyx_pf_8gedlibpy_2get_method_options(CYTHON_UNUSED PyObject *
  */
   __Pyx_XDECREF(__pyx_r);
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 152, __pyx_L1_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 153, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     try {
       __pyx_t_2 = pyged::getMethodStringOptions();
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 152, __pyx_L1_error)
+      __PYX_ERR(0, 153, __pyx_L1_error)
     }
     __pyx_t_4 = &__pyx_t_2;
     __pyx_t_3 = __pyx_t_4->begin();
@@ -2833,9 +2835,9 @@ static PyObject *__pyx_pf_8gedlibpy_2get_method_options(CYTHON_UNUSED PyObject *
       __pyx_t_5 = *__pyx_t_3;
       ++__pyx_t_3;
       __pyx_8genexpr1__pyx_v_option = __pyx_t_5;
-      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_8genexpr1__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 152, __pyx_L1_error)
+      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_8genexpr1__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 153, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 152, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 153, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     }
   } /* exit inner scope */
@@ -2843,7 +2845,7 @@ static PyObject *__pyx_pf_8gedlibpy_2get_method_options(CYTHON_UNUSED PyObject *
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":142
+  /* "gedlibpy.pyx":143
  * 
  * 
  * def get_method_options() :             # <<<<<<<<<<<<<<
@@ -2863,7 +2865,7 @@ static PyObject *__pyx_pf_8gedlibpy_2get_method_options(CYTHON_UNUSED PyObject *
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":155
+/* "gedlibpy.pyx":156
  * 
  * 
  * def get_init_options() :             # <<<<<<<<<<<<<<
@@ -2898,7 +2900,7 @@ static PyObject *__pyx_pf_8gedlibpy_4get_init_options(CYTHON_UNUSED PyObject *__
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_init_options", 0);
 
-  /* "gedlibpy.pyx":165
+  /* "gedlibpy.pyx":166
  * 		.. note:: Prefer the list_of_init_options attribute of this module.
  * 	"""
  * 	return [option.decode('utf-8') for option in getInitStringOptions()]             # <<<<<<<<<<<<<<
@@ -2907,13 +2909,13 @@ static PyObject *__pyx_pf_8gedlibpy_4get_init_options(CYTHON_UNUSED PyObject *__
  */
   __Pyx_XDECREF(__pyx_r);
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 165, __pyx_L1_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 166, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     try {
       __pyx_t_2 = pyged::getInitStringOptions();
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 165, __pyx_L1_error)
+      __PYX_ERR(0, 166, __pyx_L1_error)
     }
     __pyx_t_4 = &__pyx_t_2;
     __pyx_t_3 = __pyx_t_4->begin();
@@ -2922,9 +2924,9 @@ static PyObject *__pyx_pf_8gedlibpy_4get_init_options(CYTHON_UNUSED PyObject *__
       __pyx_t_5 = *__pyx_t_3;
       ++__pyx_t_3;
       __pyx_8genexpr2__pyx_v_option = __pyx_t_5;
-      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_8genexpr2__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 165, __pyx_L1_error)
+      __pyx_t_6 = __Pyx_decode_cpp_string(__pyx_8genexpr2__pyx_v_option, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 166, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 165, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_6))) __PYX_ERR(0, 166, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     }
   } /* exit inner scope */
@@ -2932,7 +2934,7 @@ static PyObject *__pyx_pf_8gedlibpy_4get_init_options(CYTHON_UNUSED PyObject *__
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":155
+  /* "gedlibpy.pyx":156
  * 
  * 
  * def get_init_options() :             # <<<<<<<<<<<<<<
@@ -2952,7 +2954,7 @@ static PyObject *__pyx_pf_8gedlibpy_4get_init_options(CYTHON_UNUSED PyObject *__
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":168
+/* "gedlibpy.pyx":169
  * 
  * 
  * def get_dummy_node() :             # <<<<<<<<<<<<<<
@@ -2982,7 +2984,7 @@ static PyObject *__pyx_pf_8gedlibpy_6get_dummy_node(CYTHON_UNUSED PyObject *__py
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_dummy_node", 0);
 
-  /* "gedlibpy.pyx":177
+  /* "gedlibpy.pyx":178
  * 		.. note:: A dummy node is used when a node isn't associated to an other node.
  * 	"""
  * 	return getDummyNode()             # <<<<<<<<<<<<<<
@@ -2994,15 +2996,15 @@ static PyObject *__pyx_pf_8gedlibpy_6get_dummy_node(CYTHON_UNUSED PyObject *__py
     __pyx_t_1 = pyged::getDummyNode();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 177, __pyx_L1_error)
+    __PYX_ERR(0, 178, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 177, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 178, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":168
+  /* "gedlibpy.pyx":169
  * 
  * 
  * def get_dummy_node() :             # <<<<<<<<<<<<<<
@@ -3021,12 +3023,12 @@ static PyObject *__pyx_pf_8gedlibpy_6get_dummy_node(CYTHON_UNUSED PyObject *__py
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":187
+/* "gedlibpy.pyx":189
  * 
  * 
  * 	def __cinit__(self):             # <<<<<<<<<<<<<<
+ * # 		self.c_env = PyGEDEnv()
  * 		self.c_env = new PyGEDEnv()
- * 
  */
 
 /* Python wrapper */
@@ -3051,9 +3053,9 @@ static int __pyx_pf_8gedlibpy_6GEDEnv___cinit__(struct __pyx_obj_8gedlibpy_GEDEn
   pyged::PyGEDEnv *__pyx_t_1;
   __Pyx_RefNannySetupContext("__cinit__", 0);
 
-  /* "gedlibpy.pyx":188
- * 
+  /* "gedlibpy.pyx":191
  * 	def __cinit__(self):
+ * # 		self.c_env = PyGEDEnv()
  * 		self.c_env = new PyGEDEnv()             # <<<<<<<<<<<<<<
  * 
  * 
@@ -3062,16 +3064,16 @@ static int __pyx_pf_8gedlibpy_6GEDEnv___cinit__(struct __pyx_obj_8gedlibpy_GEDEn
     __pyx_t_1 = new pyged::PyGEDEnv();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 188, __pyx_L1_error)
+    __PYX_ERR(0, 191, __pyx_L1_error)
   }
   __pyx_v_self->c_env = __pyx_t_1;
 
-  /* "gedlibpy.pyx":187
+  /* "gedlibpy.pyx":189
  * 
  * 
  * 	def __cinit__(self):             # <<<<<<<<<<<<<<
+ * # 		self.c_env = PyGEDEnv()
  * 		self.c_env = new PyGEDEnv()
- * 
  */
 
   /* function exit code */
@@ -3085,7 +3087,7 @@ static int __pyx_pf_8gedlibpy_6GEDEnv___cinit__(struct __pyx_obj_8gedlibpy_GEDEn
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":191
+/* "gedlibpy.pyx":194
  * 
  * 
  * 	def __dealloc__(self):             # <<<<<<<<<<<<<<
@@ -3108,7 +3110,7 @@ static void __pyx_pf_8gedlibpy_6GEDEnv_2__dealloc__(struct __pyx_obj_8gedlibpy_G
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__dealloc__", 0);
 
-  /* "gedlibpy.pyx":192
+  /* "gedlibpy.pyx":195
  * 
  * 	def __dealloc__(self):
  * 		del self.c_env             # <<<<<<<<<<<<<<
@@ -3117,7 +3119,7 @@ static void __pyx_pf_8gedlibpy_6GEDEnv_2__dealloc__(struct __pyx_obj_8gedlibpy_G
  */
   delete __pyx_v_self->c_env;
 
-  /* "gedlibpy.pyx":191
+  /* "gedlibpy.pyx":194
  * 
  * 
  * 	def __dealloc__(self):             # <<<<<<<<<<<<<<
@@ -3129,7 +3131,7 @@ static void __pyx_pf_8gedlibpy_6GEDEnv_2__dealloc__(struct __pyx_obj_8gedlibpy_G
   __Pyx_RefNannyFinishContext();
 }
 
-/* "gedlibpy.pyx":195
+/* "gedlibpy.pyx":203
  * 
  * 
  * 	def is_initialized(self) :             # <<<<<<<<<<<<<<
@@ -3158,7 +3160,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_4is_initialized(struct __pyx_obj_8ge
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("is_initialized", 0);
 
-  /* "gedlibpy.pyx":204
+  /* "gedlibpy.pyx":212
  * 			.. note:: This function exists for internals verifications but you can use it for your code.
  * 		"""
  * 		return self.c_env.isInitialized()             # <<<<<<<<<<<<<<
@@ -3170,15 +3172,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_4is_initialized(struct __pyx_obj_8ge
     __pyx_t_1 = __pyx_v_self->c_env->isInitialized();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 204, __pyx_L1_error)
+    __PYX_ERR(0, 212, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_PyBool_FromLong(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 204, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyBool_FromLong(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 212, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":195
+  /* "gedlibpy.pyx":203
  * 
  * 
  * 	def is_initialized(self) :             # <<<<<<<<<<<<<<
@@ -3197,7 +3199,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_4is_initialized(struct __pyx_obj_8ge
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":207
+/* "gedlibpy.pyx":215
  * 
  * 
  * 	def restart_env(self) :             # <<<<<<<<<<<<<<
@@ -3224,7 +3226,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_6restart_env(struct __pyx_obj_8gedli
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("restart_env", 0);
 
-  /* "gedlibpy.pyx":214
+  /* "gedlibpy.pyx":222
  * 			.. note:: You can now delete and add somes graphs after initialization so you can avoid this function.
  * 		"""
  * 		self.c_env.restartEnv()             # <<<<<<<<<<<<<<
@@ -3235,10 +3237,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_6restart_env(struct __pyx_obj_8gedli
     __pyx_v_self->c_env->restartEnv();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 214, __pyx_L1_error)
+    __PYX_ERR(0, 222, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":207
+  /* "gedlibpy.pyx":215
  * 
  * 
  * 	def restart_env(self) :             # <<<<<<<<<<<<<<
@@ -3258,7 +3260,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_6restart_env(struct __pyx_obj_8gedli
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":217
+/* "gedlibpy.pyx":225
  * 
  * 
  * 	def load_GXL_graphs(self, path_folder, path_XML, node_type, edge_type) :             # <<<<<<<<<<<<<<
@@ -3304,23 +3306,23 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_9load_GXL_graphs(PyObject *__pyx_v_s
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_path_XML)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 1); __PYX_ERR(0, 217, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 1); __PYX_ERR(0, 225, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_type)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 2); __PYX_ERR(0, 217, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 2); __PYX_ERR(0, 225, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edge_type)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 3); __PYX_ERR(0, 217, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, 3); __PYX_ERR(0, 225, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "load_GXL_graphs") < 0)) __PYX_ERR(0, 217, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "load_GXL_graphs") < 0)) __PYX_ERR(0, 225, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
       goto __pyx_L5_argtuple_error;
@@ -3337,7 +3339,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_9load_GXL_graphs(PyObject *__pyx_v_s
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 217, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("load_GXL_graphs", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 225, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.load_GXL_graphs", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -3362,14 +3364,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_8load_GXL_graphs(struct __pyx_obj_8g
   bool __pyx_t_7;
   __Pyx_RefNannySetupContext("load_GXL_graphs", 0);
 
-  /* "gedlibpy.pyx":233
+  /* "gedlibpy.pyx":241
  * 			.. note:: You can call this function multiple times if you want, but not after an init call.
  * 		"""
  * 		self.c_env.loadGXLGraph(path_folder.encode('utf-8'), path_XML.encode('utf-8'), node_type, edge_type)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_path_folder, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 233, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_path_folder, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -3383,12 +3385,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_8load_GXL_graphs(struct __pyx_obj_8g
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 233, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 233, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_path_XML, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 233, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_path_XML, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -3402,21 +3404,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_8load_GXL_graphs(struct __pyx_obj_8g
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 233, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 233, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 241, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_v_node_type); if (unlikely((__pyx_t_6 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 233, __pyx_L1_error)
-  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_edge_type); if (unlikely((__pyx_t_7 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 233, __pyx_L1_error)
+  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_v_node_type); if (unlikely((__pyx_t_6 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 241, __pyx_L1_error)
+  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_edge_type); if (unlikely((__pyx_t_7 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 241, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->loadGXLGraph(__pyx_t_4, __pyx_t_5, __pyx_t_6, __pyx_t_7);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 233, __pyx_L1_error)
+    __PYX_ERR(0, 241, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":217
+  /* "gedlibpy.pyx":225
  * 
  * 
  * 	def load_GXL_graphs(self, path_folder, path_XML, node_type, edge_type) :             # <<<<<<<<<<<<<<
@@ -3439,7 +3441,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_8load_GXL_graphs(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":236
+/* "gedlibpy.pyx":244
  * 
  * 
  * 	def graph_ids(self) :             # <<<<<<<<<<<<<<
@@ -3468,7 +3470,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_10graph_ids(struct __pyx_obj_8gedlib
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("graph_ids", 0);
 
-  /* "gedlibpy.pyx":245
+  /* "gedlibpy.pyx":253
  * 			.. note:: Prefer this function if you have huges structures with lots of graphs.
  * 		"""
  * 		return self.c_env.getGraphIds()             # <<<<<<<<<<<<<<
@@ -3480,15 +3482,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_10graph_ids(struct __pyx_obj_8gedlib
     __pyx_t_1 = __pyx_v_self->c_env->getGraphIds();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 245, __pyx_L1_error)
+    __PYX_ERR(0, 253, __pyx_L1_error)
   }
-  __pyx_t_2 = __pyx_convert_pair_to_py_size_t____size_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 245, __pyx_L1_error)
+  __pyx_t_2 = __pyx_convert_pair_to_py_size_t____size_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 253, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":236
+  /* "gedlibpy.pyx":244
  * 
  * 
  * 	def graph_ids(self) :             # <<<<<<<<<<<<<<
@@ -3507,7 +3509,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_10graph_ids(struct __pyx_obj_8gedlib
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":248
+/* "gedlibpy.pyx":256
  * 
  * 
  * 	def get_all_graph_ids(self) :             # <<<<<<<<<<<<<<
@@ -3536,7 +3538,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_12get_all_graph_ids(struct __pyx_obj
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_all_graph_ids", 0);
 
-  /* "gedlibpy.pyx":257
+  /* "gedlibpy.pyx":265
  * 			.. note:: The last ID is equal to (number of graphs - 1). The order correspond to the loading order.
  * 		"""
  * 		return self.c_env.getAllGraphIds()             # <<<<<<<<<<<<<<
@@ -3548,15 +3550,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_12get_all_graph_ids(struct __pyx_obj
     __pyx_t_1 = __pyx_v_self->c_env->getAllGraphIds();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 257, __pyx_L1_error)
+    __PYX_ERR(0, 265, __pyx_L1_error)
   }
-  __pyx_t_2 = __pyx_convert_vector_to_py_size_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 257, __pyx_L1_error)
+  __pyx_t_2 = __pyx_convert_vector_to_py_size_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 265, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":248
+  /* "gedlibpy.pyx":256
  * 
  * 
  * 	def get_all_graph_ids(self) :             # <<<<<<<<<<<<<<
@@ -3575,7 +3577,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_12get_all_graph_ids(struct __pyx_obj
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":260
+/* "gedlibpy.pyx":268
  * 
  * 
  * 	def get_graph_class(self, id) :             # <<<<<<<<<<<<<<
@@ -3605,7 +3607,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_14get_graph_class(struct __pyx_obj_8
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_class", 0);
 
-  /* "gedlibpy.pyx":272
+  /* "gedlibpy.pyx":280
  * 			.. note:: An empty string can be a class.
  * 		"""
  * 		return self.c_env.getGraphClass(id)             # <<<<<<<<<<<<<<
@@ -3613,20 +3615,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_14get_graph_class(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 272, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 280, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphClass(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 272, __pyx_L1_error)
+    __PYX_ERR(0, 280, __pyx_L1_error)
   }
-  __pyx_t_3 = __pyx_convert_PyBytes_string_to_py_std__in_string(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 272, __pyx_L1_error)
+  __pyx_t_3 = __pyx_convert_PyBytes_string_to_py_std__in_string(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 280, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":260
+  /* "gedlibpy.pyx":268
  * 
  * 
  * 	def get_graph_class(self, id) :             # <<<<<<<<<<<<<<
@@ -3645,7 +3647,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_14get_graph_class(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":275
+/* "gedlibpy.pyx":283
  * 
  * 
  * 	def get_graph_name(self, id) :             # <<<<<<<<<<<<<<
@@ -3675,7 +3677,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_16get_graph_name(struct __pyx_obj_8g
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_name", 0);
 
-  /* "gedlibpy.pyx":287
+  /* "gedlibpy.pyx":295
  * 			.. note:: An empty string can be a name.
  * 		"""
  * 		return self.c_env.getGraphName(id).decode('utf-8')             # <<<<<<<<<<<<<<
@@ -3683,20 +3685,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_16get_graph_name(struct __pyx_obj_8g
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 287, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 295, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphName(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 287, __pyx_L1_error)
+    __PYX_ERR(0, 295, __pyx_L1_error)
   }
-  __pyx_t_3 = __Pyx_decode_cpp_string(__pyx_t_2, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 287, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_decode_cpp_string(__pyx_t_2, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 295, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":275
+  /* "gedlibpy.pyx":283
  * 
  * 
  * 	def get_graph_name(self, id) :             # <<<<<<<<<<<<<<
@@ -3715,7 +3717,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_16get_graph_name(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":290
+/* "gedlibpy.pyx":298
  * 
  * 
  * 	def add_graph(self, name="", classe="") :             # <<<<<<<<<<<<<<
@@ -3763,7 +3765,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_19add_graph(PyObject *__pyx_v_self,
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_graph") < 0)) __PYX_ERR(0, 290, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_graph") < 0)) __PYX_ERR(0, 298, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -3780,7 +3782,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_19add_graph(PyObject *__pyx_v_self,
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_graph", 0, 0, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 290, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_graph", 0, 0, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 298, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_graph", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -3804,7 +3806,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_18add_graph(struct __pyx_obj_8gedlib
   size_t __pyx_t_6;
   __Pyx_RefNannySetupContext("add_graph", 0);
 
-  /* "gedlibpy.pyx":304
+  /* "gedlibpy.pyx":312
  * 			.. note:: You can call this function without parameters. You can also use this function after initialization, call init() after you're finished your modifications.
  * 		"""
  * 		return self.c_env.addGraph(name.encode('utf-8'), classe.encode('utf-8'))             # <<<<<<<<<<<<<<
@@ -3812,7 +3814,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_18add_graph(struct __pyx_obj_8gedlib
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_name, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 304, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_name, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -3826,12 +3828,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_18add_graph(struct __pyx_obj_8gedlib
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 304, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 304, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_classe, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 304, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_classe, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -3845,24 +3847,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_18add_graph(struct __pyx_obj_8gedlib
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 304, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 304, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_6 = __pyx_v_self->c_env->addGraph(__pyx_t_4, __pyx_t_5);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 304, __pyx_L1_error)
+    __PYX_ERR(0, 312, __pyx_L1_error)
   }
-  __pyx_t_1 = __Pyx_PyInt_FromSize_t(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 304, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_FromSize_t(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 312, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":290
+  /* "gedlibpy.pyx":298
  * 
  * 
  * 	def add_graph(self, name="", classe="") :             # <<<<<<<<<<<<<<
@@ -3883,7 +3885,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_18add_graph(struct __pyx_obj_8gedlib
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":307
+/* "gedlibpy.pyx":315
  * 
  * 
  * 	def add_node(self, graph_id, node_id, node_label):             # <<<<<<<<<<<<<<
@@ -3926,17 +3928,17 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_21add_node(PyObject *__pyx_v_self, P
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_id)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, 1); __PYX_ERR(0, 307, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, 1); __PYX_ERR(0, 315, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_label)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, 2); __PYX_ERR(0, 307, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, 2); __PYX_ERR(0, 315, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_node") < 0)) __PYX_ERR(0, 307, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_node") < 0)) __PYX_ERR(0, 315, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
       goto __pyx_L5_argtuple_error;
@@ -3951,7 +3953,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_21add_node(PyObject *__pyx_v_self, P
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 307, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_node", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 315, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_node", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -3975,15 +3977,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_20add_node(struct __pyx_obj_8gedlibp
   std::map<std::string,std::string>  __pyx_t_6;
   __Pyx_RefNannySetupContext("add_node", 0);
 
-  /* "gedlibpy.pyx":321
+  /* "gedlibpy.pyx":329
  * 			.. note:: You can also use this function after initialization, but only on a newly added graph. Call init() after you're finished your modifications.
  * 		"""
  * 		self.c_env.addNode(graph_id, node_id.encode('utf-8'), encode_your_map(node_label))             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 321, __pyx_L1_error)
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_id, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 321, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 329, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_id, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -3997,12 +3999,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_20add_node(struct __pyx_obj_8gedlibp
   }
   __pyx_t_2 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 321, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 321, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 321, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
@@ -4016,19 +4018,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_20add_node(struct __pyx_obj_8gedlibp
   }
   __pyx_t_2 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_node_label) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_node_label);
   __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 321, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_6 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 321, __pyx_L1_error)
+  __pyx_t_6 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 329, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   try {
     __pyx_v_self->c_env->addNode(__pyx_t_1, __pyx_t_5, __pyx_t_6);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 321, __pyx_L1_error)
+    __PYX_ERR(0, 329, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":307
+  /* "gedlibpy.pyx":315
  * 
  * 
  * 	def add_node(self, graph_id, node_id, node_label):             # <<<<<<<<<<<<<<
@@ -4051,7 +4053,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_20add_node(struct __pyx_obj_8gedlibp
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":324
+/* "gedlibpy.pyx":332
  * 
  * 
  * 	def add_edge(self, graph_id, tail, head, edge_label, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -4101,19 +4103,19 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_23add_edge(PyObject *__pyx_v_self, P
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tail)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 1); __PYX_ERR(0, 324, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 1); __PYX_ERR(0, 332, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_head)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 2); __PYX_ERR(0, 324, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 2); __PYX_ERR(0, 332, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edge_label)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 3); __PYX_ERR(0, 324, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, 3); __PYX_ERR(0, 332, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  4:
@@ -4123,7 +4125,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_23add_edge(PyObject *__pyx_v_self, P
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_edge") < 0)) __PYX_ERR(0, 324, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_edge") < 0)) __PYX_ERR(0, 332, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -4145,7 +4147,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_23add_edge(PyObject *__pyx_v_self, P
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 324, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_edge", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 332, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_edge", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -4171,15 +4173,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_22add_edge(struct __pyx_obj_8gedlibp
   bool __pyx_t_8;
   __Pyx_RefNannySetupContext("add_edge", 0);
 
-  /* "gedlibpy.pyx":342
+  /* "gedlibpy.pyx":350
  * 			.. note:: You can also use this function after initialization, but only on a newly added graph. Call init() after you're finished your modifications.
  * 		"""
  * 		self.c_env.addEdge(graph_id, tail.encode('utf-8'), head.encode('utf-8'), encode_your_map(edge_label), ignore_duplicates)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 342, __pyx_L1_error)
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tail, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 350, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tail, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -4193,12 +4195,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_22add_edge(struct __pyx_obj_8gedlibp
   }
   __pyx_t_2 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 342, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_head, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_head, __pyx_n_s_encode); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -4212,12 +4214,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_22add_edge(struct __pyx_obj_8gedlibp
   }
   __pyx_t_2 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 342, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
@@ -4231,20 +4233,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_22add_edge(struct __pyx_obj_8gedlibp
   }
   __pyx_t_2 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_edge_label) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_edge_label);
   __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 342, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_2); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 350, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_v_ignore_duplicates); if (unlikely((__pyx_t_8 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 342, __pyx_L1_error)
+  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_v_ignore_duplicates); if (unlikely((__pyx_t_8 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 350, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->addEdge(__pyx_t_1, __pyx_t_5, __pyx_t_6, __pyx_t_7, __pyx_t_8);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 342, __pyx_L1_error)
+    __PYX_ERR(0, 350, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":324
+  /* "gedlibpy.pyx":332
  * 
  * 
  * 	def add_edge(self, graph_id, tail, head, edge_label, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -4267,7 +4269,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_22add_edge(struct __pyx_obj_8gedlibp
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":345
+/* "gedlibpy.pyx":353
  * 
  * 
  * 	def add_symmetrical_edge(self, graph_id, tail, head, edge_label) :             # <<<<<<<<<<<<<<
@@ -4313,23 +4315,23 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_25add_symmetrical_edge(PyObject *__p
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tail)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 1); __PYX_ERR(0, 345, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 1); __PYX_ERR(0, 353, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_head)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 2); __PYX_ERR(0, 345, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 2); __PYX_ERR(0, 353, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edge_label)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 3); __PYX_ERR(0, 345, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, 3); __PYX_ERR(0, 353, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_symmetrical_edge") < 0)) __PYX_ERR(0, 345, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_symmetrical_edge") < 0)) __PYX_ERR(0, 353, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
       goto __pyx_L5_argtuple_error;
@@ -4346,7 +4348,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_25add_symmetrical_edge(PyObject *__p
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 345, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_symmetrical_edge", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 353, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_symmetrical_edge", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -4374,14 +4376,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_24add_symmetrical_edge(struct __pyx_
   std::map<std::string,std::string>  __pyx_t_7;
   __Pyx_RefNannySetupContext("add_symmetrical_edge", 0);
 
-  /* "gedlibpy.pyx":361
+  /* "gedlibpy.pyx":369
  * 			.. note:: You can also use this function after initialization, but only on a newly added graph. Call init() after you're finished your modifications.
  * 		"""
  * 		tailB = tail.encode('utf-8')             # <<<<<<<<<<<<<<
  * 		headB = head.encode('utf-8')
  * 		edgeLabelB = encode_your_map(edge_label)
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tail, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 361, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tail, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 369, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -4395,20 +4397,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_24add_symmetrical_edge(struct __pyx_
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 369, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_tailB = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":362
+  /* "gedlibpy.pyx":370
  * 		"""
  * 		tailB = tail.encode('utf-8')
  * 		headB = head.encode('utf-8')             # <<<<<<<<<<<<<<
  * 		edgeLabelB = encode_your_map(edge_label)
  * 		self.c_env.addEdge(graph_id, tailB, headB, edgeLabelB, True)
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_head, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_head, __pyx_n_s_encode); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 370, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -4422,20 +4424,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_24add_symmetrical_edge(struct __pyx_
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_kp_u_utf_8);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 362, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 370, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_headB = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":363
+  /* "gedlibpy.pyx":371
  * 		tailB = tail.encode('utf-8')
  * 		headB = head.encode('utf-8')
  * 		edgeLabelB = encode_your_map(edge_label)             # <<<<<<<<<<<<<<
  * 		self.c_env.addEdge(graph_id, tailB, headB, edgeLabelB, True)
  * 		self.c_env.addEdge(graph_id, headB, tailB, edgeLabelB, True)
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 363, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 371, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -4449,49 +4451,49 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_24add_symmetrical_edge(struct __pyx_
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edge_label) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edge_label);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 363, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 371, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_edgeLabelB = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":364
+  /* "gedlibpy.pyx":372
  * 		headB = head.encode('utf-8')
  * 		edgeLabelB = encode_your_map(edge_label)
  * 		self.c_env.addEdge(graph_id, tailB, headB, edgeLabelB, True)             # <<<<<<<<<<<<<<
  * 		self.c_env.addEdge(graph_id, headB, tailB, edgeLabelB, True)
  * 
  */
-  __pyx_t_4 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_4 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 364, __pyx_L1_error)
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_v_tailB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 364, __pyx_L1_error)
-  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_headB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 364, __pyx_L1_error)
-  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_v_edgeLabelB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 364, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_4 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 372, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_v_tailB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 372, __pyx_L1_error)
+  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_headB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 372, __pyx_L1_error)
+  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_v_edgeLabelB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 372, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->addEdge(__pyx_t_4, __pyx_t_5, __pyx_t_6, __pyx_t_7, 1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 364, __pyx_L1_error)
+    __PYX_ERR(0, 372, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":365
+  /* "gedlibpy.pyx":373
  * 		edgeLabelB = encode_your_map(edge_label)
  * 		self.c_env.addEdge(graph_id, tailB, headB, edgeLabelB, True)
  * 		self.c_env.addEdge(graph_id, headB, tailB, edgeLabelB, True)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_4 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_4 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 365, __pyx_L1_error)
-  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_headB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 365, __pyx_L1_error)
-  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_v_tailB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 365, __pyx_L1_error)
-  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_v_edgeLabelB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 365, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_4 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 373, __pyx_L1_error)
+  __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_headB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 373, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_string_from_py_std__in_string(__pyx_v_tailB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 373, __pyx_L1_error)
+  __pyx_t_7 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_v_edgeLabelB); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 373, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->addEdge(__pyx_t_4, __pyx_t_6, __pyx_t_5, __pyx_t_7, 1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 365, __pyx_L1_error)
+    __PYX_ERR(0, 373, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":345
+  /* "gedlibpy.pyx":353
  * 
  * 
  * 	def add_symmetrical_edge(self, graph_id, tail, head, edge_label) :             # <<<<<<<<<<<<<<
@@ -4517,7 +4519,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_24add_symmetrical_edge(struct __pyx_
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":368
+/* "gedlibpy.pyx":376
  * 
  * 
  * 	def clear_graph(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4545,22 +4547,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_26clear_graph(struct __pyx_obj_8gedl
   size_t __pyx_t_1;
   __Pyx_RefNannySetupContext("clear_graph", 0);
 
-  /* "gedlibpy.pyx":377
+  /* "gedlibpy.pyx":385
  * 			.. note:: Call init() after you're finished your modifications.
  * 		"""
  * 		self.c_env.clearGraph(graph_id)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 377, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 385, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->clearGraph(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 377, __pyx_L1_error)
+    __PYX_ERR(0, 385, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":368
+  /* "gedlibpy.pyx":376
  * 
  * 
  * 	def clear_graph(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4580,7 +4582,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_26clear_graph(struct __pyx_obj_8gedl
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":380
+/* "gedlibpy.pyx":388
  * 
  * 
  * 	def get_graph_internal_id(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4610,7 +4612,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_28get_graph_internal_id(struct __pyx
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_internal_id", 0);
 
-  /* "gedlibpy.pyx":392
+  /* "gedlibpy.pyx":400
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return self.c_env.getGraphInternalId(graph_id)             # <<<<<<<<<<<<<<
@@ -4618,20 +4620,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_28get_graph_internal_id(struct __pyx
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 392, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 400, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphInternalId(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 392, __pyx_L1_error)
+    __PYX_ERR(0, 400, __pyx_L1_error)
   }
-  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 392, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 400, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":380
+  /* "gedlibpy.pyx":388
  * 
  * 
  * 	def get_graph_internal_id(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4650,7 +4652,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_28get_graph_internal_id(struct __pyx
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":395
+/* "gedlibpy.pyx":403
  * 
  * 
  * 	def get_graph_num_nodes(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4680,7 +4682,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_30get_graph_num_nodes(struct __pyx_o
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_num_nodes", 0);
 
-  /* "gedlibpy.pyx":407
+  /* "gedlibpy.pyx":415
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return self.c_env.getGraphNumNodes(graph_id)             # <<<<<<<<<<<<<<
@@ -4688,20 +4690,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_30get_graph_num_nodes(struct __pyx_o
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 407, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 415, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphNumNodes(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 407, __pyx_L1_error)
+    __PYX_ERR(0, 415, __pyx_L1_error)
   }
-  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 407, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 415, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":395
+  /* "gedlibpy.pyx":403
  * 
  * 
  * 	def get_graph_num_nodes(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4720,7 +4722,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_30get_graph_num_nodes(struct __pyx_o
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":410
+/* "gedlibpy.pyx":418
  * 
  * 
  * 	def get_graph_num_edges(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4750,7 +4752,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_32get_graph_num_edges(struct __pyx_o
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_num_edges", 0);
 
-  /* "gedlibpy.pyx":422
+  /* "gedlibpy.pyx":430
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return self.c_env.getGraphNumEdges(graph_id)             # <<<<<<<<<<<<<<
@@ -4758,20 +4760,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_32get_graph_num_edges(struct __pyx_o
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 422, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 430, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphNumEdges(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 422, __pyx_L1_error)
+    __PYX_ERR(0, 430, __pyx_L1_error)
   }
-  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 422, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_FromSize_t(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 430, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":410
+  /* "gedlibpy.pyx":418
  * 
  * 
  * 	def get_graph_num_edges(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4790,7 +4792,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_32get_graph_num_edges(struct __pyx_o
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":425
+/* "gedlibpy.pyx":433
  * 
  * 
  * 	def get_original_node_ids(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4825,7 +4827,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_34get_original_node_ids(struct __pyx
   PyObject *__pyx_t_7 = NULL;
   __Pyx_RefNannySetupContext("get_original_node_ids", 0);
 
-  /* "gedlibpy.pyx":437
+  /* "gedlibpy.pyx":445
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return [gid.decode('utf-8') for gid in self.c_env.getGraphOriginalNodeIds(graph_id)]             # <<<<<<<<<<<<<<
@@ -4834,14 +4836,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_34get_original_node_ids(struct __pyx
  */
   __Pyx_XDECREF(__pyx_r);
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 437, __pyx_L1_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 445, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 437, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 445, __pyx_L1_error)
     try {
       __pyx_t_3 = __pyx_v_self->c_env->getGraphOriginalNodeIds(__pyx_t_2);
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 437, __pyx_L1_error)
+      __PYX_ERR(0, 445, __pyx_L1_error)
     }
     __pyx_t_5 = &__pyx_t_3;
     __pyx_t_4 = __pyx_t_5->begin();
@@ -4850,9 +4852,9 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_34get_original_node_ids(struct __pyx
       __pyx_t_6 = *__pyx_t_4;
       ++__pyx_t_4;
       __pyx_8genexpr3__pyx_v_gid = __pyx_t_6;
-      __pyx_t_7 = __Pyx_decode_cpp_string(__pyx_8genexpr3__pyx_v_gid, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 437, __pyx_L1_error)
+      __pyx_t_7 = __Pyx_decode_cpp_string(__pyx_8genexpr3__pyx_v_gid, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 445, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_7);
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 437, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 445, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     }
   } /* exit inner scope */
@@ -4860,7 +4862,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_34get_original_node_ids(struct __pyx
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":425
+  /* "gedlibpy.pyx":433
  * 
  * 
  * 	def get_original_node_ids(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4880,7 +4882,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_34get_original_node_ids(struct __pyx
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":440
+/* "gedlibpy.pyx":448
  * 
  * 
  * 	def get_graph_node_labels(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4918,7 +4920,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
   PyObject *__pyx_t_10 = NULL;
   __Pyx_RefNannySetupContext("get_graph_node_labels", 0);
 
-  /* "gedlibpy.pyx":452
+  /* "gedlibpy.pyx":460
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return [decode_your_map(node_label) for node_label in self.c_env.getGraphNodeLabels(graph_id)]             # <<<<<<<<<<<<<<
@@ -4927,14 +4929,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
  */
   __Pyx_XDECREF(__pyx_r);
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 452, __pyx_L1_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 460, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 452, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 460, __pyx_L1_error)
     try {
       __pyx_t_3 = __pyx_v_self->c_env->getGraphNodeLabels(__pyx_t_2);
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 452, __pyx_L1_error)
+      __PYX_ERR(0, 460, __pyx_L1_error)
     }
     __pyx_t_5 = &__pyx_t_3;
     __pyx_t_4 = __pyx_t_5->begin();
@@ -4943,9 +4945,9 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
       __pyx_t_6 = *__pyx_t_4;
       ++__pyx_t_4;
       __pyx_8genexpr4__pyx_v_node_label = __pyx_t_6;
-      __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 452, __pyx_L1_error)
+      __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 460, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_9 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_8genexpr4__pyx_v_node_label); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 452, __pyx_L1_error)
+      __pyx_t_9 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_8genexpr4__pyx_v_node_label); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 460, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_9);
       __pyx_t_10 = NULL;
       if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_8))) {
@@ -4960,10 +4962,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
       __pyx_t_7 = (__pyx_t_10) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_10, __pyx_t_9) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_9);
       __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 452, __pyx_L1_error)
+      if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 460, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_7);
       __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 452, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 460, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     }
   } /* exit inner scope */
@@ -4971,7 +4973,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":440
+  /* "gedlibpy.pyx":448
  * 
  * 
  * 	def get_graph_node_labels(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -4994,7 +4996,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_36get_graph_node_labels(struct __pyx
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":455
+/* "gedlibpy.pyx":463
  * 
  * 
  * 	def get_graph_edges(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -5027,7 +5029,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_38get_graph_edges(struct __pyx_obj_8
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_graph_edges", 0);
 
-  /* "gedlibpy.pyx":467
+  /* "gedlibpy.pyx":475
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return decode_graph_edges(self.c_env.getGraphEdges(graph_id))             # <<<<<<<<<<<<<<
@@ -5035,16 +5037,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_38get_graph_edges(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_graph_edges); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 467, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_graph_edges); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 475, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 467, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 475, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getGraphEdges(__pyx_t_3);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 467, __pyx_L1_error)
+    __PYX_ERR(0, 475, __pyx_L1_error)
   }
-  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e_______std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 467, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e_______std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 475, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -5059,14 +5061,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_38get_graph_edges(struct __pyx_obj_8
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 467, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 475, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":455
+  /* "gedlibpy.pyx":463
  * 
  * 
  * 	def get_graph_edges(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -5088,7 +5090,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_38get_graph_edges(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":470
+/* "gedlibpy.pyx":478
  * 
  * 
  * 	def get_graph_adjacence_matrix(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -5118,7 +5120,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_40get_graph_adjacence_matrix(struct
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("get_graph_adjacence_matrix", 0);
 
-  /* "gedlibpy.pyx":482
+  /* "gedlibpy.pyx":490
  * 			.. note:: These functions allow to collect all the graph's informations.
  * 		"""
  * 		return self.c_env.getGraphAdjacenceMatrix(graph_id)             # <<<<<<<<<<<<<<
@@ -5126,20 +5128,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_40get_graph_adjacence_matrix(struct
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 482, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_graph_id); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 490, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->getGraphAdjacenceMatrix(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 482, __pyx_L1_error)
+    __PYX_ERR(0, 490, __pyx_L1_error)
   }
-  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 482, __pyx_L1_error)
+  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 490, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":470
+  /* "gedlibpy.pyx":478
  * 
  * 
  * 	def get_graph_adjacence_matrix(self, graph_id) :             # <<<<<<<<<<<<<<
@@ -5158,7 +5160,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_40get_graph_adjacence_matrix(struct
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":485
+/* "gedlibpy.pyx":493
  * 
  * 
  * 	def set_edit_cost(self, edit_cost, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
@@ -5203,7 +5205,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_43set_edit_cost(PyObject *__pyx_v_se
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_edit_cost") < 0)) __PYX_ERR(0, 485, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_edit_cost") < 0)) __PYX_ERR(0, 493, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -5219,7 +5221,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_43set_edit_cost(PyObject *__pyx_v_se
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("set_edit_cost", 0, 1, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 485, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("set_edit_cost", 0, 1, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 493, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.set_edit_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -5245,28 +5247,28 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
   std::vector<double>  __pyx_t_7;
   __Pyx_RefNannySetupContext("set_edit_cost", 0);
 
-  /* "gedlibpy.pyx":497
+  /* "gedlibpy.pyx":505
  * 			.. note:: Try to make sure the edit cost function exists with list_of_edit_cost_options, raise an error otherwise.
  * 		"""
  * 		if edit_cost in list_of_edit_cost_options:             # <<<<<<<<<<<<<<
  * 			edit_cost_b = edit_cost.encode('utf-8')
  * 			self.c_env.setEditCost(edit_cost_b, edit_cost_constant)
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_edit_cost_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 497, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_edit_cost_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 505, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_edit_cost, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 497, __pyx_L1_error)
+  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_edit_cost, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 505, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __pyx_t_3 = (__pyx_t_2 != 0);
   if (likely(__pyx_t_3)) {
 
-    /* "gedlibpy.pyx":498
+    /* "gedlibpy.pyx":506
  * 		"""
  * 		if edit_cost in list_of_edit_cost_options:
  * 			edit_cost_b = edit_cost.encode('utf-8')             # <<<<<<<<<<<<<<
  * 			self.c_env.setEditCost(edit_cost_b, edit_cost_constant)
  * 		else:
  */
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_edit_cost, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 498, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_edit_cost, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 506, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
@@ -5280,29 +5282,29 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 498, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 506, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __pyx_v_edit_cost_b = __pyx_t_1;
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":499
+    /* "gedlibpy.pyx":507
  * 		if edit_cost in list_of_edit_cost_options:
  * 			edit_cost_b = edit_cost.encode('utf-8')
  * 			self.c_env.setEditCost(edit_cost_b, edit_cost_constant)             # <<<<<<<<<<<<<<
  * 		else:
  * 			raise EditCostError("This edit cost function doesn't exist, please see list_of_edit_cost_options for selecting a edit cost function")
  */
-    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_edit_cost_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 499, __pyx_L1_error)
-    __pyx_t_7 = __pyx_convert_vector_from_py_double(__pyx_v_edit_cost_constant); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 499, __pyx_L1_error)
+    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_edit_cost_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 507, __pyx_L1_error)
+    __pyx_t_7 = __pyx_convert_vector_from_py_double(__pyx_v_edit_cost_constant); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 507, __pyx_L1_error)
     try {
       __pyx_v_self->c_env->setEditCost(__pyx_t_6, __pyx_t_7);
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 499, __pyx_L1_error)
+      __PYX_ERR(0, 507, __pyx_L1_error)
     }
 
-    /* "gedlibpy.pyx":497
+    /* "gedlibpy.pyx":505
  * 			.. note:: Try to make sure the edit cost function exists with list_of_edit_cost_options, raise an error otherwise.
  * 		"""
  * 		if edit_cost in list_of_edit_cost_options:             # <<<<<<<<<<<<<<
@@ -5312,7 +5314,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
     goto __pyx_L3;
   }
 
-  /* "gedlibpy.pyx":501
+  /* "gedlibpy.pyx":509
  * 			self.c_env.setEditCost(edit_cost_b, edit_cost_constant)
  * 		else:
  * 			raise EditCostError("This edit cost function doesn't exist, please see list_of_edit_cost_options for selecting a edit cost function")             # <<<<<<<<<<<<<<
@@ -5320,7 +5322,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
  * 
  */
   /*else*/ {
-    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_EditCostError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 501, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_EditCostError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 509, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
@@ -5334,16 +5336,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_This_edit_cost_function_doesn_t) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_This_edit_cost_function_doesn_t);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 501, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 509, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_Raise(__pyx_t_1, 0, 0, 0);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __PYX_ERR(0, 501, __pyx_L1_error)
+    __PYX_ERR(0, 509, __pyx_L1_error)
   }
   __pyx_L3:;
 
-  /* "gedlibpy.pyx":485
+  /* "gedlibpy.pyx":493
  * 
  * 
  * 	def set_edit_cost(self, edit_cost, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
@@ -5367,7 +5369,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_42set_edit_cost(struct __pyx_obj_8ge
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":504
+/* "gedlibpy.pyx":512
  * 
  * 
  * 	def set_personal_edit_cost(self, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
@@ -5405,7 +5407,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_45set_personal_edit_cost(PyObject *_
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_personal_edit_cost") < 0)) __PYX_ERR(0, 504, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_personal_edit_cost") < 0)) __PYX_ERR(0, 512, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -5419,7 +5421,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_45set_personal_edit_cost(PyObject *_
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("set_personal_edit_cost", 0, 0, 1, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 504, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("set_personal_edit_cost", 0, 0, 1, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 512, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.set_personal_edit_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -5438,22 +5440,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_44set_personal_edit_cost(struct __py
   std::vector<double>  __pyx_t_1;
   __Pyx_RefNannySetupContext("set_personal_edit_cost", 0);
 
-  /* "gedlibpy.pyx":514
+  /* "gedlibpy.pyx":522
  * 			.. note::You have to modify the C++ function to use it. Please see the documentation to add your Edit Cost function.
  * 		"""
  * 		self.c_env.setPersonalEditCost(edit_cost_constant)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_1 = __pyx_convert_vector_from_py_double(__pyx_v_edit_cost_constant); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 514, __pyx_L1_error)
+  __pyx_t_1 = __pyx_convert_vector_from_py_double(__pyx_v_edit_cost_constant); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 522, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->setPersonalEditCost(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 514, __pyx_L1_error)
+    __PYX_ERR(0, 522, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":504
+  /* "gedlibpy.pyx":512
  * 
  * 
  * 	def set_personal_edit_cost(self, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
@@ -5473,7 +5475,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_44set_personal_edit_cost(struct __py
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":517
+/* "gedlibpy.pyx":525
  * 
  * 
  * 	def init(self, init_option='EAGER_WITHOUT_SHUFFLED_COPIES', print_to_stdout=False) :             # <<<<<<<<<<<<<<
@@ -5521,7 +5523,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_47init(PyObject *__pyx_v_self, PyObj
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "init") < 0)) __PYX_ERR(0, 517, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "init") < 0)) __PYX_ERR(0, 525, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -5538,7 +5540,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_47init(PyObject *__pyx_v_self, PyObj
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("init", 0, 0, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 517, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("init", 0, 0, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 525, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.init", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -5564,28 +5566,28 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
   bool __pyx_t_7;
   __Pyx_RefNannySetupContext("init", 0);
 
-  /* "gedlibpy.pyx":528
+  /* "gedlibpy.pyx":536
  * 			.. note:: Try to make sure the option exists with list_of_init_options or choose no options, raise an error otherwise.
  * 		"""
  * 		if init_option in list_of_init_options:             # <<<<<<<<<<<<<<
  * 			init_option_b = init_option.encode('utf-8')
  * 			self.c_env.initEnv(init_option_b, print_to_stdout)
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_init_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 528, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_init_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 536, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_init_option, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 528, __pyx_L1_error)
+  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_init_option, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 536, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __pyx_t_3 = (__pyx_t_2 != 0);
   if (likely(__pyx_t_3)) {
 
-    /* "gedlibpy.pyx":529
+    /* "gedlibpy.pyx":537
  * 		"""
  * 		if init_option in list_of_init_options:
  * 			init_option_b = init_option.encode('utf-8')             # <<<<<<<<<<<<<<
  * 			self.c_env.initEnv(init_option_b, print_to_stdout)
  * 		else:
  */
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_init_option, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 529, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_init_option, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 537, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
@@ -5599,29 +5601,29 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 529, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 537, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __pyx_v_init_option_b = __pyx_t_1;
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":530
+    /* "gedlibpy.pyx":538
  * 		if init_option in list_of_init_options:
  * 			init_option_b = init_option.encode('utf-8')
  * 			self.c_env.initEnv(init_option_b, print_to_stdout)             # <<<<<<<<<<<<<<
  * 		else:
  * 			raise InitError("This init option doesn't exist, please see list_of_init_options for selecting an option. You can choose any options.")
  */
-    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_init_option_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 530, __pyx_L1_error)
-    __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_print_to_stdout); if (unlikely((__pyx_t_7 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 530, __pyx_L1_error)
+    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_init_option_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 538, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_print_to_stdout); if (unlikely((__pyx_t_7 == ((bool)-1)) && PyErr_Occurred())) __PYX_ERR(0, 538, __pyx_L1_error)
     try {
       __pyx_v_self->c_env->initEnv(__pyx_t_6, __pyx_t_7);
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 530, __pyx_L1_error)
+      __PYX_ERR(0, 538, __pyx_L1_error)
     }
 
-    /* "gedlibpy.pyx":528
+    /* "gedlibpy.pyx":536
  * 			.. note:: Try to make sure the option exists with list_of_init_options or choose no options, raise an error otherwise.
  * 		"""
  * 		if init_option in list_of_init_options:             # <<<<<<<<<<<<<<
@@ -5631,7 +5633,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
     goto __pyx_L3;
   }
 
-  /* "gedlibpy.pyx":532
+  /* "gedlibpy.pyx":540
  * 			self.c_env.initEnv(init_option_b, print_to_stdout)
  * 		else:
  * 			raise InitError("This init option doesn't exist, please see list_of_init_options for selecting an option. You can choose any options.")             # <<<<<<<<<<<<<<
@@ -5639,7 +5641,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
  * 
  */
   /*else*/ {
-    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_InitError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 532, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_InitError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 540, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
@@ -5653,16 +5655,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_This_init_option_doesn_t_exist_p) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_This_init_option_doesn_t_exist_p);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 532, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 540, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_Raise(__pyx_t_1, 0, 0, 0);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __PYX_ERR(0, 532, __pyx_L1_error)
+    __PYX_ERR(0, 540, __pyx_L1_error)
   }
   __pyx_L3:;
 
-  /* "gedlibpy.pyx":517
+  /* "gedlibpy.pyx":525
  * 
  * 
  * 	def init(self, init_option='EAGER_WITHOUT_SHUFFLED_COPIES', print_to_stdout=False) :             # <<<<<<<<<<<<<<
@@ -5686,7 +5688,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_46init(struct __pyx_obj_8gedlibpy_GE
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":535
+/* "gedlibpy.pyx":543
  * 
  * 
  * 	def set_method(self, method, options="") :             # <<<<<<<<<<<<<<
@@ -5731,7 +5733,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_49set_method(PyObject *__pyx_v_self,
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_method") < 0)) __PYX_ERR(0, 535, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "set_method") < 0)) __PYX_ERR(0, 543, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -5747,7 +5749,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_49set_method(PyObject *__pyx_v_self,
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("set_method", 0, 1, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 535, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("set_method", 0, 1, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 543, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.set_method", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -5773,28 +5775,28 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
   std::string __pyx_t_7;
   __Pyx_RefNannySetupContext("set_method", 0);
 
-  /* "gedlibpy.pyx":547
+  /* "gedlibpy.pyx":555
  * 			.. note:: Try to make sure the edit cost function exists with list_of_method_options, raise an error otherwise. Call init_method() after your set.
  * 		"""
  * 		if method in list_of_method_options:             # <<<<<<<<<<<<<<
  * 			method_b = method.encode('utf-8')
  * 			self.c_env.setMethod(method_b, options.encode('utf-8'))
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_method_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 547, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_list_of_method_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 555, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_method, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 547, __pyx_L1_error)
+  __pyx_t_2 = (__Pyx_PySequence_ContainsTF(__pyx_v_method, __pyx_t_1, Py_EQ)); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 555, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __pyx_t_3 = (__pyx_t_2 != 0);
   if (likely(__pyx_t_3)) {
 
-    /* "gedlibpy.pyx":548
+    /* "gedlibpy.pyx":556
  * 		"""
  * 		if method in list_of_method_options:
  * 			method_b = method.encode('utf-8')             # <<<<<<<<<<<<<<
  * 			self.c_env.setMethod(method_b, options.encode('utf-8'))
  * 		else:
  */
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_method, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 548, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_method, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 556, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
@@ -5808,21 +5810,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 548, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 556, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __pyx_v_method_b = __pyx_t_1;
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":549
+    /* "gedlibpy.pyx":557
  * 		if method in list_of_method_options:
  * 			method_b = method.encode('utf-8')
  * 			self.c_env.setMethod(method_b, options.encode('utf-8'))             # <<<<<<<<<<<<<<
  * 		else:
  * 			raise MethodError("This method doesn't exist, please see list_of_method_options for selecting a method")
  */
-    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_method_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 549, __pyx_L1_error)
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_options, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 549, __pyx_L1_error)
+    __pyx_t_6 = __pyx_convert_string_from_py_std__in_string(__pyx_v_method_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 557, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_options, __pyx_n_s_encode); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 557, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
@@ -5836,19 +5838,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 549, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 557, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_7 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 549, __pyx_L1_error)
+    __pyx_t_7 = __pyx_convert_string_from_py_std__in_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 557, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     try {
       __pyx_v_self->c_env->setMethod(__pyx_t_6, __pyx_t_7);
     } catch(...) {
       __Pyx_CppExn2PyErr();
-      __PYX_ERR(0, 549, __pyx_L1_error)
+      __PYX_ERR(0, 557, __pyx_L1_error)
     }
 
-    /* "gedlibpy.pyx":547
+    /* "gedlibpy.pyx":555
  * 			.. note:: Try to make sure the edit cost function exists with list_of_method_options, raise an error otherwise. Call init_method() after your set.
  * 		"""
  * 		if method in list_of_method_options:             # <<<<<<<<<<<<<<
@@ -5858,7 +5860,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
     goto __pyx_L3;
   }
 
-  /* "gedlibpy.pyx":551
+  /* "gedlibpy.pyx":559
  * 			self.c_env.setMethod(method_b, options.encode('utf-8'))
  * 		else:
  * 			raise MethodError("This method doesn't exist, please see list_of_method_options for selecting a method")             # <<<<<<<<<<<<<<
@@ -5866,7 +5868,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
  * 
  */
   /*else*/ {
-    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_MethodError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 551, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_MethodError); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 559, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
@@ -5880,16 +5882,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
     }
     __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_5, __pyx_kp_u_This_method_doesn_t_exist_please) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_kp_u_This_method_doesn_t_exist_please);
     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 551, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 559, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_Raise(__pyx_t_1, 0, 0, 0);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __PYX_ERR(0, 551, __pyx_L1_error)
+    __PYX_ERR(0, 559, __pyx_L1_error)
   }
   __pyx_L3:;
 
-  /* "gedlibpy.pyx":535
+  /* "gedlibpy.pyx":543
  * 
  * 
  * 	def set_method(self, method, options="") :             # <<<<<<<<<<<<<<
@@ -5913,7 +5915,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_48set_method(struct __pyx_obj_8gedli
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":554
+/* "gedlibpy.pyx":562
  * 
  * 
  * 	def init_method(self) :             # <<<<<<<<<<<<<<
@@ -5940,7 +5942,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_50init_method(struct __pyx_obj_8gedl
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("init_method", 0);
 
-  /* "gedlibpy.pyx":561
+  /* "gedlibpy.pyx":569
  * 			.. note:: Call this function after set the method. You can't launch computation or change the method after that.
  * 		"""
  * 		self.c_env.initMethod()             # <<<<<<<<<<<<<<
@@ -5951,10 +5953,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_50init_method(struct __pyx_obj_8gedl
     __pyx_v_self->c_env->initMethod();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 561, __pyx_L1_error)
+    __PYX_ERR(0, 569, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":554
+  /* "gedlibpy.pyx":562
  * 
  * 
  * 	def init_method(self) :             # <<<<<<<<<<<<<<
@@ -5974,7 +5976,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_50init_method(struct __pyx_obj_8gedl
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":564
+/* "gedlibpy.pyx":572
  * 
  * 
  * 	def get_init_time(self) :             # <<<<<<<<<<<<<<
@@ -6003,7 +6005,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_52get_init_time(struct __pyx_obj_8ge
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_init_time", 0);
 
-  /* "gedlibpy.pyx":571
+  /* "gedlibpy.pyx":579
  * 			:rtype: double
  * 		"""
  * 		return self.c_env.getInitime()             # <<<<<<<<<<<<<<
@@ -6015,15 +6017,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_52get_init_time(struct __pyx_obj_8ge
     __pyx_t_1 = __pyx_v_self->c_env->getInitime();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 571, __pyx_L1_error)
+    __PYX_ERR(0, 579, __pyx_L1_error)
   }
-  __pyx_t_2 = PyFloat_FromDouble(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 571, __pyx_L1_error)
+  __pyx_t_2 = PyFloat_FromDouble(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 579, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":564
+  /* "gedlibpy.pyx":572
  * 
  * 
  * 	def get_init_time(self) :             # <<<<<<<<<<<<<<
@@ -6042,7 +6044,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_52get_init_time(struct __pyx_obj_8ge
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":574
+/* "gedlibpy.pyx":582
  * 
  * 
  * 	def run_method(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6082,11 +6084,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_55run_method(PyObject *__pyx_v_self,
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("run_method", 1, 2, 2, 1); __PYX_ERR(0, 574, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("run_method", 1, 2, 2, 1); __PYX_ERR(0, 582, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "run_method") < 0)) __PYX_ERR(0, 574, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "run_method") < 0)) __PYX_ERR(0, 582, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6099,7 +6101,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_55run_method(PyObject *__pyx_v_self,
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("run_method", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 574, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("run_method", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 582, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.run_method", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6119,23 +6121,23 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_54run_method(struct __pyx_obj_8gedli
   size_t __pyx_t_2;
   __Pyx_RefNannySetupContext("run_method", 0);
 
-  /* "gedlibpy.pyx":586
+  /* "gedlibpy.pyx":594
  * 			.. note:: This function only compute the distance between two graphs, without returning a result. Use the differents function to see the result between the two graphs.
  * 		"""
  * 		self.c_env.runMethod(g, h)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 586, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 586, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 594, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 594, __pyx_L1_error)
   try {
     __pyx_v_self->c_env->runMethod(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 586, __pyx_L1_error)
+    __PYX_ERR(0, 594, __pyx_L1_error)
   }
 
-  /* "gedlibpy.pyx":574
+  /* "gedlibpy.pyx":582
  * 
  * 
  * 	def run_method(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6155,7 +6157,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_54run_method(struct __pyx_obj_8gedli
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":589
+/* "gedlibpy.pyx":597
  * 
  * 
  * 	def get_upper_bound(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6195,11 +6197,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_57get_upper_bound(PyObject *__pyx_v_
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_upper_bound", 1, 2, 2, 1); __PYX_ERR(0, 589, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_upper_bound", 1, 2, 2, 1); __PYX_ERR(0, 597, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_upper_bound") < 0)) __PYX_ERR(0, 589, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_upper_bound") < 0)) __PYX_ERR(0, 597, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6212,7 +6214,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_57get_upper_bound(PyObject *__pyx_v_
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_upper_bound", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 589, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_upper_bound", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 597, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_upper_bound", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6234,7 +6236,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_56get_upper_bound(struct __pyx_obj_8
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_upper_bound", 0);
 
-  /* "gedlibpy.pyx":604
+  /* "gedlibpy.pyx":612
  * 			.. note:: The upper bound is equivalent to the result of the pessimist edit distance cost. Methods are heuristics so the library can't compute the real perfect result because it's NP-Hard problem.
  * 		"""
  * 		return self.c_env.getUpperBound(g, h)             # <<<<<<<<<<<<<<
@@ -6242,21 +6244,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_56get_upper_bound(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 604, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 604, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 612, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 612, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getUpperBound(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 604, __pyx_L1_error)
+    __PYX_ERR(0, 612, __pyx_L1_error)
   }
-  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 604, __pyx_L1_error)
+  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 612, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":589
+  /* "gedlibpy.pyx":597
  * 
  * 
  * 	def get_upper_bound(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6275,7 +6277,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_56get_upper_bound(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":607
+/* "gedlibpy.pyx":615
  * 
  * 
  * 	def get_lower_bound(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6315,11 +6317,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_59get_lower_bound(PyObject *__pyx_v_
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_lower_bound", 1, 2, 2, 1); __PYX_ERR(0, 607, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_lower_bound", 1, 2, 2, 1); __PYX_ERR(0, 615, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_lower_bound") < 0)) __PYX_ERR(0, 607, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_lower_bound") < 0)) __PYX_ERR(0, 615, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6332,7 +6334,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_59get_lower_bound(PyObject *__pyx_v_
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_lower_bound", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 607, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_lower_bound", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 615, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_lower_bound", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6354,7 +6356,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_58get_lower_bound(struct __pyx_obj_8
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_lower_bound", 0);
 
-  /* "gedlibpy.pyx":622
+  /* "gedlibpy.pyx":630
  * 			.. note:: This function can be ignored, because lower bound doesn't have a crucial utility.
  * 		"""
  * 		return self.c_env.getLowerBound(g, h)             # <<<<<<<<<<<<<<
@@ -6362,21 +6364,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_58get_lower_bound(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 622, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 622, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 630, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 630, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getLowerBound(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 622, __pyx_L1_error)
+    __PYX_ERR(0, 630, __pyx_L1_error)
   }
-  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 622, __pyx_L1_error)
+  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 630, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":607
+  /* "gedlibpy.pyx":615
  * 
  * 
  * 	def get_lower_bound(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6395,7 +6397,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_58get_lower_bound(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":625
+/* "gedlibpy.pyx":633
  * 
  * 
  * 	def get_forward_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6435,11 +6437,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_61get_forward_map(PyObject *__pyx_v_
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_forward_map", 1, 2, 2, 1); __PYX_ERR(0, 625, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_forward_map", 1, 2, 2, 1); __PYX_ERR(0, 633, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_forward_map") < 0)) __PYX_ERR(0, 625, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_forward_map") < 0)) __PYX_ERR(0, 633, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6452,7 +6454,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_61get_forward_map(PyObject *__pyx_v_
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_forward_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 625, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_forward_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 633, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_forward_map", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6474,7 +6476,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_60get_forward_map(struct __pyx_obj_8
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_forward_map", 0);
 
-  /* "gedlibpy.pyx":640
+  /* "gedlibpy.pyx":648
  * 			.. note:: I don't know how to connect the two map to reconstruct the adjacence matrix. Please come back when I know how it's work !
  * 		"""
  * 		return self.c_env.getForwardMap(g, h)             # <<<<<<<<<<<<<<
@@ -6482,21 +6484,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_60get_forward_map(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 640, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 640, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 648, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 648, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getForwardMap(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 640, __pyx_L1_error)
+    __PYX_ERR(0, 648, __pyx_L1_error)
   }
-  __pyx_t_4 = __pyx_convert_vector_to_py_npy_uint64(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 640, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_vector_to_py_npy_uint64(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 648, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":625
+  /* "gedlibpy.pyx":633
  * 
  * 
  * 	def get_forward_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6515,7 +6517,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_60get_forward_map(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":643
+/* "gedlibpy.pyx":651
  * 
  * 
  * 	def get_backward_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6555,11 +6557,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_63get_backward_map(PyObject *__pyx_v
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_backward_map", 1, 2, 2, 1); __PYX_ERR(0, 643, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_backward_map", 1, 2, 2, 1); __PYX_ERR(0, 651, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_backward_map") < 0)) __PYX_ERR(0, 643, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_backward_map") < 0)) __PYX_ERR(0, 651, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6572,7 +6574,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_63get_backward_map(PyObject *__pyx_v
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_backward_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 643, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_backward_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 651, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_backward_map", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6594,7 +6596,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_62get_backward_map(struct __pyx_obj_
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_backward_map", 0);
 
-  /* "gedlibpy.pyx":658
+  /* "gedlibpy.pyx":666
  * 			.. note:: I don't know how to connect the two map to reconstruct the adjacence matrix. Please come back when I know how it's work !
  * 		"""
  * 		return self.c_env.getBackwardMap(g, h)             # <<<<<<<<<<<<<<
@@ -6602,21 +6604,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_62get_backward_map(struct __pyx_obj_
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 658, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 658, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 666, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 666, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getBackwardMap(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 658, __pyx_L1_error)
+    __PYX_ERR(0, 666, __pyx_L1_error)
   }
-  __pyx_t_4 = __pyx_convert_vector_to_py_npy_uint64(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 658, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_vector_to_py_npy_uint64(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 666, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":643
+  /* "gedlibpy.pyx":651
  * 
  * 
  * 	def get_backward_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6635,7 +6637,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_62get_backward_map(struct __pyx_obj_
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":661
+/* "gedlibpy.pyx":669
  * 
  * 
  * 	def get_node_image(self, g, h, node_id) :             # <<<<<<<<<<<<<<
@@ -6678,17 +6680,17 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_65get_node_image(PyObject *__pyx_v_s
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, 1); __PYX_ERR(0, 661, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, 1); __PYX_ERR(0, 669, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_id)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, 2); __PYX_ERR(0, 661, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, 2); __PYX_ERR(0, 669, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_image") < 0)) __PYX_ERR(0, 661, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_image") < 0)) __PYX_ERR(0, 669, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
       goto __pyx_L5_argtuple_error;
@@ -6703,7 +6705,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_65get_node_image(PyObject *__pyx_v_s
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 661, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_node_image", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 669, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_node_image", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6726,7 +6728,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_64get_node_image(struct __pyx_obj_8g
   PyObject *__pyx_t_5 = NULL;
   __Pyx_RefNannySetupContext("get_node_image", 0);
 
-  /* "gedlibpy.pyx":679
+  /* "gedlibpy.pyx":687
  * 
  * 		"""
  * 		return self.c_env.getNodeImage(g, h, node_id)             # <<<<<<<<<<<<<<
@@ -6734,22 +6736,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_64get_node_image(struct __pyx_obj_8g
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 679, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 679, __pyx_L1_error)
-  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_node_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 679, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 687, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 687, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_node_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 687, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getNodeImage(__pyx_t_1, __pyx_t_2, __pyx_t_3);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 679, __pyx_L1_error)
+    __PYX_ERR(0, 687, __pyx_L1_error)
   }
-  __pyx_t_5 = __Pyx_PyInt_FromSize_t(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 679, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyInt_FromSize_t(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 687, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_r = __pyx_t_5;
   __pyx_t_5 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":661
+  /* "gedlibpy.pyx":669
  * 
  * 
  * 	def get_node_image(self, g, h, node_id) :             # <<<<<<<<<<<<<<
@@ -6768,7 +6770,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_64get_node_image(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":682
+/* "gedlibpy.pyx":690
  * 
  * 
  * 	def get_node_pre_image(self, g, h, node_id) :             # <<<<<<<<<<<<<<
@@ -6811,17 +6813,17 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_67get_node_pre_image(PyObject *__pyx
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, 1); __PYX_ERR(0, 682, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, 1); __PYX_ERR(0, 690, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_id)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, 2); __PYX_ERR(0, 682, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, 2); __PYX_ERR(0, 690, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_pre_image") < 0)) __PYX_ERR(0, 682, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_pre_image") < 0)) __PYX_ERR(0, 690, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
       goto __pyx_L5_argtuple_error;
@@ -6836,7 +6838,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_67get_node_pre_image(PyObject *__pyx
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 682, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_node_pre_image", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 690, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_node_pre_image", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6859,7 +6861,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_66get_node_pre_image(struct __pyx_ob
   PyObject *__pyx_t_5 = NULL;
   __Pyx_RefNannySetupContext("get_node_pre_image", 0);
 
-  /* "gedlibpy.pyx":700
+  /* "gedlibpy.pyx":708
  * 
  * 		"""
  * 		return self.c_env.getNodePreImage(g, h, node_id)             # <<<<<<<<<<<<<<
@@ -6867,22 +6869,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_66get_node_pre_image(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 700, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 700, __pyx_L1_error)
-  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_node_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 700, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 708, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 708, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_node_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 708, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getNodePreImage(__pyx_t_1, __pyx_t_2, __pyx_t_3);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 700, __pyx_L1_error)
+    __PYX_ERR(0, 708, __pyx_L1_error)
   }
-  __pyx_t_5 = __Pyx_PyInt_FromSize_t(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 700, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyInt_FromSize_t(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 708, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_r = __pyx_t_5;
   __pyx_t_5 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":682
+  /* "gedlibpy.pyx":690
  * 
  * 
  * 	def get_node_pre_image(self, g, h, node_id) :             # <<<<<<<<<<<<<<
@@ -6901,7 +6903,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_66get_node_pre_image(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":703
+/* "gedlibpy.pyx":711
  * 
  * 
  * 	def get_induced_cost(self, g, h) :             # <<<<<<<<<<<<<<
@@ -6941,11 +6943,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_69get_induced_cost(PyObject *__pyx_v
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_induced_cost", 1, 2, 2, 1); __PYX_ERR(0, 703, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_induced_cost", 1, 2, 2, 1); __PYX_ERR(0, 711, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_induced_cost") < 0)) __PYX_ERR(0, 703, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_induced_cost") < 0)) __PYX_ERR(0, 711, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -6958,7 +6960,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_69get_induced_cost(PyObject *__pyx_v
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_induced_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 703, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_induced_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 711, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_induced_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -6980,7 +6982,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_68get_induced_cost(struct __pyx_obj_
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_induced_cost", 0);
 
-  /* "gedlibpy.pyx":719
+  /* "gedlibpy.pyx":727
  * 
  * 		"""
  * 		return self.c_env.getInducedCost(g, h)             # <<<<<<<<<<<<<<
@@ -6988,21 +6990,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_68get_induced_cost(struct __pyx_obj_
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 719, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 719, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 727, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 727, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getInducedCost(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 719, __pyx_L1_error)
+    __PYX_ERR(0, 727, __pyx_L1_error)
   }
-  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 719, __pyx_L1_error)
+  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 727, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":703
+  /* "gedlibpy.pyx":711
  * 
  * 
  * 	def get_induced_cost(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7021,7 +7023,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_68get_induced_cost(struct __pyx_obj_
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":722
+/* "gedlibpy.pyx":730
  * 
  * 
  * 	def get_node_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7061,11 +7063,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_71get_node_map(PyObject *__pyx_v_sel
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_map", 1, 2, 2, 1); __PYX_ERR(0, 722, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_map", 1, 2, 2, 1); __PYX_ERR(0, 730, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_map") < 0)) __PYX_ERR(0, 722, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_map") < 0)) __PYX_ERR(0, 730, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -7078,7 +7080,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_71get_node_map(PyObject *__pyx_v_sel
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_node_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 722, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_node_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 730, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_node_map", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -7125,41 +7127,41 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   Py_ssize_t __pyx_t_17;
   __Pyx_RefNannySetupContext("get_node_map", 0);
 
-  /* "gedlibpy.pyx":737
+  /* "gedlibpy.pyx":745
  * 			.. note:: This function creates datas so use it if necessary, however you can understand how assignement works with this example.
  * 		"""
  * 		map_as_relation = self.c_env.getNodeMap(g, h)             # <<<<<<<<<<<<<<
  * 		induced_cost = self.c_env.getInducedCost(g, h) # @todo: the C++ implementation for this function in GedLibBind.ipp re-call get_node_map() once more, this is not neccessary.
  * 		source_map = [item.first if item.first < len(map_as_relation) else np.inf for item in map_as_relation] # item.first < len(map_as_relation) is not exactly correct.
  */
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 737, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 737, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 745, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 745, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getNodeMap(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 737, __pyx_L1_error)
+    __PYX_ERR(0, 745, __pyx_L1_error)
   }
   __pyx_v_map_as_relation = __pyx_t_3;
 
-  /* "gedlibpy.pyx":738
+  /* "gedlibpy.pyx":746
  * 		"""
  * 		map_as_relation = self.c_env.getNodeMap(g, h)
  * 		induced_cost = self.c_env.getInducedCost(g, h) # @todo: the C++ implementation for this function in GedLibBind.ipp re-call get_node_map() once more, this is not neccessary.             # <<<<<<<<<<<<<<
  * 		source_map = [item.first if item.first < len(map_as_relation) else np.inf for item in map_as_relation] # item.first < len(map_as_relation) is not exactly correct.
  * # 		print(source_map)
  */
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 738, __pyx_L1_error)
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 738, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 746, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 746, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getInducedCost(__pyx_t_2, __pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 738, __pyx_L1_error)
+    __PYX_ERR(0, 746, __pyx_L1_error)
   }
   __pyx_v_induced_cost = __pyx_t_4;
 
-  /* "gedlibpy.pyx":739
+  /* "gedlibpy.pyx":747
  * 		map_as_relation = self.c_env.getNodeMap(g, h)
  * 		induced_cost = self.c_env.getInducedCost(g, h) # @todo: the C++ implementation for this function in GedLibBind.ipp re-call get_node_map() once more, this is not neccessary.
  * 		source_map = [item.first if item.first < len(map_as_relation) else np.inf for item in map_as_relation] # item.first < len(map_as_relation) is not exactly correct.             # <<<<<<<<<<<<<<
@@ -7167,7 +7169,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
  * 		target_map = [item.second if item.second < len(map_as_relation) else np.inf for item in map_as_relation]
  */
   { /* enter inner scope */
-    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 739, __pyx_L1_error)
+    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 747, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_6 = __pyx_v_map_as_relation.begin();
     for (;;) {
@@ -7175,32 +7177,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
       __pyx_t_7 = *__pyx_t_6;
       ++__pyx_t_6;
       __pyx_8genexpr5__pyx_v_item = __pyx_t_7;
-      __pyx_t_9 = __pyx_convert_vector_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_map_as_relation); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 739, __pyx_L1_error)
+      __pyx_t_9 = __pyx_convert_vector_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_map_as_relation); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 747, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_10 = PyObject_Length(__pyx_t_9); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 739, __pyx_L1_error)
+      __pyx_t_10 = PyObject_Length(__pyx_t_9); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 747, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
       if (((__pyx_8genexpr5__pyx_v_item.first < __pyx_t_10) != 0)) {
-        __pyx_t_9 = __Pyx_PyInt_FromSize_t(__pyx_8genexpr5__pyx_v_item.first); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 739, __pyx_L1_error)
+        __pyx_t_9 = __Pyx_PyInt_FromSize_t(__pyx_8genexpr5__pyx_v_item.first); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 747, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_9);
         __pyx_t_8 = __pyx_t_9;
         __pyx_t_9 = 0;
       } else {
-        __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 739, __pyx_L1_error)
+        __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 747, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 739, __pyx_L1_error)
+        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 747, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_11);
         __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
         __pyx_t_8 = __pyx_t_11;
         __pyx_t_11 = 0;
       }
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 739, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 747, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
     }
   } /* exit inner scope */
   __pyx_v_source_map = ((PyObject*)__pyx_t_5);
   __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":741
+  /* "gedlibpy.pyx":749
  * 		source_map = [item.first if item.first < len(map_as_relation) else np.inf for item in map_as_relation] # item.first < len(map_as_relation) is not exactly correct.
  * # 		print(source_map)
  * 		target_map = [item.second if item.second < len(map_as_relation) else np.inf for item in map_as_relation]             # <<<<<<<<<<<<<<
@@ -7208,7 +7210,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
  * 		num_node_source = len([item for item in source_map if item != np.inf])
  */
   { /* enter inner scope */
-    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 741, __pyx_L1_error)
+    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 749, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_6 = __pyx_v_map_as_relation.begin();
     for (;;) {
@@ -7216,32 +7218,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
       __pyx_t_7 = *__pyx_t_6;
       ++__pyx_t_6;
       __pyx_8genexpr6__pyx_v_item = __pyx_t_7;
-      __pyx_t_11 = __pyx_convert_vector_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_map_as_relation); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 741, __pyx_L1_error)
+      __pyx_t_11 = __pyx_convert_vector_to_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_map_as_relation); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 749, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_10 = PyObject_Length(__pyx_t_11); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 741, __pyx_L1_error)
+      __pyx_t_10 = PyObject_Length(__pyx_t_11); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 749, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
       if (((__pyx_8genexpr6__pyx_v_item.second < __pyx_t_10) != 0)) {
-        __pyx_t_11 = __Pyx_PyInt_FromSize_t(__pyx_8genexpr6__pyx_v_item.second); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 741, __pyx_L1_error)
+        __pyx_t_11 = __Pyx_PyInt_FromSize_t(__pyx_8genexpr6__pyx_v_item.second); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 749, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_11);
         __pyx_t_8 = __pyx_t_11;
         __pyx_t_11 = 0;
       } else {
-        __Pyx_GetModuleGlobalName(__pyx_t_11, __pyx_n_s_np); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 741, __pyx_L1_error)
+        __Pyx_GetModuleGlobalName(__pyx_t_11, __pyx_n_s_np); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 749, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_inf); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 741, __pyx_L1_error)
+        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_inf); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 749, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_9);
         __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
         __pyx_t_8 = __pyx_t_9;
         __pyx_t_9 = 0;
       }
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 741, __pyx_L1_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 749, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
     }
   } /* exit inner scope */
   __pyx_v_target_map = ((PyObject*)__pyx_t_5);
   __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":743
+  /* "gedlibpy.pyx":751
  * 		target_map = [item.second if item.second < len(map_as_relation) else np.inf for item in map_as_relation]
  * # 		print(target_map)
  * 		num_node_source = len([item for item in source_map if item != np.inf])             # <<<<<<<<<<<<<<
@@ -7249,30 +7251,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
  * 		num_node_target = len([item for item in target_map if item != np.inf])
  */
   { /* enter inner scope */
-    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 743, __pyx_L9_error)
+    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L9_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_8 = __pyx_v_source_map; __Pyx_INCREF(__pyx_t_8); __pyx_t_10 = 0;
     for (;;) {
       if (__pyx_t_10 >= PyList_GET_SIZE(__pyx_t_8)) break;
       #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      __pyx_t_9 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_10); __Pyx_INCREF(__pyx_t_9); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __pyx_t_9 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_10); __Pyx_INCREF(__pyx_t_9); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 751, __pyx_L9_error)
       #else
-      __pyx_t_9 = PySequence_ITEM(__pyx_t_8, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __pyx_t_9 = PySequence_ITEM(__pyx_t_8, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 751, __pyx_L9_error)
       __Pyx_GOTREF(__pyx_t_9);
       #endif
       __Pyx_XDECREF_SET(__pyx_8genexpr7__pyx_v_item, __pyx_t_9);
       __pyx_t_9 = 0;
-      __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 751, __pyx_L9_error)
       __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 751, __pyx_L9_error)
       __Pyx_GOTREF(__pyx_t_11);
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_t_9 = PyObject_RichCompare(__pyx_8genexpr7__pyx_v_item, __pyx_t_11, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __pyx_t_9 = PyObject_RichCompare(__pyx_8genexpr7__pyx_v_item, __pyx_t_11, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 751, __pyx_L9_error)
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 743, __pyx_L9_error)
+      __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 751, __pyx_L9_error)
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
       if (__pyx_t_12) {
-        if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_8genexpr7__pyx_v_item))) __PYX_ERR(0, 743, __pyx_L9_error)
+        if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_8genexpr7__pyx_v_item))) __PYX_ERR(0, 751, __pyx_L9_error)
       }
     }
     __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
@@ -7283,11 +7285,11 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     goto __pyx_L1_error;
     __pyx_L13_exit_scope:;
   } /* exit inner scope */
-  __pyx_t_10 = PyList_GET_SIZE(__pyx_t_5); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 743, __pyx_L1_error)
+  __pyx_t_10 = PyList_GET_SIZE(__pyx_t_5); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 751, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __pyx_v_num_node_source = __pyx_t_10;
 
-  /* "gedlibpy.pyx":745
+  /* "gedlibpy.pyx":753
  * 		num_node_source = len([item for item in source_map if item != np.inf])
  * # 		print(num_node_source)
  * 		num_node_target = len([item for item in target_map if item != np.inf])             # <<<<<<<<<<<<<<
@@ -7295,30 +7297,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
  * 
  */
   { /* enter inner scope */
-    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 745, __pyx_L16_error)
+    __pyx_t_5 = PyList_New(0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 753, __pyx_L16_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_8 = __pyx_v_target_map; __Pyx_INCREF(__pyx_t_8); __pyx_t_10 = 0;
     for (;;) {
       if (__pyx_t_10 >= PyList_GET_SIZE(__pyx_t_8)) break;
       #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      __pyx_t_9 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_10); __Pyx_INCREF(__pyx_t_9); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __pyx_t_9 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_10); __Pyx_INCREF(__pyx_t_9); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 753, __pyx_L16_error)
       #else
-      __pyx_t_9 = PySequence_ITEM(__pyx_t_8, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __pyx_t_9 = PySequence_ITEM(__pyx_t_8, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 753, __pyx_L16_error)
       __Pyx_GOTREF(__pyx_t_9);
       #endif
       __Pyx_XDECREF_SET(__pyx_8genexpr8__pyx_v_item, __pyx_t_9);
       __pyx_t_9 = 0;
-      __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 753, __pyx_L16_error)
       __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_inf); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 753, __pyx_L16_error)
       __Pyx_GOTREF(__pyx_t_11);
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_t_9 = PyObject_RichCompare(__pyx_8genexpr8__pyx_v_item, __pyx_t_11, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __pyx_t_9 = PyObject_RichCompare(__pyx_8genexpr8__pyx_v_item, __pyx_t_11, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 753, __pyx_L16_error)
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 745, __pyx_L16_error)
+      __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 753, __pyx_L16_error)
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
       if (__pyx_t_12) {
-        if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_8genexpr8__pyx_v_item))) __PYX_ERR(0, 745, __pyx_L16_error)
+        if (unlikely(__Pyx_ListComp_Append(__pyx_t_5, (PyObject*)__pyx_8genexpr8__pyx_v_item))) __PYX_ERR(0, 753, __pyx_L16_error)
       }
     }
     __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
@@ -7329,22 +7331,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     goto __pyx_L1_error;
     __pyx_L20_exit_scope:;
   } /* exit inner scope */
-  __pyx_t_10 = PyList_GET_SIZE(__pyx_t_5); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 745, __pyx_L1_error)
+  __pyx_t_10 = PyList_GET_SIZE(__pyx_t_5); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 753, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __pyx_v_num_node_target = __pyx_t_10;
 
-  /* "gedlibpy.pyx":748
+  /* "gedlibpy.pyx":756
  * # 		print(num_node_target)
  * 
  * 		node_map = NodeMap(num_node_source, num_node_target)             # <<<<<<<<<<<<<<
  * # 		print(node_map.get_forward_map(), node_map.get_backward_map())
  * 		for i in range(len(source_map)):
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_NodeMap); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 748, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_NodeMap); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 756, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_9 = PyInt_FromSsize_t(__pyx_v_num_node_source); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 748, __pyx_L1_error)
+  __pyx_t_9 = PyInt_FromSsize_t(__pyx_v_num_node_source); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 756, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_11 = PyInt_FromSsize_t(__pyx_v_num_node_target); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 748, __pyx_L1_error)
+  __pyx_t_11 = PyInt_FromSsize_t(__pyx_v_num_node_target); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 756, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_11);
   __pyx_t_13 = NULL;
   __pyx_t_14 = 0;
@@ -7361,7 +7363,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_8)) {
     PyObject *__pyx_temp[3] = {__pyx_t_13, __pyx_t_9, __pyx_t_11};
-    __pyx_t_5 = __Pyx_PyFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 748, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 756, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_13); __pyx_t_13 = 0;
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
@@ -7371,7 +7373,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_8)) {
     PyObject *__pyx_temp[3] = {__pyx_t_13, __pyx_t_9, __pyx_t_11};
-    __pyx_t_5 = __Pyx_PyCFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 748, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyCFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 756, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_13); __pyx_t_13 = 0;
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
@@ -7379,7 +7381,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   } else
   #endif
   {
-    __pyx_t_15 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_15)) __PYX_ERR(0, 748, __pyx_L1_error)
+    __pyx_t_15 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_15)) __PYX_ERR(0, 756, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_15);
     if (__pyx_t_13) {
       __Pyx_GIVEREF(__pyx_t_13); PyTuple_SET_ITEM(__pyx_t_15, 0, __pyx_t_13); __pyx_t_13 = NULL;
@@ -7390,7 +7392,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     PyTuple_SET_ITEM(__pyx_t_15, 1+__pyx_t_14, __pyx_t_11);
     __pyx_t_9 = 0;
     __pyx_t_11 = 0;
-    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_15, NULL); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 748, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_15, NULL); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 756, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
   }
@@ -7398,30 +7400,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   __pyx_v_node_map = __pyx_t_5;
   __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":750
+  /* "gedlibpy.pyx":758
  * 		node_map = NodeMap(num_node_source, num_node_target)
  * # 		print(node_map.get_forward_map(), node_map.get_backward_map())
  * 		for i in range(len(source_map)):             # <<<<<<<<<<<<<<
  * 			node_map.add_assignment(source_map[i], target_map[i])
  * 		node_map.set_induced_cost(induced_cost)
  */
-  __pyx_t_10 = PyList_GET_SIZE(__pyx_v_source_map); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 750, __pyx_L1_error)
+  __pyx_t_10 = PyList_GET_SIZE(__pyx_v_source_map); if (unlikely(__pyx_t_10 == ((Py_ssize_t)-1))) __PYX_ERR(0, 758, __pyx_L1_error)
   __pyx_t_16 = __pyx_t_10;
   for (__pyx_t_17 = 0; __pyx_t_17 < __pyx_t_16; __pyx_t_17+=1) {
     __pyx_v_i = __pyx_t_17;
 
-    /* "gedlibpy.pyx":751
+    /* "gedlibpy.pyx":759
  * # 		print(node_map.get_forward_map(), node_map.get_backward_map())
  * 		for i in range(len(source_map)):
  * 			node_map.add_assignment(source_map[i], target_map[i])             # <<<<<<<<<<<<<<
  * 		node_map.set_induced_cost(induced_cost)
  * 
  */
-    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_add_assignment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_add_assignment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 759, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_15 = __Pyx_GetItemInt_List(__pyx_v_source_map, __pyx_v_i, Py_ssize_t, 1, PyInt_FromSsize_t, 1, 1, 1); if (unlikely(!__pyx_t_15)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __pyx_t_15 = __Pyx_GetItemInt_List(__pyx_v_source_map, __pyx_v_i, Py_ssize_t, 1, PyInt_FromSsize_t, 1, 1, 1); if (unlikely(!__pyx_t_15)) __PYX_ERR(0, 759, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_15);
-    __pyx_t_11 = __Pyx_GetItemInt_List(__pyx_v_target_map, __pyx_v_i, Py_ssize_t, 1, PyInt_FromSsize_t, 1, 1, 1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __pyx_t_11 = __Pyx_GetItemInt_List(__pyx_v_target_map, __pyx_v_i, Py_ssize_t, 1, PyInt_FromSsize_t, 1, 1, 1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 759, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_11);
     __pyx_t_9 = NULL;
     __pyx_t_14 = 0;
@@ -7438,7 +7440,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_8)) {
       PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_t_15, __pyx_t_11};
-      __pyx_t_5 = __Pyx_PyFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L1_error)
+      __pyx_t_5 = __Pyx_PyFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 759, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_5);
       __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
@@ -7448,7 +7450,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_8)) {
       PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_t_15, __pyx_t_11};
-      __pyx_t_5 = __Pyx_PyCFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L1_error)
+      __pyx_t_5 = __Pyx_PyCFunction_FastCall(__pyx_t_8, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 759, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_5);
       __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
@@ -7456,7 +7458,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     } else
     #endif
     {
-      __pyx_t_13 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 751, __pyx_L1_error)
+      __pyx_t_13 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 759, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_13);
       if (__pyx_t_9) {
         __Pyx_GIVEREF(__pyx_t_9); PyTuple_SET_ITEM(__pyx_t_13, 0, __pyx_t_9); __pyx_t_9 = NULL;
@@ -7467,7 +7469,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
       PyTuple_SET_ITEM(__pyx_t_13, 1+__pyx_t_14, __pyx_t_11);
       __pyx_t_15 = 0;
       __pyx_t_11 = 0;
-      __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_13, NULL); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L1_error)
+      __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_13, NULL); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 759, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_5);
       __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
     }
@@ -7475,16 +7477,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   }
 
-  /* "gedlibpy.pyx":752
+  /* "gedlibpy.pyx":760
  * 		for i in range(len(source_map)):
  * 			node_map.add_assignment(source_map[i], target_map[i])
  * 		node_map.set_induced_cost(induced_cost)             # <<<<<<<<<<<<<<
  * 
  * 		return node_map
  */
-  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_set_induced_cost); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 752, __pyx_L1_error)
+  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_set_induced_cost); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 760, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_13 = PyFloat_FromDouble(__pyx_v_induced_cost); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 752, __pyx_L1_error)
+  __pyx_t_13 = PyFloat_FromDouble(__pyx_v_induced_cost); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 760, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_13);
   __pyx_t_11 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_8))) {
@@ -7499,12 +7501,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   __pyx_t_5 = (__pyx_t_11) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_11, __pyx_t_13) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_13);
   __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
   __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
-  if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 752, __pyx_L1_error)
+  if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 760, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":754
+  /* "gedlibpy.pyx":762
  * 		node_map.set_induced_cost(induced_cost)
  * 
  * 		return node_map             # <<<<<<<<<<<<<<
@@ -7516,7 +7518,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   __pyx_r = __pyx_v_node_map;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":722
+  /* "gedlibpy.pyx":730
  * 
  * 
  * 	def get_node_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7545,7 +7547,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_70get_node_map(struct __pyx_obj_8ged
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":757
+/* "gedlibpy.pyx":765
  * 
  * 
  * 	def get_assignment_matrix(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7585,11 +7587,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_73get_assignment_matrix(PyObject *__
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_assignment_matrix", 1, 2, 2, 1); __PYX_ERR(0, 757, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_assignment_matrix", 1, 2, 2, 1); __PYX_ERR(0, 765, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_assignment_matrix") < 0)) __PYX_ERR(0, 757, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_assignment_matrix") < 0)) __PYX_ERR(0, 765, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -7602,7 +7604,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_73get_assignment_matrix(PyObject *__
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_assignment_matrix", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 757, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_assignment_matrix", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 765, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_assignment_matrix", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -7624,7 +7626,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_72get_assignment_matrix(struct __pyx
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_assignment_matrix", 0);
 
-  /* "gedlibpy.pyx":772
+  /* "gedlibpy.pyx":780
  * 			.. note:: This function creates datas so use it if necessary.
  * 		"""
  * 		return self.c_env.getAssignmentMatrix(g, h)             # <<<<<<<<<<<<<<
@@ -7632,21 +7634,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_72get_assignment_matrix(struct __pyx
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 772, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 772, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 780, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 780, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getAssignmentMatrix(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 772, __pyx_L1_error)
+    __PYX_ERR(0, 780, __pyx_L1_error)
   }
-  __pyx_t_4 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_int_3e___(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 772, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_int_3e___(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 780, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":757
+  /* "gedlibpy.pyx":765
  * 
  * 
  * 	def get_assignment_matrix(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7665,7 +7667,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_72get_assignment_matrix(struct __pyx
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":775
+/* "gedlibpy.pyx":783
  * 
  * 
  * 	def get_all_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7705,11 +7707,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_75get_all_map(PyObject *__pyx_v_self
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_all_map", 1, 2, 2, 1); __PYX_ERR(0, 775, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_all_map", 1, 2, 2, 1); __PYX_ERR(0, 783, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_all_map") < 0)) __PYX_ERR(0, 775, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_all_map") < 0)) __PYX_ERR(0, 783, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -7722,7 +7724,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_75get_all_map(PyObject *__pyx_v_self
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_all_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 775, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_all_map", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 783, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_all_map", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -7744,7 +7746,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_74get_all_map(struct __pyx_obj_8gedl
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_all_map", 0);
 
-  /* "gedlibpy.pyx":790
+  /* "gedlibpy.pyx":798
  * 			.. note:: This function duplicates data so please don't use it. I also don't know how to connect the two map to reconstruct the adjacence matrix. Please come back when I know how it's work !
  * 		"""
  * 		return self.c_env.getAllMap(g, h)             # <<<<<<<<<<<<<<
@@ -7752,21 +7754,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_74get_all_map(struct __pyx_obj_8gedl
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 790, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 790, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 798, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 798, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getAllMap(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 790, __pyx_L1_error)
+    __PYX_ERR(0, 798, __pyx_L1_error)
   }
-  __pyx_t_4 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_npy_uint64_3e___(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 790, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_npy_uint64_3e___(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 798, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":775
+  /* "gedlibpy.pyx":783
  * 
  * 
  * 	def get_all_map(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7785,7 +7787,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_74get_all_map(struct __pyx_obj_8gedl
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":793
+/* "gedlibpy.pyx":801
  * 
  * 
  * 	def get_runtime(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7825,11 +7827,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_77get_runtime(PyObject *__pyx_v_self
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_runtime", 1, 2, 2, 1); __PYX_ERR(0, 793, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_runtime", 1, 2, 2, 1); __PYX_ERR(0, 801, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_runtime") < 0)) __PYX_ERR(0, 793, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_runtime") < 0)) __PYX_ERR(0, 801, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -7842,7 +7844,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_77get_runtime(PyObject *__pyx_v_self
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_runtime", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 793, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_runtime", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 801, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_runtime", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -7864,7 +7866,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_76get_runtime(struct __pyx_obj_8gedl
   PyObject *__pyx_t_4 = NULL;
   __Pyx_RefNannySetupContext("get_runtime", 0);
 
-  /* "gedlibpy.pyx":808
+  /* "gedlibpy.pyx":816
  * 			.. note:: Python is a bit longer than C++ due to the functions's encapsulate.
  * 		"""
  * 		return self.c_env.getRuntime(g,h)             # <<<<<<<<<<<<<<
@@ -7872,21 +7874,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_76get_runtime(struct __pyx_obj_8gedl
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyInt_As_size_t(__pyx_v_g); if (unlikely((__pyx_t_1 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 816, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_As_size_t(__pyx_v_h); if (unlikely((__pyx_t_2 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 816, __pyx_L1_error)
   try {
     __pyx_t_3 = __pyx_v_self->c_env->getRuntime(__pyx_t_1, __pyx_t_2);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 808, __pyx_L1_error)
+    __PYX_ERR(0, 816, __pyx_L1_error)
   }
-  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 808, __pyx_L1_error)
+  __pyx_t_4 = PyFloat_FromDouble(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 816, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __pyx_r = __pyx_t_4;
   __pyx_t_4 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":793
+  /* "gedlibpy.pyx":801
  * 
  * 
  * 	def get_runtime(self, g, h) :             # <<<<<<<<<<<<<<
@@ -7905,7 +7907,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_76get_runtime(struct __pyx_obj_8gedl
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":811
+/* "gedlibpy.pyx":819
  * 
  * 
  * 	def quasimetric_cost(self) :             # <<<<<<<<<<<<<<
@@ -7934,7 +7936,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_78quasimetric_cost(struct __pyx_obj_
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("quasimetric_cost", 0);
 
-  /* "gedlibpy.pyx":825
+  /* "gedlibpy.pyx":833
  * 			.. warning:: run_method() between the same two graph must be called before this function.
  * 		"""
  * 		return self.c_env.quasimetricCosts()             # <<<<<<<<<<<<<<
@@ -7946,15 +7948,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_78quasimetric_cost(struct __pyx_obj_
     __pyx_t_1 = __pyx_v_self->c_env->quasimetricCosts();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 825, __pyx_L1_error)
+    __PYX_ERR(0, 833, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_PyBool_FromLong(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 825, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyBool_FromLong(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 833, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":811
+  /* "gedlibpy.pyx":819
  * 
  * 
  * 	def quasimetric_cost(self) :             # <<<<<<<<<<<<<<
@@ -7973,7 +7975,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_78quasimetric_cost(struct __pyx_obj_
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":828
+/* "gedlibpy.pyx":836
  * 
  * 
  * 	def hungarian_LSAP(self, matrix_cost) :             # <<<<<<<<<<<<<<
@@ -8003,7 +8005,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_80hungarian_LSAP(struct __pyx_obj_8g
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("hungarian_LSAP", 0);
 
-  /* "gedlibpy.pyx":839
+  /* "gedlibpy.pyx":847
  * 			.. seealso:: hungarian_LSAPE()
  * 		"""
  * 		return self.c_env.hungarianLSAP(matrix_cost)             # <<<<<<<<<<<<<<
@@ -8011,20 +8013,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_80hungarian_LSAP(struct __pyx_obj_8g
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __pyx_convert_vector_from_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_v_matrix_cost); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 839, __pyx_L1_error)
+  __pyx_t_1 = __pyx_convert_vector_from_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_v_matrix_cost); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 847, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->hungarianLSAP(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 839, __pyx_L1_error)
+    __PYX_ERR(0, 847, __pyx_L1_error)
   }
-  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 839, __pyx_L1_error)
+  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_size_t_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 847, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":828
+  /* "gedlibpy.pyx":836
  * 
  * 
  * 	def hungarian_LSAP(self, matrix_cost) :             # <<<<<<<<<<<<<<
@@ -8043,7 +8045,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_80hungarian_LSAP(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":842
+/* "gedlibpy.pyx":850
  * 
  * 
  * 	def hungarian_LSAPE(self, matrix_cost) :             # <<<<<<<<<<<<<<
@@ -8073,7 +8075,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_82hungarian_LSAPE(struct __pyx_obj_8
   PyObject *__pyx_t_3 = NULL;
   __Pyx_RefNannySetupContext("hungarian_LSAPE", 0);
 
-  /* "gedlibpy.pyx":853
+  /* "gedlibpy.pyx":861
  * 			.. seealso:: hungarian_LSAP()
  * 		"""
  * 		return self.c_env.hungarianLSAPE(matrix_cost)             # <<<<<<<<<<<<<<
@@ -8081,20 +8083,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_82hungarian_LSAPE(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __pyx_convert_vector_from_py_std_3a__3a_vector_3c_double_3e___(__pyx_v_matrix_cost); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 853, __pyx_L1_error)
+  __pyx_t_1 = __pyx_convert_vector_from_py_std_3a__3a_vector_3c_double_3e___(__pyx_v_matrix_cost); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 861, __pyx_L1_error)
   try {
     __pyx_t_2 = __pyx_v_self->c_env->hungarianLSAPE(__pyx_t_1);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 853, __pyx_L1_error)
+    __PYX_ERR(0, 861, __pyx_L1_error)
   }
-  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_double_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 853, __pyx_L1_error)
+  __pyx_t_3 = __pyx_convert_vector_to_py_std_3a__3a_vector_3c_double_3e___(__pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 861, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_r = __pyx_t_3;
   __pyx_t_3 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":842
+  /* "gedlibpy.pyx":850
  * 
  * 
  * 	def hungarian_LSAPE(self, matrix_cost) :             # <<<<<<<<<<<<<<
@@ -8113,7 +8115,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_82hungarian_LSAPE(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":856
+/* "gedlibpy.pyx":864
  * 
  * 
  * 	def add_random_graph(self, name, classe, list_of_nodes, list_of_edges, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -8163,19 +8165,19 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_85add_random_graph(PyObject *__pyx_v
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_classe)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 1); __PYX_ERR(0, 856, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 1); __PYX_ERR(0, 864, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_list_of_nodes)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 2); __PYX_ERR(0, 856, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 2); __PYX_ERR(0, 864, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_list_of_edges)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 3); __PYX_ERR(0, 856, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, 3); __PYX_ERR(0, 864, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  4:
@@ -8185,7 +8187,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_85add_random_graph(PyObject *__pyx_v
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_random_graph") < 0)) __PYX_ERR(0, 856, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_random_graph") < 0)) __PYX_ERR(0, 864, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -8207,7 +8209,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_85add_random_graph(PyObject *__pyx_v
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 856, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_random_graph", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 864, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_random_graph", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -8239,14 +8241,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   PyObject *__pyx_t_11 = NULL;
   __Pyx_RefNannySetupContext("add_random_graph", 0);
 
-  /* "gedlibpy.pyx":876
+  /* "gedlibpy.pyx":884
  * 
  * 		"""
  * 		id = self.add_graph(name, classe)             # <<<<<<<<<<<<<<
  * 		for node in list_of_nodes:
  * 			self.add_node(id, node[0], node[1])
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 876, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 884, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   __pyx_t_4 = 0;
@@ -8263,7 +8265,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_name, __pyx_v_classe};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 876, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 884, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -8271,13 +8273,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_name, __pyx_v_classe};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 876, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 884, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 876, __pyx_L1_error)
+    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 884, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -8288,7 +8290,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     __Pyx_INCREF(__pyx_v_classe);
     __Pyx_GIVEREF(__pyx_v_classe);
     PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_classe);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 876, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 884, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   }
@@ -8296,7 +8298,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   __pyx_v_id = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":877
+  /* "gedlibpy.pyx":885
  * 		"""
  * 		id = self.add_graph(name, classe)
  * 		for node in list_of_nodes:             # <<<<<<<<<<<<<<
@@ -8307,26 +8309,26 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     __pyx_t_1 = __pyx_v_list_of_nodes; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
     __pyx_t_7 = NULL;
   } else {
-    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_list_of_nodes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 877, __pyx_L1_error)
+    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_list_of_nodes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 885, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 877, __pyx_L1_error)
+    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 885, __pyx_L1_error)
   }
   for (;;) {
     if (likely(!__pyx_t_7)) {
       if (likely(PyList_CheckExact(__pyx_t_1))) {
         if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 877, __pyx_L1_error)
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 885, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 877, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 885, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       } else {
         if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 877, __pyx_L1_error)
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 885, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 877, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 885, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       }
@@ -8336,7 +8338,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 877, __pyx_L1_error)
+          else __PYX_ERR(0, 885, __pyx_L1_error)
         }
         break;
       }
@@ -8345,18 +8347,18 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     __Pyx_XDECREF_SET(__pyx_v_node, __pyx_t_2);
     __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":878
+    /* "gedlibpy.pyx":886
  * 		id = self.add_graph(name, classe)
  * 		for node in list_of_nodes:
  * 			self.add_node(id, node[0], node[1])             # <<<<<<<<<<<<<<
  * 		for edge in list_of_edges:
  * 			self.add_edge(id, edge[0], edge[1], edge[2], ignore_duplicates)
  */
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 878, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 886, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_node, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 878, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_node, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 886, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_8 = __Pyx_GetItemInt(__pyx_v_node, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 878, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_GetItemInt(__pyx_v_node, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 886, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_8);
     __pyx_t_9 = NULL;
     __pyx_t_4 = 0;
@@ -8373,7 +8375,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_5)) {
       PyObject *__pyx_temp[4] = {__pyx_t_9, __pyx_v_id, __pyx_t_3, __pyx_t_8};
-      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 878, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 886, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
@@ -8383,7 +8385,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
       PyObject *__pyx_temp[4] = {__pyx_t_9, __pyx_v_id, __pyx_t_3, __pyx_t_8};
-      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 878, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 886, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
@@ -8391,7 +8393,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     } else
     #endif
     {
-      __pyx_t_10 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 878, __pyx_L1_error)
+      __pyx_t_10 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 886, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_10);
       if (__pyx_t_9) {
         __Pyx_GIVEREF(__pyx_t_9); PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_t_9); __pyx_t_9 = NULL;
@@ -8405,14 +8407,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
       PyTuple_SET_ITEM(__pyx_t_10, 2+__pyx_t_4, __pyx_t_8);
       __pyx_t_3 = 0;
       __pyx_t_8 = 0;
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_10, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 878, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_10, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 886, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
     }
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":877
+    /* "gedlibpy.pyx":885
  * 		"""
  * 		id = self.add_graph(name, classe)
  * 		for node in list_of_nodes:             # <<<<<<<<<<<<<<
@@ -8422,7 +8424,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":879
+  /* "gedlibpy.pyx":887
  * 		for node in list_of_nodes:
  * 			self.add_node(id, node[0], node[1])
  * 		for edge in list_of_edges:             # <<<<<<<<<<<<<<
@@ -8433,26 +8435,26 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     __pyx_t_1 = __pyx_v_list_of_edges; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
     __pyx_t_7 = NULL;
   } else {
-    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_list_of_edges); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 879, __pyx_L1_error)
+    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_list_of_edges); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 887, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 879, __pyx_L1_error)
+    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 887, __pyx_L1_error)
   }
   for (;;) {
     if (likely(!__pyx_t_7)) {
       if (likely(PyList_CheckExact(__pyx_t_1))) {
         if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 879, __pyx_L1_error)
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 887, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 879, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       } else {
         if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 879, __pyx_L1_error)
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 887, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 879, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       }
@@ -8462,7 +8464,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 879, __pyx_L1_error)
+          else __PYX_ERR(0, 887, __pyx_L1_error)
         }
         break;
       }
@@ -8471,20 +8473,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     __Pyx_XDECREF_SET(__pyx_v_edge, __pyx_t_2);
     __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":880
+    /* "gedlibpy.pyx":888
  * 			self.add_node(id, node[0], node[1])
  * 		for edge in list_of_edges:
  * 			self.add_edge(id, edge[0], edge[1], edge[2], ignore_duplicates)             # <<<<<<<<<<<<<<
  * 		return id
  * 
  */
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 880, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 888, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 880, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 888, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_8 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 880, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 888, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_edge, 2, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 880, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_edge, 2, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 888, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __pyx_t_9 = NULL;
     __pyx_t_4 = 0;
@@ -8501,7 +8503,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_5)) {
       PyObject *__pyx_temp[6] = {__pyx_t_9, __pyx_v_id, __pyx_t_10, __pyx_t_8, __pyx_t_3, __pyx_v_ignore_duplicates};
-      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 5+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 880, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 5+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 888, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
@@ -8512,7 +8514,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
       PyObject *__pyx_temp[6] = {__pyx_t_9, __pyx_v_id, __pyx_t_10, __pyx_t_8, __pyx_t_3, __pyx_v_ignore_duplicates};
-      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 5+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 880, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 5+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 888, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
@@ -8521,7 +8523,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
     } else
     #endif
     {
-      __pyx_t_11 = PyTuple_New(5+__pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 880, __pyx_L1_error)
+      __pyx_t_11 = PyTuple_New(5+__pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 888, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_11);
       if (__pyx_t_9) {
         __Pyx_GIVEREF(__pyx_t_9); PyTuple_SET_ITEM(__pyx_t_11, 0, __pyx_t_9); __pyx_t_9 = NULL;
@@ -8541,14 +8543,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
       __pyx_t_10 = 0;
       __pyx_t_8 = 0;
       __pyx_t_3 = 0;
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_11, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 880, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_11, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 888, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
     }
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":879
+    /* "gedlibpy.pyx":887
  * 		for node in list_of_nodes:
  * 			self.add_node(id, node[0], node[1])
  * 		for edge in list_of_edges:             # <<<<<<<<<<<<<<
@@ -8558,7 +8560,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":881
+  /* "gedlibpy.pyx":889
  * 		for edge in list_of_edges:
  * 			self.add_edge(id, edge[0], edge[1], edge[2], ignore_duplicates)
  * 		return id             # <<<<<<<<<<<<<<
@@ -8570,7 +8572,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   __pyx_r = __pyx_v_id;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":856
+  /* "gedlibpy.pyx":864
  * 
  * 
  * 	def add_random_graph(self, name, classe, list_of_nodes, list_of_edges, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -8599,7 +8601,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_84add_random_graph(struct __pyx_obj_
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":884
+/* "gedlibpy.pyx":892
  * 
  * 
  * 	def add_nx_graph(self, g, classe, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -8643,7 +8645,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_87add_nx_graph(PyObject *__pyx_v_sel
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_classe)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("add_nx_graph", 0, 2, 3, 1); __PYX_ERR(0, 884, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("add_nx_graph", 0, 2, 3, 1); __PYX_ERR(0, 892, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
@@ -8653,7 +8655,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_87add_nx_graph(PyObject *__pyx_v_sel
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_nx_graph") < 0)) __PYX_ERR(0, 884, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "add_nx_graph") < 0)) __PYX_ERR(0, 892, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -8671,7 +8673,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_87add_nx_graph(PyObject *__pyx_v_sel
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("add_nx_graph", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 884, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("add_nx_graph", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 892, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.add_nx_graph", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -8706,16 +8708,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   PyObject *__pyx_t_14 = NULL;
   __Pyx_RefNannySetupContext("add_nx_graph", 0);
 
-  /* "gedlibpy.pyx":898
+  /* "gedlibpy.pyx":906
  * 
  * 		"""
  * 		id = self.add_graph(g.name, classe)             # <<<<<<<<<<<<<<
  * 		for node in g.nodes:
  * 			self.add_node(id, str(node), g.nodes[node])
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 898, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 906, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 898, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 906, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_4 = NULL;
   __pyx_t_5 = 0;
@@ -8732,7 +8734,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_3, __pyx_v_classe};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 898, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 906, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
@@ -8741,14 +8743,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_3, __pyx_v_classe};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 898, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 906, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   } else
   #endif
   {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 898, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 906, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     if (__pyx_t_4) {
       __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_4); __pyx_t_4 = NULL;
@@ -8759,7 +8761,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     __Pyx_GIVEREF(__pyx_v_classe);
     PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_v_classe);
     __pyx_t_3 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 898, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 906, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
@@ -8767,22 +8769,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   __pyx_v_id = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":899
+  /* "gedlibpy.pyx":907
  * 		"""
  * 		id = self.add_graph(g.name, classe)
  * 		for node in g.nodes:             # <<<<<<<<<<<<<<
  * 			self.add_node(id, str(node), g.nodes[node])
  * 		for edge in g.edges:
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_nodes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 899, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_nodes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 907, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
     __pyx_t_2 = __pyx_t_1; __Pyx_INCREF(__pyx_t_2); __pyx_t_7 = 0;
     __pyx_t_8 = NULL;
   } else {
-    __pyx_t_7 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 899, __pyx_L1_error)
+    __pyx_t_7 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 907, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_8 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 899, __pyx_L1_error)
+    __pyx_t_8 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 907, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   for (;;) {
@@ -8790,17 +8792,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
       if (likely(PyList_CheckExact(__pyx_t_2))) {
         if (__pyx_t_7 >= PyList_GET_SIZE(__pyx_t_2)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 899, __pyx_L1_error)
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 907, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 899, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 907, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       } else {
         if (__pyx_t_7 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 899, __pyx_L1_error)
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 907, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 899, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 907, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       }
@@ -8810,7 +8812,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 899, __pyx_L1_error)
+          else __PYX_ERR(0, 907, __pyx_L1_error)
         }
         break;
       }
@@ -8819,20 +8821,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     __Pyx_XDECREF_SET(__pyx_v_node, __pyx_t_1);
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":900
+    /* "gedlibpy.pyx":908
  * 		id = self.add_graph(g.name, classe)
  * 		for node in g.nodes:
  * 			self.add_node(id, str(node), g.nodes[node])             # <<<<<<<<<<<<<<
  * 		for edge in g.edges:
  * 			self.add_edge(id, str(edge[0]), str(edge[1]), g.get_edge_data(edge[0], edge[1]), ignore_duplicates)
  */
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 900, __pyx_L1_error)
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 908, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_3 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_node); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 900, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_node); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 908, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_nodes); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 900, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_nodes); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 908, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_9 = __Pyx_PyObject_GetItem(__pyx_t_4, __pyx_v_node); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 900, __pyx_L1_error)
+    __pyx_t_9 = __Pyx_PyObject_GetItem(__pyx_t_4, __pyx_v_node); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 908, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_9);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __pyx_t_4 = NULL;
@@ -8850,7 +8852,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_6)) {
       PyObject *__pyx_temp[4] = {__pyx_t_4, __pyx_v_id, __pyx_t_3, __pyx_t_9};
-      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 3+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 900, __pyx_L1_error)
+      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 3+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 908, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
       __Pyx_GOTREF(__pyx_t_1);
       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
@@ -8860,7 +8862,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
       PyObject *__pyx_temp[4] = {__pyx_t_4, __pyx_v_id, __pyx_t_3, __pyx_t_9};
-      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 3+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 900, __pyx_L1_error)
+      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 3+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 908, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
       __Pyx_GOTREF(__pyx_t_1);
       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
@@ -8868,7 +8870,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     } else
     #endif
     {
-      __pyx_t_10 = PyTuple_New(3+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 900, __pyx_L1_error)
+      __pyx_t_10 = PyTuple_New(3+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 908, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_10);
       if (__pyx_t_4) {
         __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_t_4); __pyx_t_4 = NULL;
@@ -8882,14 +8884,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
       PyTuple_SET_ITEM(__pyx_t_10, 2+__pyx_t_5, __pyx_t_9);
       __pyx_t_3 = 0;
       __pyx_t_9 = 0;
-      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_10, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 900, __pyx_L1_error)
+      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_10, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 908, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_1);
       __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
     }
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":899
+    /* "gedlibpy.pyx":907
  * 		"""
  * 		id = self.add_graph(g.name, classe)
  * 		for node in g.nodes:             # <<<<<<<<<<<<<<
@@ -8899,22 +8901,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   }
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":901
+  /* "gedlibpy.pyx":909
  * 		for node in g.nodes:
  * 			self.add_node(id, str(node), g.nodes[node])
  * 		for edge in g.edges:             # <<<<<<<<<<<<<<
  * 			self.add_edge(id, str(edge[0]), str(edge[1]), g.get_edge_data(edge[0], edge[1]), ignore_duplicates)
  * 		return id
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_edges); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 901, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_edges); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 909, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
     __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_7 = 0;
     __pyx_t_8 = NULL;
   } else {
-    __pyx_t_7 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 901, __pyx_L1_error)
+    __pyx_t_7 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 909, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_8 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 901, __pyx_L1_error)
+    __pyx_t_8 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 909, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   for (;;) {
@@ -8922,17 +8924,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
       if (likely(PyList_CheckExact(__pyx_t_1))) {
         if (__pyx_t_7 >= PyList_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_7); __Pyx_INCREF(__pyx_t_2); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 901, __pyx_L1_error)
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_7); __Pyx_INCREF(__pyx_t_2); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 909, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 901, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 909, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       } else {
         if (__pyx_t_7 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_7); __Pyx_INCREF(__pyx_t_2); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 901, __pyx_L1_error)
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_7); __Pyx_INCREF(__pyx_t_2); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 909, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 901, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 909, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       }
@@ -8942,7 +8944,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 901, __pyx_L1_error)
+          else __PYX_ERR(0, 909, __pyx_L1_error)
         }
         break;
       }
@@ -8951,30 +8953,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     __Pyx_XDECREF_SET(__pyx_v_edge, __pyx_t_2);
     __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":902
+    /* "gedlibpy.pyx":910
  * 			self.add_node(id, str(node), g.nodes[node])
  * 		for edge in g.edges:
  * 			self.add_edge(id, str(edge[0]), str(edge[1]), g.get_edge_data(edge[0], edge[1]), ignore_duplicates)             # <<<<<<<<<<<<<<
  * 		return id
  * 
  */
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_10); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_10); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_9);
     __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_3 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_10); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_10); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_get_edge_data); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_g, __pyx_n_s_get_edge_data); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_11 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_11 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_11);
-    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 902, __pyx_L1_error)
+    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 910, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_12);
     __pyx_t_13 = NULL;
     __pyx_t_5 = 0;
@@ -8991,7 +8993,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_4)) {
       PyObject *__pyx_temp[3] = {__pyx_t_13, __pyx_t_11, __pyx_t_12};
-      __pyx_t_10 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_10 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_13); __pyx_t_13 = 0;
       __Pyx_GOTREF(__pyx_t_10);
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
@@ -9001,7 +9003,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_4)) {
       PyObject *__pyx_temp[3] = {__pyx_t_13, __pyx_t_11, __pyx_t_12};
-      __pyx_t_10 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_10 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_13); __pyx_t_13 = 0;
       __Pyx_GOTREF(__pyx_t_10);
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
@@ -9009,7 +9011,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     } else
     #endif
     {
-      __pyx_t_14 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_14 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_14);
       if (__pyx_t_13) {
         __Pyx_GIVEREF(__pyx_t_13); PyTuple_SET_ITEM(__pyx_t_14, 0, __pyx_t_13); __pyx_t_13 = NULL;
@@ -9020,7 +9022,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
       PyTuple_SET_ITEM(__pyx_t_14, 1+__pyx_t_5, __pyx_t_12);
       __pyx_t_11 = 0;
       __pyx_t_12 = 0;
-      __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_14, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_14, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_10);
       __Pyx_DECREF(__pyx_t_14); __pyx_t_14 = 0;
     }
@@ -9040,7 +9042,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_6)) {
       PyObject *__pyx_temp[6] = {__pyx_t_4, __pyx_v_id, __pyx_t_9, __pyx_t_3, __pyx_t_10, __pyx_v_ignore_duplicates};
-      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 5+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 5+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
@@ -9051,7 +9053,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
       PyObject *__pyx_temp[6] = {__pyx_t_4, __pyx_v_id, __pyx_t_9, __pyx_t_3, __pyx_t_10, __pyx_v_ignore_duplicates};
-      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 5+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_5, 5+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
@@ -9060,7 +9062,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
     } else
     #endif
     {
-      __pyx_t_14 = PyTuple_New(5+__pyx_t_5); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_14 = PyTuple_New(5+__pyx_t_5); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_14);
       if (__pyx_t_4) {
         __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_14, 0, __pyx_t_4); __pyx_t_4 = NULL;
@@ -9080,14 +9082,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
       __pyx_t_9 = 0;
       __pyx_t_3 = 0;
       __pyx_t_10 = 0;
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_14, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 902, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_14, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 910, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_14); __pyx_t_14 = 0;
     }
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":901
+    /* "gedlibpy.pyx":909
  * 		for node in g.nodes:
  * 			self.add_node(id, str(node), g.nodes[node])
  * 		for edge in g.edges:             # <<<<<<<<<<<<<<
@@ -9097,7 +9099,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":903
+  /* "gedlibpy.pyx":911
  * 		for edge in g.edges:
  * 			self.add_edge(id, str(edge[0]), str(edge[1]), g.get_edge_data(edge[0], edge[1]), ignore_duplicates)
  * 		return id             # <<<<<<<<<<<<<<
@@ -9109,7 +9111,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   __pyx_r = __pyx_v_id;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":884
+  /* "gedlibpy.pyx":892
  * 
  * 
  * 	def add_nx_graph(self, g, classe, ignore_duplicates=True) :             # <<<<<<<<<<<<<<
@@ -9141,7 +9143,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_86add_nx_graph(struct __pyx_obj_8ged
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":906
+/* "gedlibpy.pyx":914
  * 
  * 
  * 	def compute_ged_on_two_graphs(self, g1, g2, edit_cost, method, options, init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -9194,25 +9196,25 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_89compute_ged_on_two_graphs(PyObject
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_g2)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 1); __PYX_ERR(0, 906, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 1); __PYX_ERR(0, 914, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edit_cost)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 2); __PYX_ERR(0, 906, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 2); __PYX_ERR(0, 914, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_method)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 3); __PYX_ERR(0, 906, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 3); __PYX_ERR(0, 914, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  4:
         if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_options)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 4); __PYX_ERR(0, 906, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, 4); __PYX_ERR(0, 914, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  5:
@@ -9222,7 +9224,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_89compute_ged_on_two_graphs(PyObject
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_ged_on_two_graphs") < 0)) __PYX_ERR(0, 906, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_ged_on_two_graphs") < 0)) __PYX_ERR(0, 914, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -9246,7 +9248,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_89compute_ged_on_two_graphs(PyObject
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 906, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("compute_ged_on_two_graphs", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 914, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.compute_ged_on_two_graphs", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -9274,14 +9276,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("compute_ged_on_two_graphs", 0);
 
-  /* "gedlibpy.pyx":929
+  /* "gedlibpy.pyx":937
  * 
  * 		"""
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
  * 			self.restart_env()
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 929, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 937, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9295,21 +9297,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 929, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 937, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 929, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 937, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   if (__pyx_t_4) {
 
-    /* "gedlibpy.pyx":930
+    /* "gedlibpy.pyx":938
  * 		"""
  * 		if self.is_initialized() :
  * 			self.restart_env()             # <<<<<<<<<<<<<<
  * 
  * 		g = self.add_nx_graph(g1, "")
  */
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 930, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 938, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __pyx_t_3 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9323,12 +9325,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     }
     __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 930, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 938, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":929
+    /* "gedlibpy.pyx":937
  * 
  * 		"""
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
@@ -9337,14 +9339,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
  */
   }
 
-  /* "gedlibpy.pyx":932
+  /* "gedlibpy.pyx":940
  * 			self.restart_env()
  * 
  * 		g = self.add_nx_graph(g1, "")             # <<<<<<<<<<<<<<
  * 		h = self.add_nx_graph(g2, "")
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 932, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 940, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   __pyx_t_5 = 0;
@@ -9361,7 +9363,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_g1, __pyx_kp_u_};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 932, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 940, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9369,13 +9371,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_g1, __pyx_kp_u_};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 932, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 940, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 932, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 940, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -9386,7 +9388,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_kp_u_);
     __Pyx_GIVEREF(__pyx_kp_u_);
     PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_kp_u_);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 932, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 940, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
@@ -9394,14 +9396,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __pyx_v_g = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":933
+  /* "gedlibpy.pyx":941
  * 
  * 		g = self.add_nx_graph(g1, "")
  * 		h = self.add_nx_graph(g2, "")             # <<<<<<<<<<<<<<
  * 
  * 		self.set_edit_cost(edit_cost)
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 933, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 941, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_6 = NULL;
   __pyx_t_5 = 0;
@@ -9418,7 +9420,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g2, __pyx_kp_u_};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 933, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 941, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9426,13 +9428,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g2, __pyx_kp_u_};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 933, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 941, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 933, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 941, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     if (__pyx_t_6) {
       __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
@@ -9443,7 +9445,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_kp_u_);
     __Pyx_GIVEREF(__pyx_kp_u_);
     PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_kp_u_);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 933, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 941, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   }
@@ -9451,14 +9453,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __pyx_v_h = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":935
+  /* "gedlibpy.pyx":943
  * 		h = self.add_nx_graph(g2, "")
  * 
  * 		self.set_edit_cost(edit_cost)             # <<<<<<<<<<<<<<
  * 		self.init(init_option)
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 935, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 943, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9472,19 +9474,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edit_cost) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edit_cost);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 935, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 943, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":936
+  /* "gedlibpy.pyx":944
  * 
  * 		self.set_edit_cost(edit_cost)
  * 		self.init(init_option)             # <<<<<<<<<<<<<<
  * 
  * 		self.set_method(method, options)
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 936, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 944, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9498,19 +9500,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_init_option) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_init_option);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 936, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 944, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":938
+  /* "gedlibpy.pyx":946
  * 		self.init(init_option)
  * 
  * 		self.set_method(method, options)             # <<<<<<<<<<<<<<
  * 		self.init_method()
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 938, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 946, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   __pyx_t_5 = 0;
@@ -9527,7 +9529,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_method, __pyx_v_options};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 938, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 946, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9535,13 +9537,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_method, __pyx_v_options};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 938, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 946, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 938, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 946, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -9552,21 +9554,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_v_options);
     __Pyx_GIVEREF(__pyx_v_options);
     PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_v_options);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 938, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 946, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":939
+  /* "gedlibpy.pyx":947
  * 
  * 		self.set_method(method, options)
  * 		self.init_method()             # <<<<<<<<<<<<<<
  * 
  * 		resDistance = 0
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 939, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 947, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9580,12 +9582,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   }
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_6) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 939, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 947, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":941
+  /* "gedlibpy.pyx":949
  * 		self.init_method()
  * 
  * 		resDistance = 0             # <<<<<<<<<<<<<<
@@ -9595,26 +9597,26 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __Pyx_INCREF(__pyx_int_0);
   __pyx_v_resDistance = __pyx_int_0;
 
-  /* "gedlibpy.pyx":942
+  /* "gedlibpy.pyx":950
  * 
  * 		resDistance = 0
  * 		resMapping = []             # <<<<<<<<<<<<<<
  * 		self.run_method(g, h)
  * 		resDistance = self.get_upper_bound(g, h)
  */
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 942, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 950, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_v_resMapping = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":943
+  /* "gedlibpy.pyx":951
  * 		resDistance = 0
  * 		resMapping = []
  * 		self.run_method(g, h)             # <<<<<<<<<<<<<<
  * 		resDistance = self.get_upper_bound(g, h)
  * 		resMapping = self.get_node_map(g, h)
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 943, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 951, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_6 = NULL;
   __pyx_t_5 = 0;
@@ -9631,7 +9633,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 951, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9639,13 +9641,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 951, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 951, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     if (__pyx_t_6) {
       __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
@@ -9656,21 +9658,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_v_h);
     __Pyx_GIVEREF(__pyx_v_h);
     PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_v_h);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 951, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   }
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":944
+  /* "gedlibpy.pyx":952
  * 		resMapping = []
  * 		self.run_method(g, h)
  * 		resDistance = self.get_upper_bound(g, h)             # <<<<<<<<<<<<<<
  * 		resMapping = self.get_node_map(g, h)
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_upper_bound); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 944, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_upper_bound); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 952, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   __pyx_t_5 = 0;
@@ -9687,7 +9689,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 944, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 952, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9695,13 +9697,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 944, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 952, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 952, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -9712,7 +9714,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_v_h);
     __Pyx_GIVEREF(__pyx_v_h);
     PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_v_h);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 944, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 952, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
@@ -9720,14 +9722,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __Pyx_DECREF_SET(__pyx_v_resDistance, __pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":945
+  /* "gedlibpy.pyx":953
  * 		self.run_method(g, h)
  * 		resDistance = self.get_upper_bound(g, h)
  * 		resMapping = self.get_node_map(g, h)             # <<<<<<<<<<<<<<
  * 
  * 		return resDistance, resMapping
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_node_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 945, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_node_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 953, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_6 = NULL;
   __pyx_t_5 = 0;
@@ -9744,7 +9746,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 945, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 953, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -9752,13 +9754,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_g, __pyx_v_h};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 945, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 953, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 945, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 953, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     if (__pyx_t_6) {
       __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
@@ -9769,7 +9771,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
     __Pyx_INCREF(__pyx_v_h);
     __Pyx_GIVEREF(__pyx_v_h);
     PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_v_h);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 945, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 953, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   }
@@ -9777,7 +9779,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __Pyx_DECREF_SET(__pyx_v_resMapping, __pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":947
+  /* "gedlibpy.pyx":955
  * 		resMapping = self.get_node_map(g, h)
  * 
  * 		return resDistance, resMapping             # <<<<<<<<<<<<<<
@@ -9785,7 +9787,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 947, __pyx_L1_error)
+  __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 955, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_INCREF(__pyx_v_resDistance);
   __Pyx_GIVEREF(__pyx_v_resDistance);
@@ -9797,7 +9799,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":906
+  /* "gedlibpy.pyx":914
  * 
  * 
  * 	def compute_ged_on_two_graphs(self, g1, g2, edit_cost, method, options, init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -9823,7 +9825,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_88compute_ged_on_two_graphs(struct _
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":950
+/* "gedlibpy.pyx":958
  * 
  * 
  * 	def compute_edit_distance_on_nx_graphs(self, dataset, classes, edit_cost, method, options, init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -9876,25 +9878,25 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_91compute_edit_distance_on_nx_graphs
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_classes)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 1); __PYX_ERR(0, 950, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 1); __PYX_ERR(0, 958, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edit_cost)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 2); __PYX_ERR(0, 950, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 2); __PYX_ERR(0, 958, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_method)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 3); __PYX_ERR(0, 950, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 3); __PYX_ERR(0, 958, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  4:
         if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_options)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 4); __PYX_ERR(0, 950, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, 4); __PYX_ERR(0, 958, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  5:
@@ -9904,7 +9906,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_91compute_edit_distance_on_nx_graphs
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_edit_distance_on_nx_graphs") < 0)) __PYX_ERR(0, 950, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_edit_distance_on_nx_graphs") < 0)) __PYX_ERR(0, 958, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -9928,7 +9930,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_91compute_edit_distance_on_nx_graphs
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 950, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_nx_graphs", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 958, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.compute_edit_distance_on_nx_graphs", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -9964,14 +9966,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   PyObject *__pyx_t_12 = NULL;
   __Pyx_RefNannySetupContext("compute_edit_distance_on_nx_graphs", 0);
 
-  /* "gedlibpy.pyx":974
+  /* "gedlibpy.pyx":982
  * 
  * 		"""
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
  * 			self.restart_env()
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 974, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 982, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -9985,21 +9987,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 974, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 982, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 974, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 982, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   if (__pyx_t_4) {
 
-    /* "gedlibpy.pyx":975
+    /* "gedlibpy.pyx":983
  * 		"""
  * 		if self.is_initialized() :
  * 			self.restart_env()             # <<<<<<<<<<<<<<
  * 
  * 		print("Loading graphs in progress...")
  */
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 975, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 983, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __pyx_t_3 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -10013,12 +10015,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     }
     __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 975, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 983, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":974
+    /* "gedlibpy.pyx":982
  * 
  * 		"""
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
@@ -10027,18 +10029,18 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
  */
   }
 
-  /* "gedlibpy.pyx":977
+  /* "gedlibpy.pyx":985
  * 			self.restart_env()
  * 
  * 		print("Loading graphs in progress...")             # <<<<<<<<<<<<<<
  * 		for graph in dataset :
  * 			self.add_nx_graph(graph, classes)
  */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 977, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 985, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":978
+  /* "gedlibpy.pyx":986
  * 
  * 		print("Loading graphs in progress...")
  * 		for graph in dataset :             # <<<<<<<<<<<<<<
@@ -10049,26 +10051,26 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     __pyx_t_1 = __pyx_v_dataset; __Pyx_INCREF(__pyx_t_1); __pyx_t_5 = 0;
     __pyx_t_6 = NULL;
   } else {
-    __pyx_t_5 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_dataset); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 978, __pyx_L1_error)
+    __pyx_t_5 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_dataset); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 986, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_6 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 978, __pyx_L1_error)
+    __pyx_t_6 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 986, __pyx_L1_error)
   }
   for (;;) {
     if (likely(!__pyx_t_6)) {
       if (likely(PyList_CheckExact(__pyx_t_1))) {
         if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_5); __Pyx_INCREF(__pyx_t_2); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 978, __pyx_L1_error)
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_5); __Pyx_INCREF(__pyx_t_2); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 986, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 978, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 986, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       } else {
         if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_5); __Pyx_INCREF(__pyx_t_2); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 978, __pyx_L1_error)
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_5); __Pyx_INCREF(__pyx_t_2); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 986, __pyx_L1_error)
         #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 978, __pyx_L1_error)
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 986, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_2);
         #endif
       }
@@ -10078,7 +10080,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 978, __pyx_L1_error)
+          else __PYX_ERR(0, 986, __pyx_L1_error)
         }
         break;
       }
@@ -10087,14 +10089,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     __Pyx_XDECREF_SET(__pyx_v_graph, __pyx_t_2);
     __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":979
+    /* "gedlibpy.pyx":987
  * 		print("Loading graphs in progress...")
  * 		for graph in dataset :
  * 			self.add_nx_graph(graph, classes)             # <<<<<<<<<<<<<<
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")
  */
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 979, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_nx_graph); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 987, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __pyx_t_7 = NULL;
     __pyx_t_8 = 0;
@@ -10111,7 +10113,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_3)) {
       PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_graph, __pyx_v_classes};
-      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 979, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 987, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
       __Pyx_GOTREF(__pyx_t_2);
     } else
@@ -10119,13 +10121,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
       PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_graph, __pyx_v_classes};
-      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 979, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 987, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
       __Pyx_GOTREF(__pyx_t_2);
     } else
     #endif
     {
-      __pyx_t_9 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 979, __pyx_L1_error)
+      __pyx_t_9 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 987, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_9);
       if (__pyx_t_7) {
         __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_7); __pyx_t_7 = NULL;
@@ -10136,14 +10138,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       __Pyx_INCREF(__pyx_v_classes);
       __Pyx_GIVEREF(__pyx_v_classes);
       PyTuple_SET_ITEM(__pyx_t_9, 1+__pyx_t_8, __pyx_v_classes);
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_9, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 979, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_9, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 987, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
     }
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":978
+    /* "gedlibpy.pyx":986
  * 
  * 		print("Loading graphs in progress...")
  * 		for graph in dataset :             # <<<<<<<<<<<<<<
@@ -10153,14 +10155,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":980
+  /* "gedlibpy.pyx":988
  * 		for graph in dataset :
  * 			self.add_nx_graph(graph, classes)
  * 		listID = self.graph_ids()             # <<<<<<<<<<<<<<
  * 		print("Graphs loaded ! ")
  * 		print("Number of graphs = " + str(listID[1]))
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_graph_ids); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 980, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_graph_ids); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 988, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -10174,51 +10176,51 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 980, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 988, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_listID = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":981
+  /* "gedlibpy.pyx":989
  * 			self.add_nx_graph(graph, classes)
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")             # <<<<<<<<<<<<<<
  * 		print("Number of graphs = " + str(listID[1]))
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 981, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 989, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":982
+  /* "gedlibpy.pyx":990
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")
  * 		print("Number of graphs = " + str(listID[1]))             # <<<<<<<<<<<<<<
  * 
  * 		self.set_edit_cost(edit_cost)
  */
-  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 982, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 990, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 982, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 990, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Number_of_graphs, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 982, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Number_of_graphs, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 990, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 982, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 990, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":984
+  /* "gedlibpy.pyx":992
  * 		print("Number of graphs = " + str(listID[1]))
  * 
  * 		self.set_edit_cost(edit_cost)             # <<<<<<<<<<<<<<
  * 		print("Initialization in progress...")
  * 		self.init(init_option)
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 984, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 992, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -10232,30 +10234,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_v_edit_cost) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_edit_cost);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 984, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 992, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":985
+  /* "gedlibpy.pyx":993
  * 
  * 		self.set_edit_cost(edit_cost)
  * 		print("Initialization in progress...")             # <<<<<<<<<<<<<<
  * 		self.init(init_option)
  * 		print("Initialization terminated !")
  */
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__6, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 985, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__6, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 993, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":986
+  /* "gedlibpy.pyx":994
  * 		self.set_edit_cost(edit_cost)
  * 		print("Initialization in progress...")
  * 		self.init(init_option)             # <<<<<<<<<<<<<<
  * 		print("Initialization terminated !")
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 986, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 994, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -10269,30 +10271,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_v_init_option) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_init_option);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 986, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 994, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":987
+  /* "gedlibpy.pyx":995
  * 		print("Initialization in progress...")
  * 		self.init(init_option)
  * 		print("Initialization terminated !")             # <<<<<<<<<<<<<<
  * 
  * 		self.set_method(method, options)
  */
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 987, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 995, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":989
+  /* "gedlibpy.pyx":997
  * 		print("Initialization terminated !")
  * 
  * 		self.set_method(method, options)             # <<<<<<<<<<<<<<
  * 		self.init_method()
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 989, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 997, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_3 = NULL;
   __pyx_t_8 = 0;
@@ -10309,7 +10311,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_1)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_method, __pyx_v_options};
-    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 989, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 997, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_2);
   } else
@@ -10317,13 +10319,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_1)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_method, __pyx_v_options};
-    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 989, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 997, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_2);
   } else
   #endif
   {
-    __pyx_t_9 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 989, __pyx_L1_error)
+    __pyx_t_9 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 997, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_9);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -10334,21 +10336,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     __Pyx_INCREF(__pyx_v_options);
     __Pyx_GIVEREF(__pyx_v_options);
     PyTuple_SET_ITEM(__pyx_t_9, 1+__pyx_t_8, __pyx_v_options);
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_9, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 989, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_9, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 997, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":990
+  /* "gedlibpy.pyx":998
  * 
  * 		self.set_method(method, options)
  * 		self.init_method()             # <<<<<<<<<<<<<<
  * 
  * 		resDistance = [[]]
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 990, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 998, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_9 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -10362,21 +10364,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __pyx_t_2 = (__pyx_t_9) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_9) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
   __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 990, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 998, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":992
+  /* "gedlibpy.pyx":1000
  * 		self.init_method()
  * 
  * 		resDistance = [[]]             # <<<<<<<<<<<<<<
  * 		resMapping = [[]]
  * 		for g in range(listID[0], listID[1]) :
  */
-  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 992, __pyx_L1_error)
+  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1000, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 992, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1000, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_GIVEREF(__pyx_t_2);
   PyList_SET_ITEM(__pyx_t_1, 0, __pyx_t_2);
@@ -10384,16 +10386,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   __pyx_v_resDistance = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":993
+  /* "gedlibpy.pyx":1001
  * 
  * 		resDistance = [[]]
  * 		resMapping = [[]]             # <<<<<<<<<<<<<<
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  */
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 993, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1001, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 993, __pyx_L1_error)
+  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1001, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_GIVEREF(__pyx_t_1);
   PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
@@ -10401,18 +10403,18 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   __pyx_v_resMapping = ((PyObject*)__pyx_t_2);
   __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":994
+  /* "gedlibpy.pyx":1002
  * 		resDistance = [[]]
  * 		resMapping = [[]]
  * 		for g in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :
  */
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 994, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1002, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 994, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1002, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 994, __pyx_L1_error)
+  __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1002, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_9);
   __Pyx_GIVEREF(__pyx_t_2);
   PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_2);
@@ -10420,16 +10422,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   PyTuple_SET_ITEM(__pyx_t_9, 1, __pyx_t_1);
   __pyx_t_2 = 0;
   __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_9, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 994, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_9, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1002, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
   if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
     __pyx_t_9 = __pyx_t_1; __Pyx_INCREF(__pyx_t_9); __pyx_t_5 = 0;
     __pyx_t_6 = NULL;
   } else {
-    __pyx_t_5 = -1; __pyx_t_9 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 994, __pyx_L1_error)
+    __pyx_t_5 = -1; __pyx_t_9 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1002, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_9);
-    __pyx_t_6 = Py_TYPE(__pyx_t_9)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 994, __pyx_L1_error)
+    __pyx_t_6 = Py_TYPE(__pyx_t_9)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1002, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   for (;;) {
@@ -10437,17 +10439,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       if (likely(PyList_CheckExact(__pyx_t_9))) {
         if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_9)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_9, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 994, __pyx_L1_error)
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_9, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1002, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_9, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 994, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_9, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1002, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       } else {
         if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_9)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_9, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 994, __pyx_L1_error)
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_9, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1002, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_9, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 994, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_9, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1002, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       }
@@ -10457,7 +10459,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 994, __pyx_L1_error)
+          else __PYX_ERR(0, 1002, __pyx_L1_error)
         }
         break;
       }
@@ -10466,38 +10468,38 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     __Pyx_XDECREF_SET(__pyx_v_g, __pyx_t_1);
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":995
+    /* "gedlibpy.pyx":1003
  * 		resMapping = [[]]
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")             # <<<<<<<<<<<<<<
  * 			for h in range(listID[0], listID[1]) :
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  */
-    __pyx_t_1 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_g); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 995, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_g); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1003, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Computation_between_graph, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 995, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Computation_between_graph, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1003, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_t_2, __pyx_kp_u_with_all_the_others_including_h); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 995, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_t_2, __pyx_kp_u_with_all_the_others_including_h); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1003, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 995, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1003, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":996
+    /* "gedlibpy.pyx":1004
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  * 				self.run_method(g, h)
  */
-    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 996, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1004, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 996, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1004, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 996, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1004, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __Pyx_GIVEREF(__pyx_t_2);
     PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
@@ -10505,16 +10507,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_t_1);
     __pyx_t_2 = 0;
     __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 996, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1004, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
       __pyx_t_3 = __pyx_t_1; __Pyx_INCREF(__pyx_t_3); __pyx_t_10 = 0;
       __pyx_t_11 = NULL;
     } else {
-      __pyx_t_10 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 996, __pyx_L1_error)
+      __pyx_t_10 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1004, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_11 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 996, __pyx_L1_error)
+      __pyx_t_11 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1004, __pyx_L1_error)
     }
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     for (;;) {
@@ -10522,17 +10524,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         if (likely(PyList_CheckExact(__pyx_t_3))) {
           if (__pyx_t_10 >= PyList_GET_SIZE(__pyx_t_3)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_1 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_1); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 996, __pyx_L1_error)
+          __pyx_t_1 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_1); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 1004, __pyx_L1_error)
           #else
-          __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 996, __pyx_L1_error)
+          __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1004, __pyx_L1_error)
           __Pyx_GOTREF(__pyx_t_1);
           #endif
         } else {
           if (__pyx_t_10 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_1); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 996, __pyx_L1_error)
+          __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_10); __Pyx_INCREF(__pyx_t_1); __pyx_t_10++; if (unlikely(0 < 0)) __PYX_ERR(0, 1004, __pyx_L1_error)
           #else
-          __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 996, __pyx_L1_error)
+          __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_10); __pyx_t_10++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1004, __pyx_L1_error)
           __Pyx_GOTREF(__pyx_t_1);
           #endif
         }
@@ -10542,7 +10544,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
           PyObject* exc_type = PyErr_Occurred();
           if (exc_type) {
             if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 996, __pyx_L1_error)
+            else __PYX_ERR(0, 1004, __pyx_L1_error)
           }
           break;
         }
@@ -10551,14 +10553,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       __Pyx_XDECREF_SET(__pyx_v_h, __pyx_t_1);
       __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":998
+      /* "gedlibpy.pyx":1006
  * 			for h in range(listID[0], listID[1]) :
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  * 				self.run_method(g, h)             # <<<<<<<<<<<<<<
  * 				resDistance[g][h] = self.get_upper_bound(g, h)
  * 				resMapping[g][h] = self.get_node_map(g, h)
  */
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 998, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1006, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __pyx_t_7 = NULL;
       __pyx_t_8 = 0;
@@ -10575,7 +10577,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCALL
       if (PyFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 998, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1006, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
@@ -10583,13 +10585,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCCALL
       if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 998, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1006, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
       #endif
       {
-        __pyx_t_12 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 998, __pyx_L1_error)
+        __pyx_t_12 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1006, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_12);
         if (__pyx_t_7) {
           __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_7); __pyx_t_7 = NULL;
@@ -10600,21 +10602,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         __Pyx_INCREF(__pyx_v_h);
         __Pyx_GIVEREF(__pyx_v_h);
         PyTuple_SET_ITEM(__pyx_t_12, 1+__pyx_t_8, __pyx_v_h);
-        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 998, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1006, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
       }
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":999
+      /* "gedlibpy.pyx":1007
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  * 				self.run_method(g, h)
  * 				resDistance[g][h] = self.get_upper_bound(g, h)             # <<<<<<<<<<<<<<
  * 				resMapping[g][h] = self.get_node_map(g, h)
  * 
  */
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_upper_bound); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 999, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_upper_bound); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1007, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __pyx_t_12 = NULL;
       __pyx_t_8 = 0;
@@ -10631,7 +10633,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCALL
       if (PyFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_12, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 999, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1007, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
@@ -10639,13 +10641,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCCALL
       if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_12, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 999, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1007, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
       #endif
       {
-        __pyx_t_7 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 999, __pyx_L1_error)
+        __pyx_t_7 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1007, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_7);
         if (__pyx_t_12) {
           __Pyx_GIVEREF(__pyx_t_12); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_12); __pyx_t_12 = NULL;
@@ -10656,25 +10658,25 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         __Pyx_INCREF(__pyx_v_h);
         __Pyx_GIVEREF(__pyx_v_h);
         PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_8, __pyx_v_h);
-        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_7, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 999, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_7, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1007, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
       }
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_resDistance, __pyx_v_g); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 999, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_resDistance, __pyx_v_g); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1007, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
-      if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_v_h, __pyx_t_1) < 0)) __PYX_ERR(0, 999, __pyx_L1_error)
+      if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_v_h, __pyx_t_1) < 0)) __PYX_ERR(0, 1007, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":1000
+      /* "gedlibpy.pyx":1008
  * 				self.run_method(g, h)
  * 				resDistance[g][h] = self.get_upper_bound(g, h)
  * 				resMapping[g][h] = self.get_node_map(g, h)             # <<<<<<<<<<<<<<
  * 
  * 		print("Finish ! The return contains edit distances and NodeMap but you can check the result with graphs'ID until you restart the environment")
  */
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_node_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1000, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_node_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1008, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __pyx_t_7 = NULL;
       __pyx_t_8 = 0;
@@ -10691,7 +10693,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCALL
       if (PyFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1000, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1008, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
@@ -10699,13 +10701,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
       #if CYTHON_FAST_PYCCALL
       if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_7, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1000, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1008, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
       #endif
       {
-        __pyx_t_12 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1000, __pyx_L1_error)
+        __pyx_t_12 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1008, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_12);
         if (__pyx_t_7) {
           __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_7); __pyx_t_7 = NULL;
@@ -10716,18 +10718,18 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
         __Pyx_INCREF(__pyx_v_h);
         __Pyx_GIVEREF(__pyx_v_h);
         PyTuple_SET_ITEM(__pyx_t_12, 1+__pyx_t_8, __pyx_v_h);
-        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1000, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1008, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
       }
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_resMapping, __pyx_v_g); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1000, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_resMapping, __pyx_v_g); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1008, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
-      if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_v_h, __pyx_t_1) < 0)) __PYX_ERR(0, 1000, __pyx_L1_error)
+      if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_v_h, __pyx_t_1) < 0)) __PYX_ERR(0, 1008, __pyx_L1_error)
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":996
+      /* "gedlibpy.pyx":1004
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
@@ -10737,7 +10739,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
     }
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":994
+    /* "gedlibpy.pyx":1002
  * 		resDistance = [[]]
  * 		resMapping = [[]]
  * 		for g in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
@@ -10747,18 +10749,18 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   }
   __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
 
-  /* "gedlibpy.pyx":1002
+  /* "gedlibpy.pyx":1010
  * 				resMapping[g][h] = self.get_node_map(g, h)
  * 
  * 		print("Finish ! The return contains edit distances and NodeMap but you can check the result with graphs'ID until you restart the environment")             # <<<<<<<<<<<<<<
  * 		return resDistance, resMapping
  * 
  */
-  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__8, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1002, __pyx_L1_error)
+  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__8, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1010, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_9);
   __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
 
-  /* "gedlibpy.pyx":1003
+  /* "gedlibpy.pyx":1011
  * 
  * 		print("Finish ! The return contains edit distances and NodeMap but you can check the result with graphs'ID until you restart the environment")
  * 		return resDistance, resMapping             # <<<<<<<<<<<<<<
@@ -10766,7 +10768,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1003, __pyx_L1_error)
+  __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1011, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_9);
   __Pyx_INCREF(__pyx_v_resDistance);
   __Pyx_GIVEREF(__pyx_v_resDistance);
@@ -10778,7 +10780,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   __pyx_t_9 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":950
+  /* "gedlibpy.pyx":958
  * 
  * 
  * 	def compute_edit_distance_on_nx_graphs(self, dataset, classes, edit_cost, method, options, init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -10808,7 +10810,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_90compute_edit_distance_on_nx_graphs
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1006
+/* "gedlibpy.pyx":1014
  * 
  * 
  * 	def compute_edit_distance_on_GXl_graphs(self, path_folder, path_XML, edit_cost, method, options="", init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -10862,19 +10864,19 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_93compute_edit_distance_on_GXl_graph
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_path_XML)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 1); __PYX_ERR(0, 1006, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 1); __PYX_ERR(0, 1014, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edit_cost)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 2); __PYX_ERR(0, 1006, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 2); __PYX_ERR(0, 1014, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  3:
         if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_method)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 3); __PYX_ERR(0, 1006, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, 3); __PYX_ERR(0, 1014, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  4:
@@ -10890,7 +10892,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_93compute_edit_distance_on_GXl_graph
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_edit_distance_on_GXl_graphs") < 0)) __PYX_ERR(0, 1006, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_edit_distance_on_GXl_graphs") < 0)) __PYX_ERR(0, 1014, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -10915,7 +10917,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_93compute_edit_distance_on_GXl_graph
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1006, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("compute_edit_distance_on_GXl_graphs", 0, 4, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1014, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.compute_edit_distance_on_GXl_graphs", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -10948,14 +10950,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   PyObject *__pyx_t_12 = NULL;
   __Pyx_RefNannySetupContext("compute_edit_distance_on_GXl_graphs", 0);
 
-  /* "gedlibpy.pyx":1030
+  /* "gedlibpy.pyx":1038
  * 		"""
  * 
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
  * 			self.restart_env()
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1030, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_is_initialized); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1038, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -10969,21 +10971,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1030, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1038, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 1030, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 1038, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   if (__pyx_t_4) {
 
-    /* "gedlibpy.pyx":1031
+    /* "gedlibpy.pyx":1039
  * 
  * 		if self.is_initialized() :
  * 			self.restart_env()             # <<<<<<<<<<<<<<
  * 
  * 		print("Loading graphs in progress...")
  */
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1031, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_restart_env); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1039, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __pyx_t_3 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -10997,12 +10999,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     }
     __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1031, __pyx_L1_error)
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1039, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":1030
+    /* "gedlibpy.pyx":1038
  * 		"""
  * 
  * 		if self.is_initialized() :             # <<<<<<<<<<<<<<
@@ -11011,25 +11013,25 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
  */
   }
 
-  /* "gedlibpy.pyx":1033
+  /* "gedlibpy.pyx":1041
  * 			self.restart_env()
  * 
  * 		print("Loading graphs in progress...")             # <<<<<<<<<<<<<<
  * 		self.load_GXL_graphs(path_folder, path_XML)
  * 		listID = self.graph_ids()
  */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1033, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1041, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1034
+  /* "gedlibpy.pyx":1042
  * 
  * 		print("Loading graphs in progress...")
  * 		self.load_GXL_graphs(path_folder, path_XML)             # <<<<<<<<<<<<<<
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_load_GXL_graphs); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1034, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_load_GXL_graphs); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1042, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   __pyx_t_5 = 0;
@@ -11046,7 +11048,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_path_folder, __pyx_v_path_XML};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1034, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1042, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
@@ -11054,13 +11056,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
     PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_path_folder, __pyx_v_path_XML};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1034, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1042, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_GOTREF(__pyx_t_1);
   } else
   #endif
   {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1034, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1042, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     if (__pyx_t_3) {
       __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
@@ -11071,21 +11073,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     __Pyx_INCREF(__pyx_v_path_XML);
     __Pyx_GIVEREF(__pyx_v_path_XML);
     PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_v_path_XML);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1034, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1042, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1035
+  /* "gedlibpy.pyx":1043
  * 		print("Loading graphs in progress...")
  * 		self.load_GXL_graphs(path_folder, path_XML)
  * 		listID = self.graph_ids()             # <<<<<<<<<<<<<<
  * 		print("Graphs loaded ! ")
  * 		print("Number of graphs = " + str(listID[1]))
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_graph_ids); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1035, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_graph_ids); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1043, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -11099,51 +11101,51 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_6) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1035, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1043, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_listID = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1036
+  /* "gedlibpy.pyx":1044
  * 		self.load_GXL_graphs(path_folder, path_XML)
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")             # <<<<<<<<<<<<<<
  * 		print("Number of graphs = " + str(listID[1]))
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1036, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1044, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1037
+  /* "gedlibpy.pyx":1045
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")
  * 		print("Number of graphs = " + str(listID[1]))             # <<<<<<<<<<<<<<
  * 
  * 		self.set_edit_cost(edit_cost)
  */
-  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1037, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1045, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1037, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1045, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Number_of_graphs, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1037, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Number_of_graphs, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1045, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1037, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1045, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1039
+  /* "gedlibpy.pyx":1047
  * 		print("Number of graphs = " + str(listID[1]))
  * 
  * 		self.set_edit_cost(edit_cost)             # <<<<<<<<<<<<<<
  * 		print("Initialization in progress...")
  * 		self.init(init_option)
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1039, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_edit_cost); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1047, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -11157,30 +11159,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __pyx_t_2 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_6, __pyx_v_edit_cost) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_edit_cost);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1039, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1047, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1040
+  /* "gedlibpy.pyx":1048
  * 
  * 		self.set_edit_cost(edit_cost)
  * 		print("Initialization in progress...")             # <<<<<<<<<<<<<<
  * 		self.init(init_option)
  * 		print("Initialization terminated !")
  */
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__6, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1040, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__6, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1048, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1041
+  /* "gedlibpy.pyx":1049
  * 		self.set_edit_cost(edit_cost)
  * 		print("Initialization in progress...")
  * 		self.init(init_option)             # <<<<<<<<<<<<<<
  * 		print("Initialization terminated !")
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1041, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1049, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -11194,30 +11196,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __pyx_t_2 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_6, __pyx_v_init_option) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_init_option);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1041, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1049, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1042
+  /* "gedlibpy.pyx":1050
  * 		print("Initialization in progress...")
  * 		self.init(init_option)
  * 		print("Initialization terminated !")             # <<<<<<<<<<<<<<
  * 
  * 		self.set_method(method, options)
  */
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1042, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1050, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1044
+  /* "gedlibpy.pyx":1052
  * 		print("Initialization terminated !")
  * 
  * 		self.set_method(method, options)             # <<<<<<<<<<<<<<
  * 		self.init_method()
  * 
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1044, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_set_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1052, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_6 = NULL;
   __pyx_t_5 = 0;
@@ -11234,7 +11236,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   #if CYTHON_FAST_PYCALL
   if (PyFunction_Check(__pyx_t_1)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_method, __pyx_v_options};
-    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1044, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1052, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_2);
   } else
@@ -11242,13 +11244,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   #if CYTHON_FAST_PYCCALL
   if (__Pyx_PyFastCFunction_Check(__pyx_t_1)) {
     PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_method, __pyx_v_options};
-    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1044, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_1, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1052, __pyx_L1_error)
     __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_GOTREF(__pyx_t_2);
   } else
   #endif
   {
-    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1044, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1052, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     if (__pyx_t_6) {
       __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
@@ -11259,21 +11261,21 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     __Pyx_INCREF(__pyx_v_options);
     __Pyx_GIVEREF(__pyx_v_options);
     PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_v_options);
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_3, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1044, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_3, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1052, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1045
+  /* "gedlibpy.pyx":1053
  * 
  * 		self.set_method(method, options)
  * 		self.init_method()             # <<<<<<<<<<<<<<
  * 
  * 		#res = []
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1045, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_init_method); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1053, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
@@ -11287,23 +11289,23 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1045, __pyx_L1_error)
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1053, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":1048
+  /* "gedlibpy.pyx":1056
  * 
  * 		#res = []
  * 		for g in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :
  */
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1056, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1056, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1056, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_GIVEREF(__pyx_t_2);
   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
@@ -11311,16 +11313,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_t_1);
   __pyx_t_2 = 0;
   __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1056, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
     __pyx_t_3 = __pyx_t_1; __Pyx_INCREF(__pyx_t_3); __pyx_t_7 = 0;
     __pyx_t_8 = NULL;
   } else {
-    __pyx_t_7 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1048, __pyx_L1_error)
+    __pyx_t_7 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1056, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_8 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1048, __pyx_L1_error)
+    __pyx_t_8 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1056, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   for (;;) {
@@ -11328,17 +11330,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
       if (likely(PyList_CheckExact(__pyx_t_3))) {
         if (__pyx_t_7 >= PyList_GET_SIZE(__pyx_t_3)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 1048, __pyx_L1_error)
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1048, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1056, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       } else {
         if (__pyx_t_7 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 1048, __pyx_L1_error)
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_7); __Pyx_INCREF(__pyx_t_1); __pyx_t_7++; if (unlikely(0 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
         #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1048, __pyx_L1_error)
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_3, __pyx_t_7); __pyx_t_7++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1056, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         #endif
       }
@@ -11348,7 +11350,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1048, __pyx_L1_error)
+          else __PYX_ERR(0, 1056, __pyx_L1_error)
         }
         break;
       }
@@ -11357,38 +11359,38 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     __Pyx_XDECREF_SET(__pyx_v_g, __pyx_t_1);
     __pyx_t_1 = 0;
 
-    /* "gedlibpy.pyx":1049
+    /* "gedlibpy.pyx":1057
  * 		#res = []
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")             # <<<<<<<<<<<<<<
  * 			for h in range(listID[0], listID[1]) :
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  */
-    __pyx_t_1 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_g); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1049, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_g); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1057, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Computation_between_graph, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1049, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyUnicode_Concat(__pyx_kp_u_Computation_between_graph, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1057, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_t_2, __pyx_kp_u_with_all_the_others_including_h); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1049, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_t_2, __pyx_kp_u_with_all_the_others_including_h); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1057, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1049, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_print, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1057, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-    /* "gedlibpy.pyx":1050
+    /* "gedlibpy.pyx":1058
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  * 				self.run_method(g,h)
  */
-    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1050, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_listID, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1058, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1050, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_listID, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1058, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_6 = PyTuple_New(2); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1050, __pyx_L1_error)
+    __pyx_t_6 = PyTuple_New(2); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1058, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_GIVEREF(__pyx_t_2);
     PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_2);
@@ -11396,16 +11398,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     PyTuple_SET_ITEM(__pyx_t_6, 1, __pyx_t_1);
     __pyx_t_2 = 0;
     __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1050, __pyx_L1_error)
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1058, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
       __pyx_t_6 = __pyx_t_1; __Pyx_INCREF(__pyx_t_6); __pyx_t_9 = 0;
       __pyx_t_10 = NULL;
     } else {
-      __pyx_t_9 = -1; __pyx_t_6 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1050, __pyx_L1_error)
+      __pyx_t_9 = -1; __pyx_t_6 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1058, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
-      __pyx_t_10 = Py_TYPE(__pyx_t_6)->tp_iternext; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1050, __pyx_L1_error)
+      __pyx_t_10 = Py_TYPE(__pyx_t_6)->tp_iternext; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1058, __pyx_L1_error)
     }
     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     for (;;) {
@@ -11413,17 +11415,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
         if (likely(PyList_CheckExact(__pyx_t_6))) {
           if (__pyx_t_9 >= PyList_GET_SIZE(__pyx_t_6)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_1 = PyList_GET_ITEM(__pyx_t_6, __pyx_t_9); __Pyx_INCREF(__pyx_t_1); __pyx_t_9++; if (unlikely(0 < 0)) __PYX_ERR(0, 1050, __pyx_L1_error)
+          __pyx_t_1 = PyList_GET_ITEM(__pyx_t_6, __pyx_t_9); __Pyx_INCREF(__pyx_t_1); __pyx_t_9++; if (unlikely(0 < 0)) __PYX_ERR(0, 1058, __pyx_L1_error)
           #else
-          __pyx_t_1 = PySequence_ITEM(__pyx_t_6, __pyx_t_9); __pyx_t_9++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1050, __pyx_L1_error)
+          __pyx_t_1 = PySequence_ITEM(__pyx_t_6, __pyx_t_9); __pyx_t_9++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1058, __pyx_L1_error)
           __Pyx_GOTREF(__pyx_t_1);
           #endif
         } else {
           if (__pyx_t_9 >= PyTuple_GET_SIZE(__pyx_t_6)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_6, __pyx_t_9); __Pyx_INCREF(__pyx_t_1); __pyx_t_9++; if (unlikely(0 < 0)) __PYX_ERR(0, 1050, __pyx_L1_error)
+          __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_6, __pyx_t_9); __Pyx_INCREF(__pyx_t_1); __pyx_t_9++; if (unlikely(0 < 0)) __PYX_ERR(0, 1058, __pyx_L1_error)
           #else
-          __pyx_t_1 = PySequence_ITEM(__pyx_t_6, __pyx_t_9); __pyx_t_9++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1050, __pyx_L1_error)
+          __pyx_t_1 = PySequence_ITEM(__pyx_t_6, __pyx_t_9); __pyx_t_9++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1058, __pyx_L1_error)
           __Pyx_GOTREF(__pyx_t_1);
           #endif
         }
@@ -11433,7 +11435,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
           PyObject* exc_type = PyErr_Occurred();
           if (exc_type) {
             if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 1050, __pyx_L1_error)
+            else __PYX_ERR(0, 1058, __pyx_L1_error)
           }
           break;
         }
@@ -11442,14 +11444,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
       __Pyx_XDECREF_SET(__pyx_v_h, __pyx_t_1);
       __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":1052
+      /* "gedlibpy.pyx":1060
  * 			for h in range(listID[0], listID[1]) :
  * 				#print("Computation between graph " + str(g) + " and graph " + str(h))
  * 				self.run_method(g,h)             # <<<<<<<<<<<<<<
  * 				#res.append((get_upper_bound(g,h), get_node_map(g,h), get_runtime(g,h)))
  * 
  */
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1052, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_run_method); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1060, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
       __pyx_t_11 = NULL;
       __pyx_t_5 = 0;
@@ -11466,7 +11468,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
       #if CYTHON_FAST_PYCALL
       if (PyFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_11, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1052, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1060, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
@@ -11474,13 +11476,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
       #if CYTHON_FAST_PYCCALL
       if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
         PyObject *__pyx_temp[3] = {__pyx_t_11, __pyx_v_g, __pyx_v_h};
-        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1052, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1060, __pyx_L1_error)
         __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
         __Pyx_GOTREF(__pyx_t_1);
       } else
       #endif
       {
-        __pyx_t_12 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1052, __pyx_L1_error)
+        __pyx_t_12 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1060, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_12);
         if (__pyx_t_11) {
           __Pyx_GIVEREF(__pyx_t_11); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_11); __pyx_t_11 = NULL;
@@ -11491,14 +11493,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
         __Pyx_INCREF(__pyx_v_h);
         __Pyx_GIVEREF(__pyx_v_h);
         PyTuple_SET_ITEM(__pyx_t_12, 1+__pyx_t_5, __pyx_v_h);
-        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1052, __pyx_L1_error)
+        __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_12, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1060, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_1);
         __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
       }
       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-      /* "gedlibpy.pyx":1050
+      /* "gedlibpy.pyx":1058
  * 		for g in range(listID[0], listID[1]) :
  * 			print("Computation between graph " + str(g) + " with all the others including himself.")
  * 			for h in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
@@ -11508,7 +11510,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
     }
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
 
-    /* "gedlibpy.pyx":1048
+    /* "gedlibpy.pyx":1056
  * 
  * 		#res = []
  * 		for g in range(listID[0], listID[1]) :             # <<<<<<<<<<<<<<
@@ -11518,29 +11520,29 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   }
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-  /* "gedlibpy.pyx":1057
+  /* "gedlibpy.pyx":1065
  * 		#return res
  * 
  * 		print ("Finish ! You can check the result with each ID of graphs ! There are in the return")             # <<<<<<<<<<<<<<
  * 		print ("Please don't restart the environment or recall this function, you will lose your results !")
  * 		return listID
  */
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__9, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1057, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__9, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1065, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-  /* "gedlibpy.pyx":1058
+  /* "gedlibpy.pyx":1066
  * 
  * 		print ("Finish ! You can check the result with each ID of graphs ! There are in the return")
  * 		print ("Please don't restart the environment or recall this function, you will lose your results !")             # <<<<<<<<<<<<<<
  * 		return listID
  * 
  */
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__10, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1058, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_print, __pyx_tuple__10, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1066, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-  /* "gedlibpy.pyx":1059
+  /* "gedlibpy.pyx":1067
  * 		print ("Finish ! You can check the result with each ID of graphs ! There are in the return")
  * 		print ("Please don't restart the environment or recall this function, you will lose your results !")
  * 		return listID             # <<<<<<<<<<<<<<
@@ -11552,7 +11554,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   __pyx_r = __pyx_v_listID;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1006
+  /* "gedlibpy.pyx":1014
  * 
  * 
  * 	def compute_edit_distance_on_GXl_graphs(self, path_folder, path_XML, edit_cost, method, options="", init_option="EAGER_WITHOUT_SHUFFLED_COPIES") :             # <<<<<<<<<<<<<<
@@ -11579,7 +11581,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_92compute_edit_distance_on_GXl_graph
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1062
+/* "gedlibpy.pyx":1070
  * 
  * 
  * 	def get_num_node_labels(self):             # <<<<<<<<<<<<<<
@@ -11608,7 +11610,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_94get_num_node_labels(struct __pyx_o
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_num_node_labels", 0);
 
-  /* "gedlibpy.pyx":1071
+  /* "gedlibpy.pyx":1079
  * 			.. note:: If 1 is returned, the nodes are unlabeled.
  * 		"""
  * 		return self.c_env.getNumNodeLabels()             # <<<<<<<<<<<<<<
@@ -11620,15 +11622,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_94get_num_node_labels(struct __pyx_o
     __pyx_t_1 = __pyx_v_self->c_env->getNumNodeLabels();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1071, __pyx_L1_error)
+    __PYX_ERR(0, 1079, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1071, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1079, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1062
+  /* "gedlibpy.pyx":1070
  * 
  * 
  * 	def get_num_node_labels(self):             # <<<<<<<<<<<<<<
@@ -11647,7 +11649,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_94get_num_node_labels(struct __pyx_o
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1074
+/* "gedlibpy.pyx":1082
  * 
  * 
  * 	def get_node_label(self, label_id):             # <<<<<<<<<<<<<<
@@ -11680,7 +11682,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_96get_node_label(struct __pyx_obj_8g
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_node_label", 0);
 
-  /* "gedlibpy.pyx":1083
+  /* "gedlibpy.pyx":1091
  * 			:rtype: dict{string : string}
  * 	 	"""
  * 		return decode_your_map(self.c_env.getNodeLabel(label_id))             # <<<<<<<<<<<<<<
@@ -11688,16 +11690,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_96get_node_label(struct __pyx_obj_8g
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1083, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1091, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_label_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1083, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_label_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1091, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getNodeLabel(__pyx_t_3);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1083, __pyx_L1_error)
+    __PYX_ERR(0, 1091, __pyx_L1_error)
   }
-  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1083, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1091, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -11712,14 +11714,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_96get_node_label(struct __pyx_obj_8g
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1083, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1091, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1074
+  /* "gedlibpy.pyx":1082
  * 
  * 
  * 	def get_node_label(self, label_id):             # <<<<<<<<<<<<<<
@@ -11741,7 +11743,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_96get_node_label(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1086
+/* "gedlibpy.pyx":1094
  * 
  * 
  * 	def get_num_edge_labels(self):             # <<<<<<<<<<<<<<
@@ -11770,7 +11772,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_98get_num_edge_labels(struct __pyx_o
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_num_edge_labels", 0);
 
-  /* "gedlibpy.pyx":1095
+  /* "gedlibpy.pyx":1103
  * 			.. note:: If 1 is returned, the edges are unlabeled.
  * 	 	"""
  * 		return self.c_env.getNumEdgeLabels()             # <<<<<<<<<<<<<<
@@ -11782,15 +11784,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_98get_num_edge_labels(struct __pyx_o
     __pyx_t_1 = __pyx_v_self->c_env->getNumEdgeLabels();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1095, __pyx_L1_error)
+    __PYX_ERR(0, 1103, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1095, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyInt_FromSize_t(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1103, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1086
+  /* "gedlibpy.pyx":1094
  * 
  * 
  * 	def get_num_edge_labels(self):             # <<<<<<<<<<<<<<
@@ -11809,7 +11811,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_98get_num_edge_labels(struct __pyx_o
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1098
+/* "gedlibpy.pyx":1106
  * 
  * 
  * 	def get_edge_label(self, label_id):             # <<<<<<<<<<<<<<
@@ -11842,7 +11844,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_100get_edge_label(struct __pyx_obj_8
   PyObject *__pyx_t_6 = NULL;
   __Pyx_RefNannySetupContext("get_edge_label", 0);
 
-  /* "gedlibpy.pyx":1107
+  /* "gedlibpy.pyx":1115
  * 			:rtype: dict{string : string}
  * 	 	"""
  * 		return decode_your_map(self.c_env.getEdgeLabel(label_id))             # <<<<<<<<<<<<<<
@@ -11850,16 +11852,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_100get_edge_label(struct __pyx_obj_8
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1107, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1115, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_label_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1107, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyInt_As_size_t(__pyx_v_label_id); if (unlikely((__pyx_t_3 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1115, __pyx_L1_error)
   try {
     __pyx_t_4 = __pyx_v_self->c_env->getEdgeLabel(__pyx_t_3);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1107, __pyx_L1_error)
+    __PYX_ERR(0, 1115, __pyx_L1_error)
   }
-  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1107, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1115, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -11874,14 +11876,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_100get_edge_label(struct __pyx_obj_8
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1107, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1115, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1098
+  /* "gedlibpy.pyx":1106
  * 
  * 
  * 	def get_edge_label(self, label_id):             # <<<<<<<<<<<<<<
@@ -11903,7 +11905,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_100get_edge_label(struct __pyx_obj_8
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1121
+/* "gedlibpy.pyx":1129
  * # 		return self.c_env.getNumNodes(graph_id)
  * 
  * 	def get_avg_num_nodes(self):             # <<<<<<<<<<<<<<
@@ -11932,7 +11934,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_102get_avg_num_nodes(struct __pyx_ob
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_avg_num_nodes", 0);
 
-  /* "gedlibpy.pyx":1128
+  /* "gedlibpy.pyx":1136
  * 			:rtype: double
  * 		"""
  * 		return self.c_env.getAvgNumNodes()             # <<<<<<<<<<<<<<
@@ -11944,15 +11946,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_102get_avg_num_nodes(struct __pyx_ob
     __pyx_t_1 = __pyx_v_self->c_env->getAvgNumNodes();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1128, __pyx_L1_error)
+    __PYX_ERR(0, 1136, __pyx_L1_error)
   }
-  __pyx_t_2 = PyFloat_FromDouble(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1128, __pyx_L1_error)
+  __pyx_t_2 = PyFloat_FromDouble(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1136, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1121
+  /* "gedlibpy.pyx":1129
  * # 		return self.c_env.getNumNodes(graph_id)
  * 
  * 	def get_avg_num_nodes(self):             # <<<<<<<<<<<<<<
@@ -11971,7 +11973,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_102get_avg_num_nodes(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1130
+/* "gedlibpy.pyx":1138
  * 		return self.c_env.getAvgNumNodes()
  * 
  * 	def get_node_rel_cost(self, node_label_1, node_label_2):             # <<<<<<<<<<<<<<
@@ -12011,11 +12013,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_105get_node_rel_cost(PyObject *__pyx
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_label_2)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_node_rel_cost", 1, 2, 2, 1); __PYX_ERR(0, 1130, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_node_rel_cost", 1, 2, 2, 1); __PYX_ERR(0, 1138, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_rel_cost") < 0)) __PYX_ERR(0, 1130, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_node_rel_cost") < 0)) __PYX_ERR(0, 1138, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -12028,7 +12030,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_105get_node_rel_cost(PyObject *__pyx
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_node_rel_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1130, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_node_rel_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1138, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_node_rel_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -12052,7 +12054,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_104get_node_rel_cost(struct __pyx_ob
   double __pyx_t_6;
   __Pyx_RefNannySetupContext("get_node_rel_cost", 0);
 
-  /* "gedlibpy.pyx":1141
+  /* "gedlibpy.pyx":1149
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getNodeRelCost(encode_your_map(node_label_1), encode_your_map(node_label_2))             # <<<<<<<<<<<<<<
@@ -12060,7 +12062,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_104get_node_rel_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1141, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12074,12 +12076,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_104get_node_rel_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_node_label_1) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_node_label_1);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1141, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1141, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1141, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12093,24 +12095,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_104get_node_rel_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_node_label_2) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_node_label_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1141, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_5 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1141, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_6 = __pyx_v_self->c_env->getNodeRelCost(__pyx_t_4, __pyx_t_5);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1141, __pyx_L1_error)
+    __PYX_ERR(0, 1149, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1141, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1149, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1130
+  /* "gedlibpy.pyx":1138
  * 		return self.c_env.getAvgNumNodes()
  * 
  * 	def get_node_rel_cost(self, node_label_1, node_label_2):             # <<<<<<<<<<<<<<
@@ -12131,7 +12133,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_104get_node_rel_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1144
+/* "gedlibpy.pyx":1152
  * 
  * 
  * 	def get_node_del_cost(self, node_label):             # <<<<<<<<<<<<<<
@@ -12163,7 +12165,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_106get_node_del_cost(struct __pyx_ob
   double __pyx_t_5;
   __Pyx_RefNannySetupContext("get_node_del_cost", 0);
 
-  /* "gedlibpy.pyx":1153
+  /* "gedlibpy.pyx":1161
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getNodeDelCost(encode_your_map(node_label))             # <<<<<<<<<<<<<<
@@ -12171,7 +12173,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_106get_node_del_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1153, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1161, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12185,24 +12187,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_106get_node_del_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_node_label) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_node_label);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1153, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1161, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1153, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1161, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_5 = __pyx_v_self->c_env->getNodeDelCost(__pyx_t_4);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1153, __pyx_L1_error)
+    __PYX_ERR(0, 1161, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1153, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1161, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1144
+  /* "gedlibpy.pyx":1152
  * 
  * 
  * 	def get_node_del_cost(self, node_label):             # <<<<<<<<<<<<<<
@@ -12223,7 +12225,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_106get_node_del_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1156
+/* "gedlibpy.pyx":1164
  * 
  * 
  * 	def get_node_ins_cost(self, node_label):             # <<<<<<<<<<<<<<
@@ -12255,7 +12257,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_108get_node_ins_cost(struct __pyx_ob
   double __pyx_t_5;
   __Pyx_RefNannySetupContext("get_node_ins_cost", 0);
 
-  /* "gedlibpy.pyx":1165
+  /* "gedlibpy.pyx":1173
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getNodeInsCost(encode_your_map(node_label))             # <<<<<<<<<<<<<<
@@ -12263,7 +12265,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_108get_node_ins_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1165, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1173, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12277,24 +12279,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_108get_node_ins_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_node_label) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_node_label);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1165, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1165, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1173, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_5 = __pyx_v_self->c_env->getNodeInsCost(__pyx_t_4);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1165, __pyx_L1_error)
+    __PYX_ERR(0, 1173, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1165, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1156
+  /* "gedlibpy.pyx":1164
  * 
  * 
  * 	def get_node_ins_cost(self, node_label):             # <<<<<<<<<<<<<<
@@ -12315,7 +12317,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_108get_node_ins_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1168
+/* "gedlibpy.pyx":1176
  * 
  * 
  * 	def get_median_node_label(self, node_labels):             # <<<<<<<<<<<<<<
@@ -12353,7 +12355,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
   std::map<std::string,std::string>  __pyx_t_9;
   __Pyx_RefNannySetupContext("get_median_node_label", 0);
 
-  /* "gedlibpy.pyx":1177
+  /* "gedlibpy.pyx":1185
  * 			:rtype: dict{string : string}
  * 		"""
  * 		node_labels_b = [encode_your_map(node_label) for node_label in node_labels]             # <<<<<<<<<<<<<<
@@ -12361,32 +12363,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
  * 
  */
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1177, __pyx_L5_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1185, __pyx_L5_error)
     __Pyx_GOTREF(__pyx_t_1);
     if (likely(PyList_CheckExact(__pyx_v_node_labels)) || PyTuple_CheckExact(__pyx_v_node_labels)) {
       __pyx_t_2 = __pyx_v_node_labels; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
       __pyx_t_4 = NULL;
     } else {
-      __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_node_labels); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1177, __pyx_L5_error)
+      __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_node_labels); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1185, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1177, __pyx_L5_error)
+      __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1185, __pyx_L5_error)
     }
     for (;;) {
       if (likely(!__pyx_t_4)) {
         if (likely(PyList_CheckExact(__pyx_t_2))) {
           if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1177, __pyx_L5_error)
+          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1185, __pyx_L5_error)
           #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1177, __pyx_L5_error)
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1185, __pyx_L5_error)
           __Pyx_GOTREF(__pyx_t_5);
           #endif
         } else {
           if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1177, __pyx_L5_error)
+          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1185, __pyx_L5_error)
           #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1177, __pyx_L5_error)
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1185, __pyx_L5_error)
           __Pyx_GOTREF(__pyx_t_5);
           #endif
         }
@@ -12396,7 +12398,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
           PyObject* exc_type = PyErr_Occurred();
           if (exc_type) {
             if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 1177, __pyx_L5_error)
+            else __PYX_ERR(0, 1185, __pyx_L5_error)
           }
           break;
         }
@@ -12404,7 +12406,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
       }
       __Pyx_XDECREF_SET(__pyx_8genexpr9__pyx_v_node_label, __pyx_t_5);
       __pyx_t_5 = 0;
-      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1177, __pyx_L5_error)
+      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1185, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_6);
       __pyx_t_7 = NULL;
       if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
@@ -12418,10 +12420,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
       }
       __pyx_t_5 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_7, __pyx_8genexpr9__pyx_v_node_label) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_8genexpr9__pyx_v_node_label);
       __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1177, __pyx_L5_error)
+      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1185, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_5);
       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 1177, __pyx_L5_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 1185, __pyx_L5_error)
       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     }
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
@@ -12435,7 +12437,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
   __pyx_v_node_labels_b = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1178
+  /* "gedlibpy.pyx":1186
  * 		"""
  * 		node_labels_b = [encode_your_map(node_label) for node_label in node_labels]
  * 		return decode_your_map(self.c_env.getMedianNodeLabel(node_labels_b))             # <<<<<<<<<<<<<<
@@ -12443,16 +12445,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1178, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1186, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_8 = __pyx_convert_vector_from_py_std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_v_node_labels_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1178, __pyx_L1_error)
+  __pyx_t_8 = __pyx_convert_vector_from_py_std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_v_node_labels_b); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1186, __pyx_L1_error)
   try {
     __pyx_t_9 = __pyx_v_self->c_env->getMedianNodeLabel(__pyx_t_8);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1178, __pyx_L1_error)
+    __PYX_ERR(0, 1186, __pyx_L1_error)
   }
-  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_9); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1178, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_9); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1186, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12467,14 +12469,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1178, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1186, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1168
+  /* "gedlibpy.pyx":1176
  * 
  * 
  * 	def get_median_node_label(self, node_labels):             # <<<<<<<<<<<<<<
@@ -12499,7 +12501,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_110get_median_node_label(struct __py
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1181
+/* "gedlibpy.pyx":1189
  * 
  * 
  * 	def get_edge_rel_cost(self, edge_label_1, edge_label_2):             # <<<<<<<<<<<<<<
@@ -12539,11 +12541,11 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_113get_edge_rel_cost(PyObject *__pyx
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_edge_label_2)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("get_edge_rel_cost", 1, 2, 2, 1); __PYX_ERR(0, 1181, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("get_edge_rel_cost", 1, 2, 2, 1); __PYX_ERR(0, 1189, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_edge_rel_cost") < 0)) __PYX_ERR(0, 1181, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_edge_rel_cost") < 0)) __PYX_ERR(0, 1189, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -12556,7 +12558,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_113get_edge_rel_cost(PyObject *__pyx
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_edge_rel_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1181, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_edge_rel_cost", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1189, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_edge_rel_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -12580,7 +12582,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_112get_edge_rel_cost(struct __pyx_ob
   double __pyx_t_6;
   __Pyx_RefNannySetupContext("get_edge_rel_cost", 0);
 
-  /* "gedlibpy.pyx":1192
+  /* "gedlibpy.pyx":1200
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getEdgeRelCost(encode_your_map(edge_label_1), encode_your_map(edge_label_2))             # <<<<<<<<<<<<<<
@@ -12588,7 +12590,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_112get_edge_rel_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1192, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12602,12 +12604,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_112get_edge_rel_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edge_label_1) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edge_label_1);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1192, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1192, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1192, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12621,24 +12623,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_112get_edge_rel_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edge_label_2) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edge_label_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1192, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_5 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1192, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_6 = __pyx_v_self->c_env->getEdgeRelCost(__pyx_t_4, __pyx_t_5);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1192, __pyx_L1_error)
+    __PYX_ERR(0, 1200, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1192, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1200, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1181
+  /* "gedlibpy.pyx":1189
  * 
  * 
  * 	def get_edge_rel_cost(self, edge_label_1, edge_label_2):             # <<<<<<<<<<<<<<
@@ -12659,7 +12661,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_112get_edge_rel_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1195
+/* "gedlibpy.pyx":1203
  * 
  * 
  * 	def get_edge_del_cost(self, edge_label):             # <<<<<<<<<<<<<<
@@ -12691,7 +12693,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_114get_edge_del_cost(struct __pyx_ob
   double __pyx_t_5;
   __Pyx_RefNannySetupContext("get_edge_del_cost", 0);
 
-  /* "gedlibpy.pyx":1204
+  /* "gedlibpy.pyx":1212
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getEdgeDelCost(encode_your_map(edge_label))             # <<<<<<<<<<<<<<
@@ -12699,7 +12701,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_114get_edge_del_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1204, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1212, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12713,24 +12715,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_114get_edge_del_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edge_label) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edge_label);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1204, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1212, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1204, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1212, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_5 = __pyx_v_self->c_env->getEdgeDelCost(__pyx_t_4);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1204, __pyx_L1_error)
+    __PYX_ERR(0, 1212, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1204, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1212, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1195
+  /* "gedlibpy.pyx":1203
  * 
  * 
  * 	def get_edge_del_cost(self, edge_label):             # <<<<<<<<<<<<<<
@@ -12751,7 +12753,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_114get_edge_del_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1207
+/* "gedlibpy.pyx":1215
  * 
  * 
  * 	def get_edge_ins_cost(self, edge_label):             # <<<<<<<<<<<<<<
@@ -12783,7 +12785,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_116get_edge_ins_cost(struct __pyx_ob
   double __pyx_t_5;
   __Pyx_RefNannySetupContext("get_edge_ins_cost", 0);
 
-  /* "gedlibpy.pyx":1216
+  /* "gedlibpy.pyx":1224
  * 			:rtype: double
  * 	 	"""
  * 		return self.c_env.getEdgeInsCost(encode_your_map(edge_label))             # <<<<<<<<<<<<<<
@@ -12791,7 +12793,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_116get_edge_ins_cost(struct __pyx_ob
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1216, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1224, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12805,24 +12807,24 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_116get_edge_ins_cost(struct __pyx_ob
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_edge_label) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_edge_label);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1216, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1224, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1216, __pyx_L1_error)
+  __pyx_t_4 = __pyx_convert_map_from_py_std_3a__3a_string__and_std_3a__3a_string(__pyx_t_1); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1224, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   try {
     __pyx_t_5 = __pyx_v_self->c_env->getEdgeInsCost(__pyx_t_4);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1216, __pyx_L1_error)
+    __PYX_ERR(0, 1224, __pyx_L1_error)
   }
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1216, __pyx_L1_error)
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1224, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1207
+  /* "gedlibpy.pyx":1215
  * 
  * 
  * 	def get_edge_ins_cost(self, edge_label):             # <<<<<<<<<<<<<<
@@ -12843,7 +12845,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_116get_edge_ins_cost(struct __pyx_ob
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1219
+/* "gedlibpy.pyx":1227
  * 
  * 
  * 	def get_median_edge_label(self, edge_labels):             # <<<<<<<<<<<<<<
@@ -12881,7 +12883,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
   std::map<std::string,std::string>  __pyx_t_9;
   __Pyx_RefNannySetupContext("get_median_edge_label", 0);
 
-  /* "gedlibpy.pyx":1228
+  /* "gedlibpy.pyx":1236
  * 			:rtype: dict{string : string}
  * 		"""
  * 		edge_labels_b = [encode_your_map(edge_label) for edge_label in edge_labels]             # <<<<<<<<<<<<<<
@@ -12889,32 +12891,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
  * 
  */
   { /* enter inner scope */
-    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1228, __pyx_L5_error)
+    __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1236, __pyx_L5_error)
     __Pyx_GOTREF(__pyx_t_1);
     if (likely(PyList_CheckExact(__pyx_v_edge_labels)) || PyTuple_CheckExact(__pyx_v_edge_labels)) {
       __pyx_t_2 = __pyx_v_edge_labels; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
       __pyx_t_4 = NULL;
     } else {
-      __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_edge_labels); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1228, __pyx_L5_error)
+      __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_edge_labels); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1236, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1228, __pyx_L5_error)
+      __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1236, __pyx_L5_error)
     }
     for (;;) {
       if (likely(!__pyx_t_4)) {
         if (likely(PyList_CheckExact(__pyx_t_2))) {
           if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1228, __pyx_L5_error)
+          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1236, __pyx_L5_error)
           #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1228, __pyx_L5_error)
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1236, __pyx_L5_error)
           __Pyx_GOTREF(__pyx_t_5);
           #endif
         } else {
           if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
           #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1228, __pyx_L5_error)
+          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1236, __pyx_L5_error)
           #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1228, __pyx_L5_error)
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1236, __pyx_L5_error)
           __Pyx_GOTREF(__pyx_t_5);
           #endif
         }
@@ -12924,7 +12926,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
           PyObject* exc_type = PyErr_Occurred();
           if (exc_type) {
             if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 1228, __pyx_L5_error)
+            else __PYX_ERR(0, 1236, __pyx_L5_error)
           }
           break;
         }
@@ -12932,7 +12934,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
       }
       __Pyx_XDECREF_SET(__pyx_9genexpr10__pyx_v_edge_label, __pyx_t_5);
       __pyx_t_5 = 0;
-      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1228, __pyx_L5_error)
+      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_encode_your_map); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1236, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_6);
       __pyx_t_7 = NULL;
       if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
@@ -12946,10 +12948,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
       }
       __pyx_t_5 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_7, __pyx_9genexpr10__pyx_v_edge_label) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_9genexpr10__pyx_v_edge_label);
       __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1228, __pyx_L5_error)
+      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1236, __pyx_L5_error)
       __Pyx_GOTREF(__pyx_t_5);
       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 1228, __pyx_L5_error)
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 1236, __pyx_L5_error)
       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     }
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
@@ -12963,7 +12965,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
   __pyx_v_edge_labels_b = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1229
+  /* "gedlibpy.pyx":1237
  * 		"""
  * 		edge_labels_b = [encode_your_map(edge_label) for edge_label in edge_labels]
  * 		return decode_your_map(self.c_env.getMedianEdgeLabel(edge_label_b))             # <<<<<<<<<<<<<<
@@ -12971,19 +12973,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
  * 
  */
   __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1229, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1237, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_edge_label_b); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1229, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_edge_label_b); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1237, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_8 = __pyx_convert_vector_from_py_std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_t_5); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1229, __pyx_L1_error)
+  __pyx_t_8 = __pyx_convert_vector_from_py_std_3a__3a_map_3c_std_3a__3a_string_2c_std_3a__3a_string_3e___(__pyx_t_5); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1237, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   try {
     __pyx_t_9 = __pyx_v_self->c_env->getMedianEdgeLabel(__pyx_t_8);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1229, __pyx_L1_error)
+    __PYX_ERR(0, 1237, __pyx_L1_error)
   }
-  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_9); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1229, __pyx_L1_error)
+  __pyx_t_5 = __pyx_convert_map_to_py_std_3a__3a_string____std_3a__3a_string(__pyx_t_9); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1237, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -12998,14 +13000,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1229, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1237, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_r = __pyx_t_1;
   __pyx_t_1 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1219
+  /* "gedlibpy.pyx":1227
  * 
  * 
  * 	def get_median_edge_label(self, edge_labels):             # <<<<<<<<<<<<<<
@@ -13030,7 +13032,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_118get_median_edge_label(struct __py
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1232
+/* "gedlibpy.pyx":1240
  * 
  * 
  * 	def get_nx_graph(self, graph_id, adj_matrix=True, adj_lists=False, edge_list=False): # @todo             # <<<<<<<<<<<<<<
@@ -13095,7 +13097,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_121get_nx_graph(PyObject *__pyx_v_se
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_nx_graph") < 0)) __PYX_ERR(0, 1232, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "get_nx_graph") < 0)) __PYX_ERR(0, 1240, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -13117,7 +13119,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_121get_nx_graph(PyObject *__pyx_v_se
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("get_nx_graph", 0, 1, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1232, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("get_nx_graph", 0, 1, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1240, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.get_nx_graph", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -13156,16 +13158,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   PyObject *(*__pyx_t_12)(PyObject *);
   __Pyx_RefNannySetupContext("get_nx_graph", 0);
 
-  /* "gedlibpy.pyx":1255
+  /* "gedlibpy.pyx":1263
  * 			The obtained graph.
  * 		"""
  * 		graph = nx.Graph()             # <<<<<<<<<<<<<<
  * 		graph.graph['id'] = graph_id
  * 
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_nx); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1255, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_nx); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1263, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_Graph); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1255, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_Graph); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1263, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_t_2 = NULL;
@@ -13180,32 +13182,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_2) : __Pyx_PyObject_CallNoArg(__pyx_t_3);
   __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1255, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1263, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __pyx_v_graph = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1256
+  /* "gedlibpy.pyx":1264
  * 		"""
  * 		graph = nx.Graph()
  * 		graph.graph['id'] = graph_id             # <<<<<<<<<<<<<<
  * 
  * 		nb_nodes = self.get_graph_num_nodes(graph_id)
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_graph); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1256, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_graph); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1264, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_n_u_id, __pyx_v_graph_id) < 0)) __PYX_ERR(0, 1256, __pyx_L1_error)
+  if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_n_u_id, __pyx_v_graph_id) < 0)) __PYX_ERR(0, 1264, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1258
+  /* "gedlibpy.pyx":1266
  * 		graph.graph['id'] = graph_id
  * 
  * 		nb_nodes = self.get_graph_num_nodes(graph_id)             # <<<<<<<<<<<<<<
  * 		original_node_ids = self.get_original_node_ids(graph_id)
  * 		node_labels = self.get_graph_node_labels(graph_id)
  */
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_num_nodes); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1258, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_num_nodes); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1266, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_2 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -13219,20 +13221,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_2, __pyx_v_graph_id) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_graph_id);
   __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1258, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1266, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __pyx_v_nb_nodes = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1259
+  /* "gedlibpy.pyx":1267
  * 
  * 		nb_nodes = self.get_graph_num_nodes(graph_id)
  * 		original_node_ids = self.get_original_node_ids(graph_id)             # <<<<<<<<<<<<<<
  * 		node_labels = self.get_graph_node_labels(graph_id)
  * # 		print(original_node_ids)
  */
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_original_node_ids); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1259, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_original_node_ids); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1267, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_2 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -13246,20 +13248,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_2, __pyx_v_graph_id) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_graph_id);
   __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1259, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1267, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __pyx_v_original_node_ids = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1260
+  /* "gedlibpy.pyx":1268
  * 		nb_nodes = self.get_graph_num_nodes(graph_id)
  * 		original_node_ids = self.get_original_node_ids(graph_id)
  * 		node_labels = self.get_graph_node_labels(graph_id)             # <<<<<<<<<<<<<<
  * # 		print(original_node_ids)
  * # 		print(node_labels)
  */
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_node_labels); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1260, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_node_labels); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1268, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_2 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
@@ -13273,32 +13275,32 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_2, __pyx_v_graph_id) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_graph_id);
   __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1260, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1268, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __pyx_v_node_labels = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1263
+  /* "gedlibpy.pyx":1271
  * # 		print(original_node_ids)
  * # 		print(node_labels)
  * 		graph.graph['original_node_ids'] = original_node_ids             # <<<<<<<<<<<<<<
  * 
  * 		for node_id in range(0, nb_nodes):
  */
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_graph); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1263, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_graph); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1271, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_n_u_original_node_ids, __pyx_v_original_node_ids) < 0)) __PYX_ERR(0, 1263, __pyx_L1_error)
+  if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_n_u_original_node_ids, __pyx_v_original_node_ids) < 0)) __PYX_ERR(0, 1271, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1265
+  /* "gedlibpy.pyx":1273
  * 		graph.graph['original_node_ids'] = original_node_ids
  * 
  * 		for node_id in range(0, nb_nodes):             # <<<<<<<<<<<<<<
  * 			graph.add_node(node_id, **node_labels[node_id])
  * # 			graph.nodes[node_id]['original_node_id'] = original_node_ids[node_id]
  */
-  __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1265, __pyx_L1_error)
+  __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1273, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_INCREF(__pyx_int_0);
   __Pyx_GIVEREF(__pyx_int_0);
@@ -13306,16 +13308,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   __Pyx_INCREF(__pyx_v_nb_nodes);
   __Pyx_GIVEREF(__pyx_v_nb_nodes);
   PyTuple_SET_ITEM(__pyx_t_1, 1, __pyx_v_nb_nodes);
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_1, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1265, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_1, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1273, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
     __pyx_t_1 = __pyx_t_3; __Pyx_INCREF(__pyx_t_1); __pyx_t_4 = 0;
     __pyx_t_5 = NULL;
   } else {
-    __pyx_t_4 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1265, __pyx_L1_error)
+    __pyx_t_4 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1273, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_5 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1265, __pyx_L1_error)
+    __pyx_t_5 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1273, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   for (;;) {
@@ -13323,17 +13325,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
       if (likely(PyList_CheckExact(__pyx_t_1))) {
         if (__pyx_t_4 >= PyList_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1265, __pyx_L1_error)
+        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1273, __pyx_L1_error)
         #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_1, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1265, __pyx_L1_error)
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_1, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1273, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_3);
         #endif
       } else {
         if (__pyx_t_4 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1265, __pyx_L1_error)
+        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1273, __pyx_L1_error)
         #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_1, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1265, __pyx_L1_error)
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_1, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1273, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_3);
         #endif
       }
@@ -13343,7 +13345,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1265, __pyx_L1_error)
+          else __PYX_ERR(0, 1273, __pyx_L1_error)
         }
         break;
       }
@@ -13352,43 +13354,43 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
     __Pyx_XDECREF_SET(__pyx_v_node_id, __pyx_t_3);
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1266
+    /* "gedlibpy.pyx":1274
  * 
  * 		for node_id in range(0, nb_nodes):
  * 			graph.add_node(node_id, **node_labels[node_id])             # <<<<<<<<<<<<<<
  * # 			graph.nodes[node_id]['original_node_id'] = original_node_ids[node_id]
  * 
  */
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_add_node); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1266, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_add_node); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1274, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1266, __pyx_L1_error)
+    __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1274, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_INCREF(__pyx_v_node_id);
     __Pyx_GIVEREF(__pyx_v_node_id);
     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_node_id);
-    __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_node_labels, __pyx_v_node_id); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1266, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_node_labels, __pyx_v_node_id); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1274, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
     if (unlikely(__pyx_t_7 == Py_None)) {
       PyErr_SetString(PyExc_TypeError, "argument after ** must be a mapping, not NoneType");
-      __PYX_ERR(0, 1266, __pyx_L1_error)
+      __PYX_ERR(0, 1274, __pyx_L1_error)
     }
     if (likely(PyDict_CheckExact(__pyx_t_7))) {
-      __pyx_t_6 = PyDict_Copy(__pyx_t_7); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1266, __pyx_L1_error)
+      __pyx_t_6 = PyDict_Copy(__pyx_t_7); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1274, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     } else {
-      __pyx_t_6 = PyObject_CallFunctionObjArgs((PyObject*)&PyDict_Type, __pyx_t_7, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1266, __pyx_L1_error)
+      __pyx_t_6 = PyObject_CallFunctionObjArgs((PyObject*)&PyDict_Type, __pyx_t_7, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1274, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     }
-    __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_2, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1266, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_2, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1274, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
 
-    /* "gedlibpy.pyx":1265
+    /* "gedlibpy.pyx":1273
  * 		graph.graph['original_node_ids'] = original_node_ids
  * 
  * 		for node_id in range(0, nb_nodes):             # <<<<<<<<<<<<<<
@@ -13398,14 +13400,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1269
+  /* "gedlibpy.pyx":1277
  * # 			graph.nodes[node_id]['original_node_id'] = original_node_ids[node_id]
  * 
  * 		edges = self.get_graph_edges(graph_id)             # <<<<<<<<<<<<<<
  * 		for (head, tail), labels in edges.items():
  * 			graph.add_edge(head, tail, **labels)
  */
-  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_edges); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1269, __pyx_L1_error)
+  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_get_graph_edges); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1277, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_7);
   __pyx_t_6 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_7))) {
@@ -13419,13 +13421,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_7, __pyx_t_6, __pyx_v_graph_id) : __Pyx_PyObject_CallOneArg(__pyx_t_7, __pyx_v_graph_id);
   __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1269, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1277, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
   __pyx_v_edges = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1270
+  /* "gedlibpy.pyx":1278
  * 
  * 		edges = self.get_graph_edges(graph_id)
  * 		for (head, tail), labels in edges.items():             # <<<<<<<<<<<<<<
@@ -13435,9 +13437,9 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   __pyx_t_4 = 0;
   if (unlikely(__pyx_v_edges == Py_None)) {
     PyErr_Format(PyExc_AttributeError, "'NoneType' object has no attribute '%.30s'", "items");
-    __PYX_ERR(0, 1270, __pyx_L1_error)
+    __PYX_ERR(0, 1278, __pyx_L1_error)
   }
-  __pyx_t_7 = __Pyx_dict_iterator(__pyx_v_edges, 0, __pyx_n_s_items, (&__pyx_t_8), (&__pyx_t_9)); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1270, __pyx_L1_error)
+  __pyx_t_7 = __Pyx_dict_iterator(__pyx_v_edges, 0, __pyx_n_s_items, (&__pyx_t_8), (&__pyx_t_9)); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1278, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_7);
   __Pyx_XDECREF(__pyx_t_1);
   __pyx_t_1 = __pyx_t_7;
@@ -13445,7 +13447,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   while (1) {
     __pyx_t_10 = __Pyx_dict_iter_next(__pyx_t_1, __pyx_t_8, &__pyx_t_4, &__pyx_t_7, &__pyx_t_6, NULL, __pyx_t_9);
     if (unlikely(__pyx_t_10 == 0)) break;
-    if (unlikely(__pyx_t_10 == -1)) __PYX_ERR(0, 1270, __pyx_L1_error)
+    if (unlikely(__pyx_t_10 == -1)) __PYX_ERR(0, 1278, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
     __Pyx_GOTREF(__pyx_t_6);
     if ((likely(PyTuple_CheckExact(__pyx_t_7))) || (PyList_CheckExact(__pyx_t_7))) {
@@ -13454,7 +13456,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
       if (unlikely(size != 2)) {
         if (size > 2) __Pyx_RaiseTooManyValuesError(2);
         else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 1270, __pyx_L1_error)
+        __PYX_ERR(0, 1278, __pyx_L1_error)
       }
       #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
       if (likely(PyTuple_CheckExact(sequence))) {
@@ -13467,15 +13469,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
       __Pyx_INCREF(__pyx_t_2);
       __Pyx_INCREF(__pyx_t_3);
       #else
-      __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1270, __pyx_L1_error)
+      __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1278, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_3 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1270, __pyx_L1_error)
+      __pyx_t_3 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1278, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
       #endif
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     } else {
       Py_ssize_t index = -1;
-      __pyx_t_11 = PyObject_GetIter(__pyx_t_7); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1270, __pyx_L1_error)
+      __pyx_t_11 = PyObject_GetIter(__pyx_t_7); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1278, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_11);
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
       __pyx_t_12 = Py_TYPE(__pyx_t_11)->tp_iternext;
@@ -13483,7 +13485,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
       __Pyx_GOTREF(__pyx_t_2);
       index = 1; __pyx_t_3 = __pyx_t_12(__pyx_t_11); if (unlikely(!__pyx_t_3)) goto __pyx_L7_unpacking_failed;
       __Pyx_GOTREF(__pyx_t_3);
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_11), 2) < 0) __PYX_ERR(0, 1270, __pyx_L1_error)
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_11), 2) < 0) __PYX_ERR(0, 1278, __pyx_L1_error)
       __pyx_t_12 = NULL;
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
       goto __pyx_L8_unpacking_done;
@@ -13491,7 +13493,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
       __pyx_t_12 = NULL;
       if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 1270, __pyx_L1_error)
+      __PYX_ERR(0, 1278, __pyx_L1_error)
       __pyx_L8_unpacking_done:;
     }
     __Pyx_XDECREF_SET(__pyx_v_head, __pyx_t_2);
@@ -13501,16 +13503,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
     __Pyx_XDECREF_SET(__pyx_v_labels, __pyx_t_6);
     __pyx_t_6 = 0;
 
-    /* "gedlibpy.pyx":1271
+    /* "gedlibpy.pyx":1279
  * 		edges = self.get_graph_edges(graph_id)
  * 		for (head, tail), labels in edges.items():
  * 			graph.add_edge(head, tail, **labels)             # <<<<<<<<<<<<<<
  * # 		print(edges)
  * 
  */
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_add_edge); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1271, __pyx_L1_error)
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_graph, __pyx_n_s_add_edge); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1279, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_7 = PyTuple_New(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1271, __pyx_L1_error)
+    __pyx_t_7 = PyTuple_New(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1279, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
     __Pyx_INCREF(__pyx_v_head);
     __Pyx_GIVEREF(__pyx_v_head);
@@ -13520,16 +13522,16 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
     PyTuple_SET_ITEM(__pyx_t_7, 1, __pyx_v_tail);
     if (unlikely(__pyx_v_labels == Py_None)) {
       PyErr_SetString(PyExc_TypeError, "argument after ** must be a mapping, not NoneType");
-      __PYX_ERR(0, 1271, __pyx_L1_error)
+      __PYX_ERR(0, 1279, __pyx_L1_error)
     }
     if (likely(PyDict_CheckExact(__pyx_v_labels))) {
-      __pyx_t_3 = PyDict_Copy(__pyx_v_labels); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1271, __pyx_L1_error)
+      __pyx_t_3 = PyDict_Copy(__pyx_v_labels); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1279, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
     } else {
-      __pyx_t_3 = PyObject_CallFunctionObjArgs((PyObject*)&PyDict_Type, __pyx_v_labels, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1271, __pyx_L1_error)
+      __pyx_t_3 = PyObject_CallFunctionObjArgs((PyObject*)&PyDict_Type, __pyx_v_labels, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1279, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
     }
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_7, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1271, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_7, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1279, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_2);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
@@ -13538,7 +13540,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1274
+  /* "gedlibpy.pyx":1282
  * # 		print(edges)
  * 
  * 		return graph             # <<<<<<<<<<<<<<
@@ -13550,7 +13552,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   __pyx_r = __pyx_v_graph;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1232
+  /* "gedlibpy.pyx":1240
  * 
  * 
  * 	def get_nx_graph(self, graph_id, adj_matrix=True, adj_lists=False, edge_list=False): # @todo             # <<<<<<<<<<<<<<
@@ -13583,7 +13585,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_120get_nx_graph(struct __pyx_obj_8ge
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1277
+/* "gedlibpy.pyx":1285
  * 
  * 
  * 	def get_init_type(self):             # <<<<<<<<<<<<<<
@@ -13612,7 +13614,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_122get_init_type(struct __pyx_obj_8g
   PyObject *__pyx_t_2 = NULL;
   __Pyx_RefNannySetupContext("get_init_type", 0);
 
-  /* "gedlibpy.pyx":1286
+  /* "gedlibpy.pyx":1294
  * 			Initialization type in string.
  * 		"""
  * 		return self.c_env.getInitType().decode('utf-8')             # <<<<<<<<<<<<<<
@@ -13624,15 +13626,15 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_122get_init_type(struct __pyx_obj_8g
     __pyx_t_1 = __pyx_v_self->c_env->getInitType();
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1286, __pyx_L1_error)
+    __PYX_ERR(0, 1294, __pyx_L1_error)
   }
-  __pyx_t_2 = __Pyx_decode_cpp_string(__pyx_t_1, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1286, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_decode_cpp_string(__pyx_t_1, 0, PY_SSIZE_T_MAX, NULL, NULL, PyUnicode_DecodeUTF8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1294, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_r = __pyx_t_2;
   __pyx_t_2 = 0;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1277
+  /* "gedlibpy.pyx":1285
  * 
  * 
  * 	def get_init_type(self):             # <<<<<<<<<<<<<<
@@ -13651,7 +13653,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_122get_init_type(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1308
+/* "gedlibpy.pyx":1316
  * 
  * 
  * 	def load_nx_graph(self, nx_graph, graph_id, graph_name='', graph_class=''):             # <<<<<<<<<<<<<<
@@ -13699,7 +13701,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_125load_nx_graph(PyObject *__pyx_v_s
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_graph_id)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("load_nx_graph", 0, 2, 4, 1); __PYX_ERR(0, 1308, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("load_nx_graph", 0, 2, 4, 1); __PYX_ERR(0, 1316, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
@@ -13715,7 +13717,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_125load_nx_graph(PyObject *__pyx_v_s
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "load_nx_graph") < 0)) __PYX_ERR(0, 1308, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "load_nx_graph") < 0)) __PYX_ERR(0, 1316, __pyx_L3_error)
       }
     } else {
       switch (PyTuple_GET_SIZE(__pyx_args)) {
@@ -13736,7 +13738,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_125load_nx_graph(PyObject *__pyx_v_s
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("load_nx_graph", 0, 2, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1308, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("load_nx_graph", 0, 2, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1316, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.load_nx_graph", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -13773,7 +13775,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   __Pyx_RefNannySetupContext("load_nx_graph", 0);
   __Pyx_INCREF(__pyx_v_graph_id);
 
-  /* "gedlibpy.pyx":1331
+  /* "gedlibpy.pyx":1339
  * 			The ID of the newly loaded graph.
  * 		"""
  * 		if graph_id is None:             # <<<<<<<<<<<<<<
@@ -13784,14 +13786,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   __pyx_t_2 = (__pyx_t_1 != 0);
   if (__pyx_t_2) {
 
-    /* "gedlibpy.pyx":1332
+    /* "gedlibpy.pyx":1340
  * 		"""
  * 		if graph_id is None:
  * 			graph_id = self.add_graph(graph_name, graph_class)             # <<<<<<<<<<<<<<
  * 		else:
  * 			self.clear_graph(graph_id)
  */
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1332, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_graph); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1340, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_5 = NULL;
     __pyx_t_6 = 0;
@@ -13808,7 +13810,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_4)) {
       PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_graph_name, __pyx_v_graph_class};
-      __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1332, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1340, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
       __Pyx_GOTREF(__pyx_t_3);
     } else
@@ -13816,13 +13818,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_4)) {
       PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_graph_name, __pyx_v_graph_class};
-      __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1332, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1340, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
       __Pyx_GOTREF(__pyx_t_3);
     } else
     #endif
     {
-      __pyx_t_7 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1332, __pyx_L1_error)
+      __pyx_t_7 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1340, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_7);
       if (__pyx_t_5) {
         __Pyx_GIVEREF(__pyx_t_5); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_5); __pyx_t_5 = NULL;
@@ -13833,7 +13835,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       __Pyx_INCREF(__pyx_v_graph_class);
       __Pyx_GIVEREF(__pyx_v_graph_class);
       PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_6, __pyx_v_graph_class);
-      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_7, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1332, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_7, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1340, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     }
@@ -13841,7 +13843,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     __Pyx_DECREF_SET(__pyx_v_graph_id, __pyx_t_3);
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1331
+    /* "gedlibpy.pyx":1339
  * 			The ID of the newly loaded graph.
  * 		"""
  * 		if graph_id is None:             # <<<<<<<<<<<<<<
@@ -13851,7 +13853,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     goto __pyx_L3;
   }
 
-  /* "gedlibpy.pyx":1334
+  /* "gedlibpy.pyx":1342
  * 			graph_id = self.add_graph(graph_name, graph_class)
  * 		else:
  * 			self.clear_graph(graph_id)             # <<<<<<<<<<<<<<
@@ -13859,7 +13861,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])
  */
   /*else*/ {
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_clear_graph); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1334, __pyx_L1_error)
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_clear_graph); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1342, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
     __pyx_t_7 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
@@ -13873,29 +13875,29 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     }
     __pyx_t_3 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_7, __pyx_v_graph_id) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_v_graph_id);
     __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-    if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
+    if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1342, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   }
   __pyx_L3:;
 
-  /* "gedlibpy.pyx":1335
+  /* "gedlibpy.pyx":1343
  * 		else:
  * 			self.clear_graph(graph_id)
  * 		for node in nx_graph.nodes:             # <<<<<<<<<<<<<<
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])
  * 		for edge in nx_graph.edges:
  */
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_nodes); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_nodes); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1343, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
     __pyx_t_4 = __pyx_t_3; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
     __pyx_t_9 = NULL;
   } else {
-    __pyx_t_8 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1335, __pyx_L1_error)
+    __pyx_t_8 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1343, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_9 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1335, __pyx_L1_error)
+    __pyx_t_9 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1343, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   for (;;) {
@@ -13903,17 +13905,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       if (likely(PyList_CheckExact(__pyx_t_4))) {
         if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1335, __pyx_L1_error)
+        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1343, __pyx_L1_error)
         #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1335, __pyx_L1_error)
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1343, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_3);
         #endif
       } else {
         if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1335, __pyx_L1_error)
+        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1343, __pyx_L1_error)
         #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1335, __pyx_L1_error)
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1343, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_3);
         #endif
       }
@@ -13923,7 +13925,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1335, __pyx_L1_error)
+          else __PYX_ERR(0, 1343, __pyx_L1_error)
         }
         break;
       }
@@ -13932,20 +13934,20 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     __Pyx_XDECREF_SET(__pyx_v_node, __pyx_t_3);
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1336
+    /* "gedlibpy.pyx":1344
  * 			self.clear_graph(graph_id)
  * 		for node in nx_graph.nodes:
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])             # <<<<<<<<<<<<<<
  * 		for edge in nx_graph.edges:
  * 			self.add_edge(graph_id, str(edge[0]), str(edge[1]), nx_graph.get_edge_data(edge[0], edge[1]))
  */
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1336, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_node); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1344, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_5 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_node); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1336, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_v_node); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1344, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_nodes); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1336, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_nodes); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1344, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_t_10, __pyx_v_node); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1336, __pyx_L1_error)
+    __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_t_10, __pyx_v_node); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1344, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_11);
     __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
     __pyx_t_10 = NULL;
@@ -13963,7 +13965,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_7)) {
       PyObject *__pyx_temp[4] = {__pyx_t_10, __pyx_v_graph_id, __pyx_t_5, __pyx_t_11};
-      __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 3+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1336, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 3+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1344, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
       __Pyx_GOTREF(__pyx_t_3);
       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
@@ -13973,7 +13975,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_7)) {
       PyObject *__pyx_temp[4] = {__pyx_t_10, __pyx_v_graph_id, __pyx_t_5, __pyx_t_11};
-      __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 3+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1336, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 3+__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1344, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
       __Pyx_GOTREF(__pyx_t_3);
       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
@@ -13981,7 +13983,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     } else
     #endif
     {
-      __pyx_t_12 = PyTuple_New(3+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1336, __pyx_L1_error)
+      __pyx_t_12 = PyTuple_New(3+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1344, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_12);
       if (__pyx_t_10) {
         __Pyx_GIVEREF(__pyx_t_10); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_10); __pyx_t_10 = NULL;
@@ -13995,14 +13997,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       PyTuple_SET_ITEM(__pyx_t_12, 2+__pyx_t_6, __pyx_t_11);
       __pyx_t_5 = 0;
       __pyx_t_11 = 0;
-      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_12, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1336, __pyx_L1_error)
+      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_12, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1344, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_3);
       __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
     }
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1335
+    /* "gedlibpy.pyx":1343
  * 		else:
  * 			self.clear_graph(graph_id)
  * 		for node in nx_graph.nodes:             # <<<<<<<<<<<<<<
@@ -14012,22 +14014,22 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   }
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":1337
+  /* "gedlibpy.pyx":1345
  * 		for node in nx_graph.nodes:
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])
  * 		for edge in nx_graph.edges:             # <<<<<<<<<<<<<<
  * 			self.add_edge(graph_id, str(edge[0]), str(edge[1]), nx_graph.get_edge_data(edge[0], edge[1]))
  * 		return graph_id
  */
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_edges); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1337, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_edges); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1345, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
     __pyx_t_3 = __pyx_t_4; __Pyx_INCREF(__pyx_t_3); __pyx_t_8 = 0;
     __pyx_t_9 = NULL;
   } else {
-    __pyx_t_8 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1337, __pyx_L1_error)
+    __pyx_t_8 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1345, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_9 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1337, __pyx_L1_error)
+    __pyx_t_9 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1345, __pyx_L1_error)
   }
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   for (;;) {
@@ -14035,17 +14037,17 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       if (likely(PyList_CheckExact(__pyx_t_3))) {
         if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_3)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_8); __Pyx_INCREF(__pyx_t_4); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1337, __pyx_L1_error)
+        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_8); __Pyx_INCREF(__pyx_t_4); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1345, __pyx_L1_error)
         #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1337, __pyx_L1_error)
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1345, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_4);
         #endif
       } else {
         if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
         #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_8); __Pyx_INCREF(__pyx_t_4); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1337, __pyx_L1_error)
+        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_8); __Pyx_INCREF(__pyx_t_4); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 1345, __pyx_L1_error)
         #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1337, __pyx_L1_error)
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_3, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1345, __pyx_L1_error)
         __Pyx_GOTREF(__pyx_t_4);
         #endif
       }
@@ -14055,7 +14057,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
         PyObject* exc_type = PyErr_Occurred();
         if (exc_type) {
           if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1337, __pyx_L1_error)
+          else __PYX_ERR(0, 1345, __pyx_L1_error)
         }
         break;
       }
@@ -14064,30 +14066,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     __Pyx_XDECREF_SET(__pyx_v_edge, __pyx_t_4);
     __pyx_t_4 = 0;
 
-    /* "gedlibpy.pyx":1338
+    /* "gedlibpy.pyx":1346
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])
  * 		for edge in nx_graph.edges:
  * 			self.add_edge(graph_id, str(edge[0]), str(edge[1]), nx_graph.get_edge_data(edge[0], edge[1]))             # <<<<<<<<<<<<<<
  * 		return graph_id
  * 
  */
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(((PyObject *)__pyx_v_self), __pyx_n_s_add_edge); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_12);
-    __pyx_t_11 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_12); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_11 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_12); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_11);
     __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_12 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_12);
-    __pyx_t_5 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_12); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyUnicode_Type)), __pyx_t_12); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_get_edge_data); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_nx_graph, __pyx_n_s_get_edge_data); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_13 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_13 = __Pyx_GetItemInt(__pyx_v_edge, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_13);
-    __pyx_t_14 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 1338, __pyx_L1_error)
+    __pyx_t_14 = __Pyx_GetItemInt(__pyx_v_edge, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 1346, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_14);
     __pyx_t_15 = NULL;
     __pyx_t_6 = 0;
@@ -14104,7 +14106,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_10)) {
       PyObject *__pyx_temp[3] = {__pyx_t_15, __pyx_t_13, __pyx_t_14};
-      __pyx_t_12 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_12 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_15); __pyx_t_15 = 0;
       __Pyx_GOTREF(__pyx_t_12);
       __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
@@ -14114,7 +14116,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
       PyObject *__pyx_temp[3] = {__pyx_t_15, __pyx_t_13, __pyx_t_14};
-      __pyx_t_12 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_12 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_15); __pyx_t_15 = 0;
       __Pyx_GOTREF(__pyx_t_12);
       __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
@@ -14122,7 +14124,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     } else
     #endif
     {
-      __pyx_t_16 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_16 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_16);
       if (__pyx_t_15) {
         __Pyx_GIVEREF(__pyx_t_15); PyTuple_SET_ITEM(__pyx_t_16, 0, __pyx_t_15); __pyx_t_15 = NULL;
@@ -14133,7 +14135,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       PyTuple_SET_ITEM(__pyx_t_16, 1+__pyx_t_6, __pyx_t_14);
       __pyx_t_13 = 0;
       __pyx_t_14 = 0;
-      __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_16, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_16, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_12);
       __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
     }
@@ -14153,7 +14155,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCALL
     if (PyFunction_Check(__pyx_t_7)) {
       PyObject *__pyx_temp[5] = {__pyx_t_10, __pyx_v_graph_id, __pyx_t_11, __pyx_t_5, __pyx_t_12};
-      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 4+__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 4+__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
       __Pyx_GOTREF(__pyx_t_4);
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
@@ -14164,7 +14166,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     #if CYTHON_FAST_PYCCALL
     if (__Pyx_PyFastCFunction_Check(__pyx_t_7)) {
       PyObject *__pyx_temp[5] = {__pyx_t_10, __pyx_v_graph_id, __pyx_t_11, __pyx_t_5, __pyx_t_12};
-      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 4+__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_7, __pyx_temp+1-__pyx_t_6, 4+__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
       __Pyx_GOTREF(__pyx_t_4);
       __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
@@ -14173,7 +14175,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
     } else
     #endif
     {
-      __pyx_t_16 = PyTuple_New(4+__pyx_t_6); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_16 = PyTuple_New(4+__pyx_t_6); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_16);
       if (__pyx_t_10) {
         __Pyx_GIVEREF(__pyx_t_10); PyTuple_SET_ITEM(__pyx_t_16, 0, __pyx_t_10); __pyx_t_10 = NULL;
@@ -14190,14 +14192,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
       __pyx_t_11 = 0;
       __pyx_t_5 = 0;
       __pyx_t_12 = 0;
-      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_16, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1338, __pyx_L1_error)
+      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_16, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1346, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_4);
       __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
     }
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-    /* "gedlibpy.pyx":1337
+    /* "gedlibpy.pyx":1345
  * 		for node in nx_graph.nodes:
  * 			self.add_node(graph_id, str(node), nx_graph.nodes[node])
  * 		for edge in nx_graph.edges:             # <<<<<<<<<<<<<<
@@ -14207,7 +14209,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   }
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-  /* "gedlibpy.pyx":1339
+  /* "gedlibpy.pyx":1347
  * 		for edge in nx_graph.edges:
  * 			self.add_edge(graph_id, str(edge[0]), str(edge[1]), nx_graph.get_edge_data(edge[0], edge[1]))
  * 		return graph_id             # <<<<<<<<<<<<<<
@@ -14219,7 +14221,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   __pyx_r = __pyx_v_graph_id;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1308
+  /* "gedlibpy.pyx":1316
  * 
  * 
  * 	def load_nx_graph(self, nx_graph, graph_id, graph_name='', graph_class=''):             # <<<<<<<<<<<<<<
@@ -14251,7 +14253,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_124load_nx_graph(struct __pyx_obj_8g
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1342
+/* "gedlibpy.pyx":1350
  * 
  * 
  * 	def compute_induced_cost(self, g_id, h_id, node_map):             # <<<<<<<<<<<<<<
@@ -14294,17 +14296,17 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_127compute_induced_cost(PyObject *__
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_h_id)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, 1); __PYX_ERR(0, 1342, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, 1); __PYX_ERR(0, 1350, __pyx_L3_error)
         }
         CYTHON_FALLTHROUGH;
         case  2:
         if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_node_map)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, 2); __PYX_ERR(0, 1342, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, 2); __PYX_ERR(0, 1350, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_induced_cost") < 0)) __PYX_ERR(0, 1342, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "compute_induced_cost") < 0)) __PYX_ERR(0, 1350, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
       goto __pyx_L5_argtuple_error;
@@ -14319,7 +14321,7 @@ static PyObject *__pyx_pw_8gedlibpy_6GEDEnv_127compute_induced_cost(PyObject *__
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1342, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("compute_induced_cost", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1350, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.GEDEnv.compute_induced_cost", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -14356,26 +14358,26 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   double __pyx_t_12;
   __Pyx_RefNannySetupContext("compute_induced_cost", 0);
 
-  /* "gedlibpy.pyx":1359
+  /* "gedlibpy.pyx":1367
  * 		None.
  * 		"""
  * 		relation = []             # <<<<<<<<<<<<<<
  * 		node_map.as_relation(relation)
  * # 		print(relation)
  */
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1359, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1367, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_v_relation = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1360
+  /* "gedlibpy.pyx":1368
  * 		"""
  * 		relation = []
  * 		node_map.as_relation(relation)             # <<<<<<<<<<<<<<
  * # 		print(relation)
  * 		dummy_node = get_dummy_node()
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_as_relation); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1360, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_as_relation); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1368, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -14389,19 +14391,19 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_relation) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_relation);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1360, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1368, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1362
+  /* "gedlibpy.pyx":1370
  * 		node_map.as_relation(relation)
  * # 		print(relation)
  * 		dummy_node = get_dummy_node()             # <<<<<<<<<<<<<<
  * # 		print(dummy_node)
  * 		for i, val in enumerate(relation):
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_get_dummy_node); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1362, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_get_dummy_node); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1370, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __pyx_t_3 = NULL;
   if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
@@ -14415,13 +14417,13 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   }
   __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
   __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1362, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1370, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __pyx_v_dummy_node = __pyx_t_1;
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1364
+  /* "gedlibpy.pyx":1372
  * 		dummy_node = get_dummy_node()
  * # 		print(dummy_node)
  * 		for i, val in enumerate(relation):             # <<<<<<<<<<<<<<
@@ -14434,45 +14436,45 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   for (;;) {
     if (__pyx_t_4 >= PyList_GET_SIZE(__pyx_t_2)) break;
     #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_3 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1364, __pyx_L1_error)
+    __pyx_t_3 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_4); __Pyx_INCREF(__pyx_t_3); __pyx_t_4++; if (unlikely(0 < 0)) __PYX_ERR(0, 1372, __pyx_L1_error)
     #else
-    __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1364, __pyx_L1_error)
+    __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_4); __pyx_t_4++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1372, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     #endif
     __Pyx_XDECREF_SET(__pyx_v_val, __pyx_t_3);
     __pyx_t_3 = 0;
     __Pyx_INCREF(__pyx_t_1);
     __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
-    __pyx_t_3 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1364, __pyx_L1_error)
+    __pyx_t_3 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1372, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __Pyx_DECREF(__pyx_t_1);
     __pyx_t_1 = __pyx_t_3;
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1365
+    /* "gedlibpy.pyx":1373
  * # 		print(dummy_node)
  * 		for i, val in enumerate(relation):
  * 			val1 = dummy_node if val[0] == np.inf else val[0]             # <<<<<<<<<<<<<<
  * 			val2 = dummy_node if val[1] == np.inf else val[1]
  * 			relation[i] = tuple((val1, val2))
  */
-    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_val, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1365, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_val, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1373, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1365, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1373, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_inf); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1365, __pyx_L1_error)
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_inf); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1373, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = PyObject_RichCompare(__pyx_t_5, __pyx_t_7, Py_EQ); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1365, __pyx_L1_error)
+    __pyx_t_6 = PyObject_RichCompare(__pyx_t_5, __pyx_t_7, Py_EQ); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1373, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 1365, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 1373, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     if (__pyx_t_8) {
       __Pyx_INCREF(__pyx_v_dummy_node);
       __pyx_t_3 = __pyx_v_dummy_node;
     } else {
-      __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_val, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1365, __pyx_L1_error)
+      __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_val, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1373, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_6);
       __pyx_t_3 = __pyx_t_6;
       __pyx_t_6 = 0;
@@ -14480,30 +14482,30 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
     __Pyx_XDECREF_SET(__pyx_v_val1, __pyx_t_3);
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1366
+    /* "gedlibpy.pyx":1374
  * 		for i, val in enumerate(relation):
  * 			val1 = dummy_node if val[0] == np.inf else val[0]
  * 			val2 = dummy_node if val[1] == np.inf else val[1]             # <<<<<<<<<<<<<<
  * 			relation[i] = tuple((val1, val2))
  * # 		print(relation)
  */
-    __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_val, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1366, __pyx_L1_error)
+    __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_val, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1374, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_np); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1366, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_np); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1374, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_inf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1366, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_inf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1374, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __pyx_t_7 = PyObject_RichCompare(__pyx_t_6, __pyx_t_5, Py_EQ); __Pyx_XGOTREF(__pyx_t_7); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1366, __pyx_L1_error)
+    __pyx_t_7 = PyObject_RichCompare(__pyx_t_6, __pyx_t_5, Py_EQ); __Pyx_XGOTREF(__pyx_t_7); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1374, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_7); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 1366, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_7); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 1374, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     if (__pyx_t_8) {
       __Pyx_INCREF(__pyx_v_dummy_node);
       __pyx_t_3 = __pyx_v_dummy_node;
     } else {
-      __pyx_t_7 = __Pyx_GetItemInt(__pyx_v_val, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1366, __pyx_L1_error)
+      __pyx_t_7 = __Pyx_GetItemInt(__pyx_v_val, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1374, __pyx_L1_error)
       __Pyx_GOTREF(__pyx_t_7);
       __pyx_t_3 = __pyx_t_7;
       __pyx_t_7 = 0;
@@ -14511,14 +14513,14 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
     __Pyx_XDECREF_SET(__pyx_v_val2, __pyx_t_3);
     __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1367
+    /* "gedlibpy.pyx":1375
  * 			val1 = dummy_node if val[0] == np.inf else val[0]
  * 			val2 = dummy_node if val[1] == np.inf else val[1]
  * 			relation[i] = tuple((val1, val2))             # <<<<<<<<<<<<<<
  * # 		print(relation)
  * 		induced_cost = self.c_env.computeInducedCost(g_id, h_id, relation)
  */
-    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1367, __pyx_L1_error)
+    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1375, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_3);
     __Pyx_INCREF(__pyx_v_val1);
     __Pyx_GIVEREF(__pyx_v_val1);
@@ -14526,10 +14528,10 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
     __Pyx_INCREF(__pyx_v_val2);
     __Pyx_GIVEREF(__pyx_v_val2);
     PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_v_val2);
-    if (unlikely(PyObject_SetItem(__pyx_v_relation, __pyx_v_i, __pyx_t_3) < 0)) __PYX_ERR(0, 1367, __pyx_L1_error)
+    if (unlikely(PyObject_SetItem(__pyx_v_relation, __pyx_v_i, __pyx_t_3) < 0)) __PYX_ERR(0, 1375, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
 
-    /* "gedlibpy.pyx":1364
+    /* "gedlibpy.pyx":1372
  * 		dummy_node = get_dummy_node()
  * # 		print(dummy_node)
  * 		for i, val in enumerate(relation):             # <<<<<<<<<<<<<<
@@ -14540,34 +14542,34 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1369
+  /* "gedlibpy.pyx":1377
  * 			relation[i] = tuple((val1, val2))
  * # 		print(relation)
  * 		induced_cost = self.c_env.computeInducedCost(g_id, h_id, relation)             # <<<<<<<<<<<<<<
  * 		node_map.set_induced_cost(induced_cost)
  * 
  */
-  __pyx_t_9 = __Pyx_PyInt_As_size_t(__pyx_v_g_id); if (unlikely((__pyx_t_9 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1369, __pyx_L1_error)
-  __pyx_t_10 = __Pyx_PyInt_As_size_t(__pyx_v_h_id); if (unlikely((__pyx_t_10 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1369, __pyx_L1_error)
-  __pyx_t_11 = __pyx_convert_vector_from_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_relation); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1369, __pyx_L1_error)
+  __pyx_t_9 = __Pyx_PyInt_As_size_t(__pyx_v_g_id); if (unlikely((__pyx_t_9 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1377, __pyx_L1_error)
+  __pyx_t_10 = __Pyx_PyInt_As_size_t(__pyx_v_h_id); if (unlikely((__pyx_t_10 == (size_t)-1) && PyErr_Occurred())) __PYX_ERR(0, 1377, __pyx_L1_error)
+  __pyx_t_11 = __pyx_convert_vector_from_py_std_3a__3a_pair_3c_size_t_2c_size_t_3e___(__pyx_v_relation); if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 1377, __pyx_L1_error)
   try {
     __pyx_t_12 = __pyx_v_self->c_env->computeInducedCost(__pyx_t_9, __pyx_t_10, __pyx_t_11);
   } catch(...) {
     __Pyx_CppExn2PyErr();
-    __PYX_ERR(0, 1369, __pyx_L1_error)
+    __PYX_ERR(0, 1377, __pyx_L1_error)
   }
   __pyx_v_induced_cost = __pyx_t_12;
 
-  /* "gedlibpy.pyx":1370
+  /* "gedlibpy.pyx":1378
  * # 		print(relation)
  * 		induced_cost = self.c_env.computeInducedCost(g_id, h_id, relation)
  * 		node_map.set_induced_cost(induced_cost)             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_set_induced_cost); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1370, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_node_map, __pyx_n_s_set_induced_cost); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1378, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = PyFloat_FromDouble(__pyx_v_induced_cost); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1370, __pyx_L1_error)
+  __pyx_t_3 = PyFloat_FromDouble(__pyx_v_induced_cost); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1378, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __pyx_t_7 = NULL;
   if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
@@ -14582,12 +14584,12 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_126compute_induced_cost(struct __pyx
   __pyx_t_1 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_7, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
   __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1370, __pyx_L1_error)
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1378, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1342
+  /* "gedlibpy.pyx":1350
  * 
  * 
  * 	def compute_induced_cost(self, g_id, h_id, node_map):             # <<<<<<<<<<<<<<
@@ -14726,7 +14728,7 @@ static PyObject *__pyx_pf_8gedlibpy_6GEDEnv_130__setstate_cython__(CYTHON_UNUSED
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1400
+/* "gedlibpy.pyx":1408
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -14767,11 +14769,11 @@ static PyObject *__pyx_pw_8gedlibpy_13EditCostError_1__init__(PyObject *__pyx_se
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_message)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1400, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1408, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1400, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1408, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -14784,7 +14786,7 @@ static PyObject *__pyx_pw_8gedlibpy_13EditCostError_1__init__(PyObject *__pyx_se
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1400, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1408, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.EditCostError.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -14802,16 +14804,16 @@ static PyObject *__pyx_pf_8gedlibpy_13EditCostError___init__(CYTHON_UNUSED PyObj
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__init__", 0);
 
-  /* "gedlibpy.pyx":1407
+  /* "gedlibpy.pyx":1415
  * 			:type message: string
  * 		"""
  * 		self.message = message             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1407, __pyx_L1_error)
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1415, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1400
+  /* "gedlibpy.pyx":1408
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -14831,7 +14833,7 @@ static PyObject *__pyx_pf_8gedlibpy_13EditCostError___init__(CYTHON_UNUSED PyObj
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1417
+/* "gedlibpy.pyx":1425
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -14872,11 +14874,11 @@ static PyObject *__pyx_pw_8gedlibpy_11MethodError_1__init__(PyObject *__pyx_self
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_message)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1417, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1425, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1417, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1425, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -14889,7 +14891,7 @@ static PyObject *__pyx_pw_8gedlibpy_11MethodError_1__init__(PyObject *__pyx_self
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1417, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1425, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.MethodError.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -14907,16 +14909,16 @@ static PyObject *__pyx_pf_8gedlibpy_11MethodError___init__(CYTHON_UNUSED PyObjec
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__init__", 0);
 
-  /* "gedlibpy.pyx":1424
+  /* "gedlibpy.pyx":1432
  * 			:type message: string
  * 		"""
  * 		self.message = message             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1424, __pyx_L1_error)
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1432, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1417
+  /* "gedlibpy.pyx":1425
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -14936,7 +14938,7 @@ static PyObject *__pyx_pf_8gedlibpy_11MethodError___init__(CYTHON_UNUSED PyObjec
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1434
+/* "gedlibpy.pyx":1442
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -14977,11 +14979,11 @@ static PyObject *__pyx_pw_8gedlibpy_9InitError_1__init__(PyObject *__pyx_self, P
         case  1:
         if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_message)) != 0)) kw_args--;
         else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1434, __pyx_L3_error)
+          __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, 1); __PYX_ERR(0, 1442, __pyx_L3_error)
         }
       }
       if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1434, __pyx_L3_error)
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1442, __pyx_L3_error)
       }
     } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
       goto __pyx_L5_argtuple_error;
@@ -14994,7 +14996,7 @@ static PyObject *__pyx_pw_8gedlibpy_9InitError_1__init__(PyObject *__pyx_self, P
   }
   goto __pyx_L4_argument_unpacking_done;
   __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1434, __pyx_L3_error)
+  __Pyx_RaiseArgtupleInvalid("__init__", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1442, __pyx_L3_error)
   __pyx_L3_error:;
   __Pyx_AddTraceback("gedlibpy.InitError.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
   __Pyx_RefNannyFinishContext();
@@ -15012,16 +15014,16 @@ static PyObject *__pyx_pf_8gedlibpy_9InitError___init__(CYTHON_UNUSED PyObject *
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__init__", 0);
 
-  /* "gedlibpy.pyx":1441
+  /* "gedlibpy.pyx":1449
  * 			:type message: string
  * 		"""
  * 		self.message = message             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1441, __pyx_L1_error)
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_message, __pyx_v_message) < 0) __PYX_ERR(0, 1449, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1434
+  /* "gedlibpy.pyx":1442
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
@@ -15041,7 +15043,7 @@ static PyObject *__pyx_pf_8gedlibpy_9InitError___init__(CYTHON_UNUSED PyObject *
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1448
+/* "gedlibpy.pyx":1456
  * #########################################
  * 
  * def encode_your_map(map_u):             # <<<<<<<<<<<<<<
@@ -15081,19 +15083,19 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
   PyObject *__pyx_t_9 = NULL;
   __Pyx_RefNannySetupContext("encode_your_map", 0);
 
-  /* "gedlibpy.pyx":1460
+  /* "gedlibpy.pyx":1468
  * 
  * 	"""
  * 	res = {}             # <<<<<<<<<<<<<<
  * 	for key, value in map_u.items():
  * 		res[key.encode('utf-8')] = value.encode('utf-8')
  */
-  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1460, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1468, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_v_res = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1461
+  /* "gedlibpy.pyx":1469
  * 	"""
  * 	res = {}
  * 	for key, value in map_u.items():             # <<<<<<<<<<<<<<
@@ -15103,9 +15105,9 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
   __pyx_t_2 = 0;
   if (unlikely(__pyx_v_map_u == Py_None)) {
     PyErr_Format(PyExc_AttributeError, "'NoneType' object has no attribute '%.30s'", "items");
-    __PYX_ERR(0, 1461, __pyx_L1_error)
+    __PYX_ERR(0, 1469, __pyx_L1_error)
   }
-  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_u, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1461, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_u, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1469, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_XDECREF(__pyx_t_1);
   __pyx_t_1 = __pyx_t_5;
@@ -15113,7 +15115,7 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
   while (1) {
     __pyx_t_7 = __Pyx_dict_iter_next(__pyx_t_1, __pyx_t_3, &__pyx_t_2, &__pyx_t_5, &__pyx_t_6, NULL, __pyx_t_4);
     if (unlikely(__pyx_t_7 == 0)) break;
-    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1461, __pyx_L1_error)
+    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1469, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_XDECREF_SET(__pyx_v_key, __pyx_t_5);
@@ -15121,14 +15123,14 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
     __Pyx_XDECREF_SET(__pyx_v_value, __pyx_t_6);
     __pyx_t_6 = 0;
 
-    /* "gedlibpy.pyx":1462
+    /* "gedlibpy.pyx":1470
  * 	res = {}
  * 	for key, value in map_u.items():
  * 		res[key.encode('utf-8')] = value.encode('utf-8')             # <<<<<<<<<<<<<<
  * 	return res
  * 
  */
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_encode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1462, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_encode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1470, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_8 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
@@ -15142,10 +15144,10 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
     }
     __pyx_t_6 = (__pyx_t_8) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_8, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1462, __pyx_L1_error)
+    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1470, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_key, __pyx_n_s_encode); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1462, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_key, __pyx_n_s_encode); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1470, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_8);
     __pyx_t_9 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_8))) {
@@ -15159,16 +15161,16 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
     }
     __pyx_t_5 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_9, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1462, __pyx_L1_error)
+    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1470, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (unlikely(PyDict_SetItem(__pyx_v_res, __pyx_t_5, __pyx_t_6) < 0)) __PYX_ERR(0, 1462, __pyx_L1_error)
+    if (unlikely(PyDict_SetItem(__pyx_v_res, __pyx_t_5, __pyx_t_6) < 0)) __PYX_ERR(0, 1470, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1463
+  /* "gedlibpy.pyx":1471
  * 	for key, value in map_u.items():
  * 		res[key.encode('utf-8')] = value.encode('utf-8')
  * 	return res             # <<<<<<<<<<<<<<
@@ -15180,7 +15182,7 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
   __pyx_r = __pyx_v_res;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1448
+  /* "gedlibpy.pyx":1456
  * #########################################
  * 
  * def encode_your_map(map_u):             # <<<<<<<<<<<<<<
@@ -15206,7 +15208,7 @@ static PyObject *__pyx_pf_8gedlibpy_8encode_your_map(CYTHON_UNUSED PyObject *__p
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1466
+/* "gedlibpy.pyx":1474
  * 
  * 
  * def decode_your_map(map_b):             # <<<<<<<<<<<<<<
@@ -15246,19 +15248,19 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
   PyObject *__pyx_t_9 = NULL;
   __Pyx_RefNannySetupContext("decode_your_map", 0);
 
-  /* "gedlibpy.pyx":1478
+  /* "gedlibpy.pyx":1486
  * 
  * 	"""
  * 	res = {}             # <<<<<<<<<<<<<<
  * 	for key, value in map_b.items():
  * 		res[key.decode('utf-8')] = value.decode('utf-8')
  */
-  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1478, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1486, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_v_res = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1479
+  /* "gedlibpy.pyx":1487
  * 	"""
  * 	res = {}
  * 	for key, value in map_b.items():             # <<<<<<<<<<<<<<
@@ -15268,9 +15270,9 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
   __pyx_t_2 = 0;
   if (unlikely(__pyx_v_map_b == Py_None)) {
     PyErr_Format(PyExc_AttributeError, "'NoneType' object has no attribute '%.30s'", "items");
-    __PYX_ERR(0, 1479, __pyx_L1_error)
+    __PYX_ERR(0, 1487, __pyx_L1_error)
   }
-  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_b, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1479, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_b, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1487, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_XDECREF(__pyx_t_1);
   __pyx_t_1 = __pyx_t_5;
@@ -15278,7 +15280,7 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
   while (1) {
     __pyx_t_7 = __Pyx_dict_iter_next(__pyx_t_1, __pyx_t_3, &__pyx_t_2, &__pyx_t_5, &__pyx_t_6, NULL, __pyx_t_4);
     if (unlikely(__pyx_t_7 == 0)) break;
-    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1479, __pyx_L1_error)
+    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1487, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_XDECREF_SET(__pyx_v_key, __pyx_t_5);
@@ -15286,14 +15288,14 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
     __Pyx_XDECREF_SET(__pyx_v_value, __pyx_t_6);
     __pyx_t_6 = 0;
 
-    /* "gedlibpy.pyx":1480
+    /* "gedlibpy.pyx":1488
  * 	res = {}
  * 	for key, value in map_b.items():
  * 		res[key.decode('utf-8')] = value.decode('utf-8')             # <<<<<<<<<<<<<<
  * 	return res
  * 
  */
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_decode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1480, __pyx_L1_error)
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_decode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1488, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_8 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
@@ -15307,10 +15309,10 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
     }
     __pyx_t_6 = (__pyx_t_8) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_8, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1480, __pyx_L1_error)
+    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1488, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_key, __pyx_n_s_decode); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1480, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_key, __pyx_n_s_decode); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1488, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_8);
     __pyx_t_9 = NULL;
     if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_8))) {
@@ -15324,16 +15326,16 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
     }
     __pyx_t_5 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_9, __pyx_kp_u_utf_8) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_kp_u_utf_8);
     __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1480, __pyx_L1_error)
+    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1488, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (unlikely(PyDict_SetItem(__pyx_v_res, __pyx_t_5, __pyx_t_6) < 0)) __PYX_ERR(0, 1480, __pyx_L1_error)
+    if (unlikely(PyDict_SetItem(__pyx_v_res, __pyx_t_5, __pyx_t_6) < 0)) __PYX_ERR(0, 1488, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1481
+  /* "gedlibpy.pyx":1489
  * 	for key, value in map_b.items():
  * 		res[key.decode('utf-8')] = value.decode('utf-8')
  * 	return res             # <<<<<<<<<<<<<<
@@ -15345,7 +15347,7 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
   __pyx_r = __pyx_v_res;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1466
+  /* "gedlibpy.pyx":1474
  * 
  * 
  * def decode_your_map(map_b):             # <<<<<<<<<<<<<<
@@ -15371,7 +15373,7 @@ static PyObject *__pyx_pf_8gedlibpy_10decode_your_map(CYTHON_UNUSED PyObject *__
   return __pyx_r;
 }
 
-/* "gedlibpy.pyx":1484
+/* "gedlibpy.pyx":1492
  * 
  * 
  * def decode_graph_edges(map_edge_b):             # <<<<<<<<<<<<<<
@@ -15410,19 +15412,19 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
   PyObject *__pyx_t_8 = NULL;
   __Pyx_RefNannySetupContext("decode_graph_edges", 0);
 
-  /* "gedlibpy.pyx":1502
+  /* "gedlibpy.pyx":1510
  * 	This is a helper function for function `GEDEnv.get_graph_edges()`.
  * 	"""
  * 	map_edges = {}             # <<<<<<<<<<<<<<
  * 	for key, value in map_edge_b.items():
  * 		map_edges[key] = decode_your_map(value)
  */
-  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1502, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1510, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_v_map_edges = ((PyObject*)__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1503
+  /* "gedlibpy.pyx":1511
  * 	"""
  * 	map_edges = {}
  * 	for key, value in map_edge_b.items():             # <<<<<<<<<<<<<<
@@ -15432,9 +15434,9 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
   __pyx_t_2 = 0;
   if (unlikely(__pyx_v_map_edge_b == Py_None)) {
     PyErr_Format(PyExc_AttributeError, "'NoneType' object has no attribute '%.30s'", "items");
-    __PYX_ERR(0, 1503, __pyx_L1_error)
+    __PYX_ERR(0, 1511, __pyx_L1_error)
   }
-  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_edge_b, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1503, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_dict_iterator(__pyx_v_map_edge_b, 0, __pyx_n_s_items, (&__pyx_t_3), (&__pyx_t_4)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1511, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_XDECREF(__pyx_t_1);
   __pyx_t_1 = __pyx_t_5;
@@ -15442,7 +15444,7 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
   while (1) {
     __pyx_t_7 = __Pyx_dict_iter_next(__pyx_t_1, __pyx_t_3, &__pyx_t_2, &__pyx_t_5, &__pyx_t_6, NULL, __pyx_t_4);
     if (unlikely(__pyx_t_7 == 0)) break;
-    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1503, __pyx_L1_error)
+    if (unlikely(__pyx_t_7 == -1)) __PYX_ERR(0, 1511, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_XDECREF_SET(__pyx_v_key, __pyx_t_5);
@@ -15450,14 +15452,14 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
     __Pyx_XDECREF_SET(__pyx_v_value, __pyx_t_6);
     __pyx_t_6 = 0;
 
-    /* "gedlibpy.pyx":1504
+    /* "gedlibpy.pyx":1512
  * 	map_edges = {}
  * 	for key, value in map_edge_b.items():
  * 		map_edges[key] = decode_your_map(value)             # <<<<<<<<<<<<<<
  * 	return map_edges
  * 
  */
-    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1504, __pyx_L1_error)
+    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_decode_your_map); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1512, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_5);
     __pyx_t_8 = NULL;
     if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
@@ -15471,15 +15473,15 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
     }
     __pyx_t_6 = (__pyx_t_8) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_8, __pyx_v_value) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_v_value);
     __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1504, __pyx_L1_error)
+    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1512, __pyx_L1_error)
     __Pyx_GOTREF(__pyx_t_6);
     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(PyDict_SetItem(__pyx_v_map_edges, __pyx_v_key, __pyx_t_6) < 0)) __PYX_ERR(0, 1504, __pyx_L1_error)
+    if (unlikely(PyDict_SetItem(__pyx_v_map_edges, __pyx_v_key, __pyx_t_6) < 0)) __PYX_ERR(0, 1512, __pyx_L1_error)
     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
   }
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1505
+  /* "gedlibpy.pyx":1513
  * 	for key, value in map_edge_b.items():
  * 		map_edges[key] = decode_your_map(value)
  * 	return map_edges             # <<<<<<<<<<<<<<
@@ -15491,7 +15493,7 @@ static PyObject *__pyx_pf_8gedlibpy_12decode_graph_edges(CYTHON_UNUSED PyObject
   __pyx_r = __pyx_v_map_edges;
   goto __pyx_L0;
 
-  /* "gedlibpy.pyx":1484
+  /* "gedlibpy.pyx":1492
  * 
  * 
  * def decode_graph_edges(map_edge_b):             # <<<<<<<<<<<<<<
@@ -20776,12 +20778,12 @@ static int __pyx_import_star_set(PyObject *o, PyObject* py_name, char *name) {
     "UINT32_t",
     "X",
     "Y",
+    "__pyx_ctuple_2ae87__58ae9__std__in_map__lAngstd__in_pair__lAngsize_t__comma_size_t__rAng__comma_std__in_map__lAngstd__in_string__comma_std__in_string__rAng__rAng__etc__etc",
+    "__pyx_ctuple_2ae87__58ae9__std__in_map__lAngstd__in_pair__lAngsize_t__comma_size_t__rAng__comma_std__in_map__lAngstd__in_string__comma_std__in_string__rAng__rAng__etc__etc_struct",
     "__pyx_ctuple_Py_ssize_t",
     "__pyx_ctuple_Py_ssize_t__and_Py_ssize_t",
     "__pyx_ctuple_Py_ssize_t__and_Py_ssize_t_struct",
     "__pyx_ctuple_Py_ssize_t_struct",
-    "__pyx_ctuple_bfa21__6a9b6__std__in_map__lAngstd__in_pair__lAngsize_t__comma_size_t__rAng__comma_std__in_map__lAngstd__in_string__comma_std__in_string__rAng__rAng__etc__etc",
-    "__pyx_ctuple_bfa21__6a9b6__std__in_map__lAngstd__in_pair__lAngsize_t__comma_size_t__rAng__comma_std__in_map__lAngstd__in_string__comma_std__in_string__rAng__rAng__etc__etc_struct",
     "__pyx_ctuple_double",
     "__pyx_ctuple_double_struct",
     "__pyx_ctuple_size_t",
@@ -21149,9 +21151,9 @@ static __Pyx_StringTabEntry __pyx_string_tab[] = {
   {0, 0, 0, 0, 0, 0, 0}
 };
 static CYTHON_SMALL_CODE int __Pyx_InitCachedBuiltins(void) {
-  __pyx_builtin_range = __Pyx_GetBuiltinName(__pyx_n_s_range); if (!__pyx_builtin_range) __PYX_ERR(0, 750, __pyx_L1_error)
-  __pyx_builtin_print = __Pyx_GetBuiltinName(__pyx_n_s_print); if (!__pyx_builtin_print) __PYX_ERR(0, 977, __pyx_L1_error)
-  __pyx_builtin_enumerate = __Pyx_GetBuiltinName(__pyx_n_s_enumerate); if (!__pyx_builtin_enumerate) __PYX_ERR(0, 1364, __pyx_L1_error)
+  __pyx_builtin_range = __Pyx_GetBuiltinName(__pyx_n_s_range); if (!__pyx_builtin_range) __PYX_ERR(0, 758, __pyx_L1_error)
+  __pyx_builtin_print = __Pyx_GetBuiltinName(__pyx_n_s_print); if (!__pyx_builtin_print) __PYX_ERR(0, 985, __pyx_L1_error)
+  __pyx_builtin_enumerate = __Pyx_GetBuiltinName(__pyx_n_s_enumerate); if (!__pyx_builtin_enumerate) __PYX_ERR(0, 1372, __pyx_L1_error)
   __pyx_builtin_TypeError = __Pyx_GetBuiltinName(__pyx_n_s_TypeError); if (!__pyx_builtin_TypeError) __PYX_ERR(1, 2, __pyx_L1_error)
   __pyx_builtin_ValueError = __Pyx_GetBuiltinName(__pyx_n_s_ValueError); if (!__pyx_builtin_ValueError) __PYX_ERR(2, 272, __pyx_L1_error)
   __pyx_builtin_RuntimeError = __Pyx_GetBuiltinName(__pyx_n_s_RuntimeError); if (!__pyx_builtin_RuntimeError) __PYX_ERR(2, 856, __pyx_L1_error)
@@ -21166,80 +21168,80 @@ static CYTHON_SMALL_CODE int __Pyx_InitCachedConstants(void) {
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__Pyx_InitCachedConstants", 0);
 
-  /* "gedlibpy.pyx":977
+  /* "gedlibpy.pyx":985
  * 			self.restart_env()
  * 
  * 		print("Loading graphs in progress...")             # <<<<<<<<<<<<<<
  * 		for graph in dataset :
  * 			self.add_nx_graph(graph, classes)
  */
-  __pyx_tuple__4 = PyTuple_Pack(1, __pyx_kp_u_Loading_graphs_in_progress); if (unlikely(!__pyx_tuple__4)) __PYX_ERR(0, 977, __pyx_L1_error)
+  __pyx_tuple__4 = PyTuple_Pack(1, __pyx_kp_u_Loading_graphs_in_progress); if (unlikely(!__pyx_tuple__4)) __PYX_ERR(0, 985, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__4);
   __Pyx_GIVEREF(__pyx_tuple__4);
 
-  /* "gedlibpy.pyx":981
+  /* "gedlibpy.pyx":989
  * 			self.add_nx_graph(graph, classes)
  * 		listID = self.graph_ids()
  * 		print("Graphs loaded ! ")             # <<<<<<<<<<<<<<
  * 		print("Number of graphs = " + str(listID[1]))
  * 
  */
-  __pyx_tuple__5 = PyTuple_Pack(1, __pyx_kp_u_Graphs_loaded); if (unlikely(!__pyx_tuple__5)) __PYX_ERR(0, 981, __pyx_L1_error)
+  __pyx_tuple__5 = PyTuple_Pack(1, __pyx_kp_u_Graphs_loaded); if (unlikely(!__pyx_tuple__5)) __PYX_ERR(0, 989, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__5);
   __Pyx_GIVEREF(__pyx_tuple__5);
 
-  /* "gedlibpy.pyx":985
+  /* "gedlibpy.pyx":993
  * 
  * 		self.set_edit_cost(edit_cost)
  * 		print("Initialization in progress...")             # <<<<<<<<<<<<<<
  * 		self.init(init_option)
  * 		print("Initialization terminated !")
  */
-  __pyx_tuple__6 = PyTuple_Pack(1, __pyx_kp_u_Initialization_in_progress); if (unlikely(!__pyx_tuple__6)) __PYX_ERR(0, 985, __pyx_L1_error)
+  __pyx_tuple__6 = PyTuple_Pack(1, __pyx_kp_u_Initialization_in_progress); if (unlikely(!__pyx_tuple__6)) __PYX_ERR(0, 993, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__6);
   __Pyx_GIVEREF(__pyx_tuple__6);
 
-  /* "gedlibpy.pyx":987
+  /* "gedlibpy.pyx":995
  * 		print("Initialization in progress...")
  * 		self.init(init_option)
  * 		print("Initialization terminated !")             # <<<<<<<<<<<<<<
  * 
  * 		self.set_method(method, options)
  */
-  __pyx_tuple__7 = PyTuple_Pack(1, __pyx_kp_u_Initialization_terminated); if (unlikely(!__pyx_tuple__7)) __PYX_ERR(0, 987, __pyx_L1_error)
+  __pyx_tuple__7 = PyTuple_Pack(1, __pyx_kp_u_Initialization_terminated); if (unlikely(!__pyx_tuple__7)) __PYX_ERR(0, 995, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__7);
   __Pyx_GIVEREF(__pyx_tuple__7);
 
-  /* "gedlibpy.pyx":1002
+  /* "gedlibpy.pyx":1010
  * 				resMapping[g][h] = self.get_node_map(g, h)
  * 
  * 		print("Finish ! The return contains edit distances and NodeMap but you can check the result with graphs'ID until you restart the environment")             # <<<<<<<<<<<<<<
  * 		return resDistance, resMapping
  * 
  */
-  __pyx_tuple__8 = PyTuple_Pack(1, __pyx_kp_u_Finish_The_return_contains_edit); if (unlikely(!__pyx_tuple__8)) __PYX_ERR(0, 1002, __pyx_L1_error)
+  __pyx_tuple__8 = PyTuple_Pack(1, __pyx_kp_u_Finish_The_return_contains_edit); if (unlikely(!__pyx_tuple__8)) __PYX_ERR(0, 1010, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__8);
   __Pyx_GIVEREF(__pyx_tuple__8);
 
-  /* "gedlibpy.pyx":1057
+  /* "gedlibpy.pyx":1065
  * 		#return res
  * 
  * 		print ("Finish ! You can check the result with each ID of graphs ! There are in the return")             # <<<<<<<<<<<<<<
  * 		print ("Please don't restart the environment or recall this function, you will lose your results !")
  * 		return listID
  */
-  __pyx_tuple__9 = PyTuple_Pack(1, __pyx_kp_u_Finish_You_can_check_the_result); if (unlikely(!__pyx_tuple__9)) __PYX_ERR(0, 1057, __pyx_L1_error)
+  __pyx_tuple__9 = PyTuple_Pack(1, __pyx_kp_u_Finish_You_can_check_the_result); if (unlikely(!__pyx_tuple__9)) __PYX_ERR(0, 1065, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__9);
   __Pyx_GIVEREF(__pyx_tuple__9);
 
-  /* "gedlibpy.pyx":1058
+  /* "gedlibpy.pyx":1066
  * 
  * 		print ("Finish ! You can check the result with each ID of graphs ! There are in the return")
  * 		print ("Please don't restart the environment or recall this function, you will lose your results !")             # <<<<<<<<<<<<<<
  * 		return listID
  * 
  */
-  __pyx_tuple__10 = PyTuple_Pack(1, __pyx_kp_u_Please_don_t_restart_the_environ); if (unlikely(!__pyx_tuple__10)) __PYX_ERR(0, 1058, __pyx_L1_error)
+  __pyx_tuple__10 = PyTuple_Pack(1, __pyx_kp_u_Please_don_t_restart_the_environ); if (unlikely(!__pyx_tuple__10)) __PYX_ERR(0, 1066, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__10);
   __Pyx_GIVEREF(__pyx_tuple__10);
 
@@ -21339,122 +21341,122 @@ static CYTHON_SMALL_CODE int __Pyx_InitCachedConstants(void) {
   __Pyx_GOTREF(__pyx_tuple__19);
   __Pyx_GIVEREF(__pyx_tuple__19);
 
-  /* "gedlibpy.pyx":128
+  /* "gedlibpy.pyx":129
  * 
  * 
  * def get_edit_cost_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents edit cost functions and returns the result.
  */
-  __pyx_tuple__21 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__21)) __PYX_ERR(0, 128, __pyx_L1_error)
+  __pyx_tuple__21 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__21)) __PYX_ERR(0, 129, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__21);
   __Pyx_GIVEREF(__pyx_tuple__21);
-  __pyx_codeobj__22 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__21, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_edit_cost_options, 128, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__22)) __PYX_ERR(0, 128, __pyx_L1_error)
+  __pyx_codeobj__22 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__21, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_edit_cost_options, 129, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__22)) __PYX_ERR(0, 129, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":142
+  /* "gedlibpy.pyx":143
  * 
  * 
  * def get_method_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents method for edit distance computation between graphs and returns the result.
  */
-  __pyx_tuple__23 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__23)) __PYX_ERR(0, 142, __pyx_L1_error)
+  __pyx_tuple__23 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__23)) __PYX_ERR(0, 143, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__23);
   __Pyx_GIVEREF(__pyx_tuple__23);
-  __pyx_codeobj__24 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__23, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_method_options, 142, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__24)) __PYX_ERR(0, 142, __pyx_L1_error)
+  __pyx_codeobj__24 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__23, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_method_options, 143, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__24)) __PYX_ERR(0, 143, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":155
+  /* "gedlibpy.pyx":156
  * 
  * 
  * def get_init_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents initialization parameters for the environment computation for graphs and returns the result.
  */
-  __pyx_tuple__25 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__25)) __PYX_ERR(0, 155, __pyx_L1_error)
+  __pyx_tuple__25 = PyTuple_Pack(1, __pyx_n_s_option); if (unlikely(!__pyx_tuple__25)) __PYX_ERR(0, 156, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__25);
   __Pyx_GIVEREF(__pyx_tuple__25);
-  __pyx_codeobj__26 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__25, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_init_options, 155, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__26)) __PYX_ERR(0, 155, __pyx_L1_error)
+  __pyx_codeobj__26 = (PyObject*)__Pyx_PyCode_New(0, 0, 1, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__25, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_init_options, 156, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__26)) __PYX_ERR(0, 156, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":168
+  /* "gedlibpy.pyx":169
  * 
  * 
  * def get_dummy_node() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Returns the ID of a dummy node.
  */
-  __pyx_codeobj__27 = (PyObject*)__Pyx_PyCode_New(0, 0, 0, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_dummy_node, 168, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__27)) __PYX_ERR(0, 168, __pyx_L1_error)
+  __pyx_codeobj__27 = (PyObject*)__Pyx_PyCode_New(0, 0, 0, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_get_dummy_node, 169, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__27)) __PYX_ERR(0, 169, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1400
+  /* "gedlibpy.pyx":1408
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_tuple__28 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__28)) __PYX_ERR(0, 1400, __pyx_L1_error)
+  __pyx_tuple__28 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__28)) __PYX_ERR(0, 1408, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__28);
   __Pyx_GIVEREF(__pyx_tuple__28);
-  __pyx_codeobj__29 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__28, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1400, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__29)) __PYX_ERR(0, 1400, __pyx_L1_error)
+  __pyx_codeobj__29 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__28, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1408, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__29)) __PYX_ERR(0, 1408, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1417
+  /* "gedlibpy.pyx":1425
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_tuple__30 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__30)) __PYX_ERR(0, 1417, __pyx_L1_error)
+  __pyx_tuple__30 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__30)) __PYX_ERR(0, 1425, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__30);
   __Pyx_GIVEREF(__pyx_tuple__30);
-  __pyx_codeobj__31 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__30, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1417, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__31)) __PYX_ERR(0, 1417, __pyx_L1_error)
+  __pyx_codeobj__31 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__30, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1425, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__31)) __PYX_ERR(0, 1425, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1434
+  /* "gedlibpy.pyx":1442
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_tuple__32 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__32)) __PYX_ERR(0, 1434, __pyx_L1_error)
+  __pyx_tuple__32 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_message); if (unlikely(!__pyx_tuple__32)) __PYX_ERR(0, 1442, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__32);
   __Pyx_GIVEREF(__pyx_tuple__32);
-  __pyx_codeobj__33 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__32, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1434, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__33)) __PYX_ERR(0, 1434, __pyx_L1_error)
+  __pyx_codeobj__33 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__32, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_init_2, 1442, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__33)) __PYX_ERR(0, 1442, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1448
+  /* "gedlibpy.pyx":1456
  * #########################################
  * 
  * def encode_your_map(map_u):             # <<<<<<<<<<<<<<
  * 	"""
  * 		Encodes Python unicode strings in dictionnary `map` to utf-8 byte strings for C++ functions.
  */
-  __pyx_tuple__34 = PyTuple_Pack(4, __pyx_n_s_map_u, __pyx_n_s_res, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__34)) __PYX_ERR(0, 1448, __pyx_L1_error)
+  __pyx_tuple__34 = PyTuple_Pack(4, __pyx_n_s_map_u, __pyx_n_s_res, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__34)) __PYX_ERR(0, 1456, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__34);
   __Pyx_GIVEREF(__pyx_tuple__34);
-  __pyx_codeobj__35 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__34, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_encode_your_map, 1448, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__35)) __PYX_ERR(0, 1448, __pyx_L1_error)
+  __pyx_codeobj__35 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__34, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_encode_your_map, 1456, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__35)) __PYX_ERR(0, 1456, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1466
+  /* "gedlibpy.pyx":1474
  * 
  * 
  * def decode_your_map(map_b):             # <<<<<<<<<<<<<<
  * 	"""
  * 		Decodes utf-8 byte strings in `map` from C++ functions to Python unicode strings.
  */
-  __pyx_tuple__36 = PyTuple_Pack(4, __pyx_n_s_map_b, __pyx_n_s_res, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__36)) __PYX_ERR(0, 1466, __pyx_L1_error)
+  __pyx_tuple__36 = PyTuple_Pack(4, __pyx_n_s_map_b, __pyx_n_s_res, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__36)) __PYX_ERR(0, 1474, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__36);
   __Pyx_GIVEREF(__pyx_tuple__36);
-  __pyx_codeobj__37 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__36, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_decode_your_map, 1466, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__37)) __PYX_ERR(0, 1466, __pyx_L1_error)
+  __pyx_codeobj__37 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__36, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_decode_your_map, 1474, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__37)) __PYX_ERR(0, 1474, __pyx_L1_error)
 
-  /* "gedlibpy.pyx":1484
+  /* "gedlibpy.pyx":1492
  * 
  * 
  * def decode_graph_edges(map_edge_b):             # <<<<<<<<<<<<<<
  * 	"""
  * 	Decode utf-8 byte strings in graph edges `map` from C++ functions to Python unicode strings.
  */
-  __pyx_tuple__38 = PyTuple_Pack(4, __pyx_n_s_map_edge_b, __pyx_n_s_map_edges, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__38)) __PYX_ERR(0, 1484, __pyx_L1_error)
+  __pyx_tuple__38 = PyTuple_Pack(4, __pyx_n_s_map_edge_b, __pyx_n_s_map_edges, __pyx_n_s_key, __pyx_n_s_value); if (unlikely(!__pyx_tuple__38)) __PYX_ERR(0, 1492, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_tuple__38);
   __Pyx_GIVEREF(__pyx_tuple__38);
-  __pyx_codeobj__39 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__38, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_decode_graph_edges, 1484, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__39)) __PYX_ERR(0, 1484, __pyx_L1_error)
+  __pyx_codeobj__39 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__38, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_gedlibpy_pyx, __pyx_n_s_decode_graph_edges, 1492, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__39)) __PYX_ERR(0, 1492, __pyx_L1_error)
   __Pyx_RefNannyFinishContext();
   return 0;
   __pyx_L1_error:;
@@ -21507,15 +21509,15 @@ static int __Pyx_modinit_type_init_code(void) {
   __Pyx_RefNannyDeclarations
   __Pyx_RefNannySetupContext("__Pyx_modinit_type_init_code", 0);
   /*--- Type init code ---*/
-  if (PyType_Ready(&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 180, __pyx_L1_error)
+  if (PyType_Ready(&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 182, __pyx_L1_error)
   #if PY_VERSION_HEX < 0x030800B1
   __pyx_type_8gedlibpy_GEDEnv.tp_print = 0;
   #endif
   if ((CYTHON_USE_TYPE_SLOTS && CYTHON_USE_PYTYPE_LOOKUP) && likely(!__pyx_type_8gedlibpy_GEDEnv.tp_dictoffset && __pyx_type_8gedlibpy_GEDEnv.tp_getattro == PyObject_GenericGetAttr)) {
     __pyx_type_8gedlibpy_GEDEnv.tp_getattro = __Pyx_PyObject_GenericGetAttr;
   }
-  if (PyObject_SetAttr(__pyx_m, __pyx_n_s_GEDEnv, (PyObject *)&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 180, __pyx_L1_error)
-  if (__Pyx_setup_reduce((PyObject*)&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 180, __pyx_L1_error)
+  if (PyObject_SetAttr(__pyx_m, __pyx_n_s_GEDEnv, (PyObject *)&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 182, __pyx_L1_error)
+  if (__Pyx_setup_reduce((PyObject*)&__pyx_type_8gedlibpy_GEDEnv) < 0) __PYX_ERR(0, 182, __pyx_L1_error)
   __pyx_ptype_8gedlibpy_GEDEnv = &__pyx_type_8gedlibpy_GEDEnv;
   __Pyx_RefNannyFinishContext();
   return 0;
@@ -21793,588 +21795,588 @@ if (!__Pyx_RefNanny) {
   if (__Pyx_patch_abc() < 0) __PYX_ERR(0, 1, __pyx_L1_error)
   #endif
 
-  /* "gedlibpy.pyx":115
- * #############################
+  /* "gedlibpy.pyx":116
  * 
+ * import cython
  * import numpy as np             # <<<<<<<<<<<<<<
  * import networkx as nx
  * from gklearn.ged.env import NodeMap
  */
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 115, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 116, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 115, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 116, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":116
- * 
+  /* "gedlibpy.pyx":117
+ * import cython
  * import numpy as np
  * import networkx as nx             # <<<<<<<<<<<<<<
  * from gklearn.ged.env import NodeMap
  * 
  */
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_networkx, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 116, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_networkx, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 117, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_nx, __pyx_t_1) < 0) __PYX_ERR(0, 116, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_nx, __pyx_t_1) < 0) __PYX_ERR(0, 117, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":117
+  /* "gedlibpy.pyx":118
  * import numpy as np
  * import networkx as nx
  * from gklearn.ged.env import NodeMap             # <<<<<<<<<<<<<<
  * 
  * # import librariesImport
  */
-  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 117, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 118, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_INCREF(__pyx_n_s_NodeMap);
   __Pyx_GIVEREF(__pyx_n_s_NodeMap);
   PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_NodeMap);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_gklearn_ged_env, __pyx_t_1, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 117, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_gklearn_ged_env, __pyx_t_1, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 118, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_NodeMap); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 117, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_NodeMap); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 118, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_NodeMap, __pyx_t_1) < 0) __PYX_ERR(0, 117, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_NodeMap, __pyx_t_1) < 0) __PYX_ERR(0, 118, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
 
-  /* "gedlibpy.pyx":120
+  /* "gedlibpy.pyx":121
  * 
  * # import librariesImport
  * from ctypes import *             # <<<<<<<<<<<<<<
  * import os
  * lib1 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/fann/libdoublefann.so')
  */
-  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 120, __pyx_L1_error)
+  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 121, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_INCREF(__pyx_n_s__20);
   __Pyx_GIVEREF(__pyx_n_s__20);
   PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s__20);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_ctypes, __pyx_t_2, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 120, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_ctypes, __pyx_t_2, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 121, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__pyx_import_star(__pyx_t_1) < 0) __PYX_ERR(0, 120, __pyx_L1_error);
+  if (__pyx_import_star(__pyx_t_1) < 0) __PYX_ERR(0, 121, __pyx_L1_error);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":121
+  /* "gedlibpy.pyx":122
  * # import librariesImport
  * from ctypes import *
  * import os             # <<<<<<<<<<<<<<
  * lib1 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/fann/libdoublefann.so')
  * lib2 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/libsvm.3.22/libsvm.so')
  */
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_os, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 121, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_os, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 122, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_os, __pyx_t_1) < 0) __PYX_ERR(0, 121, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_os, __pyx_t_1) < 0) __PYX_ERR(0, 122, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":122
+  /* "gedlibpy.pyx":123
  * from ctypes import *
  * import os
  * lib1 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/fann/libdoublefann.so')             # <<<<<<<<<<<<<<
  * lib2 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/libsvm.3.22/libsvm.so')
  * lib3 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libnomad.so')
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_cdll); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_cdll); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_os); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_os); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_path); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_path); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_dirname); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_dirname); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_os); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_os); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_path); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_path); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_realpath); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_realpath); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_file); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_file); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = PyNumber_Add(__pyx_t_4, __pyx_kp_u_lib_fann_libdoublefann_so); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_5 = PyNumber_Add(__pyx_t_4, __pyx_kp_u_lib_fann_libdoublefann_so); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 122, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib1, __pyx_t_4) < 0) __PYX_ERR(0, 122, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib1, __pyx_t_4) < 0) __PYX_ERR(0, 123, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":123
+  /* "gedlibpy.pyx":124
  * import os
  * lib1 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/fann/libdoublefann.so')
  * lib2 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/libsvm.3.22/libsvm.so')             # <<<<<<<<<<<<<<
  * lib3 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libnomad.so')
  * lib4 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libsgtelib.so')
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_cdll); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_cdll); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_os); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_os); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_path); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_path); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_dirname); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_dirname); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_os); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_os); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_path); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_path); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_realpath); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_realpath); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_file); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_file); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = PyNumber_Add(__pyx_t_1, __pyx_kp_u_lib_libsvm_3_22_libsvm_so); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_3 = PyNumber_Add(__pyx_t_1, __pyx_kp_u_lib_libsvm_3_22_libsvm_so); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 123, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib2, __pyx_t_1) < 0) __PYX_ERR(0, 123, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib2, __pyx_t_1) < 0) __PYX_ERR(0, 124, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":124
+  /* "gedlibpy.pyx":125
  * lib1 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/fann/libdoublefann.so')
  * lib2 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/libsvm.3.22/libsvm.so')
  * lib3 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libnomad.so')             # <<<<<<<<<<<<<<
  * lib4 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libsgtelib.so')
  * 
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_cdll); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_cdll); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_os); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_os); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_path); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_path); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_dirname); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_dirname); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_os); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_os); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_path); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_path); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_realpath); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_realpath); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_file); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_file); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyNumber_Add(__pyx_t_4, __pyx_kp_u_lib_nomad_libnomad_so); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_2 = PyNumber_Add(__pyx_t_4, __pyx_kp_u_lib_nomad_libnomad_so); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 124, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib3, __pyx_t_4) < 0) __PYX_ERR(0, 124, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib3, __pyx_t_4) < 0) __PYX_ERR(0, 125, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":125
+  /* "gedlibpy.pyx":126
  * lib2 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/libsvm.3.22/libsvm.so')
  * lib3 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libnomad.so')
  * lib4 = cdll.LoadLibrary(os.path.dirname(os.path.realpath(__file__)) + '/lib/nomad/libsgtelib.so')             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_cdll); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_cdll); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_LoadLibrary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_os); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_os); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_path); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_path); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_dirname); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_dirname); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_os); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_os); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_path); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_path); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_realpath); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_realpath); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_3);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_file); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_file); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = PyNumber_Add(__pyx_t_1, __pyx_kp_u_lib_nomad_libsgtelib_so); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_5 = PyNumber_Add(__pyx_t_1, __pyx_kp_u_lib_nomad_libsgtelib_so); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 125, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib4, __pyx_t_1) < 0) __PYX_ERR(0, 125, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_lib4, __pyx_t_1) < 0) __PYX_ERR(0, 126, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":128
+  /* "gedlibpy.pyx":129
  * 
  * 
  * def get_edit_cost_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents edit cost functions and returns the result.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_1get_edit_cost_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 128, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_1get_edit_cost_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 129, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_edit_cost_options, __pyx_t_1) < 0) __PYX_ERR(0, 128, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_edit_cost_options, __pyx_t_1) < 0) __PYX_ERR(0, 129, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":142
+  /* "gedlibpy.pyx":143
  * 
  * 
  * def get_method_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents method for edit distance computation between graphs and returns the result.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_3get_method_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 142, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_3get_method_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 143, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_method_options, __pyx_t_1) < 0) __PYX_ERR(0, 142, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_method_options, __pyx_t_1) < 0) __PYX_ERR(0, 143, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":155
+  /* "gedlibpy.pyx":156
  * 
  * 
  * def get_init_options() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Searchs the differents initialization parameters for the environment computation for graphs and returns the result.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_5get_init_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 155, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_5get_init_options, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 156, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_init_options, __pyx_t_1) < 0) __PYX_ERR(0, 155, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_init_options, __pyx_t_1) < 0) __PYX_ERR(0, 156, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":168
+  /* "gedlibpy.pyx":169
  * 
  * 
  * def get_dummy_node() :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Returns the ID of a dummy node.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_7get_dummy_node, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 168, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_7get_dummy_node, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 169, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_dummy_node, __pyx_t_1) < 0) __PYX_ERR(0, 168, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_get_dummy_node, __pyx_t_1) < 0) __PYX_ERR(0, 169, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":485
+  /* "gedlibpy.pyx":493
  * 
  * 
  * 	def set_edit_cost(self, edit_cost, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
  * 		"""
  * 			Sets an edit cost function to the environment, if it exists.
  */
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 485, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 493, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_k__2 = __pyx_t_1;
   __Pyx_GIVEREF(__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":504
+  /* "gedlibpy.pyx":512
  * 
  * 
  * 	def set_personal_edit_cost(self, edit_cost_constant = []) :             # <<<<<<<<<<<<<<
  * 		"""
  * 			Sets an personal edit cost function to the environment.
  */
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 504, __pyx_L1_error)
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 512, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __pyx_k__3 = __pyx_t_1;
   __Pyx_GIVEREF(__pyx_t_1);
   __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1377
+  /* "gedlibpy.pyx":1385
  * #####################################################################
  * 
  * list_of_edit_cost_options = get_edit_cost_options()             # <<<<<<<<<<<<<<
  * list_of_method_options = get_method_options()
  * list_of_init_options = get_init_options()
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_get_edit_cost_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1377, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_get_edit_cost_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1385, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = __Pyx_PyObject_CallNoArg(__pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1377, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_CallNoArg(__pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1385, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_edit_cost_options, __pyx_t_5) < 0) __PYX_ERR(0, 1377, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_edit_cost_options, __pyx_t_5) < 0) __PYX_ERR(0, 1385, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":1378
+  /* "gedlibpy.pyx":1386
  * 
  * list_of_edit_cost_options = get_edit_cost_options()
  * list_of_method_options = get_method_options()             # <<<<<<<<<<<<<<
  * list_of_init_options = get_init_options()
  * 
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_get_method_options); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_get_method_options); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_1 = __Pyx_PyObject_CallNoArg(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1378, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_CallNoArg(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1386, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_method_options, __pyx_t_1) < 0) __PYX_ERR(0, 1378, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_method_options, __pyx_t_1) < 0) __PYX_ERR(0, 1386, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1379
+  /* "gedlibpy.pyx":1387
  * list_of_edit_cost_options = get_edit_cost_options()
  * list_of_method_options = get_method_options()
  * list_of_init_options = get_init_options()             # <<<<<<<<<<<<<<
  * 
  * 
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_get_init_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_get_init_options); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1387, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = __Pyx_PyObject_CallNoArg(__pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_PyObject_CallNoArg(__pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1387, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_init_options, __pyx_t_5) < 0) __PYX_ERR(0, 1379, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_list_of_init_options, __pyx_t_5) < 0) __PYX_ERR(0, 1387, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":1386
+  /* "gedlibpy.pyx":1394
  * #####################
  * 
  * class Error(Exception):             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for error's management. This one is general.
  */
-  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1394, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_INCREF(((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0])));
   __Pyx_GIVEREF(((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0])));
   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0])));
-  __pyx_t_1 = __Pyx_CalculateMetaclass(NULL, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_CalculateMetaclass(NULL, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1394, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_1, __pyx_t_5, __pyx_n_s_Error, __pyx_n_s_Error, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_error_s_management_Th); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_1, __pyx_t_5, __pyx_n_s_Error, __pyx_n_s_Error, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_error_s_management_Th); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1394, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_1, __pyx_n_s_Error, __pyx_t_5, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_1, __pyx_n_s_Error, __pyx_t_5, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1394, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Error, __pyx_t_4) < 0) __PYX_ERR(0, 1386, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Error, __pyx_t_4) < 0) __PYX_ERR(0, 1394, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":1393
+  /* "gedlibpy.pyx":1401
  * 
  * 
  * class EditCostError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Edit Cost Error. Raise an error if an edit cost function doesn't exist in the library (not in list_of_edit_cost_options).
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_Error); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1393, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_Error); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1393, __pyx_L1_error)
+  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_GIVEREF(__pyx_t_5);
   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_5);
   __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_CalculateMetaclass(NULL, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1393, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_CalculateMetaclass(NULL, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_5, __pyx_t_1, __pyx_n_s_EditCostError, __pyx_n_s_EditCostError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Edit_Cost_Error_Raise); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1393, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_5, __pyx_t_1, __pyx_n_s_EditCostError, __pyx_n_s_EditCostError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Edit_Cost_Error_Raise); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
 
-  /* "gedlibpy.pyx":1400
+  /* "gedlibpy.pyx":1408
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_13EditCostError_1__init__, 0, __pyx_n_s_EditCostError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__29)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1400, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_13EditCostError_1__init__, 0, __pyx_n_s_EditCostError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__29)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1408, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1400, __pyx_L1_error)
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1408, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":1393
+  /* "gedlibpy.pyx":1401
  * 
  * 
  * class EditCostError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Edit Cost Error. Raise an error if an edit cost function doesn't exist in the library (not in list_of_edit_cost_options).
  */
-  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_5, __pyx_n_s_EditCostError, __pyx_t_1, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1393, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_5, __pyx_n_s_EditCostError, __pyx_t_1, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_EditCostError, __pyx_t_4) < 0) __PYX_ERR(0, 1393, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_EditCostError, __pyx_t_4) < 0) __PYX_ERR(0, 1401, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1410
+  /* "gedlibpy.pyx":1418
  * 
  * 
  * class MethodError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Method Error. Raise an error if a computation method doesn't exist in the library (not in list_of_method_options).
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Error); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1410, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Error); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1410, __pyx_L1_error)
+  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
   __Pyx_GIVEREF(__pyx_t_1);
   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_1);
   __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_CalculateMetaclass(NULL, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1410, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_CalculateMetaclass(NULL, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_1, __pyx_t_5, __pyx_n_s_MethodError, __pyx_n_s_MethodError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Method_Error_Raise_an); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1410, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_1, __pyx_t_5, __pyx_n_s_MethodError, __pyx_n_s_MethodError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Method_Error_Raise_an); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
 
-  /* "gedlibpy.pyx":1417
+  /* "gedlibpy.pyx":1425
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_11MethodError_1__init__, 0, __pyx_n_s_MethodError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__31)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1417, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_11MethodError_1__init__, 0, __pyx_n_s_MethodError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__31)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1425, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1417, __pyx_L1_error)
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1425, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":1410
+  /* "gedlibpy.pyx":1418
  * 
  * 
  * class MethodError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Method Error. Raise an error if a computation method doesn't exist in the library (not in list_of_method_options).
  */
-  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_1, __pyx_n_s_MethodError, __pyx_t_5, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1410, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_1, __pyx_n_s_MethodError, __pyx_t_5, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_MethodError, __pyx_t_4) < 0) __PYX_ERR(0, 1410, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_MethodError, __pyx_t_4) < 0) __PYX_ERR(0, 1418, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
 
-  /* "gedlibpy.pyx":1427
+  /* "gedlibpy.pyx":1435
  * 
  * 
  * class InitError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Init Error. Raise an error if an init option doesn't exist in the library (not in list_of_init_options).
  */
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_Error); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1427, __pyx_L1_error)
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_Error); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1427, __pyx_L1_error)
+  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
   __Pyx_GIVEREF(__pyx_t_5);
   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_5);
   __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_CalculateMetaclass(NULL, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1427, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_CalculateMetaclass(NULL, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_5, __pyx_t_1, __pyx_n_s_InitError, __pyx_n_s_InitError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Init_Error_Raise_an_e); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1427, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_Py3MetaclassPrepare(__pyx_t_5, __pyx_t_1, __pyx_n_s_InitError, __pyx_n_s_InitError, (PyObject *) NULL, __pyx_n_s_gedlibpy, __pyx_kp_s_Class_for_Init_Error_Raise_an_e); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_2);
 
-  /* "gedlibpy.pyx":1434
+  /* "gedlibpy.pyx":1442
  * 		:type message: string
  * 	"""
  * 	def __init__(self, message):             # <<<<<<<<<<<<<<
  * 		"""
  * 			Inits the error with its message.
  */
-  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_9InitError_1__init__, 0, __pyx_n_s_InitError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__33)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1434, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_CyFunction_NewEx(&__pyx_mdef_8gedlibpy_9InitError_1__init__, 0, __pyx_n_s_InitError___init, NULL, __pyx_n_s_gedlibpy, __pyx_d, ((PyObject *)__pyx_codeobj__33)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1442, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1434, __pyx_L1_error)
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init_2, __pyx_t_4) < 0) __PYX_ERR(0, 1442, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
 
-  /* "gedlibpy.pyx":1427
+  /* "gedlibpy.pyx":1435
  * 
  * 
  * class InitError(Error) :             # <<<<<<<<<<<<<<
  * 	"""
  * 		Class for Init Error. Raise an error if an init option doesn't exist in the library (not in list_of_init_options).
  */
-  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_5, __pyx_n_s_InitError, __pyx_t_1, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1427, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_Py3ClassCreate(__pyx_t_5, __pyx_n_s_InitError, __pyx_t_1, __pyx_t_2, NULL, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_InitError, __pyx_t_4) < 0) __PYX_ERR(0, 1427, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_InitError, __pyx_t_4) < 0) __PYX_ERR(0, 1435, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1448
+  /* "gedlibpy.pyx":1456
  * #########################################
  * 
  * def encode_your_map(map_u):             # <<<<<<<<<<<<<<
  * 	"""
  * 		Encodes Python unicode strings in dictionnary `map` to utf-8 byte strings for C++ functions.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_9encode_your_map, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1448, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_9encode_your_map, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1456, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_encode_your_map, __pyx_t_1) < 0) __PYX_ERR(0, 1448, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_encode_your_map, __pyx_t_1) < 0) __PYX_ERR(0, 1456, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1466
+  /* "gedlibpy.pyx":1474
  * 
  * 
  * def decode_your_map(map_b):             # <<<<<<<<<<<<<<
  * 	"""
  * 		Decodes utf-8 byte strings in `map` from C++ functions to Python unicode strings.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_11decode_your_map, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1466, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_11decode_your_map, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1474, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_decode_your_map, __pyx_t_1) < 0) __PYX_ERR(0, 1466, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_decode_your_map, __pyx_t_1) < 0) __PYX_ERR(0, 1474, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
-  /* "gedlibpy.pyx":1484
+  /* "gedlibpy.pyx":1492
  * 
  * 
  * def decode_graph_edges(map_edge_b):             # <<<<<<<<<<<<<<
  * 	"""
  * 	Decode utf-8 byte strings in graph edges `map` from C++ functions to Python unicode strings.
  */
-  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_13decode_graph_edges, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1484, __pyx_L1_error)
+  __pyx_t_1 = PyCFunction_NewEx(&__pyx_mdef_8gedlibpy_13decode_graph_edges, NULL, __pyx_n_s_gedlibpy); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1492, __pyx_L1_error)
   __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_decode_graph_edges, __pyx_t_1) < 0) __PYX_ERR(0, 1484, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_decode_graph_edges, __pyx_t_1) < 0) __PYX_ERR(0, 1492, __pyx_L1_error)
   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
 
   /* "gedlibpy.pyx":1
diff --git a/gklearn/gedlib/gedlibpy.cpython-36m-x86_64-linux-gnu.so b/gklearn/gedlib/gedlibpy.cpython-36m-x86_64-linux-gnu.so
index 94b9401a92d6d6d3fa7ba03c3b42d0924d074018..9d24796d0afff95aaaad76f57ff8c9a29bc1b9e1 100644
GIT binary patch
delta 118767
zcmV+W{{#S#{`C)#{`C*A*Z~yr1#t7h>-e1rdNCY_<u1&t#(9i`>;Z@D0Re~X0Ro5Y
z0Ry+}0R;TGmb(N95%8-w0000000;m8000000JxU{y95_WsxlM+000000IW9v00000
z2mk;800000aCH;_000002&^{%000002mk;800000m~|8Y000005Ue)<000002mk;8
z00000$hTp;1Z0E((wD|R1yljnmm5F@MFH2BVL$~w0ob>oKm|zw0os@9K?O<y{+BgE
z1x5k@RhNB21w{b@m%~B@NdW|x4MPRg0S1@iLj_U+3zs!S1vLQ=muo}?Rsj*0!bAl-
z3-$s40LKL_gfal4mk31#ISbhX0074Y>4P!=qnAuY1u_d<1polY1+0TI0Hl|RMFm3v
zrI*b`1v~+!mkLG&JOQVdM@9ua0jQURMg=?psh7b<1waeg0{{TW1u23u0J@h9M+HC&
z-2(ss#|7SjG62PwS4Rav3*iF*0LKNBfieKlm!d}nKnvpo0074YNP#i{+n4P}1wae(
z0{{TW1^s|B0OprCNCiL(DFgrj#|5&0G63<HgGdEH3n~Nv0LKMqfHDC5m(EB9N&yC!
z8A%070S}jCNd-dz8keX^1w#QPm+VOeLjf?CFiHhO0Xvs;N(Dm!Mwhrs1w#Q*m;6cv
zLjhcuKuZNV3xNOt0LKNfeKG)NmxN0NHvxf{yh{Z#3xNOt0LKM9eKG)Qm-$NtISY>f
z0074Y_<S+|bC*3#1vdeYmvu}9G7FCY0074YgM2aoc9*hD1vv|x0002T1wVW;0DqVE
zOa(UqoR=<51u_eq0002T1<HFe0D_lmO$9Uosh6Nl1u_e%0002T1z3AB0E?I1O$9&;
zfCT^m#{~_0G60m9C{6`B0f3idP6as&fCT^m#|4ghG60vCr%nYx3y}o?0LKMGdNKf@
zm-0>pIsuWFEl&kG3y}o?0LKNyc`^W_mvK)8KntP;0074Yd3iDbte3n`1v&wum-kNv
zISZl%0074Y`gk$`ua`Se1x^dF1polY1)+E{0JxWwPz5>xu$RwJ1u_e;1polY1u1wk
z0K1nAQ3Xy5!UX^T#|6-LG62MvWl;q>0m7G{Q3Wy!!UX^T#|2n-G62Vy-BATj3)BSw
z0LKOYb}|6amo`!bIsw#|aZ&{`3)BSw0LKM~b}|6dm#$I;E(^8*0074YPIfW?2$$zl
z1#b(s0002T1=Mvi01vleQw6{RDjrk-05RK0i3I|4G63n_H2?rejRg_X4gkjm_;NA;
zwg3PC#|7$gG5`>lOI8I=0VkK5Rs~7{DVOC|1x^7hmo!%eN&zmHe^&)g0Wp`(R|QG|
zGnW}y1x^7rmuOf8OaVBTwO9pA0XmlgSp`f1JeN;d1xx`xmz-GzN&!HZ=UD|x4ntG`
z07!`i^lmZ$=}rQdL01J6mr_~<A`5f?0074Ywr(;2Eth{<1v(3K0002T1#xaN056xs
zS_Mf9bpQYW#|0g3G5|T37+VE83v~bh0LKN`Z888mmt9*0MGJNS0074Yh;1?eMVGBx
z1xO2a0002T1vPCl07{lz1-A=!0002T1#@gN08N*|Tm?G|b^rhX#|15HG5}AP2weqI
z3wZzl0LKN;Ycc>@mt|cANeg)Z0074Yd}}fQU$?nk1t9|qc>n+a#|2PmG5}(i<zNMW
z3wi(m0LKN!XEFe6mv3POItzLL0074YfoC!RZkM!S1&j-O0002T1?pxp0C<;6Vg)!0
zdjJ3c#|5WmG5~s)j$#Fi3wr<n0LKLpW-<VMmn35aI176K0074Y&Sf$Hf0t)t1xX8g
z0002T1$t#N0D_mhV+D%~djJ3c#|7kMG602_QDg->3wr<n0LKNMWHJDUmzQJ(MGJcX
z0074YPGm9wikIhP1&RxM0002T1-WA~0F9S?Wd$?~djJ3c#|3(0G60X4yJZDZ3wr<n
z0LKL#V=@4fml$RRR113m0074YzG5-}mzQ^D1w#va0002T1#n_A0GgM@W(8RbdjJ3c
z#{~>xG60^JDQ5+V3w!_o0LKM*VKM-ym%V2NHw%0K0074YHeoUVs+a#~1&IrM0002T
z1)^Xw0IipkXazM3d;kCd#|2_wG61ia(`W@p3w!_o0LKLoU@`!*mnCThhzooG0074Y
zeP1#FwU@YQ1vU$O0002T1w3Ce0JoR<X$3<Id;kCd#|7?QG61@lM`{Iy3w!_o0LKMh
zUNQi^m)B|qFAIDC0074YBwjKAzn2kf1yBoo0002T1<7490K=DTYXw#dd;kCd#|3Cz
zG62Vy&1(fi3w!_o0LKLyT`~a5mm6#aSqpps0074Ywp=m*&X<2|1&j-Q0002T1s_~8
z0N9rnZ3Q?BeE<Le#|7M5G633_S#1T53w;0p0LKMGTQUIMm-1}|H4A+J0074Y16wix
z-<L9O1xgEj0002T1+H2$0OFU2ZUu-7eE<Le#|0Z&G63b58gB(Y3w;0p0LKN=Suz0U
zmt$`QNeg`d0074YepxaA>X)}~1&j-Q0002T1?N~Y0PUAfa0N9BeE<Le#|5WYG63(F
zj&KE23w;0p0LKMMSTX?fm+Wu_R||ar0074Y=T|ZS_m@s_1x5>f0002T1(sJb0Q#4o
zaRpineE<Le#|1H0G64RU1abw73w{6q0LKNJRx$txmzr_~I17FN0074YSXMFs3YXn-
z1&0fM0002T1;$k}01cOKa|JjHegFUf#|48`G5`;kvU3GS3w{6q0LKL~RWbk)mjiSK
zg$sTF0074YqEs>f6_=BA1vMCc0002T1!7b(02fvO07!`iQ&chl>3SBI!F&Z1m-BQ5
zB@2E40074Y5>zq(8kZw=1&9lN0002T1%FdA03DaMbp<pFegFUf#|1!BG5{Z!^>qbM
z3w{6q0LKO8QZfJ|mr!;ER11Cp0074YhEg&BCzr2w1w{*f0002T1vXML04kUNb_H4s
zegFUf#|6<*G5{`@WOoHn3xNOt0LKM(Q8EBHm%n!fItzgS0074YF;OxAI+p}^1yBo*
z0002T1=dh907RExcm+BOj{pDw#|4p4G5|)Gq<95T3!DG|0LKM3P%;2em;ZPLIt!ct
z0074Y@lP@UQkO(|1yBpA0002T1(r`T09==!c?CKPsQ>@~#|2SOG5}te<aq@`3+({_
z0LKLcPci^#moa(;Gz$O%0074Y#ZEE+b(e5@1%3-@1polY1w2kN0D_nKdId2HX$1fP
z#{~dRG5~~^HG2h33uy%a0LKNWO)>z7myLS`hzn{30074Y6-_b#k(VHR1uhF}1polY
z1>8(B0F#$ld<94gY6SoQ#|4Q@G60sBtb7HG3u^@c0LKOLOELhUmqUF8F$-%20074Y
zwo5VqqL+Vt1x*WU1polY1zt-s0Hv4LeFZ@aas>bY#|08gG61iaBYp)z3vvYj0LKN!
zN-_Ynmv4RrNeglX0074Ya!N7)wwJ?x1wIRM1polY1u0500JxV7e+4)Tas>bY#|7p|
zG61@lPk#lq3wZ?q0LKL;NiqP(mk)piFbjDF0074Y>qs&H$(Kri1vCqJ1polY1*}Li
z0M3_*fCWzrc?AFf#|2VIG62w*<A4QN3xEXx0LKOIM=}7~mqvjFSPOs!0074Yj7Ks6
z+?T0=1xgEf1polY1vf`B0Optefdx1Vc?AFf#|898G63k8L4pOe3yK8*0LKM2Mlt~P
zmV*T{3yK8*0LKORMKS>Qmpy|8Itz*g0074Yvqdrh{Fj1*1xX8v1polY1z$xn0RETO
zg9TU%kp%z%#|8XEG5`pdIfMmK3y}o?0LKNOL^1#hmz9JCN(+hw0074YOGGjN6qo9R
z1vm?e1polY1qehk02Y@mg$1<>n*{&>#|213G5{l&?S%y}3!4Q10LKLlLoxs*mnen>
zGz*&r0074Y&O$N(Dwk)51y2i`1polY1$06(04$fmh6Pv)q6Gi|#|05WG5|K0B!>l2
z3!()80LKNjK{5b1mw<-_N(-9>0074YU_mkfLYLKt1vm?v1polY1sy>$07RD<hy{5I
zumu1B#|4={G5}GRoQMTF3$O(M0LKMXKr#SRm*I#7Pz%BZ0074Y`9CrMU6(wG1v(4D
z1polY1++gh0AH7Zi3Ly#)CB+l#|2qGG5~3p+=&G`3)BSw0LKLsKQaJomm!J;N&$11
zY>EX$0d=>riUkHF2GB_W0KU+dy{-jD0n(QQuLVZ|)R#=J1waAPw~em_C=D!t1polY
z1$7KE0L3`~0KUaJ002md1z`*_09aZ`jSX{a4gg4vJwcfc0J?ye;k*U@8ixe{0KVQi
z0074YN((XohXnut#|1nKG63F}`Mm{45P$^$0LKOD3Niq=IRF5@xR+4A1yvb<1polY
z1%(PS0JJ#(09i<h1$7EC0A9Ydx4ymwhY2i^1polY1?2!T04_QJ0KP6d002md1<?R9
z09aZ`jSb^p4gg4vJ?UQ#0J@QvamWS!8khwD0KP{$0074Yy8kf%m<0d;#|5naF#t!G
ziOB^=5RnA{0LKMd|1kg|IsgE^A(!CE1yvc51polY1@-<h03A9209i<h1?B!R0A9Ww
zmrlwB_bj3X0074Y<^3@LmpT9dzLz=x07!`i(fu(1SXxMp4dYx607#8J2VD*Tx}ukH
z%LV=#sRaN4zOXs~0LKNo{4oHj1polY1+4rr0I-*d%mqgfq6Gi|#|2vaF#w7>006#<
zm*C6=RT-iM0074Y_4_dZhB^QMSxAWm<@+%JUcQF6PR#|z1{E_&0074Y8uc*%G5`Po
z#{~@aF#t1{Dbxid3+GAz0LKO0^f3VEmsr#VTop4(0074Ydh{^>G5`Po#|3QkF#t1{
z$<zfd0qK|c)CFV#>z7^C1xXyfN&o=I1&H%80LlOW07!`qb@MR*#|3lqF#x`|!qo-(
z0|Atmz1anN0hX6I+67+*j!XamzK)lj+65c|l$W~N1;PPdmwnp>dI4aU_}c|v23Sl0
z0KQn4UEBp60bZAS+y%k`7MDHU1$qG(m$=;pUj`6L006!ammS^(903-WI^G520jZZB
z-vxmItCyhP1y=^6O8@}AqnG{P1snmXw;JFDf&>+sNdN%H1(M`301p5F0LKM{<S_u6
zm$Bpp{|0zD006#tm$u~vfd+Ir006#pw>{<s69EQ8IRF5@Lzh<P1%U=XIRF5@KeyfI
z1zH3w_5%O_#|87;F#xD4006$IDgXdTi3Q)>F#uRvNR18gI}QLyjXg(i4gk9Lw`1%D
ztqltf1ONcX1scyW0O7YA`UTno3l0PT0LKM2%rOAami+~83lanX0LKNq%P{~emwEjK
zG!hO30074YeakTb-YNh9NQnh*mw^2R7Pr;?1wsK3_5%O_#|0D0F#wM$006#^mnZ)P
zDGL<@0074Y+Q~5hdY4=O1#1f&1ONcX1!T!F0QHyA{{<@x5(EGM#|19QF#t1_2L>h!
z5d;7L#|8b!F#r#hI0hvP5(EGM#|6#EF#sBsXa-da69fPN#|3T3F#uAP$Oa_~69fPN
z#|1seF#uqd_y$c25(EGM#|7obF#tK2Pyq%t3l9VU0LKNT$1woxmyiJlB?}Y;0074Y
zb;mIPZI{3S1|<s;1ONcX1x3d(07RGY0R}Y-4+H=J#{~t)F#!6PFaic03l#(a0LKO0
z#xVeRmskP@C<_$?0074Ys>U$@eV2;@2CWMf1ONcX1@gr)0DzZ60|qh+7X$zR#|5>;
zF#wF0fCC043l{_c0LKM|#W4Vym#_l{atj;;0074Y2*oh~=a(%61~3a41ONcX1<k}U
z0H&8{1O_Dw83X_T#|53lF#xfbm;?qT3mF6e0LKMw#4!NBm&gPLCkq+`0074YJH#;n
z%a{5D1~LmA1ONcX1^&Y^0M?f`1qO2qBm@8e#|41HF#yk(wFL%03mpUi0LKM3!!ZC3
zm;VI@9}67>0074Y48t)19G58u215%Z1ONcX1-`;D0KS)Z1_mJuAOrva#|4tYF#sr+
zqXq^#3>*Xi0LKMQ!Z82<ESKp920sfR1ONcX1p>k`04kR=2L@Xi_5%O_#|50hF#vxm
z003D?i3N<oF#uk^f0w5R1}Y061ONcX1zEu{04$f>2L><;AOrva#|0k2F#t4|7YGJ8
z3n2sm0LKN}z%c-6w^|4WB>@=r0{{TW1*^X?0C_3^0AEOn1)IMy0KR#bmk9<t3nT;p
z0LKMbzcB#am)!{lPZ%Kt002cX(nyH~_P#LyNs9)@1@XQy0Be^-3I;t3Bm@8e#|5dr
zF#zb7j0y%H3n2sm0LKM?zA*r9m$nK9DFGpu;0gv70dbf23I<X|_5%O_#|7cNF#vNa
z002Q*NQni|y)gj3b1DD;L0akO;Q#+fjRn{o4ggGxY$Om!i)<tq#|5>$F#sY20074Y
zr@b)%eV2|41||z41ONcX1$MnL0DqUi3kF6DA_M>c#|0(5F#v&=5DW%*G$I550LKNQ
zyfFZYEC2vV$3!F$00000>E-SJ|45Ampc@VV#|4MHF#sF{0074YdAu<I@0Yv`29XOS
z1ONcX1>L(b0N0mT4F)C)AOrva#|5mrF#tW6i46uZ3m^ml0LKMzyD<Pem&FYRN(&+c
z0074Y8@n+8l9w6|1~o4t1ONcX1>3qY0FW#I07!%A0qNK4|Nlsh1>PAB0LKNzx-kGM
z1ONcX1-H5}02P;h4hA0!DFgrj#|4eLF#rmes161%3m^ml0LKMWx-kG*m*NfvJQ5-V
z002md1qZq@0LKLdx-kH*moN_o@DP#%0074Y3b`==BP0L-z9W|~5C$6olDA9{2IvY4
zsRIB2#|8GSF#uP$KN|+l0Sm1I0074Y9jq|`*S8oQ208)@tOEc5#|7A_F#wjAT^|N8
z6{!ON0LKNLsWAXXBme-uMkD|Lx~Z3<9|mg+sRIB2#|0g!F#uGz7$63H0uYh|0074Y
zm8LNO3nTylz6+P7A_gD}tpfl6#|34kF#z3{&>{v#3#|hH0LKLsrZE7|mmwntBnzwq
z0074Y<)twIo|jM~246Xn0{{TW1$(7200ATb09i<h1#6`-0A9WUBme;EZ0!I4NR0(_
z1P%bl1zV*t0Jj4G0LKMQr7-}#m*^t~EfuT-0074Y6{Rr%g(Ls~zJ(+J0J^M~DkKJL
zA*}-d0LKNDq%i=#Bme+0?L{&FzP=;?07!`ie55e|x~-SCBnC$dtOEc5#|15<F#x2O
z2qgx93$g<M0LKNQqcH$Hmy;z1Di^N<0074YZKE*&_ap!SG3~zhBme-qub0#%1|bWq
z0{{TW1vR5F0I!xN1}zn;0{{TW1^uEi0AeHn0KQ@*006qGmqI26n+vZ40074YO`<UX
z1()b11|<us0{{TW1s$R>0C1NWCk7=Kumb=9#|7=7F#r)I001%Vz7Zt=0J^Z3R3`=@
z3#|hH0LKNjp)mmHmw+b*Eeo&%0074YeW5V`AD6f%2BjCW0{{TW1<jx_04pT`05R>p
zD<uE`y0VvaC<b&3w*vqG#|1*5F#x2O^C$*77qbHZ0LKOWpD_SqB>(_1?Y?6r006qP
zmqaNBAq%qu0074Y#h)<%ahGr@23`xe0{{TW1zevo0Nj_^DF!SHvI77B#|0&yF#t-J
z5Gn>G3$_CQ0LKOOo-qK2mq02891F4o0074Y&Ym#<RhMTf1~n$P0{{TW1&*FE0FxyE
z07#2$Bp6JKd?XOQlO+HENQnh-o-qKrx0k{y24D-h0{{TW1qhxo02P-lD+V+sxB~zH
z#|6oqF#yXY002mfY$O;=i+m&yzRM*507!`itDP|by118zD+VYFw*vqG#|3+xF#xic
zx+?}r8o2`i0LKL<oiP9eCIA3Pi)<tqz62%!0J^!C9V`Ye3%COS0LKO2oG}3EmsTtW
zUl@`D0074Yb(}E(`y&7VUr31sWt=epzWbNWECwbEy8{3K#|1Z>F#tZ7{wxM23%COS
z0LKLeoG}3PmoF^_Vmp!p0074YnVT^H^&<cPL0L$N1&Et50KWAj002Q+>4@V0|45Am
zkopY(#|3wrF#x#(0074YX`3+sBbUrA1|bW%0{{TW1wETF04|sKEe2f~k^=w$#|71z
zF#zu)002Q-NQni+nlS*r@0V^a1{Mpl0{{TW1*V!Y07RFME(R(Kw*vqG#|3ekF#xQW
z#4ZLY3%COS0LKMAnlS+6m-sFQDGRv+0074Y37Rng9k(nm1`PoaBn1Ef#|812F#xSK
z006$Nmq{=N8v!J@WiSTN3kyI60074Y^^7q9c(*`21|<UvKm`B*#|6cSF#vj(a6bku
z3qS<`0LKN5i7^0tm#9AmEek*e0074YRf#bGewW}s2F(jU1polY1$~Gy07RF#Kn7I{
zJ_P^(#|0gTF#sr+7(oUj3qS<`0LKOIhcN)6w?;t*4*?541polY1=WTz05+FkLIyYv
zKm`B*O^NQu1&@X?0I-*)LI!FJJOuy%#|0XOF#z?K8$$+y3p@n?0LKN5g)so+m#9Mq
zMGHIy0074YJ%up<1D9Yz1{9YJL<TGiKm`B*#|83)F#wR4LqrBD3p@n?0LKNrgfRdP
zmtaB$6qk%d1}h6d1polY1#^Tk0GgM=L<UF@Bn1Ef#|0pSF#wn~006$2w;DwT@&OA%
z1polY1s;Mi0Q{F1M+UDGLInT-NQnhsfiVEb1zmwL0RA?YF3=Acmw!A44wwH&21^+v
z1polY1>k@&0EjgJ09i<h1<-&o0A9X`mt#l<B@;pg002md1)YE~0LKNJfH437H<z18
z215%$1polY1x$c30LYi=NCqMcKLr2)#|0mNF#t-J7D)!y3ql0|0LKMBe=z_Zm-k5q
zG%`X3002md1^Rw50LKORelY+8Hvj<XSpNV2NR0*m-3<W81?GM+06YZ%0LKN`elY<3
zmt9H*9}7PP0074Yu6{89OqYjB26_%c1poj=G5^N}CVnvh1-B1N1}p&!LInT-#|7_v
zF#rgcMob1;0YaCdOa>U2F--;;m*7AK1eexK20|P{1poj|iRVa(1qysI08EQyBv8i%
z1AH+63^xFmAV&rkmr6|r)C)od0074Y40|yEFPAA!21E`*1poj=G4{s=x_U7HI+uG+
z24xIF1polY1xk7`0Ae?na6ASSmQMyk4nhS007WtX#|7niF#vOyP)`PE3qJ(_0LKM#
zc`*QJmq0uQ6_?mg21pA+1polY1qyjF08W=HPzG}gLInT-#|4adF#so*sZa(;D?$YT
z0LKM6crgH3Hvj<XD*ylgNR0(R*9`#21t)kh06_%+0LKLxcrgIbmmE<BVGBV80074Y
zt#>g1&X<T$28s(o1polY1sHcR0PMFNQU+xT3-AO00LKNWY%u_kw~br|_5v)}1ONcX
z1*d5-075eW0KP&q002md1(Rtp09aZ`jSZ#94FE`uJweJ10J_+>uU`g40SoU00074Y
zH)t^caF_mI26_wc1ONcX1+r%`0CJauVFt1b?*sq<#|8CfF#v&=Kw<_>3-AO00LKNL
zW-$PcmzZJ(RSWq90074YJ!UZg_m}u$2DJ<F1ONcX1$|{P0GgM$V+JJ)@&o_?#|2Gg
zF#xES=wk*o3-bg30LKLlWibG?mndWgH4F0u0074Y&15kE!<T4e1~m)%1ONcX1&w4e
z05q4VWCk@0^aKC^#|2GfF#yn)=wt?13;F~A0LKODV=(|WmrG>^G7Izs0074Ytz$6&
z-j|4F1}h8t1ONcX1$SdH06UkyWd<(`_XGd{#|1rOF#zkA_+<u53;F~A0LKOAVle<e
zmriB|Gz<3x0074YsA4ey?U#*a24f5Q1ONcX1v+9e07jSlW(GJ5`UC&~#|8IcF#tlB
zK4%6s3;F~A0LKNlVKD$omwsmkGz<9z0074YcwsRB|ChaI244`^1ONcX1rA{`04y^A
z0KP1jENBMl3;qNE0LKLtU@-ubmmp~dgA4rx0074YhhH%OVwbLI2HHmc1ONa?i)<t~
zMKS+Ki3OEjF#yK}lU^|ZiZlQKMKShFiPJ?f!08+E|Nlsh1(CfC07;47#|3p>F#!1l
z0074YXI?P?2A9ri1|SRh1ONcX1w39c00@`)Y6dJ4{{#R4NQngiUNHd21^!(z0Lhm(
zYX&I`{R992#|6q=F#vFvYikBh3;qNE0LKM!T`>TFx5R4(>;Vh@1ONcX1$tXC0D_mh
zZ3ZtA{{#R4NQnhITQLB~1vy(W0Lz#BZ3aOS{{#R4NQnjLS}_2}1?E~Y0Na;OZU#*Y
z{{#R4#|4gBF#y$<r)~x^68{7M07!`iOj<Dj#|27SF#y<?>uv@&3;zTF0LKLgS}_3L
zw=Hi52LTKD1ONcX1@>4m03??`a0X=({saI3NQni3STO*{1%FsE0I-*~a0Z|Y{saI3
z#|6|^F#x!?VQ~f^0vi4V002mfY$P~EG4@D_1)Nne0LKNIRWSglx0`bYO#v*v0{{TW
z1wvFY0L>-<0KUy8002md1uj%E09aZ`jSWDv4FE`uJ!a<(0J^@n6m<qA6b{=1002#i
z&&LJsKQRD|mq&>Ps0-Eu0074YJ3lc13Ag%*23rCP+XDaq#|70qF#vP7V2lPU0uI{)
z002#i&&LIvIxzr}mzj<RR}R|)002#i{l^72Ixzs9m;R0hFALTK0074Y{W&oJ_?I}3
z22~5z0{{TW1)VuD00x(sj|Qj^z5@UN#|7#*F#xqD006$Vmr#%fQ487w0074Yi8wI;
zNSCaT29+7U0{{TW1?e|20IMbd09i<h1=%+-0A9YUms*hq8w=S30074YwKp*UH<y5s
z1|1960{{TW1&uc`02G&~kp?IZ+XDaqO^MIP1y?sQ0ECy`kp?dd*8>0m#|0fXF#sZ$
z7?K9MJiY?}0LKMbHZcIDCIA3mNQnhVHZcIcr6vFXL0IV!;{X3hjRjDZ4FJalH#RW<
z+yejr#|0@iF#xqE0GC0J1{RkjlLjyg+yejr#|73kF#yB2Uy}ws0Ytt70074Yjx;d<
zpC$kRL0L$N1${Iz0KT6l002Q+=|$lG|45Amg_8{c#|3OOF#z2I0074YUNkWP&?f)@
zNQnhcG%=T9lLi)-_LK%DGTs9K0LKLmG%*0%CjbCQi3R^NF#zkXNR0&%lMMh&iS5S)
z^)oR5-U9#t#|7szF#z3{PL&2y3*G|&0LKN2Gcf?*m#dWqggw3k0074Y`!X>AnI-@L
zL0d?P1?n;}0KS<f003L*v)=#zNR0*akqrP$iQUHq*D^5x-U9#t#|6nUF#zM2Y?cO0
z4cY?$0LKM#GBE%~Cjgg0j|LW(+m;403*Z9)0LKL#GBE(}ml&4@91GwB0074Y_AxO4
z{g*$N21N_u0{{TW1*S1E0120mmj+`C<pTf!#|1bsF#t-J{g(zUGU5XO0LKLYF);uX
zC;$LRi3Ra6F#zeE{Qv(*jTHcp4FJal=P)q<;sXEx#|7CiF#sZ$T$lzN3*!R-0LKNg
zFfjlwmxGuFHVfJV0074YaWF9eW0%C325AfB0{{TW1^q8E06CX9nFc-!<^uo##|5-6
zF#vIwftdyz3+4j=0LKN5FEIdrm#CQr91G|J0074YW-l=Si<i%t1{@3J0{{TW1wSt_
z0ArW-nFc2d<pTf!#{~^9F#uSXD4GUP4d?>^0LKNgE-?U&C;*p1j|LQ%nVJTr4eA2`
z0LKOFEinMEC;*p1j|LQ%Uz-LR4C(^_0LKNfEinMCD3@T91{s&0n+6;U=>q@&#|3aL
zF#w*I#hV614Cw;^0LKL(EinL`D3@T91{t?5oCXvD3+e*^0LKOBD=`4Mmrb1pWe)2D
z002yh%f|(ED=`4@m%yC{BOL1k002yhbR-Z+i3L3?F#t@9bR;0h1vV=&0PeT{od#q9
z3+e*^0LKN#DKP-Cmu{a1oGb1F0074YASp2bR4D)e=}_wb|45AmCx{II#{~{4F#zQQ
z0074Y04Xs5QkONL27?Rh0{{TW1#&1c0P~l_pavrW?3d@D1|0zcmj$5)P7dn>002yh
z%f|)OCour_mtmm>Lk{i(002yh%f|(SCouqRm$IP-NDJ-*0074YFefnpVwVM?23ij8
z0{{R_iOa_Y$R;rWLziu$2GkJj0{{R_i*zI)#|7jiF#raaQKJSo3+w{`0LKNQB{2X5
zmy@FgBOL4l002yhbR-Z+i3MsUF#t@9bR;0h1!E;K01lVZqXvKr@B;t<#|79VF#yPy
zU8Dv{3-1E}0LKM{BryQ2m$9S<N(=7;0074YFC;Mls+R_(27L?f0{{TW1*ao10F0N8
zr3P6J?*jk;Oo_|K1wbP)0DG79r3OR}@B;t<Oo_|K1@0m-0JoP%rUpd|@B;t<#|54u
zF#yn)m!<|q3-1E}0LKMRA~67zm*=JiMGNl(0074Y{vj~{rI$CS21N_+0{{TW1+F16
z0GgMFrv^?7>H`1(#|2d(F#z6|;HL&F3+e*^0LKL%Au#~Lmlmi7K?~{w0074Y(;zVb
zz?WjE23QO50{{TW1#loS0O*&+s0K<4@B;t<#|0Z8F#zV58mR_D4)6m208EL?#|6qC
zF#zG0YpDiE3-AK~0LKM)A29&dm%gb684L0Q0074YQy(z^2bbch1}h8m0{{TW1sxwT
z027xOss>jJ?gIb-#|66{F#uebda4FF9P9%C08EK=BoIi61#KQN08EQ?Bp}BHW*#vB
zGndb*29y)*0{{R_iOWojbR;0h1-~6J02h~bs|G+5>;nJ*MKR<|i*zI)#|3O1F#r{}
z$*Ts$3JZz@0074Yof0tsJhzy(25kciiUR-u#|0%1F#sW#5V{5}3ycE*0LKOG4lw|K
zmq@w>RSS#*0074YjSevYftRSd1~?0h0{{TW1xgMv0E3t7x&|@}iUR-u#{~@zF#ryi
zD7yww3ycE*0LKNg4KV<PmxH?oB@2oJ0074Yf(<bM5SO#N2D%H20{{TW1-%S0085v6
zyaq7~i~|4w#|425F#xHTw7do>3yK2(0LKMQ3^4#Pm*~6(I1Y>h002#i^~VJV3^4$=
zmoB{qeie)Z002pe$4HF@k!lS9#|3l?F#wmB!Mz3{6^sJ_07;3*NR0(sY7GF#1wIQg
z0HK%ny#^u+iUR-u#|04!F#sx;5V{5xw>iEB_5lly0{{TW1^Wmw00Wmgzy_=mjspMy
zNQnhM2r&T11w9Bc00ozD!v+|a5Wxl;3ylK+0LKOY2QdK8mo~u$EfS6c002md1;Ph0
z0LKNu2QdH#mvF-d7?%*i1__sz!3HJ^ivs`v#|3T&F#u7Q$H4}@3yuQ-0LKNU1~C96
zmyf~*GBu6^002md1!M*>0LKMl1~C8%BLD#Do&Nv-NR0()WeosHiQ2~nO$IRlivs`v
z#|1wIF#uDS_QD2e5{?4^07!`i#|1F}#|6a&F#rsgaKi?x3yuQ-0LKLn1u+0smnOsp
zNDGbw0074YyaX`-50`nw22Bf&0{{TW1!M#<01=nb#0EqQjspMy#|0GxF#r>nAjJmh
z3yuQ-0LKL(12F(LmlVbZKMRfn0074Y*8(vBL6=^}21W~x0{{TW1%?7K0DYIQ#s+E)
zjspMyG400%4+1d&X_qF)23iY?0{{TW1*!ot0B@HNx&{@OpvML{5{?4^07!`iO#v|g
z#|29PF#uqf>c<8v3yuQ-0LKLt0WknBmmtUnBnyoL0074Y<^VAO*OyPo23re_0{{TW
z1%LoC0M(bZ$Oc9WjRODx#|19{F#z+o2FV5z0W5q20074Y68|s&<RJh6zT_bQ07!`i
z{{AokSXxMp4G>-p07#8J$zKfsx_q}p$_AbR5PSmw0LKML{V)L1Apiir(wFkf22~k+
z0{{TW1>^iM0LUQ#09i<h1=IX60A9Yxw_wZ$O#=(i1polY1%3H20MM7X&j#2F&jkPg
z#|4`BFaWcco6rWr3(y4s0LKN*_b>p$muS%jB@5040074Yo%b*RnwOZ-1}zKF1polY
z1!eaz0MnPy(FRlt&jkPg#{~xWFaW8SFVY4k3(y4s0LKN@_Amh9w_wr+odFBa1polY
z1ugY30IQb>)CMIB&;<Yh#|8cLFaYGYIMfD31q;3f0074Yt?w`ZE0>7b2G|R|1polY
z1<LL)01lUH+6KZ4z6AgP#|8cEFaRBwINJs#3%msY0LKN*?JxlKmuTAtEepN{0074Y
zmF+M9E|;L&22=~Z1polY1vc$400fu++Xf~Jz6AgP#{~uLFaSKaFx&>60Smna0074Y
zUF$Fa2A9~~1|<u=1polY1ug3^06(_~-UbZ}3+Dv@0LKLt-!K4Iw;<{U<pB%m1polY
z1tr}u0A`mE>;}aP=mh`(#|2f~FaUIy;Oqt+3+M#^0LKL_+%N!lmk8|!B@5;S0074Y
z{o61AK9@M{22l&=1polY1)<w808O`)?FPvK3+Dv@0LKN**)RZ4muT+>hYRNg0074Y
zJJ~P*Pq+H-20johU;_XE#|84pFaQ)F006!eAOHYJi3Q%sFaTIuNR18fCk+5djXh3W
z4FI}ewPXha0t<En0074Y!^SWGx|ebS2PO-40{{TW1(L=v0HwF10td|j3w8ql0LKNo
z#4rG-mwN;UVheNw0074YO~fz&Oqb{c2lxwi0{{TW1xmv(0I8Sj1qWgabOQhY#|7QO
zFaSH3SOy0>3w8ql0LKNF!Y}~Lm!1X(NDFoY0074YM8Yrt+PCor2Sfo7U;_XE#|8Dk
zFaZ7^006%JmqiE%NDFiW0074Yoxm^vT9=pz2PF%10{{TW1#Q4E0A-iR2nSgUbprqZ
z#{~(%FaUbDEC~lK0Sk5m0074Y(Y`PM#Fu0W2igmC0{{TW1?;^r092Pr3kN9+cmn_c
z#|5~(FaRT$eG3Oz3wQ$n0LKMYy)XbJm*5KrCkuE30074YBfT&HCzlco2Tl@r0{{R>
zi3P>HFaXB|!@Mv6DVK5#2V5Co0{{TW1zfx^0Q4UK09i<h1x&m!0A9ZIm*)%zaTs6&
z0074Y(z`GK?jHaEUr31s!n-g4zV4TB4F@F}cmn_cNQvo4i(DiaNQniAyD$L91%|sY
z08W>$4F@O-cmn_c#|2ZnFaTng;tdCAJ75C<0LKO3x-bCf9{>PBSxAWm&$=)GzUUtS
z06|*m=kfplNR0*884Uo(1-`m40CfWZ0LKNex-bBEmxT@oaSL|?0074YG`cVV_?H0>
z2U`nu0{{TW1<tuJ0F{?#4+lLOcmn_cNQ+z~7)XifNQnh`xiA371$VhH06Lex4+kh3
zcmn_cNQ+z~7)XifNQnhExiA371vR-a07aG%2O<l10{{TW1qit?0QQ$H5C=#LcLM+b
z#|5>xFaQ&`fDi|@2^F#c0074Y^{y}gxeEXQzPSqk0J^f5N*V_=3$g$J0LKNat}p<%
zw}~1D`7R5L0002T1*m^80KJ!uNe5^Pi~s-t#|1TiFaW=nN(VhFi~s-t#|7<vFaWL!
z006$O3IG7Qi~s-tNQn*9elP&(S^xk4zP$<n0J@BqXi5hs3yc5&0LKNJelP&Ex0*@^
zhy)9n0002T1qgaD02Y@mPX}lVnE(I)#|4#nFaQ{rpic)qE13WQ0LKMQc`yJ63jhGV
z2MYiIx|sj~07!`qG<h%p=~@5(|GpLr006p~mk3Y?CkvSX0074Y`*<(_4YxW_2Z#g<
zrT_o{#|3C~FaU3t%~S_y3#I@70LKOOb1(pLmq1kqJu9XF0074Yt#dE{U<&{MzF-Rg
z0J^3C002md4U}^*0O?u(|Np*k3jhGRrkAKy2PX@r0002T1zU440A`okRR`S*)Bpeg
z#|3k8FaT$l!d3^%75xAJ0LKNLaWDX%4FCYXpA7&2y8V};R|h!@{Qv*}#|2MuFaVvm
z=2r(C0u>km0074YC~q(TA`Somz9J3)0J<2L7g`4;6*EZy0LKOLZZH5c0002T1>|lp
z05g|SS_d2z7y$qP#|6D^FaRkI006!z4gdhU7?*-t2lxvZ0RRBU1+Z-}04KMFTL<<5
z6(<1z0LKNdYcK$44gdhYXbu1Xx+j;AT?ZW%CjkHe#|3?BFaU23006#k4gdhUCzrxq
z2U-g!0RRBU1rBR40Bx5iUI*3{LjnK*#|15FFaU-Z006#*7XSddLzfa?2Q&*q0ssKV
z1>|Wk0ECxOUk98OHv|9x#|1rUFaT>U006#gEdT(zH<ton2Q&*e1ONcX1@vez0BM&&
zU<aHPPXqt}#|2GjFaQlM006!XE&u?!PnYsw2Q&*$1ONcX1q5g?01LM<VF!Ex6>tOq
z0LKM=W-tJ+E&u?&uPy)px^S1mVh1z}a0CDV#|1QIFaWK$0b>UTLJPkD0074YA|x;X
z`?nLN2U-FUc>(|c#|6D1FaS9j006!@mwl%P8v%KjnWqQB6^H@=0LKN-ATR(|82|vj
zR~Y~Rx`>x>s0S+*hynlr#|4idFaTH?006#N82|vfh?lUa2OSlN0ssKV1zjL80ACpZ
z0KQ)t006p(m*S`giV%4M0074Yz8^3EEExa*zATq~sRvaVc>(|c#|2p*FaRhS003D?
zi3Ld?FaTb@D3|J~2PX@N0ssKV1sNYO0A05nss~X55Q_o;0LKNr954W}82|vjv6p?U
z2O9y4mzk^w!WEhV0074Y&l@lR&lvy!zRwu|0J@r&aIFU`6`BG70LKN78!!OS82|vj
z&=~*#x|)};tp^<yngRd-#|2#*FaXyX006$%82|vfnwR3O2Z|7j0ssKV1-=?E0HhfJ
z0KTM`e69yo8H)k{0LKMc8ZZE$82|uTNQnhW8ZZD}zMz-tt_LRzngRd-#|0T0FaXuJ
z9Ipr20VSOR0074YH5f1eG#UT^zBC#D07!`iAs8?KSXxMp4KsHP0J@!*8nFkt6|VvS
z0LKMa7cc-*8UO&kQyKsOy04e$u?I61uL1x7#{~};FaT5<006#J8UO&gua_{g2Q?M1
z0ssKV1;rLH096_Q0KQck006qLmwK`XBNeX#0074YjTSHfTp9oXzFZmr0J^W2va$!L
z5S;=50LKN;6)*rL8UO&kC6{Wm2UQuJ0ssKV1#1;B03jLx09i<h1zQy`0A9W!m)Nrh
zP8gj60074Y1Qjp<8yWxrUr31s^b{}vz8jZ5v<E2*uL1x7#|6U_FaTP&a<m7-0VT2m
z0074Y^%5`uxf%cfzPTCz07!`i;Sw+aSXxMp4fAmf0J^f5TDAwd6~qDn0LKLv5-<SR
z8UO&k*ct!;y2O_#w+Ay7!~y^S#|6(3FaX&a006$(8UO&g#Fucl2Q?MM0ssKV1%(kX
z0NNS=0KVE9006qgm%6tHBNfB~0074YO%X5v;Tiw{zTp}G0J_AN^0x=55V8UQ0LKN8
z5HJ9$8UO&ks+X#`2UQue0ssKV1uGCR0HzuM09i<h1sf1B0A9YPml(MRP8hNR0074Y
z#1Ak4pc()GUr31sv=1-<zMz+Wxd$l=!~y^S#|48AFaY1TvbhIa0VT%*0074Y9S$%6
zGaCQ^zB3yD07!`i2@Ws-SXxMp4I6C?0J_JQGP?)K71sg)0LKL{4KM&v8vp>lQ5ygN
zy4RNwyazQE*8%_l#|7mKFaS~;006#H8vp>h*Oywn2O|~N0ssKV1+5G)09hLV0KQop
z006qzmy)~(iV(*F0074Y4h%2=CK~_%z9yF}y$4kp#{vKV#|5biFaRPO003D?i3OPp
zFaTb@BA1%I2PX^H0ssKV1!)T~09Uuny$4YNB_RU<0LKLl3NQet9smHoryc+RNQnjc
z2`~UyT1brz3uz1hx*?ZBzX!<`FarPp#|0k=FaX6K006$l9smHkFqa^}2Q?Ki0{{TW
z1=R>J0LC5w0KUc^006o$mukQVBNZ?M0074Yod_@h%^m;%zRey00J<=jqQD1=5FrBq
z0LKOX2QUDd9smHonwLDm2UQs%0{{TW1(^ph0G1vA09i<h1&IeR0A9Y9m#V=BCkrqG
z0074YSqCrx%eUOY2Pp>%5=;O9NQng?05F$vVh9zNB*_Pf0q2*l$p<(A0+;8>2Pp{U
zO8@{d?Y`xg7|I9xEdT-l0LKLv{x1M96#xLfFckm*NQngl{x1MnT1brz6<rJfNR2&f
znG68B00Or_%Li=%6$%0X0LKM={4W4X6#xLfN)-SAx(b)V%m<4Q00IC2#|7Q{F90GH
z006!smtD;VGYbj=0074Yn)@#RPq&-R2bu;I2>}2A#|8EDF972W006$@4FCYT371OL
z2U80P0RRBU1&#DC0O7Z&)CVa63poM+0LKMc^Dh8Ix7^hS)d~yi0ssKV1$pQ%01=nG
z;0G}a>jD4(#|1s;F8~a;_}~Zj0T9{(0074YJmxO|s~Z3SzN?r1;s+Z6+Ls#R2O|sX
z0ssKV1?}Z801lT(;|E9!>jD4(#|4_@F905wo8t%g3+n;^0LKNF<Szggm!9MYKMU&u
z0074YP2?{C7?<ee2RjSt0ssKV1qb9W0O6M|<p(_r=>h-%#|6LRF970~cI5{>3+e&@
z0LKM)<1Ya6m%il(Jqzjr0074YFXJx&^OpwZ2QLfj0ssKV1@+=D00fsn<_C-r?E(M*
zOpAOZ2*(9k;x7O?m)zzDED-Gi002yjd?X0R1s&oq05z8w=Lani?E(M*OpAOZ2*(BB
z;V%F;msRHnQxMt$0074Yg5fU!o*MuFzMhx3=La7U?E(M*OpAOZ2*(9Z;V%G3m*?jP
zVj0>30074Y;ovU-ni~KBSxAWm(cmutUcQ=_V(14x5bXj008EQ~BnZa^h2Sp$KbNrR
z2SXI?0ssIp+enE8FyJo$#|19nF91Q82k8eP3+(~`0LKLa;4c6zmow=HI}7gu0074Y
zyx%VXZI^lJ2ShRN0ssKV1#RCi0BRfn05R=Ii3MEWF97SYNR0(<J`4cI1y$cK0Pg|-
z0LKMK-!A}Ym+t8Y84K?M0074YBi}CoWtS4_2Q&-s0ssKV1?1i@0ArU?>IX{;?*ae-
z#|4hwF924Tr|JhE3-JN~0LKMn-Y)=?m(l76OAGM=0074Y4&E;Sk(Vdy2P+Hl0ssKV
z1=!s$0G5|s>jxYQ@d5w<#|5$7F94XAgzE=A3-<y50LKMs-7f$zm(1%2DGT=k0074Y
zHQg@&C6??5whQ+H0074Yblfih9hbrE2U`pG0ssKV1rFRV027xd?FTLk_W}R_#|72f
zF8~6UVC@Gn4EF*60LKNH+b;kB9haT$2OA6b0ssKV1#;Uj0RFeb?FSD53-|&60LKMb
z+Ajc5m)-6MYBl%*0074Y<Jm6&K^*`9NQnj1*)IU;3I6~8NR0*KI1B(yiT%d~$Js9c
z_yPa`#|63BF91K6eD4P&3-$s40LKN2*)IU{m#gmwAq(^Z0074YUfC}I<CoX(2Pq5m
z0ssKV1uEGu0N9rc@CVro^a20?#|1{%F96V&@9+m@3-kg20LKN^*DnCZmtXM*Y7F!O
z0074YVb?DJwH%kQjs_T)?ePa22lN5}0LKL)m(cMCAD1rj2Q~}z0ssKV1;*Ad0I`>E
z@&`K$@d5w<#|41aF94>Owekln3-JN~0LKMM)-M2rm+bNf9}DjS0074YAJ#7bc$XIQ
z2Tu#>0ssKV1;y1b0Oyx*^9MZ(>H+`&#|3@WF97(Lxbp`w3-JN~0LKME)h_^>xAgM|
z1qcDkm-_n$R004^mr(o%J{1fD0074YoXjr(-5me`zTF)F0J;pfqx=Uf0RhaH(ESH7
z0su{y1O5jz2E<GN0KUYRJ^lwt0RoqY{s(?6{Q>|0#|0(IF95e4006$X9RL7Gi3JhL
zF929tNR166DGUHejXeQp3;??Qmofhb7y-(cO8*DG6$}Fa0LKL)$u9ub9RL8n)*S!<
zx(t;a2r3l}0{{TW1>?vs0M#7;0KU~7006oSm0SoW2Kr0@0KWQ_hzNfW{Q>|0#|0zE
zF950?006$KmlpvDRT=#P0074YzQ->BrX2tPSxAWmuE#F`UcRQ6g#idB3k(AQ0LKM>
z$1ecex3>WZ1PlQ-m%I%KMim|d0074YD7!BJRUQBUzEvIo0J<KR7Y+zl0XUa$4hVh_
z4+8)I#|1sQF90td006!(m;DY18vzfO84n1;6&?cs0LKMSxi0`s9smHoO&$OMx*nJC
z4+tw29s>XX#|00$F91#+006#D9smHk9+xl>2qzpcO#lGCFiijex;+2@07!`q#JDd2
z>3IGB|GqYta}Wr45Dx<Y0LKMCxGw-A9smHoBA5FR2vr#m0{{TW1=_bS03IFy09i<h
z1<JQC0A9WxmuV3QCkq|}0074Ynzt_iQ@5KD2r~f{fdK#j#|2!rF96LE006$t5dZ+X
zftTYF2;&QZ0RRBU1!T1^0MfV769~`&3*G<#0LKN1vo8Rsx2zQiQ2`}Y0ssKV1ue2K
z07)1C0KQ2W002md1sSq009aZ`jSVXi3;?=Smm(Jk$rWS*0074YKd~<WXBYqgzGoN!
z0J>zB7zi~LWC8#H#|8DUF92v5006#d7ytmeWS2@92qP6_0ssKV1--B@0B;xo0KRV+
z006pVmx34wiV#%-0074Y9<VO}JQx4~zC4#483<JwRRRD2#|62sF90|g003D?i3O>z
zF92S?IG2hU2qz0<0ssKV1$nP80ByIt83<7UC1(Nv0LKL#t}g(=7ytmi!59DlNQngr
zt}g&sT1brz8x0Hqx@VU%8wkl2b^-tZ#|1B~F96>d006$<7ytmec9#$w2sIUU0ssKV
z1?8+S0N@w^0KVWD006pnms%VMBNcW60074Yt*kEq=NJG0zULSK0J?UUk{k$%5N84a
z0LKLmtS<nx7ytmiw3jR$2vr$p0ssKV1*xkq0I(PU09i<h1(~Za0A9YZmzo_2Cku80
z0074YX{#>)<+sco2z?0wUYANI2p<8amue>na0^`p0074YP@yjXhqvV?2p|CqS_J?A
z#|0&zF91fj5Ge@d0W3xZ0074YHJ&d3^fv$izVtT$07!`iA)YS)SXxMp4Kw`<07#8J
znLG>tx<;2FD+oOcT?GID#|6xtF93~~X)6fq3s(gI0LKN@oG$<%mtZUiO$%5B0074Y
zeVi`<ESF#`2pX5tEC?$LSOow8#|15%F91515Ge>7w<9eG1OXXd1pojs?MR6Q{F*NS
zNQ*=y2*(BWnlAvXmq0EEG7DS<0074YxSB5jaF=~92+#{%1polY1>Ko10DYHOF9>oA
zS_J?A#|2=SF91}R)h`I<4qXKR08NSP#|3PdF93g+$uJ1#3ta^O0LKM-moET;m%A|t
zQVv}O002#i{l^6wmoET>mm4w&K>=QuT`~w80i2h4G6-B6UIhRENQnhdmM;KEi$o*{
zMKSHi1xA)H0G^lcG6+5aUY9O22p0jMmqs%PVhdgc0074Yf|V}-qnEQY2vQ$k1poj@
zi3K2)F91l3L?j4BG3`u=WF#QR1re1m0K1nYGze7-Uj+aF#|5#JF95^0gfs{#1_4gD
zt~v-d0S2{C006$Vm+CtRw+81=006${m!~`kV+QC?006$|mk>P&O%O%}0074Y&5JJp
z<u?ETzU7x~JqT49Mg;%>#|3DMF96{;003D?i3M1TF92S?;g{Pz2#^>?1polY1;2_f
z0NXbJ0AEOn1+R)P0KVIogFXnX3t<HS0LKOUi7x=zmpVTPK?`980074YuZb@J*q4Ss
z2u1<Wm(4#2WD8*h0074Y`G_w7+Lt^)2q_C<1polY1;vOj0P2@;KnN@gVg&#I#|4gv
zF97V9r$7ib3t|NT0LKMRh%W&9m*+qT?*Sv1*FgxB3&~Fa05R=FG336<mv%x3fC1Q-
z_d*DZ0X~<5LkKwm*q6OS2qys{m*qnUN(*5H0074Y`-Cq5-<LW>2;B=o0{{TW1rvlX
z01vk!MF<8A0iTyCQ3zQ94!46*2<8O=beC#Z2#^8exBFKJ6af`a0{{TW1siZL08k$Q
z0KQNk006pAmn2yTFBMM%0074Y*KaQXS04ZXzE>Xr0J=|?Xjup|0gsoQSqLcza!mjL
zG3~x`m(5uSE&<_}{8<Q_0sfbmS_p>=PXhn|#{~;+F921yDq9FW2MK*n002md1-@pN
zF;NH>w}xT}UIhWkm*r;&R002&I%o(#6<q@W0LKNfU@rij9{>QpogV-Ix?Q)9Xb5=$
z0n3*NY6w{Y|CeHF2sH-7P5=PD!<U_E2uT6~PnYOw2zpXdP5=N!G2}7rNQni$T`vIX
z`}Y6;W)B~P000000KUmi006o@0000?iQ(yh{Qv(*jSaHA3joIjnO!dcT>}6B#|4XB
zF94Pw006$09{>QlU6-_L2qp&kP5=PD`Iq2p2rCFsP5=Ng+rChj6Kn`k0d1FMYzV;%
zT>}6B#{~pjF94gjF>MGUD++E(002md1=27tmr#QS6t{7h2tEP<eV4GB2yX#?mmr%6
zUItW3006#JmxP-LVg_VN006#Zx9pn;<N^!G1ONcX1zIXE01B7eo(NS7$OHfY#|8Q+
zF8~dfJD&(x703ht0LKNFDK7y0F#rI*{4oFky2zKPp9qCY$OHfY#|8Z;F8~!X006!f
zG5`R&$OHfYNQn*PC@%o%y8i$F4<CR40002K3Nioyy2u0o0O_Ru|NmwWAA|q^0002K
z4Ke@#y2zJzpa>@m$OHfY#|4ZiF8~v_sh|ia0t>tZ0074YS|%?5u9w@Q2vrNb1ONcX
z1^Oi~0I`=lqX<|PyaWIM#|4)qF94u1006$AF#rI%yqBn>2!%?#1ONcX1^py10J$*$
z0KT~~006qY1ONa?i4Ei=F97Mf{{R0EAAkS=006$OF#rI%yaWIM>7@Su|7H&#ga7~l
z006$RF#rI%yq9;R2qz1?1ONcX1&kyw0JpcPqzD`Y6|e*V0LKMoATIz!F#rI*L@@vW
zy0Dkps0eE;r33%~#|7mdF8~oS006!bF#rHai3QOgF929tNR18SehUCdjXi0>3jn&M
zmvN~G;uWw20074Y*B&nbF);uDzA-TX0J^Z3XsQS$3$O$L0LKNL9xnhtmzb&uy$i4e
z0074Y)g3PYLYH8x2we-X1ONcX1#KNK08*F8s|bY)umk`A#|7OSF921SSgZ&o6|e*V
z0LKNb94`P}F#rI*T`>Rvy0Dj$tO$@0r33%~#{~!+F8~BF006!OmolvgRT`xP0074Y
zqZ=;(05JdnSxAWmlN&DpUcLY^m!Yi)OAD|B0074YJ{vCpTetVE2pIwkl>`6)#|0W0
zF97<N8?Xpf3zY-_0LKNo7%u?*mwT`XSQV870074YR~Rn<>@NTSzU(gm0J@cz=&%Td
zN|giv0LKNr7cT$>FaQ9)1uy^rx|IX~07!`qq!%v$>AL>^{|_I400000zWOf!0J@a~
z008Nv{{R1G4<Cd800000zWpx%0J@czx3LH(3zY-_0LKMP7cT$<x9PD6CjtwJ1ONcX
z1sWAE0G^i{v<OuTi39)u#|63+F94yJd$b5x6^R4@0LKMa6fXdfF8~0(kS_oLx`~(Q
zv<QVti39)u#|6C;F94}8006$JF8~0#i39)uNQn)k6E6Viy8i$F4<CR40002Ko-Y6Z
zx`_k;0O_Ru|NmwWAA|q^0002Kp)UXcx`~&!wFoB*i39)u#|2CiF94^v>9q(P1QmS*
z0074YB@ZtEG%o-EzBDfY0J?pb8oCH;EOG<@0LKNT4le)!F8~0(0WSamNQniJ4le*$
zT1brzqiqWSNR2(AunPdXa+k5X2;voe1ONcX1(yvk03j~`0KOqF006pum#Di4B@2B7
z0074YT@5b)FPGT62)zq^1ONcX1(ggh05X@Lya-(jeFOji#|146F91T92)zh}3w;Ct
z0LKNL3oiggmzcc>B^7-H0074YZ3`~|O)mfdzD+Lx0J?pb(!B_f5OM?n0LKN$3NHZk
zE&u?&^p|qJ2vr$!1ONcX1!D>?0PrpV09i<h1yc$y0A9ZEm*Bn#ObdMk0074Y{s}Jt
zOSd<_2#gFBLj(W-#|3BoE&#19006$NEdT(zLzmdg2vZ9~1ONcX1^@gm0I!!e%m^n7
zLj(W-#|6#&E&#TdXv_$u3qu3|0LKL#{4M~+ml(|mE(=2h0074Y<ohlFw6{^s2!a9?
zTLb_A#|2mUE&xw1006#EE&u?!TbJn12vZAN1ONcX1@HJS08y7l&<H0BTLb_A#|6Fk
zE&y1Uc+d!?3tI#L0LKLl_$~lvmnhK)E(==(0074Y)b}m`R<~i%2-XJyXty}o2r2>r
zJ(pwK2t5H<m!aDTDF{1H001%DzB{+l+Xz_%3)BPv0LKOY=q><>w>IDihXD)J1ONcX
z1#Rap0GF4@;RwhJ)C2$k#|4_^E&!UBo8kx}71RU(0LKMw<}LtUG5`R+UNQgxy407_
z;s|OB)C2$k#|7@?E&yhiN8<=cKGg&O0LKNJ<t_lQG5`R+urdGuy43^#07!`ig5@p%
z=_&vJ|45AmkwgmsG400%cI7Sr)dT<l#|3ERE&#Nb&Ep7T3)KVw0LKOT<SqcRmpkML
zN(<El0074YrQ|LEyqA#V2xtq{1ONcX1v2C=0LzyH<p?1Q)C2$k#{~xDE&ybgFXae2
z3)BPv0LKNv<1PS!mvrR_NDI{j0074YZsRThy0^#W2>b&Jp#%T`#|3BKE&zGA&glq`
z0Slo70074Y&D|~lhnHyU2*?Ye1ONcX1^V1B0E(A8><A(ip#%T`#|6#YE&xt2006#D
zFaQ9$p_g*(2x<$V1ONcX1y0;909Kdh><CO1q67c{#|8J>E&!k~006$AFaQ9$qL)bR
z2pkKd1ONcX1;*Ph0I8R6?FcIiq67c{#|4nvE&!sJrR@kR3!(%70LKMe+b#gOm)h+J
zZ403U0074Y<Jv9&RF_ij2s#U)1ONcX1)kb20CAU>?g&T=q67c{#|2B;E&!*u>h1_O
z1rY540074Y3e_$EXcPbdzG#;)^avXP?YB(y2m}KQ`2hd`#|7HaE&z+STlWY90TulL
z0074Y%+M|X;1mD=zTgx90J{B`bNC3r75M=G0LKOQ&n^Io6aWCei4*_;y7`w#`3PAR
z{Q&>~#|4ScE&$CG006$t6aWCa{g<}+2qzW#0RRBU1x?Q`0GAX10KS(L006rBm-6`t
zGYk3w0074Y1<x)3rI#@J2sIe{0RRBU1;x%T0J0PS08EMNzOobm0J{5^efkI>GW!7l
z0LKN5&MpA66aWBBiR-?z6aWCa`vCv|NQn(`&MpAy%l!ZU4<A4PzKaw90J`~?&-w^w
z5bXf~0LKON%`N~~6aWCeSeHio2x$xb0RRBU1$)gd0Na<k`v{~9`vCv|#|6yHE&#Z<
zY5WMx2Nezl0074Y^~5d!KQ#aVzCSep0J;vBN(2dN3l0SU0LKM(#4Z3mx4#4lCI~D$
z0RRBU1x36r0CW!k0KRk&002md1u?uX09aZ`jSWK^3jj!sJ!T^d0J=Mu5e*6Y6-faA
z0LKL%yDk8e4*&qZlMes@x=EKH4hbC<NdW)=#|7=WE&!Ad006#}4*&qVNtaR%2}27>
z0RRBU1(muk0HT+m4hcOANdW)=#|2HgE&!#M=ne^66-faA0LKOAxh?>w4*&qZrw;%C
zx=EK<4+&KZNdW)=#|3@4E&#QcxDN^S3rPV00LKM=xGn&=m$(oKMG!jy0074YEx0ZK
zWe)%VzGasU5eZcpI{^Ry#|6l@E&yQ<003D?i3PZ~E&yJ>VV8Rm2~HS00RRBU1!1=?
z09y|L0AEOn1yQ#y0KQw7;SmW}3rPV00LKOLwk`m^w?h&MIsqj{0ssKV1+BF%0N57*
z0KV84002md1(~%j09aZ`jSZ_53jn%Cm!cC1$rVxp0074YzqBp@^A`XBzVjCV0J>6_
zfD{Qe6;c8K0LKM&v@QVj7XSdh^cMgCx>A?Q6bXV5Mgjl;#|7TAE&$6H006$rms}MI
zRT)MC0074Yd9yA6#}@zqSxAWmX|pZ>UcSed$`uJG3sM3A0LKM6vn~MnxBL|ejY$=*
z0RRBU1)*s!0Gkp30KS_N006qKm!4k<9~G_v0074YZD}q5q7ncAzM>KU0J^T1(q9QU
z3$6hG0LKL&X)XYuw-sOslmZoy0RRBU1%YNR01pxX0KN|r006p>m%d^N9~F=R0074Y
zO=d0t77_pez7`Sy0J@Nu@?r@%3y=W-0LKLYW-b5}w>4u4lmZo>0RRBU1z}?@096tI
z0KQcc006q5m)>Ow9~GYg0074YEn_YKToM2PzFZOj0J@)-5@rcF3!ec10LKO3VlDt$
zw^e2dlLrgT0RRBU1wvac092RrZ3%)4%mDxZ#|5=oE&x`SfNlxB3(Nrk0LKOGSuOx&
zmq>33JqyeM0074YrCBZjXqS+037s6w0RRBU1^rkq07Mf20KP;M002md1?gBW0J_YV
zNpJ}_9n1j$0LKNYSS|oY69526i3OWjE&vaNX#M~HzDAdwa0wv+RF}VS2_y^50RRBU
z1xHve086*-a0&GU73u*10LKMKQ!W4t6aWCe3lsnVy6TtsbqO66>Hz=%#|0HrE&vh~
z006!c6aWCa>X#~Z3GNH(0RRBU1rky&01&q$cL|RL3%vmV0LKM*OfCTTm%V%mXbZgo
z0074Y1xzjg`Ij(#2|X3P0RRBU1-(ly0Ot|_0KVrE006qZmx6r>HaEQi0074Ya!W1%
z=@I|{NQn(&OD+Hpg!=yf|Gw!G006qZ0RRB$2><{8W)B}k006%C5&!_Yy_fHO2`3A^
z0RRBU1r$pz0PeRTehGR70a=%vg9&L1+W`Oo#|0%tE&#cg5QGVr231P{0KQe1t%M0S
z0ausggb7y_+W`Oo#|7_0E&#9-006$Q6953Z+m}#<2|Ekh0RRBU1)oDM0J4{sg$XeR
zEK2|YzATr`g$XDN+W`Oo#|1S*E&#}uh6!l_SC?vr2`mfS0RRBU1(ZTA0JpcHh6z#!
z6?6dr0LKM4J1zio5dZ+ba}fXlx^$NajtLzVbO8VW#{~sDE&y~9006#p5dZ+XbeB4g
z2~#3*0RRBU1)VxB03s0p07Wt6MKSiiA`t)pNQn)DIxYaZahI=-2__410RRBU1ywpO
z05F%}jtNZ{asdDU#|8a4E&x3d001%VzC95D0J?IQM2`s}3vvMf0LKNzIW7Q8mvD~>
zLl|=b0074YcR4NqSP=jKMKR>QSP=jKx^tJyj|q|sa{&MV#|6zeE&yhiXpjjd6?6dr
z0LKNLI4%Hn5dZ+bbrApnx^$PKkO?CdbO8VW#|33LE&z5B006#r5dZ+XbeGzY2|*Qg
z0RRBU1rInb0EH0%0KSD0006pmmoSkD9Tjx}0074Y-8U`(h7kY&zJ?J10J?RTVvz|W
z6?Fjs0LKNTH!c8&5dZ+bhY<h(x^<VDkqILebpZeX#|3RSE&zxT006#-5dZ+Xb(hkS
z2`&qD0RRBU1u8c#0F9Rmk_kZzbO8VW#|7RtE&zMCSCR=|0W4tw0074YbTuvj{}2EG
zzW)#a07!`iU^OlPSXxMp4RVwU07#8JpxX)nx?#8ClL@B*5McoT0LKNPGcEw`5C8za
z?w6UB2~`<k0RRBU1w1n@0O}9`09i<h1uQcz0A9Z8mj{*!P8eYU0074Y)iN#s<PZP=
zUr31s#WF4czT}r}mI+W9VF3UD#|3RNE&$#T002Q*NQnhpGA;nV-nZ743GxLjOaTA@
z#|2_6E&w(V006!=5C8y3i3Lt9E&y0sNR16(iwXcpjXlhh3IMuHxAdF|KLHh90RRBU
z1@|g009OzI0KQic006pPmq?xo9Ti>y0074Y#VRfUSP%dJzE}_d0J>h6dY%bG3tj;L
z0LKMwDlPzKm&l$8MGIa50074Y9V#vWX}1`k30DCSOaTA@#|676E&w19006!qmw%uM
z9u-~z0074YjVLYvYY+eczH1Nw0J>h6vY-hg3tj;L0LKMYC@uhbm*AiY9SdFo0074Y
zEhsJke3uBJ2}>1T0RRBU1=lAo0Dcev0KR?@006pPmuR60B@12w0074YohL2;jF*_9
z37rdG0RRBU1@$H_0FReIq6sS%UI73A#|6J8E&!1b006#`5C8zWUYCHP2~`<P0RRBU
z1yv?402UAc09i<h1w|$<0A9Wpm+qnoP8du90074Y?j<e&5D)+WUr31s-X$&oz7UsG
zqX|$MOaTA@#|4HZE&vJ;002Q*NQnh?B`yHI3YWd332_Tv0RRBU1^gr~0G78oqzQHg
z5Gnxx0LKM;9xedM4gdhY$(O&X2^#?_x7(@-@&pw)0RRBU1zi~~0O<|@0KVxC006o;
zm*TGpLkl<o0074Y1sN^?_m?oR2|Wuq0RRBU1-%$90Q#4Bun9#KH~|0v#|3Q|E&%)v
z006%H4gdhUIG56}2|Eio0RRBU1sxbJ01B5Fu?Zy$H~|0v#|7;dE&vUeNU;ee6*vI^
z0LKNr7cKx54*&qZ6%POax;U4Du?cJtDggih#|1nWE&#X=006$Ym;SK{?F%>o0074Y
zKNc<k6SwxV2}%bL4FUiF#|7#RE&!($006$Hmr%M18vzZsY`O{D0t*-d0074Y;tMVS
z%C}R!319&g7y<wQ#|3u^E&#|C006$o6#xLb7?;St3APm+0ssKV1*{1!03Q|r0KOj<
z006oimy^E<G8GvD0074YWeF|-))fE%zSb200J<5M+P?`;6&wNp0LKLZ2`&H(761Ug
z3>E+Yx*V52zzH7<83F(R#|6y@E&$}0Xut_I7#acq0LKN52rdBe6#xKCiR-@c6#xLb
z8ke=e2`)4m0ssKV1xE-j0P+<808EMNzVa0S0J<6i002md4J!yP0O@=D|NmwWA4C8E
zzRDE<0J<2L7r_aO5DfwV0LKN02QC1X6#xLfmY1@@3E&GH0ssKV1(*gd02-H_!U;+X
z8Ug?S#|1+ME&%tp^1=zi0W2Q^0074Yb_6Z}coqNvzIYY@07!`iVgxP#SXxMp4Rm7)
z07#8JC~OJ<x*xaT#0mZb6)yq+0LKMP04@Lm7XSdh0v7-Px-XaW$O+IDD*^xj#|3Bq
zEdZGo006$3761UcE0@^G2|x=g0ssKV1rq-)0GpR1$_arLF9HAn#|4J|EdcKp006%4
z761UcFPFK>2_zRR0ssKV1x@}f0H+oJ05RLXrxpMJx-6IU$_XJ0ECK)k#|0JsEdaHb
zAj=6g6)ge)0LKN@{Vf2(761Ug!WIAkx-FM#%LyYDEdl@l#|54JEdaw7006$j761Uc
zEtjIp2_p+F0ssKV1!es$0MVDw%LyF|Edl@l#|1t8EdbP)_{#~75FY{n0LKNS{4D@z
z761UgXqT4E2}lbs0ssKV1x)-c0R5Ng%n8#AEdl@l#|3WtEdbiL$IS`uA1vVj0074Y
zb=xff4h;YRz77ol07!`iVcRVLSXxMp4RbmQ07#8Jv4aW#y5YCs?FlRa74ZN70LKL#
z+ARPs4FCYXE)4(xy78AH?+G&t@c;k-#|72dEdV^1VDAY%3-JH|0LKN5*)0G+m#FUv
zH5Ksy0074YP1!8~K@9)^zCjHD0J`y)^6v>-3-JH|0LKO0*ew80mss!#r3>)@0074Y
zE!ZspQkMwv2{RS(0002T1@G4_08|YC0KQZW006r2mr(HuD;4no0074Yuh%UARSf_D
zzEuqX0J`y)knssc5a9p-0LKMY*DV133;+PW{g>tO2~`>40002T1@P7_0Qn3609i<h
z1>n{#0A9ZNms9cyP8i_;0074YiPkLu^9%q0Ur31sdDblezVny4@(EBG;Q#;t#|0tQ
zEdcEd002Q*NQngy)-3?O?Uy3+30e#B0002T1+LXC09uEK^9hHC^9qND^9zTE^9+ZF
z^9_fG^A3lH^ACrI^ALxJ^AU%K^Afj*^AnDM4KiV6I5%HkH8^2qIbWCGbQBtwv~(1S
z4KiV6I5%HkH8^2qIbWCGbQBtwWOWpgmpyb88kaA06keD9G86`f`*jqD`*jtE`*jw#
z`*j!XH7N*m6aWAK006jj6aWAK0062o6aWAK003}x6aWAK005YE6aWAK0078`5v&)7
z5v&-85v&=95v&@A5v&`B5v&}C5v(1D5v(4E5v(7F5v(AG5v(DH5v(GI5v(JJ5v(MK
z5v(PL5v(SM5v(VN5v(YO5v(bP5v(eQ5v(hR5v(kS5v(nT5v(qU5v(tV5v(wW5v(zX
z5v($Y5v((Z5v(+a5v(<b5v(?c5v(_d5v(|e5v)0f5v)3g5v)6h5v)9i5v)Cj5v)Fk
z5v)Il5v)Lm5v)On5v)Ro5v)Up5v)Xq5v)ar5v)ds5v)gt5v)ju5v)mv5v)pw5v)sx
z5v)vy5v)zO5v)#$Vt;W6chC+1v;)dQ5C9R42s!`&0Fnjo5;p*j2nhqgy*B^=0000}
z#c2cq5s?T6wGIFY0|?1C0000008=Its{j#_2nMwd00{#C$u|H10000YThZD85t9f8
zwGIGNM0>UX5tIl9wGIFY1Ng`{0000003s6GVgM192nMwd03Ra!&sqQxmIwy54gezM
zgWCWRmk0*64gezZ?Kc1sn3p&YPa*-Cms<}{EEk#x%M&*M2?J(N6aWAK002|w->(1>
zo0pysPb?Rl2o4lC00{$OPZR(E0000}4k5q*5uKOe4^I_Yo(K*UHvkC(TTc`K00000
zQ>U-M01=-E4iq;42?JP96aWAK002{3>Tv)Opa>2WHvkC(RZkQE00000Q@}Om01=@G
z4iq;42?J736aWAK002{~VRrx#qL)t)PZcer2r>mX00{#~PZR(E0000}Zak_05u^we
z1vdZ*13pg_0000008?lIvj7pLmyr-p6<MYT6$Li{2?I7y6aWAK002|^qJaPrrwEn@
zHvkC(F;5f#00000Q?_V001>DNi3K+R2?Hok6aWAK002|`NrM0psR)(_HvkC(98VMg
z00000Q_kF301>K}{}4|VEvpEY2R8r-0}@XZ0000008?hA$N&+n2$lyo00{#PPZR(E
z0000}6LwDk5v`X&5l<Cat_YR~HvkC(15XqH00000Q(SjX01>YUmIpTg2?PF46aWAK
z002{VoreGsun3k1HvkC(`A!r700000QyZik01>eWi3K+R2?Ov>6aWAK002`EphEx=
zvX`$BPZce*2$lyo00{%*P80wD0000}l=Kb&5wr-F2R8r-1Kv&)0000008?6>X8;kk
zm+=u#6<M|ji3K+R2?Nkh6aWAK002{k7wiBLw+PS&HvkC(#!eIf00000Q#*Gr01>zd
z&<8gF2?M}R6aWAK002|f+N1yxxd_k)HvkC(x=s`T00000Q@pnK01>*EUlLChExQQN
z2R8r-1FlXK0000008@)M)Bq8@2+#*N00{%AP80wD0000}!{&Sd5xtk85>FLbz6j6<
zHvkC(oK6$~00000Q}0Pv01>|k&<8gF2?Lf+6aWAK002|9P1gVszzEO>HvkC(kWLf;
z00000Q>S}&01?3m&<8gF2?L5w6aWAK002{4_k;iu!j}&dPZcf02+#*N00{$rP80wD
z0000}c|JS<5yS}c1UCQ)19wgo0000008>VwCjb$}mr)Z>6)naH@&q>k2?J?P6aWAK
z002|Y-MIh}#|ZKSHvkC(VonqQ00000Q_U#@0TIZTlM_!Y0m+xZ6HgU?$_R-CHvkC(
zKu#0@00000Q-5ah01?Xw@&q>k2?H}u6aWAK002{y9De{2%m@P*HvkC(C{7dr00000
zQ{^#@01?dyi3K+R2?H2T6aWAK002{z4p;yY&Is}ZHvkC(3r-XO00000Q=~o)0TIs#
z0~j{|2?GF56aWAK002`Q{af4s5zq)S1vdZ*1Nco80000008<Z+F#r+Kmv9tM6@StQ
zG6gpP2?OX&6aWAK002|hYkvR{(+DyJHvkC(;7t?&00000Q<ljL0TI*)G6gpP2?N+o
z6aWAK002{5&vyV3)d(^LHvkC(&`lHo00000Qzm=401?&*G6gpP2?NMY6aWAK002|s
z66XLB*9dF{HvkC(xJ?uQ00000Qyi7FMgS4m2#EzZ00{%AO%wnC0000}Z@kk05!sg@
z6;Bm^+6ajSHvkC(h)om#00000Q}&dE01?{=PX#vs2?Kad6aWAK002`F2XX)r+z5#U
zHvkC(XiXFV00000Q$dGn01@2?i3K+R2?JP76aWAK002{#?gjx7-Ux{WHvkC(NKF&~
z00000Q`U@501@8^i3K+R2?IDy6aWAK002`Q5g8o-5#R`k1vdZ*11L=t0000008<jm
zn*b5vm(UeY6@TIgi3K+R2?GdC6aWAK002|CQ!W4z;|Na$HvkC(_)HW400000Q{z+*
z01@N}c?UND2?OX%6aWAK002{lF(d#H<p_BPHvkC(*h~}v00000Q@caG01@U0c?UND
z2?NMX6aWAK002|}*x&#W=LmTRHvkC(xJ(oP00000Qyf=As{j$`2zduL00{%AOcVeB
z0000}s#aG35$TtJ7Ecvf>IiuUHvkC(h)fg!00000Q)ER)01@j5c?UND2?Kac6aWAK
z002`8KMMg7><D=WHvkC(XiO9U00000Q?XX;01@p7c?UND2?JP66aWAK002|!oH+my
z?w9TsPb?Sj2wodE00{#)OcVeB0000}B<@oH5%8BW7f%&g@d#cUHvkC(7)%rZ00000
zQ;>k>01@&C0~j{|2?GdB6aWAK002{gr$zt~^9TbNHvkC(_)8Q300000Q;{3H01@;E
z0~j{|2?OX$6aWAK002{AX#@cg^_QO)PZcfp2m=^500{%gOB4VA0000}<z-?35%&lK
z7&ib31Gq~R0000008@K8kpL0+m*E#r6<PTR0~j{|2?Ll*6aWAK002|ttqcJX`Uq?V
zHvkC(h)Waz00000Q#Rnt01^8LYy~#}2?Kab6aWAK002`6zBvF9{0M9XHvkC(XiF3T
z00000Q|Rqj01^F{PZ&=XfBpz;1vdZ*14v600000008?<b5dabY2y6v600{#)OB4VA
z0000}9S9@<5daBn1vdZ*11L)r0000008?WL!2l5f32X&700{#aOB4VA0000}aNNcK
z5dsNp1vdZ*0|-kL0000008=;h82}Lj32X&700{&5N)!M90000}90Vdo01*TUYy~#}
z2?OX#6aWAK002`dUO)g51(*LAPZcc&2~Pz#00{%gN)!M90000}-MFOy5eErR1vdZ*
z1Gq{Q0000008`9t<p2=~mq8g%6<G-hPX#vs2?Ll)6aWAK002|WK#u?s3JFgIHvkC(
zh)NUy00000Q>L4_01*obPX#vs2?Kaa6aWAK002{x4rKrl3<*yKHvkC(Xi5|S00000
zQ~XpU01*wBuNhAjEe;7@8#e$614v2~0000008_bC2LKTd39|<`00{#yN)!M90000}
zg*+et5fGR08BY}#5eZ%!HvkC(2uc(H00000Q*>;{01*<G9U4z87ZVAM9yb691K>#%
z0000008<s4?f?-KmtY!C6)hDBTn9G*2?M}M6aWAK002`_3*rC~771JjHvkC(ut^jE
z00000Q@lpV01+3Lq8d*u0T`Fh8c!8183_^}HvkC(fJqbp00000QvwN(01+Aq5+64J
z2?KCR6aWAK002|y4q^Zi8<!CqPZdKP2@)SS00{$7NfZD80000}S7%WG5giE<A2$FA
z13*a>0000008`3ZJ^&FO2@)SS00{#yNfZD80000}S*4Z$5g(Un8&52MAPLzYHvkC(
z@JJK@00000Q$lyC01+Vx?FTmi2?Nkb6aWAK002{HUwHr#A_=nxHvkC(ut*dD00000
zQ(EdE01+bzvj;Z-2?LNw6aWAK002`9gjN6%BngckHvkC(a7YvY00000Q}m_%01+h#
zjUG1u2?J0_6aWAK005Df6dX?+0VbC*98VQlCkfdfHvkC(5J(gN00000Q-Lx%01+q&
z*&jCm2?Ov)6aWAK002|jcY**BDGAvhHvkC(&_@&i00000Q-D&I01+w)*&jCm2?MZ4
z6aWAK002`Q-NFD7E0>=fPb?QK2@fGR00{$dM-%`60000}VQ-ND5iOVE98W9(E|(7-
zPZcdM2~r_900{#SM-%`60000}2|*D75iki-AvXXC1Mo%^0000008?Q!1pyH;mr)&0
z6#+7rXB|&07c&XhAvXXC1CT}(0000008@#OIsg$gm#Q64ELk-P8X`9U2?J0@6aWAK
z002`3W&r>ZHVF(0HvkC(Fh&#r00000Q~U>?01-C{3<@^@2?G#D6aWAK002{2ZGZp~
zI0^3wHvkC(@I@2=00000Q;)Uo01-KtFCI@7EjkI(2{!-<1F%ID0000008>IQ9RLwK
z2^u0d00{$-MHB!40000}2=smc5j>Z19#1R*J(r#yPb>jGm&+bcECD~4_#RIz0YH}}
zA5Sa+L6=e=Pb>jKmwz8mEdfK9tshS;0YsPDA5Sa+MVAL4Pb>jOmoy+xECENCVIWT|
z0Z5mQAWtj-Nte4IPb>jSm*^l*ECEZG6(LV70Zf-dAx|s;O_y&WPb?Qs3Ew3*00{%2
zJQM%`0000}2hvdh5l@%0Ax|s;P?z5!PZcdu2?n(e00{%|$Tt800000}Du*io5mE^T
zwGIFY1L(*%0000008<gdTmTVMmmwlg6<JgX2DJ_V2?N;3Hvj+t002`NMnV7)RS5>Q
z4gd)Q(8xCc00000Q?Fq=01;LR2DJ_V2?NN;Hvj+t002{rR2KjdR|y8S4gd)Qz{ocM
z00000Q}dn!01;T1k0MVMEm;W$wGIFY1F*<90000008`_azyJ|i2?n(e00{%A$Tt80
z0000}6bQ8d5nGqhB2N`rTnPrX4gd)Qn8-H(00000Q(V?P01;gY2DJ_V2?LPGHvj+t
z002|;iw*!0UI_-Z4gd)Qh{!hp00000Q=AOe01;ma2DJ_V2?K!0Hvj+t002{EN-F>n
zV3$85PZcd;2?n(e00{$d$Tt800000}D%Rux5n>4jwGIFY18B%M0000008{R-v;Yxf
zmw_Ws6<K5n2DJ_V2?JQjHvj+t002{!q5}XCWeEng4gd)QP{=m`00000Q<EBB01;*h
z2DJ_V2?I#THvj+t002|Ry`=yVX9)(i4gd)QK*%=$00000QylmC01;@H?;}qYEolh`
zwGIFY12D)p0000008<{#F#r*22?n(e00{#q$Tt800000}9H_|v5o?z*Bu^DtYzYRn
z4gd)Q7|1sO00000Qvlo`01<5o2DJ_V2?G$wHvj+t002{k4>kZ1ZV3jp4gd)Q2*@`8
z00000Q#CM*01<Bq2DJ_V2?GGgHvj+t002|`Te|=eaF?GXPZce32?n(e00{%|$2R}~
z0000}fX0Ua5poFzwGIFY1L(&$0000008<SZc>ob}m*FH&6<Kr%2DJ_V2?N;2Hvj+t
z002{fDw_ZibqNNw4gd)Q(8o6b00000Q?uZt01<Wx2DJ_V2?NN-Hvj+t002|(qXhvG
zcL@fy4gd)Qz{fWL00000Q@I#a01<eXPbE(kEqMtBwGIFY1F*+80000008>O7CIAt7
z2?n(e00{%A$2R}~0000}RXx)H5qp=BB~KMud<h1%4gd)Qn8!B&00000Q-n1F0TF!(
z2DJ_V2?LPFHvj+t002{t+DZTsehCJ(4gd)Qh{rbo00000Q<^mh01<x)2DJ_V2?Kz~
zHvj+t002`3mM8!bfS3OzPZceJ2?n(e00{$d$2R}~0000}EGP5;5rPQ@wGIFY18B!L
z0000008{NaG5`^Smq8{^6<LG{2DJ_V2?JQiHvj+t002{*4xj)Lg$V|=4gd)QP{%g_
z00000Qw_(u01<`>2DJ_V2?I#SHvj+t002{loCp9BhY1F?4gd)QK*u)#00000QwiJI
z01=3nuO?3wEr|&RwGIFY12D%o0000008`41zyJ}72?n(e00{#q$2R}~0000}Iw5cX
z5sR1cCQlVvj0py{4gd)Q7{@mN00000Q;;A|01=G|2DJ_V2?G$vHvj+t002|REKUFs
zjtK^}4gd)Q2*)=700000Q}HxV01=M~2DJ_V2?GGfHvj+t002{VZ=V1Wke6R4PZceZ
z2?n(e00{%|#y0=}0000}jtjE@5t0c8wGIFY1L(##0000008=7ENB|L&m!T(56<L%C
z2DJ_V2?N;1Hvj+t002{knI`}dl?ev54gd)Q(8f0a00000Q@ab001=i62DJ_V2?NN+
zHvj+t002`nZPx%1mk9>74gd)Qz{WQK00000Q#L1W01=p%4=7I+Etv@hwGIFY1F*(7
z0000008_RI^8gW=2?n(e00{%A#y0=}0000}3sgk_5u2A$C{GnxoCyZC4gd)Qn8r5%
z00000Q+0Vb01=%D2DJ_V2?LPEHvj+t002`qKllI<o(TrE4gd)Qh{iVn00000Q?85y
z01=-F2DJ_V2?Kz}Hvj+t002{k^^pJ(pqIZWPZcep2?n(e00{$d#y0=}0000}BFseq
z5uynOwGIFY18BxK0000008{2i4FD0NmjNkH6<MSS2DJ_V2?JQhHvj+t002|<gqHvj
zr3nVL4gd)QP{ua^00000Q$>~k01>7M2DJ_V2?I#RHvj+t002|go(TaFrwInN4gd)Q
zK*l!!00000Q?!by01>E{Zz)d|EvX3xwGIFY12D!n0000008`q&od6N42?n(e00{#q
z#y0=}0000}+%a_k5v!N6DNhwytO*9S4gd)Q7{)gM00000Qx6q<01>ST2DJ_V2?G$u
zHvj+t002`7etZBCt_cRU4gd)Q2*x)600000Q#%u(01>YV2DJ_V2?GGeHvj+t002`$
zCvN}|u$LbyPZce(2?n(e00{%|#Ww%|0000}f)=R&5wZyewGIFY1L(y!0000008?zh
zF#r*>mtiVT6<M?i2DJ_V2?N;0Hvj+t002`K7zzLpwFw5b4gd)Q(8V_Z00000Q}N<?
z01>tc2DJ_V2?NN*Hvj+t002|ioIL;$w+RNd4gd)Qz{NKJ00000Q{x5(0TH;D&niz9
zEx8E>wGIFY1F*$60000008`!zy#Nuq2?n(e00{%A#Ww%|0000}z_<nh5xbWWD^C?!
zya@)i4gd)Qn8h~$00000Qx2IT01>?j2DJ_V2?LPDHvj+t002|=BjErMz6l1k4gd)Q
zh{ZPm00000Q-9=l01>|l2DJ_V2?Kz|Hvj+t002{t<aYoOz?Xk3PZce}2?n(e00{$d
z#Ww%|0000}7jOUp5yA-uwGIFY18BuJ0000008^B*vj7ppm%%Gf6<Ndy2DJ_V2?JQg
zHvj+t002|HTLu9U#R&$r4gd)QP{lU@00000Qy)M)01?Is2DJ_V2?I#QHvj+t002|Q
zA}#<C#|Z|t4gd)QK*cuz00000Q>dZc01?QSFDy?LEy)Q6wGIFY12Dxm0000008{el
zmH-jT2?n(e00{#q#Ww%|0000}pSK4A5zCixEKe0#%n1gy4gd)Q7{xaL00000Q#M$8
z01?dz2DJ_V2?G$tHvj+t002{!LJ9y8&Ity!4gd)Q2*o!500000Qv%=M01?j#2DJ_V
z2?GGdHvj+t002|DBE$d@(3jsVPZcfE2?n(e00{%|#5Vu{0000}&#3_c5z+|;wGIFY
z1L(vz0000008{-TA^;K7mmw`r6<O2?2DJ_V2?N-~Hvj+t002{X=YIeZ)d>c*4gd)Q
z(8M<Y00000Q{kCK01?&+2DJ_V2?NN)Hvj+t002|Lx0nDC*9iu-4gd)Qz{EEI00000
zQvn8_01?=ik1bCXE!hbMwGIFY1F*z50000008_Jnc>odG2?n(e00{%A#5Vu{0000}
znD77q5!;v1El(9$+zAG?4gd)Qn8Y^#00000Q$r#;01@2@2DJ_V2?LPCHvj+t002`$
z`~?6J-U$Y^4gd)Qh{QJl00000Q&Ob@01@8_2DJ_V2?Kz{Hvj+t002|3<*)z|;Fmux
zPZcfU2?n(e00{$d#5Vu{0000}lM2!R5#k93wGIFY18BrI0000008=LY@&FOzmw_%%
z6<Op72DJ_V2?JQfHvj+t002{Hl?wn7<p~D04gd)QP{cO?00000Q;uD301@U12DJ_V
z2?I#PHvj+t002|;{_+43=LrV24gd)QK*Toy00000QxQid01@by?=DXjE$ImcwGIFY
z12Dul0000008_!;5djhE2?n(e00{#q#5Vu{0000}Rk~{c5$l&RFHaR&><I?74gd)Q
z7{oUK00000Q>BRY01@p82DJ_V2?G$sHvj+t002`NMk@dj?g<994gd)Q2*fu400000
zQ^ocF01@vA2DJ_V2?GGcHvj+t002{3OqKu<@Ry%2PZcfk2?n(e00{%|!#4l`0000}
zi^Pop5%LKJwGIFY1L(sy0000008`Eu009y6m*Fo@6<PEN2DJ_V2?N-}Hvj+t002{&
z%hCW5^$7;G4gd)Q(8D(X00000Q&AQ;01@^H2DJ_V2?NN(Hvj+t002|c#l`>;_X!5I
z4gd)Qz{58H00000Q;7MK01^0?PcTmvE%^xswGIFY1F*w40000008@081_2TJ2?n(e
z00{%A!#4l`0000}Jzzxu5&M^sFi#a){0RoN4gd)Qn8P;!00000Q<iNQ01^EO2DJ_V
z2?LPBHvj+t002`~Pdfk+{s{)P4gd)Qh{HDk00000Q%yR~01^KQ2DJ_V2?Kz`Hvj+t
z002{b9WDS701B2dPZTWy3I??f00{$d!#4l`0000}bo_$=5dsPZwGIFY18BoH00000
z08@X2mjDq1mq9U46<Gud2DJ_V2?JQeHvj+t002{G91sB!1queW4gd)QP{TI>00000
zQ_W|)01*ZX2DJ_V2?I#OHvj+t002`KBD??*2MPwY4gd)QK*Kix00000Q>@7i0TEf3
zuQ5**Em{f&wGIFY12Drk0000008{mczW@<i3I??f00{#q!#4l`0000}xTkLb5nPw?
zF;5j)T?z)Z4gd)Q7{fOJ00000Q~&Up01;ja2DJ_V2?G$rHvj+t002`p#a;jrUkV1b
z4gd)Q2*Wo300000Qxu&f01;pc2DJ_V2?GGbHvj+t002|QYSaJ`VV7SrPZcd<3I??f
z00{%|!Z!c_0000}-azyK5n~DlwGIFY1L(px0000008{Qj^Z*fLm!UFG6@O(42DJ_V
z2?N-|Hvj+t001IlLi7L;W(o$i4gex@Li7L;X9@<j4gey8Li7L;XbJ|k4geyOLi7L;
zX$l6l4gezlSMdN5Y6=Fm4gewvL-YU<YYGOn4gew<L-YU<YzhXo4gez2eenPhZ3+gp
z4gexne(?YiZVCpq4gex<S%2vO5pN0xwGIFxh+p&o5pW6ywGIFxe{}Q!5pfCzwGIFx
zu$%D!5poI!wGIFxH+J*@5pxO#wGIFxM|Si85p)U$wGIFxS9bIO5p@a%wGIFx2nDAA
z5q1g&wGIFxHBrX^5qAm(wGIFxeNo2%5qJs)wGIFx7YL^S5qSy*wPOwdQ|r>_01<i$
z2DJ_V2?Nl=Hvj+t002|`;O77ldkO}%4gd)Q$ig=O00000Qx9*)01<o&2DJ_V2?M~w
zHvj+t001J&Z^r-;eF_G(4geyx;^zPnehLP)4gezLa>oD>f0q|DPaYSH3Trqw00{%2
zhc^HK0000}sgRxk5sjBjG*2iO<qE1eHvkC(Aci*p00000Q(?C+01@Vwi!@It7d{K=
zI5z+Z1Av4#0000008>~GbN~@Qm&!CxCl{9sC^<I(2?KzGHvj+t002`-Z^Qr*n3o7O
zPbUFLmoPO?C;^O@Sv5~60m+wwHBTr3ewV8?PbUGwm(w**C;<kS{54M~0Zf-AHcu!4
zl9x_4PbdL8mv=T#IsxLBvNlgB0lJslHcux3w3h=nPbdNNmn=6=C>J>riaR#|2?M}u
zHvj+t002`(hWr2#f0t`FPbdM!my|b8C;=ary*E!v0g;yoI8Sf^@0WQvPbdL3m!ddN
zC;<kS%Q#Oc0dSY}I8P`6MVB2pPbUE=mqj^GCjn)bZ8=XU0jZakIZr478kfI0PbdMB
zm*_c9C;<wW5jsyO0fLu0I!`D8`j=xmPbdL@my9}3C;<eQwK`8I0eqL<I!`D8h?fUD
zPbdNa9+xpYPbUF*msvYcC;{-7f;&$r0W_DZJ5MM9dY9BYPbdMvm;F0WC;_IICOl6l
z0mPS2JWnVAZkKpGPbdLLm!UjQCl^XB|3EhY2?Ky2Hvj+t002`m*O>qjj+fs&PbU}b
zEj>Xu00{%29yb600000}Z|A-M5ipk>Jx?bAj+aF}PbdM>mu@{zCjoMomOW1=0nV4c
zJx?cDDli;EHvkC(Kp8gx00000Qy$7601@Iaqd_+S2?GEaHvj+t002`r2gm>sf-%fN
zHvkC(Ko>Uv00000QwKyo01@6X^g%ZO2?Kx@Hvj+t002_}OrHP|I+s2^PbdNPmuEgt
zCjlOpk3LT*0b-Z9K2InCx|iWTPbU|ZIX^=;00{$N4mSV*0000}2xtTW5tf%9KTj1c
z{y9HGHvkC(zzsJ500000QyKsw01^K=fX5C12?HRvHvj+t002{p?kWHgXP03=PZa@Z
zmv=u;C;`xypg&I)0nwMXKTjwDb(h~iPZa@nm-RnSC>I($Q$#la2?HPrHvj+t002|$
z{R;sR8<#XdPqP=9P##4$00{#?t~USx0000}*(?J95#5*mKu;$DB9|sXPbU|`QG`V|
z00{$tt2Y1u0000}+YJc;5y_WiK~E<E=a-B@Pcs2xm%u?!Cjobt=s`~>0fd(jLQfR|
zQ->x(PlqN#P=_W$QHLf%Qimo&Q->x(REH))Rfi@*R);1+SBEA-ScfJ;S%)S<T8Ab=
zTZbk>T!$t?U56$@UWX<^Uxy|_V236`VTUF{VuvO|V}~X}WQQg~Wrrq0W``z1XNM+2
zXon_3X@@34YKJC5YlkL6Y=<U7ZHFd8Zigm9Z-*vAaEB&BaknNya=i%+`DZ!+0018}
z&jSGw`G@dJa)<CsbBFLtbcgUub%*dvc8BmwcZcvxc!%&yd57>zdWY~!dx!8#e24H$
zeTVQ%ez)*Te-6EW_h$yR4gdpX4QK!#?%zKE5%*_800000p8@%p4*?PMX9l$n00V*z
zXaFN(01@<O<p2Nx0G|QCT2TNIwr9Gi4gdq;4QK!(VgM1gX9l$n03z<+KL8Q8XS%2k
z08DrQ5x8dtwGIGB(hX<;A~-IK01?Ay;SFd2B2+SS01?DuXU3=w002HB&Y)rd5%FgR
zwGIFP022sI#0~(T0bn?m01<^}2DJ_V111h=03&z+5rt<4wGIGGZ2%F6XH2~g03s!J
z*#HrVXV3%y08Daaa{v*FX9l$n006@QQ8Zlu5pkE_34kFHac2g#4gl%kE&yE3bUOeM
zahDMafD{3FmnI5;ApwGyND6>&e`RMn00003C=X}=BX|H2WoHJp4gey-F24W~W@kD8
z0001=0sbr701-=P&;$Sg18EOv03&Yz5ld$VwGIFu2Z~|<5ld(3whjP3B8kIC01;eg
zWdQ&HOmYAbU1vG~00000p8+5XLjV!|W;y@>00D*%XaFO001^FWh7V`}mo^K44lC_u
zIsgCw0mTn!03v!>_y7^^W{v{@03taZZ2%GOW{v{@06rp~$qWGz@|SN5fFOV8W;y@>
z00RIJXaFN~F#r+gX3zux03&lU01@YA&;$Sg0G|OkYqJ0m&t^IR0009k5NH4+WpZ)=
z5zl4@wGIFuk-U5W5zl4@wGIFPp8@_Fv;YyhW;y@>00DLoXaFCPynFx=x@HEo4geq3
zu95%|x@HEo4ggGP01>-px~P8+08DNG5xZu(s15)=OmqMd#AXJy4gdfEp8-^{WB?Je
zW;y@>00X%YXaFN+a&iC>vStRg4gen(odW?8vStRg4geqV5q1C(vStRg4ggGYWpe-#
zvt~K~0001=0peLG01=F4IsgCw0|F6f03%@l5sYSc&<+42VgM10W_N$k4ggGTa&G_;
zjb;Y54gfw(X#f$CX1b^j08DNG5s_xPs15)C0G|Q;X2<{$gl0Mb0003g5oiD-VE_??
zW_QpI03%`m5rk%U&<+3qp8=tK`v4J4WI_M{00SfxXaFO401-@N2DJ_VivhB@ga8r#
zWVQiVYT5u10%cjX1pt2l0000Yfj&k65lv)G0000YAfo*M5lv)G0000Ym(%tD5l&>v
z0000-Q4(kXA~-IK01;ng;SFd2N3;=W03vN9hyW2{WI)6Y06roBFlqo1Vq`$X4gdfE
zN7NB$03vN9hyW30WHJB%06roBFlqo1W@It|00000NAwYB03v^FB!~bJZ)7q6002HB
z05EC*5pQHN0000007n!OXaFK@B!~bJcw{mF002HB05EC*5qM-W0000007o<uXaFK@
zB!~bJj$|?b002HB05EC*5sqXs0000006rpZB!~bJkz}i)4gfwP05EC*5s_r8q7DE6
z002jy6KDV;I4*yS01=jC;SFd2M}iV)03vN9hyW3pWKqNp06roBFlqo1nPgGK4gdfE
zN1_sF03vN9hyW3qWI)6Y06roBFlqo1nq)x44gdfEN5T?l03vN9hyW3uWaR(=06roB
zFlqo1o@C_!00000N8%D_03vN9hyW3%WHJB%06roBFlv7Q5vOD_0000007n87XaFK@
zB!~bJuw>-`002HB05EC*5wK+C0000007oJdXaFK@B!~bJxnwc`002HB05EC*5xHbC
z0000007pU-XaFK@B!~bJ!er$D002HB05EC*5yE8U0000007qgIXaFK@B!~bJ%VaVD
z002HB05E@Q01?Y%G5`Po002jV6KDV;Z6t^Q5!7Vm0000!A^<RI01?z=<p2Nx002HB
zZ6t^Q5!Ylzq7DE)A^<RI01?+@MWPM>0000-`x9sYOm<;x01@D1LI3~&Om+Yf;baE2
z4gfw(Z*FA(5$a?*0000Y{!|D65$a?*0000YFH3)301@kC`PL2qOmby(01@nDIsgCw
zJ|fPbVgM2EWCpbk00000K1_CDYyc7WWI_M{08DgkZ~zhbWCpbk06rqlpke?K`(y^S
z4gdfE0G|PUpq>B`yJ7~l4gdp<6lee+Z6t^Q5xZhK0000Yfj&k65xio`0000YAfo*M
z5xjq5$^ZZWBA3(l01>@n$^ZZWK1^?JWdIS<Vmbf-03rcZ2mle%Vmbf-03t6-UH}o(
zV)@n%0001=0ir$=01=#G2DJ_V1NIbX03U56hyW3sVhZRE03v}tMgS3=Vi)KR03slw
z{Qwc2Vi)KR03w&u_5cx{V#)vj06t7_Ze=O}5wl`C0000Y0aXY95wl`C0000YFH2qk
z5wv3Y)(!vw0Hy_Oq&ol+XqT@LfE|BmV$cKt0L|ep09>X5XaEstVmbf-0MOzt09*ud
zqyQ0VVmbf-0OI2=09<t94gnEqV$cKt0KDWb0M-IxVPpUifMP<pkg@;~YGQT}Hvqro
zE&xJ26_5ZCYhvXaXaIobE&xIRm4pBhY+^Le4gdw`E&yJT2LJ%h6=(oL>w6hL01<&=
z2DJ_VhvzN;C6EUI07A~7VgM0=mt+xuHUWZ{n-PE@0fLvj5r829aF^Z@fMo%0mqZeP
zUjf0Fq!NI+0o%7;6M$F&0dlvu6o9e;0d==(7JxYce>7rt5H|n;yBuf$AA{Mq01-7}
z&;$SgAEp9m01-7}IsgCw9|Uov01-A~IsgCwA9Ugl0TDK0&;$Sg)&gQ-WB?IkVj{SZ
zvH%e`Vs;QW03sP=_y7?&Vg|Ji03v|dw*V12Vg|Ji07q;bXaFM4pke?KU}6Tf4gdg0
zejI23e<IGHVgM0gVg|Ji002jx9B2R{>w7-{5oBTpwGIG2BF>;<01;$j2DJ_V002HB
z>w7-{5oKZqwGIG2BF>;<01;(k2DJ_V0002j0&*bZ01*UY0WuwE03S0N009vMVmbf-
z03V6aYyc4iVs;QW08DUia{v(qVmbf-08D8Be-Q;@IsgCwB9@d~01*aaTM{<_M<X3*
z08DgkZ~ze)Vs;QW06rqlpke?K8e#^u4gdfEJ|afBz5o#(Vmbf-0001=0e;YI01@_K
zb`UoJ0cjm*03S0N009y9VLAW+03wOdYyc7WVRjHV08DUia{v+eVLAW+006@QGw@FU
z0ukYtA{l@pf8k+U5;p(=Ra|>Yg8&iXVLAW+03`M<09-Q~009x<VLAW+0E6}}0780e
z+5i#aVLAW+0Oj^B0AX!uWB?K3VLAW+05JD10AXcrWB?K3VLAW+0HXIU005r>hRH1e
z5x8MG00001p&n=e9{`nv01>!hG|vtIAEp9m01>!hOgaDn03!3Io&XWKVFtAh03zrr
z1OO4bVFtAh03!O(!T=GvVFtAh03tk^;s6o4VFtAh03xcA{QwcXVZFr;07rx#XaH0A
zi3b4@#g`8nfE9nmVFtAh00{%=zBd2>00000J|fwN0ss-qVFtAh06rqlpke?K&S3_%
z4gdfE0HOmBp#lIAm0>yn0004>0gqq801-T42DJ_V0Ye{X03Y2E&;SuUVFtAh03XC1
zTL2L}VFtAh03&O601-T42DJ_VB8kIC01-W52DJ_VBHMqzC;$;YVM8Bi06ro!F%bX}
zKw-M54gezXvvB|sL16~94geyY^bP<KLSY8A4gg0pA7}s~&Y)rd5lLYNwGIFPJ|fPb
zVgM0MVFtAh0000069~!?Hvpdjh$!X&5jkN7wGIFSe;;T7AKekq01-K12DJ_VBVlr9
z01-K12DKIr03&O601-K12DJ_V0H+0un9u+b1D7}(fFXYaVFtAh008(d09>5(4ge7Y
zVFtAh0NVI20BvD%X8;icVFtAh04(}00Bvh`01*RW2DJ_Ve)}!}VP*gk1Ys&6XaKGJ
zE&xKro1_2{1z~8%4giP!E&xJQGIIbC24T9W4ggsGE&yI21^@sxAZP$0iNi+#5glO$
zwGIG6{uu-j01+Lx+#7%l0Rj1!=^TI{0r{5-9e^}{=wJr54gdkiA!q;}-4W0L5$IqB
zwGIFyZ(?cy5$IqBwGIFuhLc4A5$IqBwGIGGZ2%GKU_&2h005r>ER1ph5z}Bg00001
z=ptwUBW(Z?(_lj%XaFDH-gE#F(_nYE4ge!za%TV$(_m2n002h~B4_|2X~E0@5!GNi
z0001g07oApXaFK<!OQ>=)?hjS0000-F(POHB5A?Q01?+<IsgCw07pY2XaFK<!OQ>=
z*kC#U0000-RU&8rB5A?Q01??>IsgCw07qvcXaFK<!OQ>=+F&{W0000-c_L^4B5A?Q
z01?|@IsgCw07r`=XaFK<!OQ>=++aEY0001GN1Y;Q03vC@%m5MHU^)N*002j?B4_|2
zX~E0@5#C@r00000N5LX!03vC@%m5MJU^)N*002HBR)*F95#eCH#SQ>WX#f%8U^)N*
z06rpV!OQ>=<X}1g000000JQ`z$a(+~&X=JcfExkMm%Sc<8v)Um*&cv#I>%rEt0QOt
zBW(Z?$6!MrXaE490%BIl01>ZXIsgCw0Hy^100000)Bykh000000i4ArXaH?(01>a3
zgdcz;0kM~<AAleMv6sppfF=R5m+>EfAOW(M5+Hyk0kfAmAb=nNvzJ;RfF=R7mw_OF
zAOW<Oq9A}K0kxOOAb=nNwU_E3fF=R9mk}X=AOW_QG9iE_0k@Y~A%GwOx0iY$fF=RB
zm!TnmAOX0S!Xbbr0lAmyA%GwOxt9tefF=RDmoXxMAOX6UQX+sR0lSxZB7h(PyO)|G
zfF=RFm%$={AOXCW;v#?|3%+2z#SQ>rX#f$wmk%R=Bmu#fG$Vi@0l}A5BY-G>jbH|~
z4gdjkC};p5hLc4A5shGX&<+42ZEOG$jbPlz4gepR87%-2jbJ(e001AeLuLRGjbH|~
z4gen_8+ZT_j$j704geo#_uBvwj$j704geoW(+U6)j$j704geoa%bfrbj$j704ggGV
zZ~zgHU_&2h07p|OXaFLkjQIdD5td*EwGIFPM`b8z03xG|`2Z1^U<S1g002HBqm20g
z5u#uQwGIFP0K)+R0zv>0e3vaGfFBWjU_&2h0N?{J09?c!TL2M!mt7=)9|2gGeI$S&
zAy{BTA7}tl1TO$w0r%nn5m;bR0RRA=1TO$WiNi+#5m}eMB!C`&NMJeu0004$DQEyA
zZ*Tw+NMJ)BXaFB=B!~bJNMHuF4genk_u>E%NMKO`002yMZEye)Nni%G4gg1YDQEy9
zqm20g5m8_UwGIFPJ|gRTKL8O_U<S1g06rqlpke?KRA2_R4gdfE0K)+)d>a4}I$#F1
z4ggS~1poj5002psTP1)JAv$0~A7}s#1up<x0r%nn5jtQ|0RRAh1up<XiNi+#5j&TT
zC4eP=DquPQ00045Drf*BZ*Tw+Dquq&XaFB=B!~bJDqse+4genk_u>E%Dqv9o002yM
zZEye)D_{n-4gfwP>w7-{5jJ23wGIG2BF>;<01-A|2DJ_V0002P0Uh&J01+Hu2DJ_V
zpr8c+00000NdN!<000000i3`pXaH?*Z~ze@9AHBqXaG_MF92Kt_u>E%9AHra006`W
zF91S`!$$xS9hXxkfF)lHU^)N*00AB=XaFN`Z~zeuU_&2h03U56hyW1`U<S1g03QMO
z;s6m0U{L`808DgkZ~zevU<S1g06rq?dp`gX7+?mq4gfwP&Y)rd5g1?wwGIFP005u@
zW6Ayi5%!nYCV(Ix_FqFEXaJW7F93l8iNi+#5%*s@00000VRB`201^0?7AJrt0SK2n
zCx9XW@0VUDfFUsNUqc^g00C9S0Ri{o01@wBQ2_t|0aQYX!$$xS@m~hD4giY?F8~0;
z0k)1801@ezuP1;ZAn9L2A7}spRmA}T_u>E%>0eO+000400K);f)dBz!<CpL!fFXb5
zUqc^g00C9S0Ri{o01@L~Q2_t|0aO5=0c+{+01?k$IsgCw0oE*N03&a301?k$Lmy}W
zA8jOv01?k$2DJ_V9|8B`01?k$Q2_t|OmuB<01?q&2DJ_VJ|gRTKL8QmUk0@f06rql
zpke?K-(Lo`4gdfE0K)-icDeu&#9x<?D1aXk#9u=nXaK<oF92Kt_u>E%#FxM*fF1$5
zm)|IWAb+}FLmy}W0ae8TA{%%B5xQSd0RR92R8ySv4ge9nUk0@f00B?{p8<9Cp#Tx4
zUpfE)00Gi1XaFN`Z~zgeUqc^g03U56hyW3$Uk0@f03QMO;s6n*Ur_-708DgkZ~zgg
zUk0@f06rq?dp`gXwqFLd4gfwP&Y)rd5w>3jwG9pc0002P0byNq01=><iz$F1f1qDO
zA7}spRmA}T_u>E%pkGk|000400G|QjEh_*Kj9)qc0004TE@%KFZ*Tw+j9)_^XaFB=
zB!~bJj9&(|4genk_u>E%j9*a!002yMZEye)j$a0~4gfwP>w7-{5u9HJwGIG2BF>;<
z01=#D2DJ_V0002P0l(Bs01<s(mn$lO9}#_DLmy}WMhPzfTmkpu01<tcTPlDa0dbdl
zDu5tHabH6pXaLIzF92L^B!~bJabE_t4gePlF92Kt_u>E%abHmZ002!2F92b5ZEye)
zb6*Cv4gi@7F90O~0000&>w7-{5qX#3Du5J!d0z&#4gduUF8}}l0G|P};8Oq*U0(*Z
z4gdjqFK7TGZ*Tw+U0*{VXaFA~8+ZT_U0+cF0001=0g5;n01-4_chC+10r4+r03U{v
zMF0^rUw6<B0E+=EnKb|rQD4>qVqs(e5ma9ywrZOI5j9^1wGIFxZ>#SB5jJ0U&<+59
zN7OH903yzyVgM0RUk0@f002HB>w7-{5mjFXwGIG2BF>;<01;JR2DJ_V0001=0qF8f
z01*;j2DJ_V0m?9F03TBZQ2-GVUk0@f03RiG*#Hp|Uk0@f03TOVcK{I*Uk0@f03Y6-
za{v(&Uk0@f03&T{cK{I+Uk0@f03WV@8pr?<6kj?30019J22lVJ6kj?3001J1!$$xS
z6<-Fm4gez2gBJi16<-Fm4gezTD<}XF7GDOn4gg1+FlYcG>w7-{5hGs)wGIG2BF>;<
z01+c!2DJ_V002j=FlYcG&Y)rd5iMT^wGIFPJ|gRTKL8OgUk0@f06rqlpke@j5ief`
zwGIFP005r>opj*<5%FFIwGIFQr7>s#A5#WV01@$C2DJ_VA6HX%01@$C2DJ_VA0>9#
z01@$C2DJ_VAG1Sd01@(D2DJ_VBW-JU01@(D2DJ_VAHWa{0TJ?E2DJ_VBV=!501@(D
z2DJ_V)&gQ-WB?HkUrcUu01@+lUIw)e07sTFXaFJv?SlXj^<D<G4gfwPv342&5%*pO
zwGIGBbTMcEBF>;<01^3K2DJ_V07r%~XaFM4pke?K{ayyO4gdf?BF>;<01^IP2DJ_V
z0000!BF>;<01*#g2DJ_V005r>qVEC$5z1Z$wGIFQ05WI*A5#WV01?W6UU$$A03TOV
zcK{K}UIw)e08D8B5zAh>s15)ktqWZM5zAh>s15)=BA@@v01?ezchC+1OmuB<01?h!
z2DJ_V005r>Prrfy5xiaowGIFQH!^4dA7}U501>=i2DJ_VA0>9#01>=i2DJ_VB5fpy
z01>@j2DJ_V0G|O@C~p9N5qn++wGIFQ?=olrA0>9#01<m$2DJ_VAFJ0`01<m$2DJ_V
zA1j4@01<m$IsgCw)&gQ-WB?JeULtDcA^;J5UIw)e03v7i+W--LUIw)e03wj;1pyI$
zUIw)e03x{a1pyI%UIw)e03v&*kpK~ZUIw)e07ud?XaFMXdp`g>5w>0iwGIG2BF>;<
z01>uc2DJ_V002HB>w7-{5w~6jwGIG2BF>;<01>xd2DJ_V0002P0iCGY01-f!s4jpZ
ze?VRawGIF%3oig%SCrKN5kOwh1ONbG3oig%=}yZ45kOwh1ONc73oig%R5Eh_5kX!;
z0002_3oig%i-=SJ5kX$35H|ot3@-rI0%Bog01<RvLW#pi01-o82DJ_VdkilCQ=9iJ
z01-r9#Sdrz1CfA!LQ?cA01-uAx~L8Smq;&w9|3ijXfJ>u0d=>DFMtXGe*#`Q00001
z2sUT{A3dJ#01*OS2DJ_Vivg=A=>QQdUW)-Y1S0?uFkVb@Wpn@$1711+001IPH?sf{
z1YQQU4gexnI0pd{1zrZV4geyoJ`Mp923`iW4geyA+=~Da2VMrX4gexzjP3vt2wn!Y
z4gey+Q!W4z30?-Z4gexy45gL;5ei-gwGIGBi<e$8fFA*?mwqvT9tB6sHD~~roiTtL
zMMvv3XaFMXdp`gXI9>*|4gfwP&Y)rd5jb83wGIFP06rq?dp`gXIbH^}4gfwP&Y)rd
z5jkE4wGIFP006@QX%?XX5%QP8F@PaT@?8eC4gfL@F92L6cG&<C@?8eC4ghNnF92b3
zWpn@$^IbXs007_(F91S#qNxB8^j!wE4ge|+F92R(2LJ%KHfR7s>w7-{5&xGgGJqff
z|CdcNfM@~fm$ov1AuZ`$2DJ_V-wrPTT%It801@e32DJ_VXAds`7G*eS07#<+00000
z0AL0H01@q%1T%oG0o9kXGk|M<%v}bx4gdqsIA{PPZ(?cy5zJi%wGIFuo-l_15zJi%
zwGIFxml8n$5zSo&wGIGGbZ`I>&RuuV4gfw(W&jb~UCO);06t7|Wpe-#-dzT@4gfwP
z&Y)rd5#L<~wGIFP0001=0hgoK01?4m2DJ_V0r@y+03&p801?4mchC-h03V((hX4`5
zT?Vxd005r>zo&cv5%XL+00003J2_|oBX44A01@+C2DJ_VAD%FW01@+C2DJ_VA8jOv
z01@+C2DJ_VOmuJn5%gSl&<+3qp8-NGDgY68Tn4oc009*`XaFN`a4`T8cU%Uw4ge!>
za54Z9cU%Uw4gen>sEYu95qDhW0000Vo#CJW5qDfV0000V)J;wR5qDfV0001s0Rie$
z01@0=NAfvn08C;45r<su0000?VE_?`T;%`&08DrQ5r|yn0000?Y;0lx5sF+*q7DE|
zY;0iw5sO?!q7DE|Y;1S{5sO?!q7DEeI4+9-5sh5o4QK!&R5Eja01=K{#;6Ve06t7&
z01@R}?EnA(Okn^K=3FWO001J1!$$xS=UgfP00000p8=w|4gnD{Tsi;%00Y!IXaFC{
zJyHM>GF%3=4gepN7dQYBGF-Z-4geqX*Sr7`GF<7l4genl0T2KYGhDsJ4geqe(82%_
zGhDsJ4genwc`yKf5i?x9#SQ=;By)ZM5i?vm0000Ypbkd>5j0!|wGIGBvpQ%1OlxI%
z01-D_2DJ_VB5fpy01-D_2DJ_VJ|fkwk^m7vTn4oc0000-#X4vJOmJ^=01;JOx~L8S
z06t7`Z*u?<TU@%R4gdfEp8?TLD*zD%Tn4oc00TceXaFC7&Mv+H5d~ZZwGIFuD|3DT
z5d~a20000VDORrl5d~ZZwGIFuk{37t5e8i8whjOvB6EHK5e8g)sty1?A{g7101+cx
z2DJ_VJ|fPbVgL~=Tn4oc00000p8-nGg8&ifTRH#$00DeFXaFD8u95%|>01W14geoV
zNq_(m>07;j#SQ=;_Yv0s5$RjK#SQ=;gYKyS5$any0000V&aRRG5$any0001s0b(&#
z01^3H)&gQ-WB?KTTOttY)c_IeTe_$Y03t|9fB+HfTL!fb03!Gi*8mahTL!fb03sOM
zl>ia!TL!fb03zn-KL8Q!TfMvv07vaRXaFM4pke?&5%60EwGIFPNB=u$03yzyVgM2K
zTL!fb002iHJZJzS>w7-{5&c^RwGIG2BF>;<01^FL2DJ_V002inJZJ!y5jcPwe@9<D
zXaFMXdp`gX|62yN4gfwP&Y)rd5&v5TwGIFP06rq?dp`gX09*#O4gfwP&Y)rd5dd5U
zwGIFP005r>P`Xe65z$*Z00003<veHrAJwjs01?t#2DJ_VA4f@m01?t#y~PdyANLX0
z01?t#y~PdyAI`3l01?t#IsgCwe<DapfB+HGTL!fb03!Gi*8maITL!fb06rqlpke?K
z+*<~<4gdfE*8(WFrvMSgTLB6^XaFO001?Jp2DJ_V9}wx)01?Jpx~L8S0H_6uToM2g
z!CM2L0T4h601>rYIsgCw0Zlz<03Qrff&dY<TL!fb03WmlEC3O;Te_$Ye*iur&Y)rd
z5w}|gwGIFP0M`P30v7-guUi9XJ!k+Qv<55y5wBais15)C*8+KS3IP$QTLXhVXaFA&
z>D2%cs9U<I4gdh30nBJ|01=p52DJ_V0sB2@03&Z=Y5)<KTL!fb03Rt<uK*F4TL!fb
z0M-IxVPpUiqgx`eP&@z;f0<hbwGIFxiNi+#5t&;CwGIFxsx8g{5t>^%0000->^*1z
zBF>;<01=&A2DJ_V06rqlpke?KqFV;F4gdfEp8<iv;s6miTRH#$009s`XaFN`Vrl>p
zI9mp_4gen|cG&<CI9mp_4gepqP&@z;I9t8N4giY*cw-#^5s6zOe|@5<01-J`2DJ_V
zOmuJn5j$IV&<+41ml8n$5j<N4wGIGGW&jaATgto+03tlWa{v)OTfM~&03zO=a{v)O
zTL!fb03zy=1ppC0TRH#$07soZXaFM4pke?KURwsW4gdg0uRdr1BF>;<01;_h2DJ_V
z07uV0XaFMXdp`gXUTRwgwGIG2BF>;<01;|i2DJ_V002kfK4<_U&Y)rd5p7!rwGIFP
zNAo^t03yzyVgM0!TL!fb002HB3+M;|5sX^~wGIG2BF>;<01=K`2DJ_V0002P0r)?0
z01*|J$vS``Iu~0s&kg{e5-$Kkk-U5W5g1zrwGII25-$K=KpFr5UqEO8LZzQa01+Eo
z2DJ_V9uqGBLO=5i0TCOQBs+j40U?)3JAgF-BA0(VfFJ=Pm!La<G9n{e2DJ_V%@Z#G
zD|XKX0000002crN0000007A~7VgL~%m-IV;E&&>sBRqgXRRCKCwGIFS%Rp!VAKekq
z01*IN2DJ_V0G|Q-W4Qnk^;!nC4gdoHL1+LYZ(?cy5%pRIwGIFyYh`%=5%pRIwGIGG
zZ2%GWT3Edf006@Q2)Gph5#yJ+Jb)or<5~u_4gh}?F92K+>D2%c<5~u_4gh%-F91Sv
zctrpa<XQ%`4geGuF90H+Iza#t<yyL^4gg_o01@U|QoRlUHx@4d7J?0E05re_00000
z0N@e;01@k#F+G4<0mYY$J%Igxwps?Y4gdpeLuddaZvYXtS_ZWa03&Gt5w=>os15)h
zVGm>g5w==70000VPFKzV5w}`80000V^%oxi5x8190000-H$!LuOl|-XyIQ)a4gfw(
zasUy%S_ZWa0000-Q$uI~Ol|-X!&<ti4gfw(asUy<S_ZWa0000!Ol<&v5z1Opy$%3A
zOl$xV&04ys4gdfE0G|Oa00{sQl3E6}4gdqgLuddaZvYXJS_ZWa03&Gt5t3TEs15)h
zPFKzV5tCXv0000V^%oxi5tLdw0000Y#y<)H5tdrIs15)=OmYAbn_3374gdfE!vQ9i
zod6MwS_ZWa05GTp0001&hd+Q6Hj7#YwGIII7%u>AY5)<7S_ZWa07Mxt0AX?f5sX>}
zwGIHI87}}{Fb4ntIz(syLe8LK01=Ou(LaDae;`@`QAKC~ABK}f01+Ts2DJ_VA8jOv
z01+Ts2DJ_VBXnW_5g=LywGIFu1MGYN5g=LywGIH*0%Bog01<>*BHKQI01+Wt2DJ_V
zN9aUo03wgIB>)jLS_ZWa07uJ2XaFMnJGuZ7I9fUY0000!BE%hA01-o42DJ_VJ|fPb
z1Y!UYPM48DfD?aUS_ZWa002icMQ8vb0F{IQ5qnxR&kg`0aD-C;5qw$(wGIG2BI|oU
z01<#%2DJ_VJ|fPbVgM0<S_ZWa00000J|gRTKL8PhS_ZWa06rqlpke?Kg<1x+4gdfE
z0M`QZm%soK;aLNZMQ8vY0F{IQ5#d=h&kg_|hLc4A5#fJXy~PdyA8jOv01@F?y~Pdy
zBXnW_5#d?A#SQ=f*8<lB7yuE~Sp!W*XaFAom4pBh)mb#p4gep9lSKd#)ma9$4geo*
zB!~bJ)ma9$4ge!`VgM1<Sq8NZ03riiZvYY2Sq8NZ03x*l<p2@ZSq8NZ03vXNQveax
zSq8NZ07riiMrZ&c>w7-{5#3n^wGIG2BF>;<01@3;2DJ_V002iYMrZ&c>w7-{5#Ct_
zwGIG2BF>;<01@6<2DJ_V002HB>w7-{5#Lz`wGIG2BF>;<01@9=2DJ_V0001=0eBHu
z01>iT2DJ_V0gy&$03RiG*#Hr;Sq8NZ03ZH|2LV|TvRRNuXaFCAuloQIvRS>w4geyF
z!$$xSw^;_Y4gdfX2*sxk0G|ODD;59|dRYdw4gdqeMrZ&Z-4W0L5qen$wGIFyVRB~x
z5qen$wGIFP!vW}y%K#B@mwG~gAuVuO2DJ_V{24C*ZDDd}01<Fm2DJ_Vgc>gZB8kIC
z01<In2DJ_VLc|?g01<MR$U=Z50e6?}LVzp*c9$AMfc#KgSq8NZ00XT^XaFDG5zqh;
zTv-OS4ge!za%TV$Tv-OS4geyY^bP<KU0DXT4geyF!$$xSU0DXT4gex`1YiIWURh|z
z4gdhd0h~R>01-ZyW<-D?N<LWzwGIHf94`P|>7@w)5k6T4wGIFo9WMY}n)oUJ5k6T4
zwGIHW9WMZj0W6s{01;MMLc|?g01-b~2DJ_Vo*pj%LW#pi01-cz(L{hG0YaDdM1VdW
zNLi{%XaEC|uzwdmO=tkc#03BV0001x6951aNtaPYfN=p-m%~MXH33zZ`9*+xe=AuA
zwGIFS3QK4JAKekq01+!$2DJ_VBVlr901+!$2DJ_VBWrg65i3~+wGIFxiNi+#5iD5-
zwGIFx^?L*X5iMC-wGIFPp8+x4I{*>USO&EY009e3XaFDG5zqh;(O3qx4gep-9a{hq
z(OA934geo5j8gy+(O9~u4geode|rM}5z$x%wGIG_0W6s{01+ZtOk-~V5z<(85H|oK
zNYe@c5z|-(wGIFx>^&F&5!6@)wGIFxA{%%B5!F}*wGIFxBfLHU5!YA-wGIFxt)71X
z5!qM<wGIGGYXA}1SiQv#08C^65!zV2#SQ>WZe#!v+gQ4&4ggGUYXA}4QdqjE4geyF
z!$$xS-B<>-4gg2)OK1R0aBp(}5%*ZSs15*3X#f%TSh}bV002HB>w7-{5g}OywGIG2
zBF>;<01+Ws2DJ_V0002P0e?2Y01>p8#7BT2f3#S35H|pUATI!1#2s4z5wuvn#SQ@A
zATI!IZeaisv{<^R4gh;0F92Lf(+U6)wO9tV4gl~WF92a?01>uW<r`=KPa-b>LI9P7
z01>xXG|vtIm?AF#LUY6B01>!Yy~Pdy;UX^pVQBynxmdcW4gfqOF91S`!$$xSx>yFa
zmt#nPApy*ng-C#Lf1OwcwGIFSrA=r6AG_BE01=&72DJ_VAH*G701=&7y~PdyA1#bi
z01=&7x~L8SBHa<t01=*8Dj{e9BE*}d01=;9XvYo!BAoON01=>A2DJ_VBA6L101=^B
zIsgCw0G|PAS+4*QHds0U0004wPG|rh3Irbj5jR){wGIFue>3Qr01-G?j!tL*ABnHZ
z01-J@2DJ_VAF;kV01-M^y~PdyA8%(-01-P_x~L8SA7_sY01-S`&;$SgivjSD1pyI`
zSc?I%jWYldidc&Qb5*JU5r<gT0%Bog01=m1OlxI%01-Y|2DJ_VB5fpy01-Y|2DJ_V
zOmJ^=01-b}f4Zm+03s!J*#HqhSdLC;03x|{-~bUpSdLC;07r37XaG!OZEye)WLP=?
z0000!BImW301<Lnj!tL*K1^e6Z~zf>SULaz0000069~P<4gl8zIIk!G5hz#zyH02T
zAJ#e001+uz&;$SgA1Xvf01+x!2DJ_V0G|QEgCPJBe*joI0000198YKf9~e0(01*LL
z2DJ_VA7_sY01*OM&;$SgA8DH?01*RNIsgCwivjSD1pyHoSc?I%jWYld7FbMcWqAM*
z1y}~P4ggGWZ*u?<23WeN4gdhx0{IN*01@U_0e??u03X&l&;SwVSI`6i03T$!Pyi9=
zS2_Rye*hnMd=CH-=~ud_4gen%{0;yS>Q}m`4geoVKNbKH>sPv{4geys1}p#(?N_>}
z4gezk4YmLg?pM$R0001=0a~}h01?Yq2DJ_V1LjX?03W^@PXH0jSI`6i03Qd6VgM1$
zSGuSU03Uj5+5i#ESGuSU03YbnT>ufxSGuSUAOIimgh~Jr%~#L_0019)Z3h4m%~#L_
z001BA8tMQM&X<u(fGvN$R|d5X00A*jXaFB2cG&<Cy;lac4geyF!$$xSzE=jd4gdh3
z0l`Ph01>cP2DJ_V17}cZ03&Z=Y5)<iR|d5X03V((hX4_<R|d5X08Dgn01>fQchC+1
z0G|OrfKvbwwNyF)0004aTxb9v?Pihy5w%nXwGIG_0f)o)01<zYSBn7}`pp0lgI6Mb
zNZJ4qxl{(V4geyDNZJ4qx>N?W4geyNNZJ4qyHp0X4geyXNZJ4qyi^9Y4geyhNZJ4q
zy;KIZ4gg2xP-p-m`Rg<Q5f4=cwGIFPNB2-@03yzyVgL~^RR*;V002h`QD^`n&Y)rd
z5j<4}wGIFPM;m`pXaFM4pke?KOjQQ84gdg0Em3FyBF>;<01;4C2DJ_V07pMjXaFM4
zpke?KQB?-D4gdg0QBi0BBF>;<01;tT2DJ_V07qj{XaFM4pke?KV^s#V4gdg0bx~*l
zBF>;<01<Ch2DJ_V07s`$XaH0Ai3b4@cvZ!x4gd)QP{2_)0000008@Og`v4JmRR*;V
z00{#~z&8K@00000N4ZgG03yzyVgM0*RR*;V002kpQD^{D_=yJr5q(v~rw#xK13<tx
z0000008@Og`v4JrmyAt-8-IaS2DJ_V07nl}XaFM4pke?KgjEK$4gdg0K~iV{Q}~Go
z0TG2&#itGc2?H>|Hvj+t002{buloQIg;fT%4gd)QD8M%W0000007p|&XaFM4pke?K
ziB$%*4gdg0Wm0GWBF>;<01=E;2DJ_V07sWnXaH0Ai3b4@ja9{`4u1d%10cXR00000
z08@Og`v4J*RR*;V00{#az&8K@00000N2yY103yzyVgM15RR*;V002k3QfL4o&Y)rd
z5t3B~wGIFPN6k`b03yzyVgM19RR*;V002kdQfL4o&Y)rd5tda3wGIFPNAXf<03yzy
zVgM1DRR*;V002h=Q&ngHBF>;<01=#32DJ_V07n&5XaFM4pke?KomB?44gdg0CsSwu
zBF>;<01=;62DJ_V07qF<XaH0Ai3b4@s8z+M4gd)Q5WqJ800000Q+%)c01>E{`c8lw
zDXdinwGIFPN1an>08{vh2LTbSRmG<c00{#Cz&8K@0000}e6RZe5v`X#Pk<YLu~i1O
z4gdg0!Bc1eBF>;<01>oR2DJ_V07vsvXaH0Ai3b4@wN=HZ4gd)Q@V_?z00000Q+%)c
z01>rS2DJ_V2?OZAHvj+t0000-1ypDNBF>;<01>%W2DJ_V07n;8XaFM4pke?Kyj2FZ
z4gdg0NmOV6Q}~Go0TI1b#itGu00{%&zc&B?0000}e6RZe5xtk{Pk<YL!Bqye4gdg0
zZB%FgBF>;<01?7f2DJ_V07rjRXaFM4pke?K#Z?Bi4gdg0kyK~^BF>;<01?Jj2DJ_V
z07s)#XaFM4pke?K$yElm4gdg0wNz*TBF>;<01?bp2DJ_V07u7EXaFM4pke?K%~b}q
z4gdg0*;Hr%BF>;<01?k@RR*;V002h|RcHWH_=yJr5!hA5rw#xK1JJ)W0000008@Og
z`v4KxRR*;V00{%gzc&B?00000M;%pY03yzyVgM1`RR*;V002i%RcHWH_=yJr5#3eA
zrw#xK1HivG0000008@Og`v4K$m+(-48ztdY2DJ_V07r9GXaFM4pke?K<W&Z>4gdg0
zrB!GEQ}~Go0TJbwFj0UNSLIa(wGIFY1E{|@00000002j~RcHVr&Y)rd5$RP1wGIFP
zN6A%a03yzyVgM2BRR*;V002k(RcHWH_=yJr5$#pQrw#xK1E9Y*0000008@Og`v4K`
zmz+_68-MXt2DJ_V07oBIXaFM4pke?K@>K@44gdg0F;-{*BF>;<01@?72DJ_V07pYs
zXaFM4pke?K_EiS84gdg0RaR&KBF>;<01^3B2DJ_V07qw5XaFM4pke?K{8a|E4gdg0
zc~)ouBF>;<01^FF2DJ_V07r{fXaFM4pke?K|0q=kwGIFPN4-{P08{vh2LTZXR>h|d
z00{$-zc&B?0000}e6RZe5eS!FQh*zO3|0oU4gdg00as`MQ}~Go0TB&W#itGc2?Kz?
zHvj+t002{buloQI4ORxV4gd)Qc)vFQ0000007ny7XaFM4pke?K5mpAZ4gdg0C0A$w
zBF>;<01*^c2DJ_V07q9>XaH0Ai3b4@6;{Qk4gd)QaKASI00000Q+%)c01*{dRtB{W
z00{$Vzc&B?00000M`>4R03yzyVgL~tRtB{W002jOS7-nt&Y)rd5gb+qwGIFPN3B<A
z08{vh2LTZsR>h|d00{$Nzc&B?0000}e6RZe5gnIKQ-B+PAyx*p4gdg0(N|~yBF>;<
z01+Zq2DJ_V07v6jXaFM4pke?KB~}Kt4gdg0^;c*BBF>;<01+lu2DJ_V07nN{XaFM4
zpke?KDOLux4gdg08CYllBF>;<01+%!2DJ_V07olWXaFM4pke?KEmj7#4gdg0Jy>V}
zBF>;<01+>LRtB{W002i%SZDwu&Y)rd5j$1}wGIFPM`2iK03yzyVgM0GRtB{W002jG
zSZDwu&Y)rd5mQzMwGIFPM}=5u03yzyVgM0eRtB{W002j~SZDxK_=yJr5rS65rw#xK
z15m#=0000008@Og`v4JwRtB{W00{#~zc&B?0000-07uDKXaFM4pke?KhE@i(4gdg0
z+gNA-BF>;<01=2*2DJ_V07ng3XaH0Ai3b4@kygd04gd)QK)*Kt00000Q+%)c01=Uw
zu2g^<f0b4SwGIFPM=@Dw03yzyVgM1BRtB{W002j0S!e)L_=yJr5ujGZrw#xK12Dfg
z0000008@Og`v4K3RtB{W00{#qzc&B?00000M|D|f03yzyVgM1ORtB{W002jaS!e(v
z&Y)rd5vEoKwGIFPN10h@03yzyVgM1lRtB{We*geSt669OBF>;<01>=a2DJ_V07t!9
zXaFM4pke?Ky;cUb4gdg0&sk^yBF>;<01>`c2DJ_V07n5@XaH0Ai3b4@$5zFs4gd)Q
zAip;N00000Q+%)c01?Ml2DJ_V2?H3vHvj+t0000-6Iy5hBF>;<01@3*2DJ_V07oTS
zXJ`N-&Y)rd5$RS2wGIFPM>kq%03yzyVgM29RtB{W002ixT4(?w&Y)rd5$je4wGIFP
zM_XEG03yzyVgM2BRtB{W002jgT4(@M_=yJr5%pHZrw#xK0}#JA0000008@Og`v4L3
zmo8R-8-EK|2DJ_V07tP}XaFM4pke?K7gq+g4gdg0!&+znBF>;<01+5h2DJ_V07unY
zXaFM4pke?K8CM3i4gdg0=UQk0BF>;<01+Bj2DJ_V07n^HXaH0Ai3b4@BUi<z4gd)Q
z0KYc?00000Q+%)c01+cs2DJ_V2?O}PHvj+t0GG&CfDL~+R|d5X002ilTWA0x&Y)rd
z5k*%9wGIFPM^9U503yzyVgM0FR|d5X002i}TWA0x&Y)rd5l2@BwGIFPM{`?f03yzy
zVgM0HR|d5X002jYTWA0x&Y)rd5mZ+OwGIFPN0(b@03yzyVgM0fR|d5X002j+TWA0x
z&Y)rd5o{Y*2DJ_V07tu9XaFM4pke?Kc2@?q4gdg0+gq2fSAZS@`&*aKSAZS?8<*}^
zfF6HGJ6vc0BI|oU01<~*2DJ_VJ|fPbVgM0`R|d5X0000-TU=-WBI|oU01=2+2DJ_V
zJ|fPbVgM0{R|d5X0000!B7_kT01=H>2DJ_VJ|fPbVgM11R|d5X00000p8=s4{Qwc2
zR0g#U00I77XaFBc0+s*~om2+34ge!$WoCZ>5uH^2TxbB+0%Bog01>ZLB4_v801=*4
z2DJ_VBEDu#01=*42DJ_VBES#~0TG^52DJ_VN8wy(03yzyVgM1QR0g#U002HB>w7-{
z5wKJSwGIG2BF>;<01>cM2DJ_V0000J2p`D~0G|OlIW7PZj#N4T0004bU1$IwNdkYC
z01=K<2DJ_VAHHTy01=K<2DJ_VAC+$a01=K<&;$SgAAm+=01=K<&;$Sg9|c)L01=K<
zIsgCwB5fpy01=N=2DJ_VB8kIC01=Q>IsgCwJ|fPbVgM19R0g#U0001=0c=|s01<ao
zIsgCw0j^zW08{M0CIAtDRIbhr00|xgSim;`00000B4nvl01<&yuFehs0I3Dq0`>qA
za+haWfIoj#R5}0v00E){3!upW5lvJ&00001p8>63g8&g3R5}0v00F38XaI`=8v4xu
z5l2)aeMs5>5gSwnwGIGBLtbbABF>;<01+ru2DJ_V07q3`XaFM4pke?KEmQ`z4gdg0
zXI^LkBF>;<01+}&2DJ_V07rRVXaFM4pke?KKU9APwGIFPM~hx)03yzyVgM0DR0g#U
z002HB>w7-{5lB=9wGIG2BF>;<01-%32DJ_V0001=0qo)201*UKIsgCw0mNQt0E+<{
z`pp0l6jT7B0}Av501^CCIsgCw0ipxGp5Oox^iw(j0004^12TpI0TJv|IsgCw0iOXR
zv8I0j5!+Kb000013SVddivb$?%>WVLQvja<G1r>_5wB7@00001CSPa(ivb$?%>WVB
zQvja<NNEHC5us8#00001LSJYAivb$?%>WUrQUIR;x;&r&5tUIo00001GGJ%`BW(Z?
zl~D$@4gl5yVqs(e5wlTDX#f$HQ91ws08D>!Wpn@$mr*(Z002yL01=o`WdQ&HB4z!c
z01=r{2DJ_VB73Ki01=u|2DJ_VA|Q6z01=!~2DJ_VA|-a&01=*12DJ_VBD>bl01=;2
z2DJ_VN7Y|w03yzyVgM1VQ3kaR002kzUuXa#>w7-{5wuYTwGIG2BF>;<01>oN2DN_<
z0000-6<}xpBI|oU01>rO2DJ_VJ|fPbVgM1fQ3kaR0000!BI|oU01>xQ2DJ_VJ|fPb
zVgM1hQ3kaR00000p8=UGJ^&F_Q91ws009$WXaFB&{h$C5RZ#}D4ge!|01;JD2DJ_V
zOkr$n01;PF2DJ_VBHo^J01;SG2DN_<03s!J*#Hq(Q3kaR03vN9hyW2-Q3kaR03tQo
zIsg$_Q91ws08DUia{v)qQ91ws08C|aasUxpQ91ws07t-JXaFM4pke?KY*7Za4gdg0
z(qL!+BF>;<01<;x2DJ_V07v9tXaFM4pke?KjZp@*4gdg01Yu|ZBI|oU01<zVQ3kaR
z06rqlpke?Kj!_1+4gdfEJ|fPbVgM16Q3kaR0001=0i}u<01+lpIsgCw0drw!03&Yz
z5hhUvwGIFucSZXE5hhUvwGIFuC3e{W5hhV(0RRBj0%Bog01;48Q$UlO01+oqb75!z
z2?M~ZHvj+t001J*bUOeMLs6F>U4R||h?hQHfExjbms(wbBp{Da2DJ_V0+B#d{k!S_
z5sy#?wGIFRkx1bIWprWy5s#OiU4R+^l9#z%fFJ=Mm)c!`r(AhZ2DJ_VAfyKX00000
zMgaf-000000i1zlXaH?=01<gmchC+1aU?GQZD9Zrc~Azm4gk|6F92<801<gm2DJ_V
zH6<?qss&<UWB?I^P_Lv1000000AX(c5qg*1UVtP4gqH$efGh!lmoHy{JOSgEY+r!q
z0pXWaV1VU+3Q*XF0{{WbW@rE(j^Dum5eiTSwGIG_0f)o)01<Ogivb$?%>WT@P$I43
z!2l5pP}qh803xX{!vGNsPzJRQ03r#lp8ydJP}qh803v-z+5iy`PzJRQ03wG-+5iy{
zP`aoN03wk{+5iy|P#v`n03w@6+5iy}PzJRQ03xM-NZJ4q6HtJJ0{}-OW@rE+k-U5W
z5j#)@wGIFPM^R>I08DgkZ~zfRPzJRQ06rq?dp`gXL{J8`4gfwP&Y)rd5kyc1wGIFP
z002j0W@rE+&Y)rd5m`_MwGIFPM}uZ)03z#qKL8PLPzJRQ06rqlpke?KZ%_uc4gdfE
zN26wcXaFMXdp`gXa8L%d4gfwP&Y)rd5pYljwGIFP07t`SXaFMXdp`gXbWjGh4gfwP
z&Y)rd5p+-nwGIFP06qlgz-Ise0G|QLFQWhvn@@m*0{{VLXlMW*j^Dum5t~m2wGIG_
z0f)o)01^97ivby#H2@LWPm2LrF;xH&)K80l0UG+v01@?1BK>=R01=!|2DJ_VB4p;{
z01=%}2DJ_VA_=aa01=)~fP@18B7I2O01=>12DJ_VB8N!Y01=^22DJ_VB9Tbi01={3
z2DJ_VBAZCs01=~49kmVsBBe;$01>25LI3~&B5^XW01>56LI3~&N0(=403ul<iU1Kv
zuulfH4gexlGIIbCu}`|F4gdg0$!BN)A`t1-01?Pfx~L8SN4jTd03yzyVgM1+PX@IP
z002HB&Y)rd5!FuywGIFP07vO(m%U?v9)AgFXaFMXdp`gX_fH134gfwP&Y)rd5%*69
zwGIFP07ofkXaFMXdp`gX_)iA44gfwP&Y)rd5%^CAwGIFP07pq^XaFMXdp`gX{7(k8
z4gfwP&Y)rd5&TaEwGIFP06rq?dp`gX{Z9t94gfwP&Y)rd5&cgFwGIFP006@Qo&{n4
z01+gYA7p?be<V)^wGIH`C@%oE0f)o)01=H(e%S~B00000ivb$?%>WUIPgAYp!2l5@
zPg&Ok00WV*enP1*!vGN`PX@IP09Gk40740_p8ydjPg&Ok08J_{07898+5izLPX@IP
z0I@1B078dI+5izMPr9fM0I4f4078*S+5izNPaU-mQ2?7PF91TDNZJ4qEKdft4gfhV
zF91`eNZJ4qEl*8=0ssS%(0*R9YXbnEX=ngKk-U5W5miqHwGIF@E-wHU01#*ZYt;w<
z00000;A;Z_5m%QGWq?}&UYBfTfFJ>0myKnBAOT*Nu4RBX0dtq;Wq>sSiI*H^fFJ>h
zmpx{HCJu^E2DJ_VJ|fPbVgM0}mvLr*A_0z<lxBb+0gjioW`HdLCzs=9fLsABmq2HL
zmjNZ0*k^!T0cf{1Xn;Ne0ce+WX@Ft@X}8yDfDr-#X}9TXfP4!9X_sbqfLQ@)m%w*`
zT>&kZ9(aIV0VTJ5c!2l;0WG(6dVn|qf96gGwGIFQ=zwSdAFblS01@U+6VEmPivfqj
z_W%(VPm2K>`pp0l5l<pO+Q9%3=T1TZ001HhuAcxA=uQT;4gey3NZJ4q>P`l=4geyD
zNZJ4q>rO%d001J9NZJ4q>`p=e001JJNZJ4q?M^}f001JTNZJ4q?oI}^4gg1)e}HHJ
zBI|oU01^032DJ_VJ|fPbVgM2NP6o9O0000-tAJ<#BF>;<01*OD2DJ_V07uJ!XaFMX
zdp`gX5>E!T4gfwP&Y)rd5fV=ZwGIFP06rq?dp`gX6Hf-U4gfwP&Y)rd5fe`awGIFP
z005r>w)^`45wK1MwGIFQ+kt2Re;=*l!2l7kPGsv20E+>K!}kCY;7*GH8v4xu5!+58
zK-$3o5wT7}0000Y39g?25wcDOwGIFxeMs5>5wuPQwGIFxhe+B05w%W20000Ykx1GA
z5w=c30000Yn@HLK5w}i40000YrAXQU5x7nUwGIGBjDctXBI|oU01?7Ye+IP<06rql
zpke?K!cGRY4gdfEN1lOb03yzyVgM1$P6o9O002k6foK3C>w7-{5!_A&wGIG2BF>;<
z01@0y2DJ_V002HB>w7-{5#3G(wGIG2BF>;<01@3z2DJ_V0001=0e)x)0TDS)@WKTE
z0UCp703VLu!2l6CP6o9Oe*lXChr{;(5u#3u0UG+v01=%|BCX=V01-M)@WKTEBB?OL
z01-P*2DJ_VA_=aa01-S+@WKTEB7I2O01-Y;2DJ_VB8N!Y01-b<x~L8SB9Tbi01-e=
z9kmVsBAZCs01-h>2DJ_VBBe;$01-k?_Dul*M}2~503wmRd;k$@e@+Iq4gdg0tAc0%
zOmuB<01<Oe2DJ_VJ|gRTKL8PPP6o9O06rqlpke?Kb4~`e4gdfE07t!oXaFM4pke?K
zh)xEz4gdg0-GXQUBI|oU01=)}2DJ_VJ|fPbVgM1IP6o9O0000-{eox!BI|oU01=-~
z2DJ_VJ|fPbVgM1Je@+Iq4gdfEJ|gRTKL8P<P6o9O06rqlpke?KqfQ334gdfE0G|P#
z7n1-H%1urR0{{V8glGUCj^Dum5z0*lwGIG_0f)o)01-1zivb$?%>WTDP9m-1!2l7<
zO->2}03xX{!vGP?O$N0N03r#lp8yffO->2}03v-z+5i#He@zCp4geyDNZJ4q&`r9i
z4geyNNZJ4q(M=t-4geyXNZJ4q(oF`n4geyhNZJ4q(@hxr0su$6gJ=LEk-U5W5&KOB
zwGIFPN9u!U08DgkZ~zemP6o9O06rq?dp`gX1WpFE4gfwP&Y)rd5d=;KwGIFP002k)
zgJ=LE&Y)rde-Rl@2DJ_V07o2zXaFMXdp`gXFHQ!v4gfwP&Y)rd5id>#wGIFP07pE8
zXaFMXdp`gXFir-w4gfwP&Y)rd5im{$wGIFP06rq?dp`gXG)@M!4gfwP&Y)rd5j0K)
zwGIFP005r>Ucc!85nD|d`vL#~qlIVyACBL_01;bFe+IP<0E+>K!}kCY#7&C<8v4xu
z5x-3$t>VD|5nN3e`vL$WsW8I;5nW9NwGIFx39g?25nfFg`vL$WeMs5>5nxRQwGIFx
zhe+B05n)Zbs15)kkx1GA5n@dpwGIFxn@HLK5o1jTwGIFxrAXQU5oAq50000-`-Er!
zB9Xj&e*h7TO$N0N002iRg=hdwbZu||5tU5_wGIG2BI|oU01=f<2DJ_VJ|fPbVgM19
zO$N0N00000M>~aR03yzyVgM1UO$N0N002i@g=hdG>w7-{5x`9bwGIG2BF>;<01?1V
z2DJ_V002jOg=hdG>w7-{5y4FcwGIG2BF>;<e*h7|O$N0N0000-n}uiqBI|oU01?Ga
z2DJ_VJ|fPbVgM1vO$N0N0000!1mV+X0001=0Um*@01+Kc2DJ_V0hWen03WU5!2l5*
zO;4IO0E+>K!}kCYO-+jd8v4xu5lBrUK-$3o5gtuK0000Y39g?25g$zkwGIFxeMs5>
ze-R-~2DJ_VB8N!Y01+ZhLI3~&B9Tbi01+ciLI3~&BAZCs01+fjLI3~&BBe;$01+ik
z2DJ_VM@NQe03z#qKL8OgO$N0N06rqlpke?KFHHuu4gdfEM_Gnw03yzyVgM02O$N0N
z002jMhG+mH>w7-{5lKx3wGIG2BF>;<e*h6lO$N0N0000!BI|oU01-+}2DJ_VJ|fPb
zVgM0JO$N0N00000p8+GiD*zGaOa`?M00D`IXaFCr;=uqB=S)TpHvo$Phr{;(5f@F1
z0UG+v01*;RB0$=~01@a+LI3~&A_=aa01@d-2DJ_VB7I2O01@j<2DJ_VB8N!Ye*h8e
zOhNzv03wk{+5i#lOhNzv03w@6+5i#mOhNzv03xMG+5i#nOa`?M07p89XaFMXdp`gX
z`Ai124gfwP&Y)rd5&298wGIFP07p!RXaFM4pke?K15F0C4gdg0Y=>w7BI|oU01*>S
z2DJ_VJ|fPbVgL~nO$N0N0000!e<JI9KL8OFO$N0N06rqlpke?K6io)T4gdfE0G|OH
zWO)D)u}lWF4gdjsh-d&Gt>VD|5wT2vZx8^B0f)o)01@F#ivb$?%>WVHOd>$q!2l7m
zOhNzv03r#lp8yfFOa`?M03v-z+5i!?Oa`?M03wG-+5i!@OhNzv03wk{f7$>Mw@g9+
z001JJNZJ4qxJ*I-001JTNZJ4qxl9JN4gg0jh-d&J>w7-{5yMOdwGIG2BF>;<01?AX
z2DJ_V002inh-d&J&Y)rd5zR~nwGIFPM_-6&03z#qKL8QkOa`?M06rqlpke?K-Ao3x
z4gdfEJ|gRTKL8QlOa`?M82~;a&Y)rd5#CG&wGIFP006@Qzn#wj5qp<4gn%J`drUQ^
zHUP{-F95ayhr{;(5vxr2*#!Uq0001s0UG+v01>84LO|NV01<plLI3~&D@88=fdUDx
zp8yekOa`?M002ULNZJ4qe@q6o4gj1*F90HkNZJ4qfJ{OF002UfNZJ4qflNXG0019G
zF91TDNZJ4qf=ogH003x4F91R)rAXQU5ra$ywGIHqMlS$IQ;BE*BI|oU01=H$2DJ_V
zJ|fPbVgM11m;Ho*A_13|AccS_7N<-GwGIH3M=t;+FfjlCLe8LK01>B`Uxk250jQU%
zg@7OdsF%xyfH47(m;Hr+E&-I6EQWw40hX6jhJc9yMVH-%fFXZHOvafu0J%vo0JZ^#
z!}kCYbxgS11poj5004^t8v4xu5pYaGK-$3o5k^cx0000ON-qF`0tv3401-z_2DJ_V
z07898+5izrOa`?M0EtR303wG-+5izsOhNzv078*S+5iztOhNzv01Qhn079Ec+5izu
zOhNzv098vb0759GNZJ4qO-u&04gj)CF91jTi)a8M>w7-{5m!tGwGIG2BF>;<01;Q0
zrH6nb0b!TOhkz*-aZCoa4gi5nF90PVG5`QV&Y)rd5pkCXh=58Wa!dxb4ge=jF90hJ
z+ywvt0000o000000000&&Y)rd5ptJvh=45tT9=`SfGPoAm&}NOPXS<;9EpH@0nV41
ziGU%0&PxWh4ghUWF95ayhr{;(5jRYb$Oixb0001s0UG+v01+}wQ?25`01?ki_Dul*
z1CgMALa8vr01?nj2DJ_V;!iICLJ6*)01?qk_Dul*#!xQ+LVZZu01?wm2DJ_VDp4;0
zLWfA&01?znx~L8S4^l4xLXk+?01?$o9kmWM0I5<h079Ec+5i#OO9r(L04`H6079im
z+5i#PODX^W00C4l0A6so0RS0~XaGWyynFx=|CerxfExk<OqYv_fLZ|vm*$FqAOQ%M
z2aA9p0SK2Ui-0!)9hYE>fHeU#mzIlwAOSO%w~K%#4m3;#wGIG2BF>;<01-5o>x+OQ
z0XUZqjDR2kIF~PsfGq*hmtBm2U;)pUzKnom0TY)ajetr4&!=vUfC2%}w{MStsR03y
zml=_OA%BodPntFWiCQlJwgHF3_W%*VOIhCq000000E+<{`pp0lyGueq+Q9%3kxN1V
z008J(F93l839g?25t2&=wGIFPLVZZu01=c+2DJ_VY+Ek?B8N!Y01=f-LI3~&LXk+?
z01=i;LI3~&?^`bbLYqk101=l<LI3~&I9x9PLMWw3+5i!lO9r(L0H|Cq07o~LXaFMX
zdp`gXqDuy~4gfwP&Y)rd5u%sjk$@rrtd|6mfGHNdO9r(L0C!z4042~f002VHpke?K
zyq80gfJy<qmyVKvAOXFXuabZ<0i~DWl7K1!sh13sfF=Q}mot-qmVX*c2DJ_V0XCUv
z03RRW!2l5&OEn8O0E+>K!}kCYflG@48v4xu5qwJ`j^Dum5gSVewGIFxmpVZJ5gbc4
znP>na!5eA-5gkj4I0pbC39g?25gtnhwGIFxeMs5>5g<zjwGIFxhe+B05g|)D0000Y
zkx1GA5h6<lwGIGBr+=7e03yzyVgM0ZO9r(L002k1m}meZ&Y)rd5nD?JwGIFPN7<NY
z03z#qKL8PKO9r(L06rqlpke?KZc7HW4gdfENBNj&03z#qKL8PZO9r(L06rqlpke?K
zeM<(l4gdfEM;V!D03z#qKL8PaO9r(L06rqlpke?KeoF?m4j2Fc06rq?dp`gXf=dRq
z4gfwP&Y)rd5rUWBlz<fhte5qafFXaZN-Nqm0GwYh0E+>K!}kCY6HAK$8v4xu5e`d2
zj^Dum5v@uFwGIF#U@rh7mpVZJ5w1!Po@f9<!5eA-5wA+ofC2!$U@rhd39g?25wJ=I
zwGIGxVJ`qeeMs5>5wc1KwGIIMVJ`qehe+B05wl7<0001#VlMzfkx1GA5ww?;m4GAx
z?3cQgfH48>m+X~*Avyp{2DJ_VJ|fPbVgL~UO9r(L00019kRSj6v6^TALhE}!01*#M
z2DJ_VvSTj*C6FKh07A~7VgL~jmq(U>HUSWqewKhB0T7p;mVh<^6qnYPfFJ=Bm-m){
zF#*b#DVKmD0o0dEmw-C~=a+|<fRzC=mmQdZF#+k9Pndv}0W+87n1C?>_Lmo#fRX{e
zm#Ue7BLO>?&Y6HAe>+MBwGIH-W-kD?0f)o)01=!@j+h7l00000ivb$?%>WUWN<s|b
z;{XvnN(QwK00C$(08{qk;{XvoN=<+Q00WW0eM0@(LI4pyN(QwK003z(07A(fvH%f3
zN(QwK032#B0740_p8yd+N=<+Q0RL()07898+5izkN(QwKe*leZF91`ANZJ4qLrS`+
z4gdp@sC`0_NZJ4qL`u4-4gjBQF91TDNZJ4qMM^pV0004PF91TNNZJ4qMoI>?4ggJU
zF91SuGOqv;M@j~@4gkPyF90GwN!kDrNJ=^Y002`tGp_&<NlI`b000A#@O@KPN!kDr
zN=k4c000A#8<2iSUY}?HBI|oU01;?P2DJ_VJ|fPbVgM0nmk67HA^~xiE1Q5I0dbd1
zn}8qzahGkIfHwhim!+G4AOUli#+!g30dtq<n}9a~ftMegfHeV^msOmAAOV+`cAS7F
zE|^LNwGIG2BF>;<01=o<2DJ_V002iIp=bai>w7-{5t&K`wGIG2BF>;<01=s&%A9~A
z0iBoaoPZnwotFxofHDD|mpz?;AOW72UY&q80iTzZoq!+#pO?0sfGq()m*ky*mjOMO
ze4c=p0ZEq>pMb~#N|)WAfXD%6m!_bASpjUep@2*Q0e!bkqJWeG0e!dPrGN(k0e-jY
zrhth90e-i1s(=dt0e_c(tAO4Cf0tjZfYkwiw>7PRX9EF$mz}VHN&$bj=&*n{0s()w
zAhUoA0RexvEVO_N0RexvIJJNa0RexvM7Dr40s()wc({NI0Rc+4gt>qV0Rc(3kh*{i
z0RcUioV$Qs0g9LUyMRmqO1DS6fC~WuNw-tIfMx*!J-52PfOr7`{g*SqfFXbVNyeEr
z07aTF0JZ^#!}kCYElPOZ1poj5004^t8v4xu5hzMRK-$3o5&lU+0001_nlAu>0tv34
z01^L52DJ_V07898+5iy&N(QwK02P}r03wG-+5iy(N<sht078*S+5iy)N<sht0GOLE
z079Ec+5iy*N<sht0O6Z20759GNZJ4q1xg0B4gfx!F91hT#b^K`>w7-{5f4fRwGIG2
zBF>;<01*$D`@n!A0U4Jb!GI|iDM|*l4gd|EF90PFGXMZW&Y)rd5h<5l!GKC4DoO^m
z4gj;AF90jf-30&u0000o000000000&&Y)rd5h|C-!GJ9R5|{VEfGPnNmm<P|E&&*q
zQNn<H0l=5e!hj)vz)7s7HUJWzF95ayhr{;(5${QhuLS@A0001s0UG+v01@d)LJ6*)
z01?4S2DJ_Vho3J1LVZZu01?AU2DJ_V(w{E?LWfA&01?DV2DJ_VR-i8cLXk+?01?GW
z2DJ_V0--MeN5aQw03z#qKL8QWNd~nJ06rqlpke?K&q)R<wGIFP0A4T(000)qXaGX%
zdp`gX>PZH*4ghqaF90Pl3IG5?&Y)rd5$czi!+=U6>q!Q+4gmC_F90iIuLS@A0000f
z000000000&&Y)rd5$l%(#DFFa>`4Z-4gfwP&Y)rd5$u;Z#DE<E(3e)kfGz>jmxaWD
zRsn*S+{Az(e}YK{wGIHOqAvip0f)o)01>lE(BlOF00000wgDRY%>WUwNwnkz00000
z08>V)w*V1?NxG;G00WVre<G5O9sm)9NzeoU0740_p8yesNpK+m07s)Q03v-z+5i!U
zNzeoU02aXTX8_{j1poj5004kT000q%Nsr2C00hqPmwd&5)&ZKAQ^tUV0i>7M#(>iS
zq_?ccfQbVEq_;N9fE@w>nwM0~fIWXDNd~nJ00UmqXaFA{;lThAB}p{p4giY*hr{;(
z5k^Ui0UG+v01-k-A_=aa01+lh2DJ_VB7I2O01+rj2DJ_VM?unP03z#qKL8OjNd~nJ
z06rqlpke?KGD!xt4gdfEJ|gRTKL8O!Nd~nJ06rqlpke?KLrDg;4gdfE0H1#WN(X2F
z5hpwbwGIFQS=DF&AN-UK0TCxW2DJ_VA7R&M01+oV2DJ_Vivfqj_W%)?Jc|Ju`pp0l
zkvt+UP-*}XC_DzW4gezk+Cl&kDLe+X4gey_9kKusDm(_Y4gewvuAcxAD?A3Z4gey3
zNZJ4qEj$Lb4geyDNZJ4qE<As_s15)kkx1GA5idNts15)kn@HLK5imSD0000YrAXQU
z5ivXlwGIFxaWby}5i&dmwGIFxK1tdD5i>kG0000YIWw;S5i~powGIGBO4DcnBI|oU
z01;6<2DJ_VJ|fPbVgM0QJO;H60000-c++SAOmuB<01;X|2DJ_VJ|dSq&ww14D$jrl
zMO-`vwGIG2BI|oU01;d~2DJ_VJ|fPbVgM0bJO;H600000N7&P708{vh2LTagJjJID
z00{$#z&8K@0000}e6RZe5oVWp&wv|$Y&-_F4gfw(bZu||5pO&OwGIG2BF>;<01<CI
z2DJ_V0000-Ce&yEBI|oU01<OM2DJ_VJ|fPbVgM0yJO;H60000-IMiqWBF>;<01<XP
z2DJ_V07pvHXaFM4pke?KcsvHR4gdg0T-0a)BF>;<01<jT2DJ_V07rb(XaFJ{>w7-{
z5raGiwGIG2BF>;<01<;c2DJ_V002jvmv7L39|63VkI;Z0MBLP903z#qKL8PwJO;H6
z06rqlpke?KlspEt4gdfENBq=i03z#qKL8PxJO;H606rqlpke?Kl{^Nu4gdfEM;z5?
z0GIC2fE#~DJk@9bBI|oU01=uz2DJ_VJ|fPbVgM1EJO;H60000!BI|oU01=x!2DJ_V
zJ|fPbVgM1FJO;H600000paO3MJ^&FMJO;H60FdSe0000009^|J0000000EqF<7fa}
z{FDv>5ga@QwGIHdtS<muVb^E?5ga@QwGIIBtS=-0LJ6*)01+KL2DJ_V9j-3`7GBb5
z0I}u<00000005u>01+UUaM6IO0Z*43(twZwB$uqxfKLIMxBSw8xB~%Ix3$)QcmV-x
zw=UR#M*;zImv-8KDFK3)pxS`f0fo1C+kmkHe}X#(wGIFQ*yd;eAN-UK0TF^b2DJ_V
zACQL601<*a2DJ_Vivfqj_W%(GJc|Ju`pp0l|2rZ+{9^zSgF6Pb4gezk+Cl&kggXYc
z4gey_9kKusg*yhd4gewvuAcxAhC2qe4gey3NZJ4qh&u+g4geyDNZJ4qi95Qe4geyN
ze@NN@5sEvys15)kn@HLK5sN!I0000YrAXQU5sW(qwGIFxaWby}5sf<rwGIFxK1tdD
z5so`L0000YIWw;S5sy0twGIFxS4r9c5s*6uwGIGBY2;`CBI|oU01>S_2DJ_VJ|fPb
zVgM1XI|j850000-m*i*wOmuB<01>u3msQ|^B7eo?XaG!fZEye)xH|^54gfwP>w7-{
z5x6@BwGIG2BF>;<01>!52DJ_V0000-^W<m%A`9pU0TIYM2DJ_VK1_6NZ~zg@I|j85
z06rqlpke?K%R2_O4gdfE07nz$XaFMXdp`gX&pQUS4gfwP&Y)rd5zjjYwGIFP07oU|
zXnz1A&Y)rd5z;#bwGIFPM^oi!03r+M2muk<I|j8506t7~ZEye)+&c!f4gfwP&Y)rd
z5!^clwGIFP002jG<!AsR>w7-{5#T!pwGIG2BF>;<01@Cj2DJ_V002jY<!AsR&Y)rd
z5#u`swGIFPN0;Sj03yzyVgM24I|j8507w8wspV(@BF>;<01@Up2DJ_V07tv!XaFM4
zpke?K=sO0r4gdg0+vR8gBI|oU01@#!2DJ_VJ|fPbVgM2GI|j850000-`{kF<;eZ|i
z8<+0kfFA)nmk#29A4FT`XaFMXdp`gX13U(`4gfwP&Y)rd5d%C1wGIFP07rY~XaFMX
zdp`gX1Uv?{4gfwP&Y)rd5d=I2wGIFP07skVXaJXO;(!}PN4w@|03z#qKL8O4JO;H6
z06rqlpke?K2|Nb14gdfEJ|gRTKL8O5JO;H606rqlpke?K3Ook24gdfE0H6Y1nS1~d
zcbAdkfFV71I|j850IAk509=rU&j1m3I|j850OHm!0740_p8yefI|j8505{k#02XoM
zXaJGS1^@s6005wn000qtm+#_$w*jk{r{jQ-0VJ0=<bY2BnzxALfOZ4{!ngnEfE)n<
z)3+SzfU*RC-8u%f4gdjt_Gkbf{FDv>5#2flwGIFuF}Lji5#2flwGIG_0f)o)01;w4
zivb$?%>WTvJ0d>(V*nA}ItH~403!X`LI4rpItH~403yj9vH%g_ItH~403r#lp8yfz
zItH~403v-z+5i#bItH~403wG-+5i#cI=ZM103wlpNZJ4q<vO~k4geyXNZJ4q<~lk6
z001JTNZJ4q=Q;+p4gex?GOqv;=sE_q4gexPN!kDr={h<9001I6Gp_&<>N*Cs4gexo
zN!kDr>pBLt4gg0E^k@Jg>w7-{5ePd5wGIG2BF>;<01*f~2DJ_V002ik^k@J~bZu||
z5fM9=`tN`uQ)={R08DgkZ~ze#I|j8506rq?dp`gX6FUaA4gfwP&Y)rd5feKGwGIFP
z002jr^k@Jg3+M;|5hFVWwGIG2OmuB<01+lT2DJ_VJ|fPbVgL~)mqGA=6HO{R2DJ_V
zJ|fPbVgL~;I|j850000-%JgUeBF>;<01+)a2DJ_V07v-rXaFJ$=m-H3G&=^h4gfw(
zbZu||5jQ&qwGIG2BF>;<01-Erq40nce>*z{wGIG2BF>;<01-Po2DJ_V002iS^=JSh
z&Y)rd5k5NxwGIFPM?Cdt03yzyVgM09I|j85002i$^=JSh&Y)rd5kWf!wGIFPM_~16
z03yzyVgM0CI|j85002jT^=JSh>w7-{5llM<wGIG2BF>;<01-?(2DJ_V002jzfAwep
zBI|oU01;a|2DJ_VJ|fPbVgM0aI|j850000-!1ZVVBI|oU01;d}2DJ_VJ|fPbVgM0b
zI|j850000-;Pq$#BI|oU01;g~2DJ_VJ|fPbVgM0cI|j850000-0QP7ABI|oU01;k0
z2DJ_VJ|fPbVgM0dI|j850000-e<1c~03z#qKL8P5I|j8506rqlpke?KUpof14gdfE
zM?m&y03z#qKL8P9I|j8506rqlpke?KV><@54gdfEM_~4703z#qKL8PAI|j8506rql
zpke?KWIG164gdfEJ|gRTKL8PBI|j8506rqlpke?KWjhA74gdfE0H6Y11rp!@5z?17
z@_->d(mDpU4gf>=F92LIx9tEC(mDpU4gh@kF91RbuAcxA(>eyV4gk^lF8~%6^JoBJ
z)&>9o0000Wl>h({)|Zs>fVTk#mrV13kO3r@-SdD?0h+fh^ni8*0Ufub_<$S%0WP<}
z`GB$nR-`%xwGIFQ5CdrdAN-UK0THA+2DJ_VA3Zs001>1*2DJ_VAMtEA01>1*2DJ_V
zivfqj_W%*VI*S1s`pp0ly*eTZuAcxAr8)++4gdh40y$Hn01;`I$^L*Le`z`fwGIH?
z8884`J{V&F5otOGwGIH08ZZD{1Oy)d5otOGwGIIA8ZZF10mGZ201=!zsmcWa00000
zwgGlZr~na|I$_HN000000JZ^nCrtnmmO22-1poj5002TgIcfkAYB~nB4geV(FaSdF
zY&QTAYdQwC4gjGXFaUu91_`d801<4LZ~uT5Ja9Tx2WbEa1HiU70000008_EPIsg%I
zI;u)&00WV*e<k1-002TF<O=~2b2_@H4gf11FaSbxXHftVbeF*YfFA*dx8VPPWB~!4
zmqh`AQ~|J;ngM}a2(UT@0Z{`Ash9o%fhPf%mnH&%!2zO|=mLQ@S35cewGIFQy9a3i
zAN-UK0TDYo2DJ_VA3Zs001-Pn2DJ_VAMtEA01-Pn2DJ_Vivfqj_W%)BI*S1s`pp0l
zR5~IFuAcxAJURxo4gdh40<EqX01*H>mvsYyAb$Wl2DJ_Vogpv)Ts|0M01*H>2DJ_V
zQ6ew^Tm%Fk01*H>2DJ_Vup%%3wgJPNp#TvxI;qD6000000JZ^kN~i!4FFIk!1poj5
z006cDc_&Q(5iL3Z$OQlZ0000&JvnLs5dk^|wGIH;BQO9$@oYB$5dt~}wGIGbBrpJh
z0vrjhp8yd9ItH~4003Tq761S&3TXgSE9jX35eJtr1c4PM2s)}tXaEC|uzw}c761T3
zBIFAJ5ehoGs15+@BrpI%b7xTi5et`V1c4s`9k+`Ffn)&zG?(xNfm8uUmqiAFTnI-x
z1p!e53YeGv0f85nzy^U30WO!(27$o=k(ZMPfh1PaIR>>300C4CX#gMmlnwzA(m4jT
z4geoLIcfkA(m4jT4geqVY&QTA(m4jT4giY*hr{;(5$-vQ0UG+v01@jsA_=aa01?wU
z2DJ_V0H6YG#u5M#mzNC)fgyjFIR>>309`3C09-y8V*nACIR>>301+xM09*tF9{>@T
zIR>>30B|ZW0JZ_co1p*^$~hUl1poj5006cDc1oxK5ym;uyafON0002C0eL4)01?DF
zaJ&To00000LOnTZ01=ou2DJ_VnJX{=Lh)=j01=rv2DJ_VAuKQefdYRCuAcxAnmGow
z4gdgN;1B=+#SUozQ!D7101=)!R0nAQ2?Id4Hvj+t002|5zB&LApE;^ZXaEC|uzw{m
z5dZ)}BIFAJ5urJ{s15+DEHD5<b7xTi5u!Q9s15*XEieEY98YKfU%mwZ00000IRF3v
z000005wtlG4ru@ZT2Hqu34vSz0n3+(3V~Ds-k07Afm{gQIRybx0}6+i{sDm&msksd
z5dp=QY72qE0f3h*41pwnX*mY94gdiX5orJ)JvnLs5otLR5orJ)b_%lq5otLFwGIG_
z0UG+v01<*YA_=aa01;|A2DJ_VB7I2O01<3C2DJ_VNA(bC03yzyVgM0$IR>>3002HB
z>w7-{5ra7fwGIG2BF>;<01<;Z2DJ_V0000J2=os(0K);I8_NI!5nh+k41pm%UO5J}
z4gei7FaTV33bOzaUO5J}4gh~KFaSacuAcxAUpWS~4gldXFaQ^r5NQAk<plr$0001x
zQvd)FVwW!sfk**#w|EVKUjctJIR>>300E5?X#gKRIcfkAGdU6wX#k4>8v4xu5llHE
z39g?25i~glwGIFxeMs5>5jHsnwGIGBa1&_&BF>;<01-kt2DJ_V06rq?dp`gXO*sa&
z4gfwP&Y)rd5luM;wGIFP006@QG4lce5hytZwGIHF<^=!%0001z0RWd)4}lRqC^-hT
z4gec8FaTUG*kJ$>C^-hT4gh^LFaSacuAcxADLDqU4glXYFaQ@H6KMdd<^=!%0000G
zQ~&@GESIDYfj<F1x8x6jUjcssI0m&200GDwX#gKRIcfkA0XPy7X#gLbWv~Dd0XPP=
z4geo#Q?&pQ0XPP=4genmdIJFw0XPP=4giY*hr{;(5fnL#0UG+v01*p0BJROF01*N>
z2DJ_VBK57F01*Q?2DJ_VB0$=~01*T@2DJ_VA~uJ101*W^2DJ_VA{~E4&;StzI0m&2
z03x6aXaErhI0m&203xO&3jh%aI4S@D03r#lp8ydFI0m&203v-z+5iy?I0m&203wG-
z+5iy@I0m&203wk{+5iy^I0m&203w@6+5iy_IJ&3~03xMG+5iy`I0m&203vZRuK*Dc
zI0m&203tp~+5iy|I0k>U4gexKGp_&<5;!^l001IaN!kDr6F5Qu001I$k^}$|6gWZv
z001H<HVOa{6*$<20{|j2lmq|~7C0&Z002k*7HI$?3+M;|5i&RiwGIG2OmuB<01-7f
z2DJ_VJ|fPbVgL~}I0m&200000M;#Yw03z#qKL8OqI0m&206u>r&Y)rd5ji*pwGIFP
z07owuX#gV5pke?KJU9lm4gdg0K^JKNBF>;<01-Vn2DJ_V07r8dX#i9Bi3b4@MmWW%
z4gd)QxWG3600000Q+%)c01-ww2DJ_V2?MaeHvj+t0000-p%-ZYA`9pU0TE0%2DJ_V
zK1_6NZ~zfcI0k>U4gfwP&Y)rd5l=V<wGIFP002k97ij<@>w7-{5mPt@wGIG2BF>;<
z01;C-2DJ_V002kR7ij<@&Y)rd5mq<`wGIFPM*$dV08DgkZ~zf(I0m&206rq?dp`gX
zY&ZtB4gfwP&Y)rd5o|aHwGIFP002iX7-;}ZbZu||5pRDu2DJ_VJ|gRTKL8PLI0m&2
z06rqlpke?KZ#V|E4gdfE07pR>X#gV5pke?KaySOH4gdg0a~Np=Q}~Go0TF*V#itGc
z2?MCWHvj+t002{buloQIe>euU4gd)Qpujf(0000007r!wX#gV5pke?KgE$7Y4gdg0
zml$aPBF=xHVgM0^I0m&2002j+7-;|^&Y)rd5r#MhwGIFPN4pql03yzyVgM0{I0m&2
z002kn7-;}ZbZu||5t29twGIG2BI|oU01=Wn2DJ_VJ|fPbVgM16I0m&200000M;jSw
z08{vh2LTbDIK`(900{$_z&8K@0000}e6RZe5uTU46oDImqBsV%4gdg0KN)ENBF>;<
z01>1(2DJ_V07p?7X#gV5pke?Kr8ow)4gdg0V;N}xBF>;<01>A+2DJ_V07sD-X#h-g
zZEye)u{Z{`4gfwP>w7-{5wSQ1wGIG2BF>;<01>e`2DJ_V0000-qZw%cBF>;<01?DE
z2DJ_V07tcd8EF6_&Y)rd5zaUUwGIFPN5>gy03yzyVgM1(I0m&2002kz8EF6_3+M;|
z5%M?&wGIG2OmuB<01@>#2DJ_VJ|fPbVgM2KI0m&200000M->`r03z#qKL8Q=I0m&2
z06rqlpke?K`8Wo(4gdfEM<*I-03yzyVgM2PI0m&T4gdg0IT~pIBF>;<01^E-2DJ_V
z07pw2X#gV5pke?K{x}A;4gdg0YnMJ2fgb^jmtGcu9|5bEeinfr0?QhgpB8~0LF*c6
z03z#qKL8OBIR>>306rqlpke?K5IF|54gdfEM++Ni03z#qKL8OCIR>>306rqlpke?K
z5jh664gdfEM=O{97J(lDOP3xOfgb^Dmp&JP9tB5>8)*QST^E5HMMtX}X#gVYdp`gX
z894^E4gfwP&Y)rd5g9oKwGIFP06rq?dp`gX8aW2F4gfwP&Y)rd5gIuLwGIFP005u@
z3ajw|5xbXx7l9#vyEg{44gj|{FaTUW7-IktyEg{44ghC2FaTTx1RnqqyEg{44gk_P
zFaWjz!<(T15&1XLA_xEg0002C0d`8L01@>!5i1A)00000wgGu3O#l(|H?S%Q00000
z079H)umBOfHwLv101-Mc077R|wEz*lHwLv10IoVP073&UdIJFwzBdN74ggC#FaSac
zuAcxAzc&W84glyoFaTZvd;$O&A87znE9jX35yO{H7=aZg#5cZ3X#fL}pnoNRd;$PM
zBIFAJ5ym&Vs15)$JTL%4b7xTi5yzK`7=a%F;g_lyfwTeJw_h27eE|Xbx7`|noC5)H
zmwFz7oB@xw7axJ90Re}%z950P0Rg7BcOrou0Ra`al_P;h0s$4b+$Dil0|DBXGbw?6
z0SlM0DS>DK4VM}!fn)(rmwzgOo&o2V9xH+40r<B9EP+e`0qB=cErH_!__s4If$sqU
zw3j0=fj|MXmt-)3Q~}VJxiEpQ0oSK!F@XXBPnT&kfm{JSx5hJpP5}W6w-q&kR007D
zw`w<m3IPGyw{<vyBmn{3x0*SDN&*4gx8*y5Ndf`fw=X_{0RaIOw=q9~I0OOomuf?S
z!2w2>EJT4^f4VgWwGIFQVo_-TA3Zs001>)15)o+tAO6O501>)12DJ_VAG^;30TH@2
z2DJ_VA0Sd>01>)12DJ_VAK^Y*01>)12DJ_Vivfqj_W%*6H;Vy!7?J=HJ2r~}8v4xu
z5t%n45!v7X5xX@8wGIFxg0@ot5xg}9wGIFx39g?2e-XVk2DJ_VB7I2O01>}6IsgCw
zB8N!Y01?17IsgCwB9Tbi01?482DJ_VBAZCs01?792DJ_VBBe;$01?AA2DJ_VB5^XW
z01?DBIsgCwB0fpl01?GC2DJ_VA~`d!01?JDx~L8SB3DV;01?ME9kmVsB6E@i01?PF
z2DJ_Ve<CL~3IGwwH3qc~03tG!1OO4rH3qc~03t6x5C9R&H3qc~03yhg1OO4tH3qc~
z03y?r1OO4uH3qc~03zO$1OO4vH3qc~07pwoX#gS%=m-H3>oo?o4gfw(bZu||5$-hx
zwGIG2BF>;<01@sr2DJ_V0000-c}i&jB3V>Qe*h8mHNHq`06rq?dp`gX_B95z4gfwP
z&Y)rd5%x6(wGIFP002j)N@)NhSyV~@5&Sj2NNE5*BI|oU01^H*2DJ_VJ|fPbVgM2T
zH3qc~00000N4ZLA03yzyVgL~kHU_m0002kJN@)Nh&Y)rd5fU~AwGIFPN9#&y08Dgk
ze{cX16E+654gfwP&Y)rd5fe5BwGIFP07n^1X#gS%=m-H3E;a_W4gfw(bZu||5ivFf
zwGIG2BF>;<01+`Z2DJ_V0000-IZJ5(BI|oU01-7d2DJ_VJ|fPbVgL~}HU_m00000-
zOG{}0BF>;<01-Gg2DJ_V07qR*X#gV5f1qLj5ji#nwGIFPM{i4M03yzyVgM0NHU_m0
z002jUOKAWik-U5W5oI<8wGIFPN3Tn108DgkZ~zf*HU_m006rq?dp`gXZZ-zB4gfwP
z&Y)rd5pFgHwGIFP002k9OKAWi&Y)rd5r8%ZwGIFPNAF8%03r+M2muj>HU_m0e*ivA
zbZu||5s5YiwGIG2BF>;<01=5c2DJ_V0000-K}=}?B3V>Q01=NiVo_-TM<7gT03z#q
zKL8PtHU_m006rqlpke?Kkv0ak4gdfEM>0%l03yzyVgM17HU_m0002HB&Y)rd5tKFt
zwGIFP07sKdX#gTwR7wC5nKoike`x?ma!hFeBI|oU01=xu2DJ_VJ|fPbVgM1FHU_m0
z0000-giL7wBF>;<01=%w2DJ_V06rqlpke?Ko;C)x4gdfEN2N?@03yzyVgM1cHU_m0
z002j~Olbfj&Y)rd5wtc2wGIFPN6AcS03yzyVgM1fHU_m0002kZOlbfjf6ky{01>}7
z2DJ_V07vaiX#gV5pke?K(l!RQ4gdg0|4eBBB9Xj&01@dn2DJ_V07orNX#h-gZEye)
z@-_yw4gfwP>w7-{5%M+$wGIG2BF>;<01@&w2DJ_V0000-KTT->BF>;<01*N=2DJ_V
z07p?xX#gV5pke?K2R8<_e+~cuM}JLe03r+M2muijHwLv106t7~ZEye)7B>d94gfwP
z&Y)rd5f(QFwGIFP002kPO=$okSyV~@5ga$ZNNE5^uuW+IBI|oU01+NH2DJ_VJ|fPb
zVgL~yHwLv10000-!cA!aBF>;<01+TJ2DJ_V06rqlpke?KAvXrKe+~cu07oNEX#gTw
zR7wC5CO5uFX#ht8PH6xl>w7-{5hynXwGIG2BF>;<01+rR2DJ_V002i6PH6xl&Y)rd
z5h^zZwGIFPJ|fPbVgL~<HwLv10000-HBM;&BF>;<01-Sl2DJ_V07pkoX#gV5pke?K
zJ~sxn4gdg0bxvsje<BO$2muj9HwLv106t7~ZEye)M>htw4gfwP&Y)rd5l1%$wGIFP
z002kXPH6xlSyV~@5luIAQE322rcP-9BI|oU01;0&2DJ_VJ|fPbVgM0OHwLv10000-
zxK3#RBF>;<01;6)2DJ_V07uGBX#gV5pke?KQa1**4gdf?e<IGHVgM0SHwLv10000-
zJx^%>B3V>Q01;R>bWv#lM+#4A03z#qKL8P0HwLv106rqlpke?KS~mu@4gdfEM;uRS
z03yzyVgM0bHwLv1002iWPiX)m&Y)rd5nVS1wGIFPJ|fPbVgM0dHwLv10000-Pfuw8
zBF>;<01<UJUIw)e002i}PiX)m&Y)rd5qCERwGIFPM{`eU03yzyVgM0%HwLv1002jY
zPiX)m&Y)rd5rQ`cwGIFPN2O0`03z#qKL8PnHwLv106rqlpke?Ki#G<f4gdfEN5xN<
znM;8l0_9Jaxl4f_0R@-XOMxE&C71Y1fgb@yml#Ze9|2{TI81>b0fm=XOo1N(rI&b2
zfgb|JP?woZfgXS5P-y@n>w7-{5u-N-wGIG2BF>;<01=}%2DJ_V002h?QE31o>w7-{
z5u`T;wGIG2BF>;<01>1&2DJ_V002iNQE31o>w7-{5vVr?wGIG2BF>;<01>D+2DJ_V
z002itQE31o>w7-{5vex@wGIG2BF>;<01>G-2DJ_V000#}BI|oU01>J;2DJ_VJ|fPb
zVgM1UmkUjSGXZ0lKTUxle`7TUwGIHEr!W9qJ{V&F5o0w5wGIHNtS|sv1Oy)d5o0w5
zwGIFit}p<$0mGZ201>h^y)Xs<00000wgGlZr~nbJHB~SM000000JZ^nCrtnmtTo^-
z1^@s6002V%#&rM@WHko04ggiJFaSck&jSGwWi<x14gedlFaSazQc`395oR?8wGIH`
zurL5Z;XYdc5oa|9wGIHMu`mFE0tv3401;?42DJ_V0A3(;004|sX#i6z=$QZ!Y&AS>
zX#fcWptm;w00000Q?b4}01<7MF;0OL5pgxTs15*avM>Nbb7xTi5ptJSPJtu=jhA~)
zfjI$_m$goTGy#aW>P~@a0RgR-R8WD;0kgLuQGv(-0sXh?Qi0O~0cV%KRe@*$nztub
zfqVi1ez&Apfzkm1>z8+0foK6Ix8PcVgaZLtw`E>|Y6<})m(XZ|Hh(HL2DJ_V0pD$D
z03SU$Y5)-`H4+hN0E+>K!}kCYQZ<VK8v4xu5l%HC39g?25i2zYwGIFxeMs5>5iK=v
zApig(he+B05iT_bwGIGB!fj~)BI|oU01-Gf2DJ_VJ|fPbVgM01H3qc~0000!BI|oU
z01;0$2DJ_VJ|fPbViN!nPc;U$4gdfE0K);1>v{kYAD5VEfgvFuH3qc~0ObcU09-EE
zVE_>yH3qc~03HZ20740_p8ydcm)dE88v!Gi_Gy7!0WFtPYJtE3I=2*Sfr$eFI=7x~
zfeQfvKbNX+fz<&&mw<4A?*TuzaB+bH0Rb(ybaH_U0Rb(yfOCNh0Rca^jC6sA0e??3
z2DJ_V0dRt803SU$Y5);WGZGPL03W;00|60FGX}K|03TI%vH%fJGX}K|03QZFy#NtU
zGX}K|03RtPA^;IjGX}K|0E+>K!}kCY1vQHS=;62k5xF#r0hK#b01>n_ivb$?%>WVn
zG$OQd^8gW0GX}K|03zGK;s6m*Gk*rP4gex!Lg)YyQZoj%4gezIexd*oQ!@s&4geze
zRu2FXR5J#(4geyGnlS(oRWk;)4gezZB+&p7Rx<{*4gey?l79dZS2G5+4gey<l|ld!
zSThE-4gewvuAcxASu+N;4gey3NZJ4qTQdf=4geyDNZJ4qTr&o>4geyNNPpS@5nVF|
zwGIFxn@HLK5neO8s15)krAXQU5nnSMwGIFxaWby}5nwY0wGIFxK1tdD5n(e1wGIFx
zIWw;S5n?mCs15)kS4r9c5o0qt0000YbCLuA5o9wu0000YCpHQI5oI$5wGIFxGL!@W
z5oR+TwGIGB2YhJ&Q}~Go0e=y2GsUM400{%gz&8K@0000}e6RZe5pXjGwGIFY1HiyH
z00000002iBd}#n8&Y)rd5p**KwGIFPM@M{V03r+M2muj!GX}K|06t7~ZEye)d@}~M
z4gfwP&Y)rd5qvWSwGIFP002j4d}#n8>w7-{5r8uWwGIG2BF>;<0DlpHGX}K|0000-
zd3<RABF>;<01<;T2DJ_V07r{_X#gV5pke?KkuwIh4gdg0xqN8=A`9pU0TG}x2DJ_V
zK1_6NZ~zgbGX}K|06rqlpke?KqcaAz4gdfE07uz;X#gVYdp`gXr!xk%4gfwP&Y)rd
z5vMZ-wGIFP07vV5X@3AB&Y)rd5vnr=wGIFPM;U!-03r+M2mukhGX}K|06t7~ZEye)
zzB2~34gfwP&Y)rd5xz49wGIFP002iheQ5w9>w7-{5yCSDwGIG2BF>;<01?772DJ_V
z002izeQ5w9&Y)rd5ydkGwGIFPM|pi|03r+M2muk+GX}K|0DnGAbZu||5!o{awGIG2
zBF>;<01??U2DJ_V0000-nSE&hBI|oU01@3Y2DJ_VJ|fPbVgM1{GX}K|0000-t9@wz
zBF>;<01@Cb2DJ_V07t!jX#gV5pke?K?lT6p4gdg0&wXhCBF>;<01*K+2DJ_V07v0{
zX#gV5pke?K3V$>PwGIFPNArDY03wmRd;k$4GzPT}002iJerW(qbZu||5h^qWwGIG2
zBI|oU01+xQ2DJ_VJ|fPbVgL~;GzPT}00000M>BqD03yzyVgM0DGzPT}002iterW(A
z&Y)rd5lA!!wGIFPM^}Dn03yzyVgM0LGzPT}002j6et&5IBF>;<01-_z2DJ_V07rX%
zX#gV5pke?KPBaF!4gdg0jecnWBF>;<01;0#2DJ_V07svGX#gV5pke?KSu_T<4gdg0
z&3<VBA`9pU0TF042DJ_VK1_6NZ~zf&GzPT}06rqlpke?KYcvM64gdfE07vb9X#gVY
zdp`gXZ+|ofwGIG2BF>;<01<CA2DJ_V002k-erW(A&Y)rd5ppyJwGIFPM<0J_03z#q
zKL8PgGzPT}06rqlpke?Kgfs@V4gdfEM@xTc03ul<iU1LgGzPT}03uW}a{v*KG`gq`
z06rqlpke?Kr8EY$4gdfEM~{DL03s0S)c_HvG&#Dc4gg1Pe`x?B&Y)rd5vnu>wGIFP
zM}U8603yzyVgM1cGzPT}002HB&Y)rd5w$c1wGIFP07t!-aC?C*0_%U5pL>BUe;I&j
z08DgkZ~zg%GzPT}06rq?dp`gXzcdE54gfwP&Y)rd5x+DBwGIFP002iTfN20C&Y)rd
z5#TfiwGIFPM?HXP03yzyVgM20GzPT}002i%fN20C&Y)rd5#lrkwGIFPM`3_z03yzy
zVgM22GzPT}002jGfN20C&Y)rdDiP&02DJ_V07sL6X#gVYdp`gX?lcCq4gfwP&Y)rd
z5$-ewwGIFP07tWzhJ1k^0@Hw(r+k4P0`q{E$9#bv0TY+ze1RW9Gl6LUBI|oU01*K-
z2DJ_VJ|fPbVgL~VH3qc~0000-Q-Ns!BI|oU01*N;2DJ_VJ|fPbVgL~WH3qc~0000-
zbC*Vafgb^rmu7u|9|5zMhJAq_0@Hz)r+tAQ0`q~F$9;hw0TY+zeSscjM>B$H03z#q
zKL8O7H3qc~06rqlpke?K3^fL|4gdfEM^l1n03z#qKL8O8H3qc~06rqlpke?K4K)U}
z4gdfEJ|gRTKL8O9H3qc~06rqlpke?K4mAd~4gdfE0H6Y1bszu{_LnJsfgykPG6uB{
z0H_!-09-y8V*nBMG6uB{00|s109*tF9{>^dG6uB{0E8Ve0JZ_co1p*^M>FkG1^@s6
z006cDc1oxK5koVdUj_gG0002C0eL4)01-hmAYTRm00000Lc7lc0TK5y2DJ_VbRIDP
zLRELN01@~y2DJ_VCm%5YLIx#2y#NvUG6uB{0NNih075AyA^;KkG6uB{0F59q0740_
zp8yg2m+pRn7(D+nJZ)(J2?OA_Hvj+t002|5zB&LA05f7yX#fL}kbfoMdH?`IBIFAJ
z5d|~4s15*(Au#|#b7xTi5eAn+e}N<c8<%H)fjI#om!W@wIRPh^+kb)00Z6y0fPvWs
z0V9`#hJnKYJGV!NfinUDf|q-Vfsg?tmkf%5PXU^jT#A8+0Zo_Xih+;;nYVU}fn)&z
zLbuzDfvo}ol9!2(fj|L~m&%WUcLA`MLXd&30k^mDkb(OP0YaDUoPp~B*q7Ozfh2!M
zG6uB{00BOxX#gKRIcfkAMluo+X#gKCjy?bpMluGq4giY*hr{;(5#ut80UG+v01?|V
zivhGK009xBGK&G4h0OpFpfV!A!6X0?M=}Pr4gey?J;DGHNHPYs4gewvuAcxANiqht
z4gey3NZJ4qOELzv4geyDNZJ4qOfr84wGIFxkx1GA5lu3>s15)kn@HLK5l%84wGIFx
zrAXQU5l=D(wGIFxaWby}5l}J)wGIFxK1tdD5m7P*wGIFxIWw;S5mGW)*8>0|S4r9c
z5mPcvfC2zVHKS<&B9Xj&01<vN2DJ_V07qk^X#h-gZEye)hcX7W4gfwP>w6_X01<~W
z2DJ_VJ|fPbVgM0`G6uB{00000N0p;#08{vh2LTa@mmr^k6@Q5`2DJ_V2?OxJHvj+t
z0000-!=q^cA`9pU0TGWf2DJ_VK1_6NZ~zgKG6uB{06rqlpke?Kk}?Lh4gdfE07v7a
zX#gVYdp`gXmNEvl4gfwP&Y)rd5tcFrwGIFP07vzsX#gV5pke?KnKA~o4gdg02c&5L
zBF>;<01=xq27k2<002iBq-g*m&Y)rd5uGvywGIFPM>(Wv03z#qKL8P)G6uB{06rql
zpke?Ko-zit4gdfEM_Ht403z#qKL8P-G6uB{06rqlpke?Kp)v-w4gdfEM|q@a03z#q
zKL8Q0G6uB{06rqlpke?KuQCR;4gdfEN2{c108{vh2MYlaurkG`4gd)Q=$CMyffbii
zpn((@vNFY|4gd)Q*uXac00000Q+%)c01>j6exQLHf3-3OwGIFPM;)bU03r+M2muko
zG6uB{06t7~ZEye)#WDu94gfwP&Y)rd5ydhFwGIFP002ilrD*^n>w7-{5y>(JwGIG2
zBF>;<01?SD2DJ_V002i%rD*^n&Y)rd5y~<KwGIFPM`5LD03yzyVgM1%G6uB{002jU
zrD*^nf9rcc01@0W2DJ_VJ|fPbVgM1`G6uB{0000-p`~d6BI|oU01@3X2DJ_VJ|fPb
zVgM1{G6uB{0000-!KG;cBI|oU01@6Y2DJ_VJ|fPbVgM1|G6uB{0000-;iYK+BI|oU
z01@9Z2DJ_VJ|fPbVgM1}G6uB{0000-0j6mHU?S^#KL8QnG6uB{06rqlpke?K;4%ia
z4gdfEM<J$Z03z#qKL8QrG6uB{06rqlpke?K<T3`e4gdfEJ|gRTKL8QsG6uB{06rql
zpke?K<uV4f4gdfE0K);xl?4G2JD2yNfgvF~G6uB{0OfNr09-DPJ^&FrG6uB{03CEO
z0740_p8yd&mpP(=8v#F;Ric5?0aKSKqk*jffVbA8ftCRQnzwDGfu;cgQkN8`f$jmc
zw*#nw$^rqmm(HqzasjuuL92nB0s&K(-mQUM0aCX-u7NTF0aLecuz?@~f9f#?wGIFQ
ze7I=<A3Zs001@gj5)o+tAJ#p>01@gj2DJ_Vivfqj_W%(pGK&Ek`pp0lBr+liuAcxA
z>oEqk4gey3NZJ4q?J)+m4geyDNZJ4q?lA_n4geyNNZJ4q?=c3o4geyXNZJ4q@G(t*
z0stbVNZJ4q@i8g@002jOK(}cCBI|oU01^8!2DJ_VJ|fPbVgM2QF$T2`0000-t+#0a
zQ}~Go0TKK$#itGc2?GehHvj+t002{buloQI{FhF%fg5ZBG6uB{06t7~ZEye)1u_P;
z4gfwP&Y)rd5d|^^wGIFP002k(w`l+(>w7-{5eYH|wGIG2BF>;<01*i?2DJ_V002h~
zxM=_)&Y)rd5eza0wGIFPM;`zMfck&9X#gV5pke?K4l)L{4gdg0Ke%ZCml3pq8%0N7
zxM=_)>w7-{5hgMQwGIG2BF>;<01+lK2DJ_V002HB>w7-{5hpSRwGIG2BF>;<01+oL
z2DJ_V0002P0h$sW01@ApHMD^tJ>M|~wGII8kud;V);+=i5#KQewGIH8k}&{639g?2
z5#TWfwGIGBlQ94m*S2W@^qdF)00000U<v~O5#yJWw1LL~@V8sFfj<EO4Y!-NfldJd
z4VUV-fs+C7w|=;R2>}7{w}`ocRRMpFF$T2`00GLtX#gKRIcfkAjxiDuX#gMAJ;DGH
zjxh$c4giY*hr{;(5zaA-0UG+v01?PBA_=aa01=Nd2DJ_VB7I2O01=Tf2DJ_VB8N!Y
z01=Wg2DJ_VB9Tbi01=Zh2DJ_VBAZCs01=ciO@IObBBe;$01=fjDgXcgN6A3HX#gVY
zdp`gXpD_lt4gfwP&Y)rd5uY&zwGIFP07v`3X#i9Bi3b4@pfSa#4gd)Q7{NCH00000
zQ+%)c01=><oV$S=Y^E^=wGIG2OmuB<01>G%2DJ_VJ|fPbVgM1TF$T2`00000M@hhG
z03z#qKL8P}F$T2`06rqlpke?KtuY3*4gdfEM_a&Y03yzyVgM1aF$T2`002jAz-a&?
z&Y)rd5wbA`wGIFPM~%Q~0GD69fg43ft-xsjBI|oU01?VD2DJ_VJ|fPbVgM1!F$T2`
z0000!BI|oU01?YE2DJ_VJ|fPbVgM1#F$T2`00000!vSAtSO5`&mxH{4Aw7dJ2DJ_V
zGn_F1T-H6p01<;R2DJ_V-kdQ2LJ6*)01<>S2DJ_VjGZw67A3!F0Q8^;000000I&=L
z01=0m;k<#z0hG6?y@5Xg0kOB~zJX2w0kM}jz=4wil()>lfe8Ttl(*KwfmH#29x(>B
z4gdkv$!P!|JvnLs5gsuT5orJ)+AGQc5gsuHwGIFurI}&?5gsuHwGIG_0f)o)01<LA
zivb$?%>WT>F(L`Bp8ydbF$T2`03v-z+5izDF$T2`03wG-+5izEF$T2`03wk{+5izF
zF$T2`03w@6+5izGF-?F103xM-NZJ4qB{5Ba0sta$GOqv;CNU}i002iX$Y}r~>w7-{
z5il_ZwGIG2BF>;<01+@T2DJ_V002i}$Y}sm_=yJr5iv2vrw#xK131Aq0000008@Og
z`v4I!F$T2`00{#y!8ZT^00000M~}#903r+M2mui{F$T2`06t7~ZEyf*5jrsjwGIG2
zBF>;<01-Md2DJ_V0000-ugGZtBI|oU01-Yh2DJ_VJ|fPbVgM07F$T2`0000-!N_R<
zBF>;<01-hk2DJ_V07uivX#gV5pke?KLoo)m4gdg01<7duQ}~Go0TD!(ki>x%e?&0`
zwGIFY10caS00000002ib$!P#03+M;|5lS%zwGIG2OmuB<01-_w2DJ_VJ|fPbVgM0M
zF$T2`00000M^njZ03z#qKL8O?F$T2`06rqlpke?KQ85O!4gdfEM`g)r03yzyVgM0T
zF$T2`002jK$!P#0&Y)rd5mqq<3AGLY07sX}X#kgL#eo||N4Lpo03z#qKL8PKF$T2`
z06rqlpke?KZZQV64gdfEJ|gRTKL8PLF$T2`06rqlpke?KZ!re74gdfE0K)-wJ4OHz
z<Cl%afgyk6Fb1^_0C=V`09-y8V*nB3Fb1^_07<7Y09*tF9{>^KFb1^_01&7#0JZ_c
zo1p*^7BNMo2mk;8006cDc1oxK5fU+BsR#f70002C0eL4)01*%|0I3K700000LfR|J
z01@Oc2DJ_V5UDW$LZz8v01@Rd2DJ_Vo2fAXLJ2gkp8yf&Fb1^_04Az20A2tO0{}?O
zX#i6z=$QZ!>M&FXX#fcW5V$u000000Q?b4}01@k#Va9<J5$!O#s15*fsxbgUb7xTi
z5$>0U#(^IK05P|x#(`u30T-7H$bpIhB$t=SftLX#w=c<oZUF&8x1P#@P5}Wzx9!V;
zVgUhFw@}T2P5}W`m!HmolK~}{GtYrr0VKDN&w+CR0VS6M(Sch5B)4DDfx7_#5tkd&
zfx!W#m)+BWIDa-U2DJ_V0TSG403SU$Y5);7FcJ}I03YFgq5u&#Fb1^_0E+>K!}kCY
z&@hVu8v4xu5y&u$0kkLp0TGihivgO2%>WUQFd_&j000p;Fb1^_03w<%x&RS4Fb1^_
z03r#lp8yd#Fb1^_03v-z+5izdFb1^_03wG-+5izeFn<QM4geyNNZJ4qJutec4geyX
zNZJ4qJ}@1%4geyhNZJ4qKQIQh4gex?GOqv;KrjZi4gexPN!kDrK`;ij4gexKGp_&<
zLNHm^0{|jdN!kDrLoiK%0su!%*=YbGk-U5W5pFOBwGIFPM|;_608DgkZ~zf^Fb1^_
z06rq?dpkb>5qB^KwGIG2BF>;<01<aE2DJ_V0000-t=VY+Q}~Go0TFpH#itGc2?JQc
zHvj+t002{buloQId6!Prfg68+Fb1^_06t7~ZEye)f-nZP4gfwP&Y)rd5rQxVwGIFP
z002k(*=YbG>w7-{5r!}ZwGIG2BF>;<01<{T2DJ_V002h~+GzkH&Y)rd5s5GcwGIFP
zM<3d003yzyVgM0~Fb1^_002iZ+GzkH&Y)rd5sfegwGIFPM^V~o03v_udp`gXjxYwb
z4gfwP&Y)rd5sokhwGIFP07r4!X#gVYdp`gXkuV0e4gfwP&Y)rd5s@$kwGIFP07sG9
zX#gVYdp`gXpD+fs4gfwP&Y)rd5uY#ywGIFP07t{xX#i9Bi3b4@pfJU!4gd)QNWnJ%
z00000Q+%)c01==t2DN_<00{#?!8ZT^00000NAcQe03r+M2mukVFb1^_06t7~ZEye)
zvM>g<4gfwP&Y)rd5wb7_wGIFP002i3+i3tI>w7-{5w<V}wGIG2BF>;<01>t@2DJ_V
z002iL+i3tI&Y)rd5w|b~wGIFPM>X4N03yzyVgM1kFb1^_0006<RhKB&fgb^Nmq^!v
z9|4t@XxD)sf3@3b03z#qKL8QTFb1^_06rqlpke?K%rFME4gdfEN7dVD03z#qKL8QU
zFb1^_06rqlpke?K%`gVF4gdfENA=rj03z#qKL8QYFb1^_06rqlpke?K(J%(J4gdfE
zJ|gRTKL8QZFb1^_06rqlpke?K(l7?K4gdfE0K);|1=r>P5i6I=*MT7+D=-GN4gf;A
zF#ufQexd*oD=-GN4gh+&F#tjduAcxAESC+~fg1rYmnqnR(g8!4{Mdo50dTjf*@2b;
z0gAUl+kvJ50YaDP+<~V7qPLLUftLaSq?ah*fpP()x0~RBoB{zumpJ2rT>(P3mg9jk
z0s%v}$>o6{0)N^s2DJ_V0oLtl03SU$Y5)=1FA@=H03R~KGyoCWF9x*^0E+>K!}kCY
z8Ze6i8v4xu5fm^Y39g?25!){YwGIFxeMs5>5#28awGIFxhe+B05#BEbwGIFxkx1GA
z5#KKcwGIFxn@HLK5#TRPfC2y_rAXQU5#cW?0000-)qm`103z#qKL8QyF9x*^06rql
zpke?K>n{ei4gdfEM+fa`08{vh2LTc6FU6-000{$V!8ZT^0000}e6RZe5$rDpwGIFY
z17N{700000002id?P&la3+M;|5%MnvwGIG2OmuB<01@>s2DJ_VJ|fPbVgM2KF9x*^
z00000M@UueX#gVYdp`gX`7Z{w4gfwP&Y)rd5&16$wGIFP07qx-X#gV5pke?K{4WNz
z4gdg0dF^QcBF>;<01^H#2DJ_V07seaX#kgN>46(XN4f2303z#qKL8OHFb1^_06rql
zpke?K7BB|24gdfEJ|gRTKL8OIFb1^_06rqlpke?K7cd634gdfE0K)-WENK7{&zFzs
zfgvH!F9x*^09Dp809-P{GyoCLF9x*^007r90740_p8yfim(l5g8v)ao?dgHX0pPc_
z>VZE20sXi2>w!)I0sWUm?SYd4;J4WAfe8Tt;J4!LfmH#2elG^K4gdip_GthgJvnLs
z5q>Wc5orJ)GQu<f5q>WQwGIG_0f)o)01>_~ivb$?%>WU&FCq!9p8yemF9x*^03v-z
z+5i!OF9x*^03wG-+5i!PF9x*^03wk{+5i!QF9x*^03w@6+5i!RFHL{~03xMG+5i!S
zFDd{407oM#^=SYi>w7-{5sxnhwGIG2BF>;<01=Nb2DJ_V002i-^=SZ8_=yJr5s;VQ
z@PQS7kS_+c4gd)QaKSeK0000007rxMX#gS%=m-H3mM;di4gfw(bZu||5t%OrwGIG2
zBF>;<01=rl2DJ_V0000-qxES3BI|oU01=%p2DJ_VJ|fPbVgM1HF9x*^0000-we@KL
zBF>;<01==s2DJ_V07u95X#gV5pke?KqAvyswGIFPN9Xlv0GGM(fg43f2li<IBI|oU
z01>(`2DJ_VJ|fPbVgM1kF9x*^0000!BI|oU01>+{2DJ_VJ|fPbVgM1lF9x*^00000
z!vU2V^8gWZm)-G!At7@w2DJ_Vn&2@2Tr$Em01<O92DJ_VMd2|3LJ6*)01<STAM$}4
z0e6=>@`1+zgtr6pfj<EOp|?ZyfldJdp_h{Nfs+A*w<h+12>}6ww=nmCRRMnvF9x*^
z00A!lX#gKRIcfkA4lfcBX#gMoYx4jR4lf3^4gep$^8Wx44lf3^4giY*hr{;(5n?Zk
z0UG+v01;d-A_=aa01*!_2DJ_VB7I2O01*){2DJ_VB8N!Y01*-|2DJ_VB9Tbi01*=}
z2DJ_VBAZCs01*@~O@IObBBfYJ+5iz1FHL{~03vZRuK*DiFDd{407r@aX#gVYdp`gX
zATI{B4gfwP&Y)rd5g;!HwGIFP07tw1X#i9Bi3b4@Auq+J4gd)Qn87yy00000Q+%)c
z01+XVT={_;e<v>nwGIG2OmuB<01+xL2DJ_VJ|fPbVgL~;F9x*^00000M+yFE03z#q
zKL8OfF9x*^06rqlpke?KE-wbP4gdfEM;rcW03yzyVgL~_F9x*^002iV{%HUr&Y)rd
z5i>6awGIFPM_>MF08{vh2LTZ@FU6-000{$#!8ZT^3IG5AQ+%)c01-5n`hgpNIxhyb
z4gfw(bZu||5j`&kwGIG2BF>;<01-Ve2DJ_V0000-t^R2MBI|oU01-hi2DJ_VJ|fPb
zVgM0AF9x*^0000-zy4_eBF>;<01-ql2DJ_V07udOX#gV5pke?KMlS}n4gdg0@&0K5
zBI|oU01;g;2DJ_VJ|fPbVgM0MT`vZ;4gdfEM-l&N03z#qKL8P4F9x*^06rqlpke?K
zUM~i<4gdfEJ|gRTKL8P5F9x*^06rqlpke?KUoQr=4gdfE0K)-@mNEbl)0gM^fgyj>
zE(Wy@0N?5{09-y8V*nA;E(Wy@0I=&Z09*tF9{>^4E(Wy@0C(&$0JZ_co1p*^1}}xS
z2mk;8006cDc1oxK5dtrvx(EOO0002C0eL4)01*H$K)MJ300000LjG&>01?zK2DJ_V
zckM9%LcQ|;01?$L2DJ_V0`4&YLJ0+~p8yfomyG*?7(Ci8R0nAQ2?H>=Hvj+t002|5
zzB&LA+b*g~XaEC|pnoOM7Xtu7BIFAJ5#27js15+x?lAyDb7xTi5#E>J`+*+;@V5Mc
zWdQ*PmuCHeiUAat@%@390Ts82{()`*0W!Dv|A9^c0W!Bo0fJ%y0Y$f}0)kEf0Y#Vj
z1A>zQ6_<?!f?EL;x9J3ea{&Prmt6*eTLBcexdwu}0RjOpmv#t(zyWfXHwl6`f0`}^
zwGIFQNDgWMA3Zs001=ul5)o+tivfqj_W%*XE{g#g`pp0lzAhpOuAcxAn=S^m4gey3
zNZJ4qoh~W>001I~NZJ4qo-PKp4gg0h4r%}*>w7-{5vVQ(wGIG2BF>;<01>Dz2DJ_V
z002HB>w7-{5x*`5wGIG2BF>;<695svE(Wy@00000!vP$#ZvYXGmjMcbAw7>S2DJ_V
zsr@kkTrSvQ01=Na2DJ_V;{7oILJ6*)01=Qb2DJ_V8~!l>7Pk#*0D)o$00000000pI
z01=ayUkZZNe>^S*wGIFQO%ZAUA3Zs001-Sc5)o+tA0r4)01-Sc2DJ_Vivfqj_W%)o
zE{g#g`pp0lcP=6cuAcxAJuU{d4gey3NZJ4qKQ0Ef4geyDNZJ4qKrRNg4geyNNZJ4q
zK`uf7001JJNZJ4qLM}~!0stbVNZJ4qLoNoj4gex?Ofs(k5kxKqwGIGBED&k{BI|oU
z01-|u2DJ_VJ|fPbVgM0NE(Wy@0000-UJz;kQ}~Go0TE9w#itGc2?MCXHvj+t002{b
zuloQIPnZ7-f*XHcE(Wy@06t7~ZEye)VJ-%>4gfwP&Y)rd5n(O{wGIFP002j<5NZG-
z>w7-{5oIn0wGIG2BF>;<01;&_2DJ_V002k65NZG-&Y)rd5oRt1wGIFPN6-*z03yzy
zVgM0oE(Wy@002ku5NZG->w7-{5qK^JwGIG2BF>;<01<z9E(Wy@0000-5D{tsBI|oU
z01<gE2DJ_VJ|fPbVgM0&E(Wy@0000-FcE41BI|oU01<jF2DJ_VJ|fPbVgM0(E(Wy@
z0000!BI|oU01<mG2DJ_VJ|fPbVgM0)E(Wy@00000!vTwYHUJSZE(Wy@0MN|{00000
z0L}vd005U54T29nF)jwR4gmW8F#ud62u}bJF)jwR4geznG5|seuAcxAGA;(S4glu>
zG5{8Q4{8A9%?JPh0001>V*>yYHJ4-!g3tj%mo*N8n*md|(GG%60RdCDAP|B#0Rcjn
zSP_Ef0YbMx5`tg>fA}p1wGIFQc^GN{A3Zs001@~t5)o+tivfqj_W%(fE{g#g`pp0l
z94;aWuAcxA`7H*u4gey3NZJ4q`z=BM001I~NZJ4q{4EBx4gg177-|3_>w7-{5e6;>
zwGIG2BF>;<01*Z*2DJ_V002HB>w7-{5gjfDwGIG2BF>;<e*h63E(Wy@00000!vR|`
zqyQ1^Ee5p?0I*;O000000JQ)B0000000ErS8EODrJvnLs5$!DowGIFo3Nip(F4$oJ
z5$!DowGIGM3Nipf39g?25$-JpwGIG_3Nio|=oe}Lv0w)P00000P!9nB5%Dd)8EOCl
z`WI>d!wND0m!1@YQUMK@@f3oZf0``@wGIFQq#J4gA3Zs001=uk5)o+tA0r4)01=uk
z2DJ_Vivfqj_W%*xEsFsf`pp0l)h!|kuAcxAn=J;l4gey3NZJ4qoh=5n4geyDNZJ4q
zo-GEo4geyNNZJ4qpDjWF001JJNZJ4qpe;>+0stbVNZJ4qp)Cfr4gex?HZrdO5uz;y
zwGIGBgBof8BI|oU01>P$2DJ_VJ|fPbVgM1WEe5p?0000-wHj&wQ}~Go0THd2Ko)`(
zf2}PBwGIFY1F*q200000002kh8fpL{3+M;|5xgx1wGIG2OmuB<01>|}2DJ_VJ|fPb
zVgM1pEe5p?00000M*|yb03z#qKL8QKEe5p?06rqlpke?K!z~824gdfEM->}t03yzy
zVgM1uEe5p?002iP8)^U|&Y)rd5yveC2DJ_V07plc8W(~d1xIHaY5<ox7lIo_M~53~
z03z#qKL8QfEe5p?06rqlpke?K*ewRN4gdfEJ|gRTKL8QgEe5p?06rqlpke?K*)0aO
z4gdfE0K)+f>`nj?kC$B+f*~P~Ee5p?0BQ>|09+#oPXG~*Ee5p?0Fw+d0740_p8ye%
zm!B7c8v&D-yBC7c0ic(R7=oJtv9~B0f=&SevA1^`f;RyHpqH;3g69FCx0f7(U;%$t
zEe5p?00Gb<Y5*TSIcfkARxJ_{X#k4>hr{;(5q>R;0UG+v01<jEA_=aa01;O$2DJ_V
zB7I2O01;U&LI3~&B8N!Y01;X(2DJ_VN3$Yo03z#qKL8PAEe5p?06rqlpke?KWGx1@
z4gdfEJ|gRTKL8PXEe5p?06rqlpkfmM5qm8LwGIFP006@QsM=%z5lfee9)ckuODzVq
z4giT3G5}mI*kJ$>ODzVq4gkUyG5|seuAcxAOqa_Zf*S!(m**aWU;%5FNFRclf3+$G
zwGIFQVJ&I^A3Zs001>q+5)o+tAGk+R01>q+2DJ_VAGw`R01>q+2DJ_VA3`ID01>q+
z2DJ_VAG55~01>q+2DJ_V9|z|T01>q+2DJ_V9~5*r01>q+2DJ_Vivfqj_W%($EsFsf
z`pp0lE-fNA0Sf^Uwkig-4geyff8pZ*5w|J^wGIFx9NNJE5x6P_wGIFx39g?25xFV`
zwGIFxeMs5>5xXh|wGIFxhe+B05xgn}wGIFxkx1GA5xpt~wGIFxn@HLK5xy!q0000Y
zrAXQU5x*)r0000YaWby}5x^=2wGIFxK1tdD5y2|Ds15)kIWw;S5yC1Rf3*$(B3DV;
z01?9~x~L8SB6E@i01?D09kmVsA}2Nq01?G12DJ_VA~KW&01?J22DJ_VN023I03r+M
z2muk+Dh9O<06t7~ZEye)*(wIL4gfwP&Y)rd5!osRwGIFP002j@C29a7>w7-{5#1^V
zwGIG2BF>;<01@3P2DJ_Ve*geS!X;_|BF>;<01@CS2DJ_V07uj%Y5*e6pke?K;wlEU
z4gdg00w!tzA`9pU0TJ>l2DJ_VK1_6NZ~zhYDh9O<06rqlpke?K^(qFn4gdfE07oJw
zY5*eZdp`gX`6>pr4gfwP&Y)rd5&0?xwGIFP07o<?Y5*e6pke?KfBY&2wGIFPM@A-U
z03yzyVgM2SDh9O<002i=CTaj8&Y)rd5fLi}wGIFPM}{V903r+M2muixD+aX=06t7~
zZEye)Br68B4gfwP&Y)rd5hN=HwGIFP002j{CTaj8SyV~@5h^RbNNE5*BI|oU01+%J
z2DJ_VJ|fPbVgL~=e=7#H4gdfE07v2`Y5*cxR7wC5GAq7FX#hSV>w7-{5i~0XwGIG2
zBF>;<01-4R2DJ_V0000-^d@QmBF>;<01-wj2DJ_V07nKVY5*e6pke?KNGk@l4gdg0
z7$<4~BF>;<01-(m2DJ_V07phAY5*b&=m-H3SStp#4gfw(e{^kd01;a&2DJ_VJ|fPb
zVgM0aD+aX=00000M`kB#03z#qKL8P5D+aX=06rqlpke?KUn>T+4gdfEM|dY{03yzy
zVgM0hD+aX=002jdCu#s9&Y)rd5pOF7wGIFPN1P{W03yzyVgM0*D+aX=002j>Cu#s9
z&Y)rd5q~QNf3*$(07t+lY5*e6pke?Kf-45K4gdg0(kE&FBF>;<01<;L2DJ_V07n2Q
zY5*b&=m-H3k}C$a4gfw(bZu||5tS<jwGIG2BF>;<01=fd2DJ_V0000-ASh}8BI|oU
z01=rh2DJ_VJ|fPbVgM1DD+aX=0000-GAL>QBF>;<e*h7jD+aX=002isC~5#A&Y)rd
z5uGarwGIFPM^-3m03yzyVgM1SD+aX=002jXC~5#A3+M;|5w|M_wGIG2OmuB<01>(?
z2DJ_VJ|fPbVgM1kD+aX=00000N2Dlf03z#qKL8QFD+aX=06rqlpke?KzAFZ`4gdfE
zN46+xSO6l<pke?K!7B!}4gdg0$S7(6BF>;<01?6~2DJ_V07u#=Y5*e6pke?K%_|1A
z4gdg02q|g+A`9pU0TJCR2DJ_VK1_6NZ~zhDD+aX=06rqlpke?K;Fqx^f)jt{E51l+
z06rq?dp`gX=qm=b4gfwP&Y)rd5$G!hwGIFP002j1DQW;BSyV~@5$-F#NNE5*BI|oU
z01@yj2DJ_VJ|fPbVgM2FD+aX=00000M|LS{03yzyVgL~WEC#g>002jbDQW;B&Y)rd
z5d<s-wGIFPN17>W03yzyVgN1?1uO=&4gdg0$SG<7A`9pU0TC1|2DJ_VK1_6NZ~ze(
zEC#g>06rqlpke?K7ngA*f)jrmEC#g>06rqlpke?K8!QI34gdfENBSvh03yzyVgL~y
zEC#g>002h}Drx{C&Y)rd5g#lDwGIFPM;<C_03yzyVgL~~EC#g>002iYDrx{C&Y)rd
z5kD*jwGIFPM?xxU03wmRd;k$vEC#g>002jDDrx{sbZu||5nn6@wGMv(J|gRTKL8P5
zEC#g>06rqlpke?KUn~Z-4gdfE07rr<Y5*e6pke?KZ7c@04gdg0lqzZfBF>;<01<F3
z2DJ_V07s@OY5*e6pke?Kb1Vk64gdg0xGHJ@BF>;<01<R72DJ_V07uFyY5*e6pke?K
zi!27U4gdg0+$w4SBF=xHVgM18EC#g>002ksDrx{Ck-U5W5vwc)wGIFPM;t3^08Dgk
zZ~zgsEC#g>06rq?dp`gXv@8a-4gfwP&Y)rd5wt7@wGIFP002iyD{25D3+M;|5ydP9
zwGIG2OmuB<01?P62DJ_VJ|fPbVgM1yEC#g>00000M|dl003sk+R7wC5&Mdx2X#hSV
z>w7-{5zs6KwGIG2BF>;<01?oaAt!<pf7UF%NNE5*BI|oU01?<M2DJ_VJ|fPbVgM1?
zEC#g>00000N4P6$03yzyVgM29EC#g>002kID{25D&Y)rd5$r4mwGIFPN8BrF03yzy
zVgM2CEC#g>002ksD{25D&Y)rd5%(+xwGIFPM*u8p03yzyVgL~VEe5p?002i4L@a6m
zBF>;<01*r=2DJ_V07oP&Y5*e6pke?K7cB<04gdg0L@a6mBI|oU01+T92DJ_VJ|fPb
zVgL~!Ee5p?0000-WS2K6f*%2dmsco)9|5G7cPN4%0>mtrm?(lCf8;D`03z#qKL8Ok
zEe5p?06rqlpke?KGc5+S4gdfEM+7Zu03z#qKL8OlEe5p?06rqlpke?KG%W_T4gdfE
zM<gw303z#qKL8OpEe5p?06rqlpke?KI4uUX4gdfEM?@`Z03z#qKL8OqEe5p?06rql
zpke?KIV}dY4gdfEB|akSdp`gXIxPmZ4gfwP&Y)rd5jrgfwGIFP005u@48HsT5lELB
zDS{z?NGb-k4glI0G5}mY7-IktNGb-k4gek*G5}lz1RnqqNGb-k4ge7wG61##!<(T1
z5v?lB>;(V-0002C0d`8L01>GwQ|tu*00000wgGu3O#l(6D&6Y^0000007AG&Q2-H1
zDh9O<0DK%W07AK)PXG~0Dh9O<06`rx0761WBZvSIODYDn4gdxoG5|uetkeJzOezMo
z4gkv@G5|sc=MMl8O)3Vp4gioJG5|spbT|MJPAUeq4gggkG5~=B39g?25l@yXf)zYd
zDveHR00{$NxHkX*0000}vA#M05mYLlPHF%Hk$`_Cz-0gcLL%e~0TEa#x~L8S8X+<O
zLUU(P01;W2Q7VEY0fd)!DuOuyikGk|f;0htm*pygv;ldSnk#~|0du!cEP|N<0jQVh
zErQGeuD7->g53cD;kT49f_DM|Gq(&gf@lT-@3(b1g4zNB*SAnSf;RyHt+$9hf`9=5
zShop3f)@(`_Lm__f=GYbCI+<*00B^1Y5*TSIcfkA+9nbaX#gMA0sa6H+9n3I4gepN
z_TvB%+9n3I4geoQBZvSI+9n3I4gepstkeJz+9n3I4genq=MMl8+9n3I4gen%bT|MJ
z+9n3I4giY*hr{;(5iu%@0UG+v01+rEBD8Vy01?|J2DJ_VA~$~l3jq<_CI+<*03!MT
zlmHRkCI+<*03y<1ng9{rCI+<*03xH|;{XxgCI+<*03saP!2l89CI+<*03r#lp8yfz
zCI+<*03v-z+5i#bCI+<*03wG-+5i#cCI+<*03wk{+5i#dCI+<*03w@6+5i#eCOQBB
z03xMG+5i#fCc1y94gex?GOqv;=q4St4gexPN!kDr=_UrX4gexKGp_&<>Lxk>001Ia
zN!kDr>m~-Z4gex^k^}$|>?XRX4gex2HVOa{?Is<y4gexDlmq|~?j{Dc4gg0xP-*}o
z3+M;|5&k9ywGIG2OmuB<01*Kv2DJ_VJ|fPbVgL~VCkB7D4gdfE07qL;Y5*eZdp`gX
z1t$iz4gfwP&Y)rd5d|j(wGIFP07q?5Y5*e6pke?K2qy-$4gdg0e^6=wBF>;<01*l&
z2DJ_V07tD*Y5*b&=m-H38Yc#|4gfw(bZu||5gjK6wGIG2BF>;<01+K02DJ_V0000-
z%}{CpBI|#9KL8OSCkC|+06rqlpke?KAtwg44gdfEN8eCt03yzyVgL~&CkC|+002kv
zP-*}o&Y)rd5hW)EwGIFPM*~r603yzyVgM02CkC|+002i7QEC7pk-U5W5nv|<wGIFP
zM?+C+08DgkZ~zf$CkC|+06rq?dp`gXX(tA?4gh~XBF>;<01;^?2DJ_V0000-aZzdj
zA`9pU0TFs92DJ_VK1_6NZ~zf~CkC|+06rqlpke?KeJ2LB4gdfE07suuY5*cxR7wC5
zgD1X7X#hSV>w7-{5rroPwGIG2BF>;<01<^J2DJ_V0000-%~5IqB3V>Q01=BPzDQ{R
zJ|chXdp`gXjVA`R4gfwP&Y)rd5sfDXwGIFP002kdQEC7p&Y)rd5uYapwGIFPNAXc=
z03yzyVgM1LCkC|+002h=QfdGq&Y)rd5uzsswGIFPM-@_P03yzyVgM1WCkC|+002ir
zQfdGq3+M;|5xXY_wGIG2OmuB<01>_?2DN_<06rqlpke?Kz9$B?4gdfE07qj|Y5*eZ
zdp`gX!Y2l`4gfwP&Y)rd5yB@1wGIFP07rFFY5*e6pke?K#U}=}4gdg0hf-<)BF>;<
z01?tB2DJ_V07scpY5*e6pke?K-zNsO4gdg0t5RwJBF>;<01@IR2DJ_V07t!2Y5;#C
z&Y)rd5#=WawGIFPN6%7f03yzyVgM25CkC|+002k*QfdGq3+M;|5%nhqwGIG2OmuB<
z01@~n2DJ_VJ|fPbVgM2NCkC|+00000M;%jY03z#qKL8Q@CkC|+06rqlpke?K{3iyr
z4gdfEM=w)q03yzyVgM2UCkC|+004hSK~rh~BF>;<01*Hv2DJ_V07p|(Y5*e6pke?K
z3n&J)4gdg0fm3P#A`9pU0TCJ~2DJ_VK1_6NZ~ze<C<e6-06rqlpke?K9ViC14gdfE
z07s!yY5*eZdp`gXAt(m54gfwP&Y)rd5g{lBwGIFP07tV^Y5*e6pke?KBq)CdwGIFP
zN5xZW03yzyVgL~(C<e6-002kVQ)&Pr&Y)rd5ilqQwGIFPM+H=B03r+M2muj5C<e6-
z06t7~ZEye)LnsEd4gfwP&Y)rd5kn{jwGIFP002ibRB8YsSyV~@5lJY%NNE5*BI|oU
z01-<l2DJ_VJ|fPbVgM0KC<cGE4gdfE07qd|Y5*cxR7wC5Q7FDhX#hSV>w7-{5mP7z
zwGIG2BF>;<01;Ct2DJ_V0000-b5v>oBF>;<01;&<2DJ_V07r#XY5*e6pke?KXD9}>
z4gdg0msDy1BF>;<01;>?2DJ_V07u1CY5*b&=m-H3cPIw64gfw(bZvid01<j92DJ_V
zJ|fPbVgM0(C<e6-00000N99y%03z#qKL8PaC<e6-06rqlpke?KekcaD4gdfENB2}}
z03yzyVgM0=C<e6-002h`RcZht&Y)rd5rQZNwGIFPM;ldY03yzyVgM1lC<e6-002iV
zRcZht&Y)rd5yU74wGMv(07pMnY5*dUynFx=+b9OL4gdg0ZB=RjOmuB<01@OU2DJ_V
zJ|gRTKL8QrC<e6-06rqlpke?K<R}KU4gdfE07rjSY5*e6pke?K@+bzi4gdg0kyUB{
zBF>;<01@>l2DJ_V07s)$Y5*e6pke?K_$UUo4gdg0wN+{WBF=xHVgM2OC<e6-002kF
zRcZht&Y)rd5fCW`wGIFPN7+?s03yzyVgL~tDF(F;002kpRcZhtk-U5W5iltRwGIFP
zM;TUX08DgkZ~zfGDF(F;06rq?dp`gXIVlFU4gfwP&Y)rd5jiOawGIFP002ivR%!qu
z3+M;|5lSfrwGMv(K1_6NZ~zfaDF(F;06rqlpke?KO(_Po4gdfE07rFJY5*cxR7wC5
zQz^bkX#hSV>w7-{5mhM$wGIG2BF>;<01;Iw2DJ_V0000-qgHAFB3V>Q01;a$zDQ{R
zJ|gRTKL8P3DF(F;06rqlpke?KT`2~&4gdfE07tb}Y5;#C&Y)rd5pO95wGIFPN5@ub
z03yzyVgM0wDF(F;002kXR%!qu&Y)rd5ppR8wGIFPM+aAG03r+M2muj-DF(F;06t7~
zZEye)g((KL4gfwP&Y)rd5rruRwGIFP002idS84zvSyV~@5sN9lNNE5*BI|oU01=HT
z2DJ_VJ|ch4pke?KjVT7T4gdfE07qk2Y5*cxR7wC5lPSJPX#hSV>w7-{5tS(hwGIG2
zBF>;<01=fb2DJ_V0000-bysQtBF>;<01>At2DJ_V07r*cY5*e6pke?KsVN4v4gdg0
znOAB6BF>;<01>Sz2DJ_V07t7=Y5*e6pke?Kt|@;8wGIFPN7+|u03r+M2mukkDF(F;
z06t7~ZEye)!6^o{4gfwP&Y)rd5y2@2wGIFP002h^SZV+wSyV~@5yvUMNNE5*BI|oU
z01?S42DJ_VJ|fPbVgM1zDF(F;00000M>SY#03umbN&pehDZWT)06rq?dp`gX(J2PC
z4gh~XBF>;<01?qC2DJ_V0000-M_6hABF>;<01@LU2DJ_V07qF^Y5*e6pke?K<tYZW
z4gdg0YglRkBF>;<01@da2DJ_V07rdTY5*e6pke?K>L~`b4gdg0k63B|BF>;<01@;l
z2DJ_V07s!%Y5*e6pke?K{wW5v4gdg0vsf}}03yzyVgL~dDh9O<002kRSZV+w>w7-{
z5fUl}wGIG2BF>;<01*-@2DJ_V002kxSeM6Cf*t`Am*!N09|1F$233L|0aKSIRe~Qu
zb6IKtBI|oU01+)J2DJ_VJ|fPbVgL~>Dh9O<0000-lUZs2BI|oU01+-K2DJ_VJ|fPb
zVgL~?Dh9O<0000-vzLZdf*%6YS(m3(f*u0%S(nFEf*oZ?6IyBjBI|oU01-7R2DJ_V
zJ|fPbVgL~}Dh9O<0000-Gg@i@BI|oU01-AS2DJ_VJ|fPbVgL~~Dh9O<0000!BI|oU
z01-DT2DJ_VJ|fPbVgM00Dh9O<00000paPQ>WB?IwmkCyaA%AZs2DJ_VzJ4+QTs|0M
z01<B{2DJ_Vkclz?Tm%Fk01<B{2DJ_VB#SZtwgJPNp#TxmCV_PZ000000JZ^kN~i!4
z&L#wP1^@s6006cDc_&Q(5zHo#bOrzb0000&)&c$i5pX63wGIG}j4}X1l=kBQ5pgC4
zwGIGOjWPg2LPjHq01<K~2DJ_V8ICdlLbI&Y01<O02DJ_V-i|T=LI>v$01<R12DJ_V
zqmMEGLKJj301<U22DJ_VXpk}hfdUDxp8yedm+w}B6+C(-jZSI+2?L<GHvj+t002|5
zzB&LAdnTVwY5)U~fPW=mf&c(QBIFAJ5q>7Rs15)vkum^6b7xTi5r3CJSArw~u$N_5
zf;0iEmzr0Cv;n1;PgsJq0iTxyS%S0ynYX!Fg0TSs&6jIig31BYw>DgY_yGY4w=Z6T
z(E$NsmjPgcXaS73YG8u#1_4sHTxWvH0s&mNCu)LG0Rb|%d252~0Rfb^W^IBe0s+vs
zj&Fju0Ri2&Msb3e4FSHF-+qE5e=sBlwGIFQ7K~~DA3Zs001+@G5)o+t9|$*C01+@G
z2DJ_VAMvoV01+@G2DJ_VA3`ID01+@G2DJ_VAG55~01+@G2DJ_V9|z|T01+@G2DJ_V
z9~5*r01+@G2DJ_Vivfqj_W%)ECW`?Y`pp0lQzjy#;o|@iF(d}H4gew?f7-zS5i%qO
zwGIFx`T&#w5i=wPwGIFx(qWna5i}$QwGIFx39g?25j7+RwGIFxeMs5>5jP|TwGIFx
zhe+B05jZ3UwGIFxkx1GA5ji9VwGIFxn@HLK5jrF~0000YrAXQU5j!M00000YaWby}
z5j-RYwGIGBb%bgFA`9pUe*qCpBnGt(06t7~ZEye)Pb3Dl4gfwP&Y)rd5l<urwGIFP
z002joglYgH>w7-{5mO`vwGIG2BF>;<01;Cp2DJ_V002j)glYgH&Y)rd5mqDywGIFP
zN4bP*03yzyVgM0XBnGt(002klglYgH3+M;|5oja^wGIG2OmuB<e*h6{BnGt(06rql
zpke?KYa|A>4gdfE07nOfY5*eZdp`gXZzKk_4gfwP&Y)rd5pN_0wGIFP07n^xY5*e6
zpke?KawG<|4gdg0D}`zRBF>;<01<N}2DJ_V07qGcY5*b&=m-H3j3frN4gfw(bZu||
z5sxGWwGIG2BF>;<e*h7WBnGt(00000M~8)K03umbN&pd+B)&*#06rq?dp`gXmm~(Y
z4gfwP&Y)rd5tk$ewGIFP002j|g=zpISyV~@5uGHyNNE5*BI|oU01=-g2DJ_VJ|fPb
zVgM1JBnGt(00000N5_R~03yzyVgM1bBnGt(002kXg=zpIf6ky{01>k!2DJ_V07nOg
zY5*b&=m-H3#Uuu`4gfw(bZu||5y&J4wGIG2BF>;<01?O}2DJ_V0000-HHK;cB3V>Q
z01?h4zDQ{RJ|gRTKL8QXBnGt(06rqlpke?K&?E-64gdfE07qknY5*cxR7wC5)+D}2
zX#hSV>w7-{e-YRu2DJ_VJ|fPbVgM1?BnGt(00000M|Fm303yzyVgM29BnGt(002ja
zhH3yJ&Y)rd5$q%ewGIFPN417(03r+M2mulOBnGt(06t7~ZEye)03`;s4gfwP&Y)rd
z5db9ywGIFP002kThH3yJ>w7-{5d<X$wGIG2BF>;<e*h5#B?h$)0000-=Z0zkBF>;<
z01*cz2DJ_V07v<TY5*e6pke?K2qgx!4gdg0Cx>bPA`9pU0TCA^2DJ_VK1_6NZ~ze+
zB?h$)06rqlpke?K8YKp`4gdfE07plMY5*eZdp`gX9wi2~4gfwP&Y)rd5gsK5wGIFP
z07qGee`)|C&Y)rd5g{c8wGIFPM{9>_03yzyVgL~$B?h$)002jshiU*K3+M;|5i%tP
zwGIG2OmuB<01-7M2DJ_VJ|fPbVgL~}B?h$)00000N5_Y103umbN&pc%CB8^$06rq?
zdp`gXJtYRU4gfwP&Y)rd5j`aawGIFP002kze}`%SB3V>Q01-nazDQ{RJ|gRTKL8O$
zB?h$)06rqlpke?KMI{Ec4gdfE07nOiY5*e6pke?KS0x6u4gdg08Hj2CBF>;<01;Uw
z2DJ_V07ol`Y5*e6pke?KS|tXx4gdg0S%_)?A`9pU0TF2>2DJ_VK1_6NZ~zf(B?h$)
ze*iur&Y)rd5o{#}wGIFP002jMh-v^L>w7-{5pX32wGIG2BF>;<01<E{2DJ_V002je
zh-v^L&Y)rd5pyL5wGIFPN1cdj03yzyVgM0zB?h$)002kJh-v^L3+M;|5t=0iwGIG2
zOmuB<01=%f2DJ_VJ|fPbVgM1HB?h$)e*gdg07v<VY5*cxR7wC5qb0sbX#hSV>w7-{
z5v3&twGIG2BF>;<01>4n2DJ_V0000-Cy8nRB3V>Q01>MtzDQ{RJ|gRTKL8P}B?h$)
z06rqlpke?KttAGv4gdfE07p5AY5*e6pke?Kza<8>4gdg0ONnX#BF>;<01?3@e+IP<
z002i_iE02M&Y)rd5yB+~wGIFPM~jJS03r+M2muk&B?h$)06t7~ZEye))g=bC4gfwP
z&Y)rd5!EFIwGIFP002k1iE02MSyV~@5!)rcNNE5*BI|oU01@3K2DJ_VJ|fPbVgM1{
zB?h$)00000N9T!Z03umbN&pe#e<i+1X#hSV>w7-{5#=QYwGIG2BF>;<01@RS2DJ_V
z0000-`H5-(BF>;<01@{k2DJ_V07naoY5*e6pke?K`6ULm4gdg0D~f6WOmuB<01^5n
z2DJ_VJ|fPbVgM2PB?h$)0000-S&C`^A`9pU0TBx(2DJ_VK1_6NZ~zewe<lXC4gfwP
z&Y)rd5e_B>wGIFP002jaifRBNSyV~@5fmoANNE5*BI|oU01*}@2DJ_VJ|fPbVgL~q
zCI+<*00000N41J-03umbN&pcYCca2%06rq?dp`gX9wr904gfwP&Y)rd5gsN6wGIFP
z002kFifRBN&Y)rd5ilkOShWrS07u!1Y5*e6pke?KGA0JK4gdg0`HE@)OmuB<01-1L
z2DJ_VJ|fPbVgL~{CI+<*0000-8H;KFBI|oU01-VV2DJ_VJ|fPbVgM06CI+<*0000-
zIhPEFf*%1{mn?^ZA3=GGY5*eZdp`gXRwf3u4gfwP&Y)rd5mqJ!wGIFP07sdNY5*eZ
zdp`gXS0)Cv4gfwP&Y)rd5mzP#wGIFP07to(jE90A1xMM7Y5<q3hk_eLNBN6t03z#q
zKL8P3CI+<*06rqlpke?KT_y&#4gdfEJ|gRTKL8P4CI+<*06rqlpke?KUM2>$4gdfE
z0H6Y|O8@~8#h1^Af+2s!BL=k&09z6>09-y8V*nAwBL=k&0Jjt~09*tF9{>@>BL=k&
z0EZSc0JZ_co1p*^C?v5E1^@s6006cDc1oxK5hNr$5C#AM0002C0eL4)01+Z2$PWep
z00000LI^il01?I`2DJ_VL>MywLh-P&01?L{2DJ_VF&Q%eLPCEdhyW4DBL=k&03I4M
z07A2@)Bq95BL=k&01F#4073`n4*(I$BL=k&0Qeg-074XWH~<mLBL=k&0OcGr0D%Gt
zuAcxA%p(T14gdgNFlzt+X_0CGQ!D7101?n5jZSI+2?OA`Hvj+t002|5zB&LA(IcNu
zY5)U~fPW?6YXBGkLL%e~0TI+Ax~L8S(j7AZLUU(P01?%f28n_s0S1>ViGnl&0VJ1U
ziGs5M`<E4pg0unkm&%HQv;pt8e~W^#0RbbIF^z)C0V%iqje@WN0av%JkAgM=0iL(y
zk%FlK0oS*VlY;dF0YtZimxB5N0j#%wnt}-e0pz!doPq)Z0UfuDo`Tf}0g{(urh+5^
z@t23Df+2tLArcX305)(l09>Qt;{XxyAqKS$0Pk=!09*>n9{>^YAqKS$0K{}N09>Z@
z+W-;qAqKS$0Q`0{0JZ^#!}kCYu_I)-1^@s6004^t8v4xu5u+nQRA^WL5%M7hwGIF@
zcQXJ&5!v7X5%VDiwGIIBcryS(g0@ot5%eJjwGMv(&3iKdLJ6*)01@>e2DJ_V!F@9T
zLVZZu01@{g2DJ_VM}9K^LWfA&01@~h2DJ_V7lShZLXk+?01^2i2DJ_V(T6htLYqk1
z01^5j2DJ_VN{cf9LZwLB01^8kIsgCwFOV|;LUA&$01^Bl2DJ_VQIazNLOw~_01^Em
zx~P8+0Q{IU075x4uK*GLAsw|20C}4;076$u+5i#%AqKS$0Q#LX077$;1OO2LA_lb%
z0BN8z0754=3IGuSA_lb%0PdtS075d91OO2NA_lb%0Dq=4075T65C9PaA_lb%01&7%
z07A%=1OO2PA_lb%00F8q0CsDv1^@s6005UIsDcs!7neAwf*=7JmszNSAORYedZ>ac
zHXtItNNE5Ak$`_k9<6EsBI|oU01+Z02DJ_VJ|fPbVgL~$A_lb%0000-Fs*6;BF>;<
z01+gY+o*y#0V$Uhse&W{E0;N`f*=7ams+WUHUTY{k*R_#0V<cise*0+K9?=3f;9m^
zmt(4eH33zZnW};*9c&^7wGIIBtup`?01#*Z-meA#00000P>cWo5pI{?s)Abqb(cG<
zf*=8Pmt3oYAOUrkeXD{u0f(2htAaHFjhE}If*=8qmkg|eAOVn<Ev$ko0hX6ptb!x~
zn3sO6f*=8ym!GVHHUXNK)~tdh0h^cntb#lNpqDwVf+PW=mtL)cAOWJ6f31Qx0i>6<
zt%4;1rI*~Tf+_)_mj|wbZUMQMcCLao0lSy1u7WiIyqDsxf+YdQmk6(dH38X|KCgl)
z9q=LswGIG~vNHe{01#*Zg|r3$00000pp5_k5%QOWuYy}5`633j4gfN<GXN{{um%7C
z0000L000000000&>w7-{5&4$^u!0}~`IjQFf;RyPms+rbH31Hnkg$R&D;gsPwGIGQ
zv@-xD;En(QVRUV901+J{2DJ_VjI=WVCE$(#07A~7VgL~xm*lX5D*+>y5V3+3GbJMi
zwGIG2BF>;<01+i42DJ_V002jtwQ2w&&Y)rd5ho)CwGIFPN36AK03yzyVgL~+mshcZ
zKmje6n6ZK+0WX)mv4S80FPGl2f+hhmmj$wdB>^&*E3$$v0WOzXvVw8}Lzm66f;Rz0
zmj$zeH33VPJF|iy0Zx}(vw|Q2PM3YNf;U=JBcD!c00WVLe@892Y5*eZdp`gXRwD+r
z4gfwP&Y)rd5mqAxwGIFP07pN#Y5*e6pke?KSR)3t4gdg0QMhUVBF>;<01;Uu2DJ_V
z07qlEY5*e6pke?KT9*#Af;j<Smq@gNBmrTUZM1?Q0b!Spw1PGPW0%FWf+YcDm+Q2G
zB>`oZ5w(IW0brLtwSsa1f0wbff;Rzym+7^FH35Q`AGU%uB#a{lwGIFTx-$R(N6xxx
z03z#qKL8PxBL=k&06rqlpke?Km6v$7f+7K=mz%bNAOWP8ytaZG0i~DMwt^r5rI+}&
zf;IuBmoK-1AOWVAP`8360jHODw}K!6r<a(wf;IuDm(RC?AOWbC@VA0C0jZZKxPl-7
zsh3E&f;a)Hmw~u~AOWhEqPT)40jrnExPl-7tC#Ayf;b|qBL=k&0JOX_04wIV1^@s6
z0015U00000002VHpke?Ktd~T&f;a)KmwmZ{AOWqHp1FcJ0j`(Vxq=`8u9x_^f;IuN
zmoK`4AOW(MP`ZLR0kfBhx`H49vzMy6f;IuPm*KjCAOW<O0=t4O0fo0PyMj^!0jHOP
zzJeitry&Nl4gfK~GXPvZ7-Iktry&Nl4geLvGXPu!1Rnqqry&Nl4gmSUGXS;$!<(T1
z5$Pdh$_4-c0002C0d`8L01@ROG|C1500000wgGu3O#l(&A;8H700000079eT;{Xw;
zAqKS$0OP?k0744N9{>@lAqKS$058Hb079lN^xFUtsv!oo4gh?@GXO#fuAcxAt04xp
z4gko)GXP#tkN^Oa#cBXkE9jX35wDj~zk(G3u$O1Qf*=vIA-bpz006@?077$TQ2-IN
zmzckTBmvEry}yDq0m!%Vzk+B10pyok!GhZXe3v)Ef;@jAAO^J#00G|2Y5*TSIcfkA
zAs`YFX#gKmjGq7zAs_~|4gent${zp`As_~|4geoVDuMtJAs_~|4geo|R}cUZAs_~|
z4gepf^xFUtAs_~|4giY*hr{;(5tAW{0UG+v01=2GB2;Kt01+Y}2DJ_VA`#i(01+b~
z2DJ_VB7%RmQveYpAO^J#03r#lp8ydhAO^J#03v-z+5izJAO^J#03wG-+5izKAO^J#
z03wk{+5izLAO^J#03w@6+5izMAUXg503xMG+5izNAO^J#03vZRuK*D&AiAgy03tp~
z+5izPARV<103tavuK*D)AO^J#03ugO+5izRAO?T64gex^k^}$|FdzoC4gex2HVOa{
zF(3xD4gg09$Z7y03+M;|5kepawGIG2OmuB<01-tX2DJ_VJ|fPbVgM0EAO^J#00000
zM>oi703umbN&pc{AihXx06rq?dp`gXO&|uf4gfwP&Y)rd5ltWlwGIFP002j2$Z7y0
zSyX>Y01;ClzDQ{RJ|gRTKL8O`AO^J#06rqlpke?KRUihn4gdfE07rMoY5*e6pke?K
zXCMZ(4gdg0iO6aIBF>;<01;^*2DJ_V07sk1Y5*dUynFx=k{|}P4gdg0$;fH|OmuB<
z01=xY2DJ_VJ|gRTKL8P%AO^J#06rqlpkjXj5t|?ewGIFP002k#$Z7y03+M;|5wIWz
zwGIG2OmuB<01>kw2DJ_VJ|fPbVgM1dAO^J#00000M@h+Q03umbN&peLAYxHz07odv
zY5*eZdp`gXyC4R&4gfwP&Y)rd5xXD;wGIFP07p8>Y5*e6pke?Ky&wj)4gdf?BF=xH
zVgM1oAO^J#0000-o5^YbB3V>Q01?9=Vo_-TM|#O>03z#qKL8QMAO^J#06rqlpke?K
z#UKW?4gdfEM~um803yzyVgM1xAO^J#002HB&Y)rd5y&70wGIFP07tFKY5*e6pke?K
z+#m+E4gdg0zsYI<BF>;<01@6G2DN_<002kP$!Y*1&Y)rd5#JyNwGIFPN8`zA03yzy
zVgM28AO^J#002kz$!Y*1k-U5W5fC8;wGIFPM<dE=08DgkZ~ze*AqKS$06rq?dp`gX
z86gI>4gfwP&Y)rd5g8!{wGIFP002i(%4z^23+M;|5ilVJwGIG2OmuB<01<yPAqKS$
z06rqlpke?KGa&}G4gdfE07tdTY5*cxR7wC5IU%1;Y5+%q%4z^2>w7-{5j!CUwGIG2
zBF>;<01-PO2DJ_V002jn%4z^2&Y)rd5j`OWwGIFPN2bbZ03yzyVgM07AqKS$002HB
z&Y)rd5kDaYwGIFP07n_iY5;#CSyV~@5kw)MPHF&0=E`aSBI|oU01-wZ2DJ_VJ|fPb
zVgM0FAqKS$0000-_{wSkBF>;<01-$b2DJ_V07nYTY5*e6pke?KNg)Qc4gdf?BF>;<
z01-+d2DJ_V002iT%W423&Y)rd5o93-wGIFPM?K4G03yzyVgM0lAqIc74gdg0Ps?fm
zBF>;<01;;)2DJ_V07qfVY5*e6pke?KaUlk^4gdg0fy-(DBI|oU01<j22DJ_VJ|fPb
zVgM0(AqKS$0000-q04FjBI|oU01=5H2DJ_VJ|fPbVgM0|AqKS$0000-!OLm@BI|oU
z01=8I2DJ_VJ|fPbVgP>;iXjHI4gdfEN8!tA03z#qKL8PnAqKS$06rqlpke?Kiy;QJ
z4gdfEM*+-g03z#qKL8PoAqKS$06rqlpke?Kj3EZK4gdfEM<L8=03z#qKL8PpAqKS$
z06rqlpke?KjUfiL4gdfEM?uVL03z#qKL8PqAqKS$06rqlpkjXj5so1SwGIFP07qfW
zY5*eZdp`gXk0A!N4gfwP&Y)rd5sx7TwGIFP07rq$Y5*eZdp`gXkRb-O4gfwP&Y)rd
z5s)DUwGIFP07s$BY5*eZdp`gXlpzMS4gfwP&Y)rd5tJbYwGIFP07t>hY5*eZdp`gX
zl_3VT4gfwP&Y&=201=fT2DJ_V002HB>w7-{5tbnawGIG2BF>;<01=iU2DJ_V0001>
z0+H4o01?5Lm&k%4f59IHwGIG0#xnq1J{V&F5y2k@wGIH+$1?z21Oy)d5y2k@wGIG6
z$uj`90mGZ201+7=8~FwR00000wgGlZr~nZaAXESc000000JZ^nCrtnm6CmLK1^@s6
z002T#jGq7z!XE~;4gl85GXO#g${zp`!yg8<4gj0WGXO$Ie=33i5yT$`wGIGa%rgK&
zdRGtt5yc+{wGIF!%`*T(ru5qY5yl?|wGII4%`*Uj0tv3401?L@2DJ_V0AA3m0030c
zY5-F!=$QZ!%O6$ZY5)lXK)E*n00000Q?b4}01?a|bWv#l1CfA#C2*_&074?<3jq<&
zAG)Xx0I$w72mnHJXHftV(3if+f+PX?m*dHTGy(LN7|MdQ0q&R0%7U~3=(mE)f|UUQ
z8n+nDf~f%kRJX#;f;j;Jl$ZR^f@lH6w`I_R=m7zxw@%W6MF9a4mz2|jXaPmH{nLUE
z2>}$B4c&s<0brNu-GVlMh93sC4gdiV<7xmOJvnLs5r!WU5orJ)8QdxW5r!WIwGIG_
z0f)o)01>bsivb$?%>WUsA0i2^p8yeu9|pA!03v-z+5i!WAMnBj03wG-+5i!XAMnBj
z03wk{+5i!Y9|pA!07vuUY5*eZdp`gXl^+JR4gfwP&Y)rd5tSb+2DJ_V002HB>w7-{
z5v?BvwGIG2BF>;<01>Sp2DJ_V0002P0oXE501<na$=-q?A$uPNwGIFZ4Kx5;8QdxW
z5qlp7wGIGk4Kx5k39g?25qy^m-+~(ff0rlUf^GqpmzLjxfdM*~7~q0v0X&z1;DTfU
zLYL;?f@T41mrdb<!vTG_5#oYC0s(!OQsjdD0e-ho<$|*U0e+VR=z>ZCez#BPf&l>m
zez#HSg4h88LYFG*g8l(Iw<+v`nF0Yqx5Ms&zySe?mxS<wXaS0s?eKzV0hYH`@q(KK
z0hYJ&^@5cG0hqUl_<}_T0hqVW|ALzY0hqUH1cR9Z0hqT02ZNOY0iBnc34?S2`<D+2
zgDwI8mp%%EDgpnOX$pgQ0mYZ%3WF*E%eM^+gLMG{0Z+Gt41-z$0gAWZ4TG5i0g1PH
z5QB09e?T7wwGIFQ+!$*BA3Zs001-eR5)o+t9~s;#01-eR2DJ_Vivfqj_W%)UABzDR
z`pp0lW*;I6uAcxAK_3RS4gey3NZJ4qLmy5G0{|k2NZJ4qL?2EH0{|kCNZJ4qMIQ#W
z4gg2L7;69`>w7-{5l<fmwGIG2BF>;<01;0gD+aX=0000!BI|oU01;;&2DJ_VJ|fPb
zVgM0m9|pA!00000paKQ4QUDP(mz5NQAw4x82DJ_V*kLpPTp8Rd01-7G2DJ_VIASya
zLJ6*)01-AH2DJ_V+F~>S7Bd)Y03k{U000000C0~201-Kt^Av+_0Z^A>6@!5R%a`O8
zgJ=QGmrWLfWC7EcvKE780sWU57lTp(4wq~fgLwg<m*y9PM**L=E*OIz0Rj4#O&Nn<
z0r<DA8H30H0r<Ce8-pnU0gjiT9D~OJT$gAagJc0-m(3l6<pEr`vmS#|0s&mN1|fs$
z0RaxT@gjpj0s#(}G9`ok0S~t@CWEsA0S}ktD1%7>54SEUg8=~n54SKXgWdrF)0Yn{
zgZu%@w+t<VnF0aRx27+HqyYg#mrF5&XaPi*v@wHd0Z_LcGJ~500Z_NOHiMM{0aCY5
zIfF$90aCY>K!ckE0aCXyMT40F0aCZiM}wIH0YtZDN`sjK0YkU{OoNRBfBYT>wGIFQ
zWmRhcA3Zs001^Bi5)o+tivfqj_W%(jABzDR`pp0lARi(LuAcxA{T>Fj4gey3NZJ4q
z{~m$g4geyDNZJ4q03Qam4gg0;Rcinu>w7-{5egp$wGIG2BF>;<01*lw2DJ_V002HB
z>w7-{5g{K2wGIG2BF>;<695q*9|pA!00000!vQN>nE(;-mmX1rAtCV|2DJ_Vt-Uk=
zTrSvQ01@#X2DJ_V=DjokLJ6*)01@(+U{Qk`0ri)BQG-$e5tq+VgRFnF9tO1z00D$o
zYXBcTIcfkAv>p->X#gLi;o|@iv>pbv4gen<+Q9%3v>pbv4giY*hr{;(5#Sz+0UG+v
z01@0CA_=aa01>qw2DJ_VB7I2O01>wyLI3~&B8N!Y01>zzLI3~&B9Tbi01>$!DgXcg
zBAZCs01>(#2DJ_VM`uG<YXBnadp`gX#2yB<4gfwP&Y)rd5yTz_wGIFP06rq?dp`gX
z-5v(D4gfwP&Y)rd5#1gJwGIFP006@QQd+?P5qXzAQ-dLYc^(F}4gfB{Gyq&a7-Ikt
zc^(F}4gmkZGyq%#1Rnqqc^(F}4gki$Gyt{%!<(T15v(3>z6SsR0002C0d`8L01>Dj
zIll)000000wgGu3O#l(59@xGI00000079eT;{XwQ9tO1z0LH;I074wv!2l6^9tO1z
z06W4o0D%I339g?25qus7wGIFPULcYI0QOmH08=aInE(-h9#jWu00{$dxi<g+0000}
zvA#M05rH16N@xHBk+6RyV3Gj<LL%e~0TF{9x~L8Su);I|LUU(P01<>9#;6VeT*EW~
z8XQk(087CK000000672v0000001=iRL0M}60a{PDTvUTy0RgSIx>bV~0Rg9%(N=@z
ze>@%rwGIFQz+7tpA3Zs001-SM5)o+tAEV*p01-SM2DJ_V9~|1j01-SM2DJ_Vivfqj
z_W%)S9*Y4Q`pp0lWF8_3uAcxAJst+N4gey3NZJ4qKORB=001I~NZJ4qKpsK>001J9
zNZJ4qK^{&D0{|kMNZJ4qLLLUS4gg1^L|khCBI|oU01-?c2DJ_VJ|fPbVgM0L9tO1z
z0000!BI|oU01;&#2DJ_VJ|fPbVgM0k9tO1z00000!vXY)aR3nk9+!GpgCKta9tO1z
z08Ym=09-y8V*n8W9tO1z0Efsl09*tF9{>>n9tO1z08Ys?0JZ_co1p*^G#+yc2mk;8
z006cDc1oxK5ilMJ4+sDN0002C0eL4)01+-8s167K00000LZjj101*Nn2DJ_VPRcX@
zLLAz`01*Qo2DJ_VugWw4LJ0+~p8ydAmmXPz7+eS*R0nAQ2?LP1Hvj+t002|5zB&LA
z2_C9SXaEC|uzw{$3IYH^BIFAJ5epu=s15*J%QOH&b7xTi5ey#2s15)K%rpQR98YKf
z3K0kZ00000IRF3v000005gxaaS%YN(0X3KST7!fEL6?17gJ=Ozx8_@eiv$5sw{l>E
zg#rOpm-u3XcmXn(abtrn0XCPQV}m9EIJd`RgKPl-LANwzgH{6pFPDpGgTVn0mrZGd
zIDf<)2DJ_V0gi5K03SU$Y5)<$9TE{~03V~_;{XxF9R{@y03RIM!2l7&9R{@y0E+>K
z!}kCY@EwZ*8v4xu5$qiz39g?25yc$_wGIFxeMs5>5yu@u0000Yhe+B05y%}v0000Y
zkx1GA5y>4sKnDOKn@HLK5y~9~wGIGBaYJrv03z#qKL8Qb9R{@y06rqlpke?K)Ex%3
z4gdfEJ|gRTKL8Q!9R{@y06rqlpke?K?HvZS4gdfE0K)+?F}nZ}iI+!egCT#39R{@y
z0PNp109-y8V*n9}9R{@y043ox09*tF9{>@F9R{@y0PNv30JZ_co1p*^ydB8-2LJ#7
z006cDc1oxK5x5;8{|5j70002C0eL4)01>tw!2SmS00000LZjj101=8E2DJ_V?BX;4
zLLAz`01=BF2DJ_VVdFFaLJ0+~p8yezm+WeT7(9?2R0nAQ2?MaXHvj+t002|5zB&LA
zksYc^XaEC|pnoL*@BsiqBIFAJ5tALds15)k<TL<6b7xTi5tNreYl9yFrng{ggJl5$
zy_d#pgM<OemrHGfXaU%_v~7cn1OeE$KyrhH0s-BZ#&m;p0sNOcb%QDb08N)?b%S>S
z#+To9gDL^amkM@+Tmj6tXLf^M0RjL`m&kX6DFFhv@^^z<0RhRkPI-e@0|B>}rhJ3J
z0sWU~eS<iExEltw4gdkMifaHLJvnLs5x5%?5orJ)qv7KK5x5%$wGIFu9NNJE5x5%$
zwGIG_0f)o)01<N?ivb1J%K#Ci9E$<U8fE|yp&W|=k^BS!5q2Dl0aV?701<N>ivb$?
z%>WT<9U_Dl2mukf8`y>e03xO&3jh(i8!7+*03w-x<?jFyyBh|z4gezEvvB|syc-6!
z4gezCl)V5Ey&Jly4geyChe`kuz8kuz4gewpdIJFwzZ(X%4gexR+Q9%3z#F=#4geyW
zFv9>5!5e^t0{|kgtsnpq!W)2u0{|kNKzaZX!y5**4gez9cc=gn#2W^+4gewvuAcxA
z#Ty2HwGIFxeMs5>5yu-s0000Yhe+B05y%@t0000Ykx1GA5y>0ah64a1n@HLK5y~4X
z0000YrAXQU5z89}wGIFxaWby}5zHGyhXVj2K1tdD5zQNbgaZH~IWw;S5zZS1wGIFx
zS4r9c5ziY2wGIFxbCLuA5zrgDs15)kCpHRy01?p}2DJ_VA~KW&01?s~IsgCwA}>D>
z01?w02DJ_VBFK~k01?z1IsgCwBGZ%v01?$22DJ_VBHok)01?(3x~L8SBI%R_01?+4
zx~L8SN6v$503yzyVgL~j90s)x002h-glhm(_=yJr5f&W9rw#xK0|>%50000008@Mg
zuloQI7MG!agBwj690s)x002iMglhmI&Y)rd5g!}|wGIFPM>d3O03yzyVgM0190s)x
z002j5glhm(_=yJr5kDNorw#xK1Ngx=0000008@Og`v4I?mkEG_8)-uv2DJ_V07s03
zYXBn7pke?KNE`;W4gdg0x`b;0OmuB<01;sv2DJ_VJ|gRTKL8P790s)x06rqlpke?K
zVH^gv4gdfE07vYEYXDRDi3b4@V;se&4gd)Q=)pGt00000Q+%)c01;!Cjevt2e`g#9
zwGIFPM-YW;03yzyVgM0o90s)x002iKg=+vJ&Y)rd5o;U<wGIFPM>K_N03yzyVgM0v
z90s)x002i+g=+vJ>w7-{5pf&_wGIG2BF>;<01<H<2DJ_V002jHg=+vJ>w7-{5p)~|
zwGIG2BF>;<01<Q?2DJ_V002jZN`-3xBF>;<01<f{2DJ_V07tZiYXD4iZEye)j~oWI
z4gfwP>w7-{5sw@OwGIG2BF>;<01=NI2DJ_V0000-=7nnjQ}~Go0TGdxae;#se~}yp
zwGIFY1JJ=Y00000002k$g=+vJ&Y)rd5tSSUwGIFPM+$~(03yzyVgM1B90s)x002iE
zhHC&K&Y)rd5t$qYwGIFPM=pkI03yzyVgM1I90s)x002i$hHC&K>w7-{5uY3ewGIG2
zBF>;<01=-Y2DJ_V002jBhHC&Kf9rcc01=`b2DJ_VJ|fPbVgM1M90s)x0000-fQD-T
zBF>;<01>Ag2DJ_V07tThYXDRDi3b4@u^h#x4gd)Q$iX)N00000Q+%)c01>eq2DJ_V
z2?M~vHvj+t0000-;D&1eA`9pU0TH|$2DJ_VK1_6NZ~zg%90s)x06rqlf1qLj5x*P;
zwGIFP002iahid>LSyV~@5yc$7NNE5^5r=C4BI|oU01?L=2DJ_VJ|fPbVgM1x90s)x
z0000-BZq4MBF>;<01?R?2DJ_V06rqlpke?K${Yr@4gdfEM}&uK03umbN&peh9KJ|t
z07qkoYXBnadp`gX(HsV~e+~dXBF>;<01?p~2DJ_V002jIhid>L&Y)rd5z`z7wGIFP
zJ|fPbVgM1;90s)x0000-mWOKqBF>;<01@RJ2DJ_V07s~YYXBn7pke?K=NtyL4gdg0
z)`x2VA`9pU0TBru2DJ_VK1_6NZ~zeu9R{@y06rqlpke?K3>^lwe+~cu002iQh-&~M
zSyV~@5fUA~NNE5^2Z(C`BI|oU01*@&2DJ_VJ|fPbVgL~o9R{@y0000-8Hj5DBF>;<
z01*})2DJ_V06rqlpke?K7aaz*4gdfEM|y~B03umbN&pcY9ll6u07qGfYXBnadp`gX
z9vud?4gfwP&Y)rde-R!X2DJ_V002j8h-&~M&Y)rd5g;7~wGIFPJ|fPbVgL~#9R{@y
z0000-jEHLhBF>;<01+}B2DJ_V07ssPYXBn7pke?KG#v)D4gdg0u!w5_BF>;<01-7E
z2DJ_V07u@4YXBk(=m-H3M;!*W4gfw(bZu||5lS5fwGIG2e<IGHVgM0J9R{@y00000
zNB)Rw03z#qKL8O<9R{@y06rqlpke?KP8|ld4gdfEM-Yi?03yzyVgM0O9R{@y002iK
ziE98N&Y)rd5mFrnwGIFPM>L6R03yzyVgM0S9R{@y002i+iE98N>w7-{5nLSxwGIG2
zBF>;<01;dr2nMwd0000-beBhkgC7BumuH289|5$NhlPV50@R6@sD*<bLG+1h03z#q
zKL8PL9R{@y06rqlpke?KZyg4;4gdfEM-+-{03z#qKL8PM9R{@y06rqlpke?Ka2*D<
z4gdfEM>LlQhJzmgRF@})gC1o^bc$;LBI|oU01<W_2DJ_VJ|fPbVgM0#9R{@y0000-
zl!|KrBI|oU01<Z`2DJ_VJ|fPbVgM0$9R{@y0000!BI|oU01<c{2DJ_VJ|fPbVgM0%
z9R{@y00000paKXh2LKU$muiNCA%A@v2DJ_V+5R*DTs|0M01<r~2DJ_VqX0DkTm%Fk
z01<r~2DJ_Vya6=;wgJPNp#TxE8<E!s000000JZ^kN~i!4tQ!jB2LJ#7006cDc_&Q(
z5vm)g;s*c#0000&qv7KK5q=v6wGIFd12q6b9NNJE5q}#7wGIHQ12q6b2?eg701<$f
z4~K&oT!b4`2WbEa1JJoQ0000008_EPIsg%c8>&iZ00WVre<g6<0RTcG<O=~2ha0-6
z4giz{H2^|$XHftVh#SVJ4ggaHH2@kMPiO!G=m!7*0000v000000000Hnzw_8gJl5$
zvA5}ngM9%3xtC{(gJ=QCm(YrXXaVH6Ig5kC0s-W={f&bG0Rah@|Bi#x0TGwekApP<
z9+wG_gNp$&mz9u%(*ZiSX_1410Rd{a?2?0n0s)w}aFv5-0s+vs*_VT?0RbJihM9vR
z1p&#osh)!Y0Rg$UtDl2W0s+al{h@;a0Rg$U|DuD)1p%v<$f$$C0os>}se?FwKpO_N
z4gdkmu4@1vJvnLs5kMOf5orJ)qv7KK5kMOTwGIFu9NNJE5kMOTwGIG_0f)o)01<2(
zivb$?%>WT-8zKp=p8yd-8wRxw03v-z+5izl8$tj803wG-+5izm8$tj803wk{+5izn
z8!7+*03w@6+5izo8wRxw07tGvu4@1y>w7-{5l|ZjwGIG2BF>;<01;3d2DJ_V002HB
z>w7-{5osF+wGIG2BF>;<01;^$2DJ_V0002P0f&Y001*Y3gsOufe+3%`wGIGkS2X}!
zJ{V&F5d|9twGIF~STz7#1Oy)d5d|9twGIFRSv3H*0mGZ201-GFU%CeX00000wgGlZ
zr~nZ(8!5X7000000JZ^nCrtnmG8@Rc2LJ#7002Ux;o|@i1{(&o4gdmLH2^{!+Q9%3
z2O9>p4ghyrH2{GE90{(U01*fq2DJ_V0A65^0RTO*YXDO#=$QZ!43|KwgB2zX8>&iZ
z00WV*e<h%g0RTcG<O=~24;#9u4gl<0H2^|$XHftV5SM$agC7ARx0|bjWdQ*>w*jq#
z76Ab>ml&>t=6}W-2DJ_V0Uxw$03SU$Y5)<&8WIs{03V~_;{XxH8V0ov03RIM!2l7)
z8V0ov03Y{O4*(Iy8V0ov0E+>K!}kCY^csr+8v4xu5%3x!39g?25yu(^wGIFxeMs5>
z5y=`t0000Yhe+B05y~1u0000Ykx1GA5z87v0000Yn_EcQ01?a@LI3~&BBe;$01?d^
z2DJ_VM*_5K03z#qKL8Qe8V0ov06rqlpke?K*BS=34gdfEJ|gRTKL8Q&8V0ov06rql
zpke?K@frrT4gdfE0K)-s2)F<dfS1^>gCT!_8V0ov0E%BV09-y8V*n9=8V0ov07zgp
z09*tF9{>@68V0ov01sg`0JZ_co1p*^z8WyK2LJ#7006cDc1oxK5xN>twg&(J0002C
z0eL4)01>zv@U;g300000LZjj101<&22DJ_V{9!c!LLAz`01<*32DJ_Vabh(9LicA@
z4*(H^8V0ov0On#f0D%GtuAcxAgc=654gdgNkd6TWZntXyQ!D7101=29tVn492?HRy
zHvj+t002|5zB&LAi5k90X#fL}uzw}sjsXBdBIFAJ5sMnSs15*HV>JLmb7xTi5sa7L
zu!A1~rkDS)gR}vlx3sZ?eE|W#mol@1jsfA9!Lx(v0lAl~w1eh<K^g|N4gdkOx@!O*
zJvnLs5kVRf5orJ)qv7KK5kVRTwGIFu9NNJE5kVRTwGIFu_f`)85kVRTwGIG_0f)o)
z01<B*ivb$?%>WT=8X^g<p8yd;8V0ov03v-z+5izm8bSa703wG-+5izn8bSa703wk{
z+5izo8bSa703w@PNZJ4qM;bx^001JTNZJ4qNE!yU4gg1%x@!O;>w7-{5mFikwGIG2
zBF>;<01;9e2DJ_V002HB>w7-{5o{U;wGIG2BF>;<01<2&2DJ_V0002P0o89Y01^9_
zY_@|TfBP8*wGIFjX*B>`J{V&F5&IbiwGIH%X*B>`1Oy)d5&IbiwGIH8YBd110mGZ2
z01-JF@UI5|00000wgGlZr~nZ)8WFJv000000JZ^nCrtnmGa9h42LJ#7002Ux;o|@i
z{22zd4giL0H2^{!+Q9%3{TT+e4gmXWH2^~Qe^w6w5&jtlwGIGqY&8IZ0tv3401^Kg
z2DJ_V0A9e20RRHNYXDO#=$QZ!0~)MIX#fcWK)N>o00000Q?b4}01*TlzDQ{R1Cg+Q
zB@m7Q074?<3jq-Z8oH<s0OV{n077$TQ2-GK8pfy&099=@02&-mXaIS$2LJ#70021v
z2mk;80000HA(xZ4gR}t~w@A2yeE|VFm&Un+jsaYkRl0-f0W_CDyMyL`z8MC!4gdi!
z!D|2?JvnLs5xyA`5orJ)qv7KK5xyA)wGIFu9NNJE5xyA)wGIG_0f)o)01@gLivb$?
z%>WVR86pX;p8yfR83wfu03v-z+5i#38A1R603wG-+5i#48A1R603wk{+5i#585sKl
z03w@6+5i#683wfu07nu-!D|2_>w7-{5zZL~wGIG2BF>;<01?g^2DJ_V002HB>w7-{
z5$72OwGIG2BF>;<01@XI2DJ_V0002P0TC?)01<<i=)8j=e}fqYwGIHDay0;4J{V&F
z5rY{9wGIGpb2R{51Oy)d5rY{9wGIFcbTt6B0mGZ201>tsBdZ4h00000wgGlZr~nbN
z8Hub1000000JZ^nCrtnmuo)<<2LJ#7002Ux;o|@igc%044gd{xH2^{!+Q9%3g&795
z4ghX-H2^{h1+Jd}5r&tIy@MD$iWyV~X#fcWV7fN|00000Q?b4}01=BBs!C`81Cg+Q
zC4h_p074?<3jq<08M>$r02_8S077$TQ2-H+m*KsG9|54Z0ltG}0Rgv{XTO7l0mGN<
zzk_H2(zjK>gTn#=(zhAIg8=~n!?zp6g8=~n!?zv9gVO;4v6u12gTVodm$S!%IDbYN
z2DJ_V0rbpk03SU$Y5);N84?j`03V~_;{Xvx83wfu03RIM!2l6P83wfu0E+>K!}kCY
zav6&O8v4xu5pEeG39g?25l0yYwGIFxeMs5>5lI<B0000Yhe+B05lR_C0000Ykx1GA
z5la~u`vL$Wn@HLK5lk5dwGIGB*F(%}03z#qKL8O{83wfu06rqlpke?KRv8Ah4gdfE
zJ|gRTKL8PL83wfu06rqlpke?KZy5%)4gdfE0K);&&u9P<3zx0PgCTzl83wfu00M$F
z09-y8V*n8g83wfu0Mmjs09*tF9{>>x83wfu0C9sg0JZ_co1p*^J{cpX2LJ#7006cD
zc1oxK5jq)(rw0H40002C0eL4)01-GDD5nPi00000LZjj101*rs2DJ_VafCGhLLAz`
z01*ut2DJ_V(u6etLJ2gkp8ydK83wfu05gR(0A9d~0RU*vYXDO#=$QZ!5*btnX#fcW
zfVwvT00000Q?b4}01*?Hfysju5fvG_s15*sg*5;|b7xTi5f+!G$%7vOD7VJRgJl5$
zKbI`agM<M~m$A!(XaQQc8q9;k0s&gL-p+#o0Rc<5;Ln2r0Rc<5;?RTB0RcIew$g*a
z0eP2r(}OsF&KL%@4gdkU*lPeEJvnLs5zZJA5orJ)qv7KK5zZI}wGIFu9NNJE5zZI}
zwGIG_0f)o)01^5aivb$?%>WVh7$OO-p8yfh7zVWt03v-z+5i#J7(xI503wG-+5i#K
z7(xI503wk{+5i#L7%Bh&03w@6+5i#M7zVWt07smE*lPeH>w7-{5#AUEwGIG2BF>;<
z01@682DJ_V002HB>w7-{5%(AdwGIG2BF>;<01@{X2DJ_V0002P0fIY401=ZI2DJ_V
z5E=*o00000^8)|?000000h~15YXDq5IcfkAlNbiI4ghA6H2_>b7-IktlNbiI4gfcj
zH2_?H1Oy)d5tA4OwGIIOk~ILf0mGZ201?I*Um6Gi00000wgGlZr~nbd7%3YF00000
z0JZ^nCrtnmz!=CH2mk;8002Ux;o|@ilo$rJ4gmg>H2^{!+Q9%3l^6!K4ghnMH2{GE
z39g?25tbMRwGIFPUNC9`04v*T08=aInE(+yniy0EX#fcWpt?5z00000Q?b4}01=xQ
zs!C`81Cg+QC2(p2074?<3jq<G7`mtq0O*u8077$TQ2-I1muc059|5qpiPeK;0RhLi
z@79AB0Rh362H1n<e^wX<wGIFQ_1$X#A3Zs001;Lg5)o+tAEV*p01;Lg2DJ_V9~|1j
z01;Lg2DJ_Vivfqj_W%)s7>fZK`pp0lei$MNuAcxAR~QDh4gey3NZJ4qSr|e9001I~
zNZJ4qS{OnA001J9NZJ4qTNo+;001JJNZJ4qTo?wm4gg2kL)~itBI|oU01;*w2DJ_V
zJ|fPbVgM0l7zVWt0000!BI|oU01<x}2DJ_VJ|fPbVgM0;7zVWt00000!vRK|LjVyQ
zm#*1^A%7bf2DJ_VhM6@0Ts|0M01+D)2DJ_VSDG~dTm%Fk01+D)2DJ_V9-B1)wgJPN
zp#Tw17*7@m000000JZ^kN~i!4N*Eaz2mk;8006cDc_&Q(5l9%g76<?U0000&qv7KK
z5gZr>wGIFtoHYPK9NNJE5gix?wGIH2oHYP}0x}7%p8yda7zVWt003TqXaWFd;A;R=
zE9jX35h55=2WbEa1HigB0000008_EPIsg$Pmw?)X6cHsDx~L8S2%R+mLUU(P01+ma
zrP_lZ0XVnC+Jj{Q0Z+Fp+=CVY0ZEra-Gk<T<rfCE4gdk&;%fjOJvnLs5#<*W5orJ)
zqv7KK5#<*KwGIFu9NNJE5#<*KwGIG_0f)o)01*lpivb$?%>WSx7$OO-p8yf&7Y4Ns
z03v-z+5i#g7eW9403wG-+5i#h7eW94006@QPb>%k5vmskwGIG)5(xkR000290{{S*
zr{04Qf2tP-wGIHLp)~+pJ{V&F5vmskwGIGxqBQ_q1Oy)d5vmskwGIG2qcs4w0mGZ2
z01@36&k_j$00000wgGlZr~nb!7nu_Y000000JZ^nCrtnm*B3Yw2><{9002Ux;o|@i
zs}}~f4gfx+H2^{!+Q9%3tQQ8g4gj;HH2{GER|&4401>Sh2DJ_V0A6qn1OQLvYXDO#
z=$QZ!u@_VaX#fcW;JP;e00000Q?b4}01>hms!C`81Cg+QC9n+y074?<3jq<d7rLkp
z04Sw3077$TQ2-IOmyO?p9|6a=t>1%X0Ri5Z6XApR0oa!d;)CXYaTf-)4gdkl=W75T
zJvnLs5pfq15orL60f)o)01=rNivb$?%>WUV7a|F+p8yeZ7Y4Ns03v-z+5i!B7b*Y%
z03wG-+5i!C7Y4Ns07tFoYXBnadp`gXe-{R|4gfwP&Y)rd5q}p3wGIFP06rq?dp`gX
zmKO%K4gfwP&Y)rw01=iK2DJ_V0002P0Uq;x01;-FgX4oCJ!TgMwGIFssWkvxF4$oJ
z5oQ+#wGIGQsWkvX39g?25oZ?$wGIG}sWku=Ip=Et(Ow4t00000&<+6r5o(v=<Ac>c
zG#3W74gdi=>1zNVJvnLs5i}PP5orL60f)o)01;Ieivb$?%>WTn7a|F+p8ydx7Y4Ns
z006@QV#JLA5h<7S<bxqSDHjH{4ghYdH2_>L*kJ$>DHjH{4gjaBH2^{huAcxADi;Q|
z4glb)H2@a==xYF>)dv6o0001B+yMX)EtgW|gSLN#76!Er00BhpYXBcTIcfkAg%%PK
zX#gLytkeJzg%$?24genq=MMl8g%$?24giY*hr{;(5f>MW0eTpc01*Tiivb$?%>WS*
z7a}YgD*zFO76!Er03r#lp8yeu76!Er03v-z+5i!W76!Er03wG-+5i!X7CHa`03wk{
z+5lM*ixxTn001JJNZJ4qj1~s94geyhNZJ4qjTQ#A4gex?GOqv;juvns001IBN!kDr
zj}~wt002in>uUg0_=yJr5t$anrw#xK11Q2b0000008@Og`v4J{m*D1u8-JV@2DJ_V
z07rl8YXBk(=m-H3triBg4gfw(bZu||5wI2pwGIG2BF>;<01>bj2DJ_V0000-pX+M?
zBI|oU01>nn2DJ_VJ|fPbVgM1e76!Er0000-vFmF9BF>;<01>wq2DJ_V07u{JYXBk(
z=m-H3$rc8+4gfw(bZu||5oXL52DJ_VJ|fPbVgM1$76!Er00000NB`?<03z#qKL8QX
z76!Er06rqlpke?K&=v-@4gdfEM-l9603yzyVgM1-76!Er002iL>}vob&Y)rd5!Dt3
zwGIFPM^)@=08{vh2LTc9m*D4v6@Tv*2DJ_V2?G$qHvj+t0000-gY0VnA`9pU0TJ~U
z2DJ_VK1_6NZ~zhb76!Er06rqlpke?K_!b7W4gdfE07s+jYXBnadp`gX{1yha4gfwP
z&Y)rd5&RYgwGIFP07td#YXBn7pke?K{}u+d4gdg0$LwnWBF>;<01*Kf1_!kc002kl
z?3cUfgB@i@2kmPBBI|oU01*=x2DJ_VJ|fPbVgL~n7Y4Ns0000-C+%whBI|oU01*@y
z2DJ_VJ|fPbVgL~o7Y4Ns0000!BI|oU01+4$2DJ_VJ|fPbVgL~s7Y4Ns00000!vXgk
zEdUWlm;LC2A%8{|2DJ_VQL8loTs|0M01-wO2DJ_VDXcXBTm%Fk01-wO2DJ_VWvw*;
zwgJPNp#Twm7AVCD000000JZ^kN~i!4c^1&b2><{9006cDc_&Q(5qB0+#0dZZ0000&
zv#itr5l0pVwGIG|t~CHc2j>p}5l9vWwGIG5uQdQd2{f*s01-(R2DJ_V^shAlUXZ>8
z0448h08=aInE(+@7E}jm00{#CyEgy;0000}vA#M05l)xU>4OvzQ5L$W4ggfJH2^|$
zXHftVQkVAWgCqfCmmBJX(E)gu?dpTr0e+Xp>w~HRj+b%lgTMigw=(U63jqO}w?6KJ
zbpZjwx3uqr<pBZIx0>;TIspOKm(}uv*#XzLuk(Wo0Rh*yyYz!i0RfM<3HE~v0RfJ;
z759S&0|Ae>9r}X{0RfJ;Df@#C0|6hGIR1k;f3y__wGIFQIt6S1A3Zs001>nm5)o+t
z9~5*r01>nm2DJ_VAMrXY01>nm2DJ_Vivfqj_W%(z7K;IT7?J=HAr^}P8v4xu5iS-Y
zN0Ma#5w#TtwGIFx39g?25w;ZuwGIFxeMs5>5x5lwwGIFxhe+B05xEsQ0000Ykx1GA
ze-XMBIsgCwBAZCs01>+t2DJ_VBBe;$01><u2DJ_VB5^XW01>?va3KHyB0fpl01>_w
zf!_`QN2dd908{vh2LTbt6~(6x00{#~!Z!c_0000}e6RZe5y%w=wGIFY13<zz00000
z002k118e{y&Y)rd5z7?@wGIFPN9O}<e*hv2=m-H3+!Y434gfw(bZu||5#JRCwGIG2
zBF>;<01@962DJ_V0000-2Lx;YBI|oU01@LA2DJ_VJ|fPbVgM226$Z5q0000-83b$q
zBF>;<01@UD2DJ_V07pv%YyeaEi3b4@3>L+w4gd)QIKnpo00000Q+%)c01*rpm+k+9
z8fp?22DJ_VK1_6NZ~ze%76!Er06rqlpke?K6&41y4gdfE07sbwYycwbdp`gX85Rb$
z4gfwP&Y)rd5g8T+wGIFP07t6?Yycw8pke?K92N$(4gdg0y##CkBF>;<01+M*2DJ_V
z07u;fm9c~!e@FcUYycwbdp`gXFct>24gfwP&Y)rd5ik}8wGIFP07o4KYycwbdp`gX
zF%|~34gfwP&Y)rd5iu49wGIFP06rq?dp`gXH5LZ74gfwP&Y)rd5j7SDwGIFP006@Q
zHESRM5pfj;wGIG)xd{LO0002S2><{900001ob(uMmkj}g2Y+!D2DJ_VLCQ4%Ts|0M
z01<H&2DJ_VLCZA&Tm%Fk01<H&2DJ_VN6a+<wgJPNp#TxA6{xxi000000JZ^kN~i!4
zs1;ng2><{9006cDc_&Q(5vCR8xd{LO0000&6m&QM5pop<wGIGq%{2f*@j5I35pxv=
zwGIFy&NTo+2_CMW01<Q*2DJ_V*3LBmUcin806hn608=aInE(-Zm$3na6()HVs!C`8
z1Cfw_B|wh_074?<3jq;*6}qSn05;Dx077$TQ2-Hsm+S$ABms+;5dwrc0g;zW0))8%
zr<Vl-gvkM|m(K%)ssX*XdIW?I0RhUFiUow10oJ!7281X90r9s>2ZV_O0r9uh3WNs%
z0rHpK3xq%c^p_qCgxLY~xA+W%3jqQ2w*(G^O#uPDw^9#;(*gm#w<Z#VO#%T8mvj__
zIDb7A2DJ_V0ZAEb03SU$Y5);E6%r9?03ZM48UPVJ6$Z5q0E+>K!}kCYU=@o28v4xu
z5nUA`39g?25k3_LwGIFxeMs5>5kM98O#uJ^!vRVI`~VRL6$Z5q0AR`o000000Gk5<
z0000000Eo^C2Rm(JvnLs5eF3pwGIFk?SC}@Ts|0M01*cj2DJ_V749_vTm%Fk01*cj
z2DJ_V$?i1(wgJPNp#Tvz6{yPx000000JZ^kN~i!4F%>V%2LJ#7006cDc_&Q(5ib?3
z$_D@d0000&|Ku6~5eO9qwGIHI?==8I39g?25eXFrwGIFn@HGHl(6|8r>KbeSQwA&O
znE(+Dm#`It6($Z9PsI)Z1Cg+QB_O#0074?<3jq-k6}qSn0C?~<077$TQ2-GVm+KXT
zBmo_l5Eg`?0Wg=c7KGLTHkXMPgkS+mx9S&!Kmh?vw<{Thw*mo7x8oax0RaI(m*pIU
z`~gh2;vIwm0Rcd_<Q{~f0RjL5mxmyPNq=<|2DJ_V0h%Xl03SU$Y5);+6cQ0>03Sjl
zhyW3F6b7{p03ZM48UPV>6b7{p0E+>K!}kCY^%RQ%dKi)b5#kh!0UG+v01@vLA~zYQ
z01<W+2DJ_VA_=aa01<Z-2DJ_VB7I2O01<f<2DJ_VB8N!Y01<i=IsgCwB9Tbi0DlpC
z6gmI^03w@6+5i!J6b7{p03xMG+5i!K6b7{p03vZRuK*E#6mTH`03tp~+5i!M6!uL4
z07nNVYyeaEi3b4@i4?`B4gd)QXu>xD00000Q+%)c01=542DJ_V2?JolHvj+t0000-
z876E1BF>;<01=E72DJ_V07pkAY<~bE3+M;|5uFqUwGIG2OmuB<01==R2DJ_VJ|fPb
zVgM1K6b7{p00000M`tE%03z#qKL8P=6b7{p06rqlpke?Kq!b3V4gdfEM|mb}03yzy
zVgM1R6b7{p002j;CTsvx_=yJr5zQ3Erw#xK16aa00000008@Og`v4Km6qijRgc@nn
z6b7{p06t7~ZEye)))WS{4gfwP&Y)rd5!Mt2wGIFP002k%CTsvA>w7-{5!w_6wGIG2
zBF>;<01?_02DJ_V002h|Cu{&B&Y)rd5#1C9wGIFPM;#|@03yzyVgM1}6b7{p002il
zmk=X_9)Cw&Cu{&B>w7-{5%ClTwGIG2BF>;<01@#N2DJ_V002jQCu{&B>w7-{5%LrU
zwGIG2BF>;<01@&O2DJ_V002HB>w7-{5%v@YwGIG2BF>;<01@^S2DJ_V0002P0R$4L
z01-A62DJ_VVDkt700000oCyE`000000i5$cY?o~#ga>~%6b7{p0LS|^09-y8V*n90
z6b7{p0B!v>09*tF9{>?H6b7{p0IB{p0JZ_co1p*^Z4`*~2mk;8006cDc1oxK5or`W
z^aua|0002C0eL4)01;;tzViqG00000LP8^m01-D72DJ_V(*HF8LjU9%01-G82DJ_V
zg#b1HLJ1zOp8yd#6b7{p05}0Q0A3J;0|0v|YyeX$=$QZ!J(m$AgcT+}6sk&S00WVb
ze<gr~0{}uI<O=~2K@_^E4gi<|HUL6%XHftVLYGV=gd_n|mvJP7(E(_eLnVaR0dAKZ
zCWNX1ez(UagbM)yi<i?UgqH!Aw`(YbECB(sx0oq}3jqPMx1=hBFaZIzx790zw*mpR
zmyIrj*#Wh;XD@^c0RgqQb1;Mf0RgqQbuoko0RgqQeKLec0RexP#WRHb0e-i_G=xC`
z0e`myHiYH@0e-jZH-rlT0e`pmID~rw0qvJ_JA^oY`x6GW4gdk1Kx_aXJvnLs5&IJo
z5orJ)yUzmw5&IJcwGIG_0f)o)01+h=ivb$?%>WT06e0<(p8yg369%;o03v-z+5i#$
z6G8w003wG-+5i#%6HW>P03wk{+5iy%6b7{p07rj7Yycwbdp`gX3KRyl4gfwP&Y)rd
z5egJ52DJ_V002HB>w7-{5h4@@wGIG2BF>;<01+Y-2DJ_V0002P0q*Po0TJ+*RXl_t
zJ@69-wGIFSAvOS9yUzmw5%3cRwGIF&A~pa*39g?25%CiSwGIH?A~paP^FM3=opT8Q
z00000@DK$65%ia{JcMci3zr@}gyaGEw*x+eQ2_z}msLN6XaNnkzCVPE1OW}VN<@T-
z0s;TG)JB9!1Am<p2DJ_V0X9%<03SU$Y5)<P6A}?=03W;00|61869%;o0E+>K!}kCY
z-xG@g8v4xu5!e$V39g?25uOtUwGIFxeMs5>5ug(WwGIFxhe+B05up<XwGIFxkx1GA
z5uy`90000Yn@HLK5u+0ZI0pbCrAXQU5u_6awGIFxaep$e01>4V2DJ_VM-@+O03z#q
zKL8Q069%;o06rqlpke?KuM-Bf4gdfEM@LU=08{vh2LTbV6UC<v00{$l!Z!c_0000}
ze6RZe5wH^mwGIFY18~AO00000002jIPiz1p3+M;|5x)}#wGIG2OmuB<01?6y2DJ_V
zJ|fPbVp0GR!V?Cy4gdfE07sQiYycwbdp`gX#uEm$4gfwP&Y)rd5yle+wGIFP07s`!
zYycw8pke?K#}fv%4gdg0xle2WBF>;<01?U)2DJ_V07uzRm#s^L9R)}EPiz2}%S(hC
zHAfjxYycwbdp`gX+Y<)04gfwP&Y)rd5!(|6wGIFP06rq?dp`gX+!F@14gfwP&Y)rd
z5!{#GON74xl9!)Mgdri469%;o01GlU09?Dz0|60|69%;o0QE990740_p8ye)m*Gr=
z8v&M={Y-?=0i&1BO@x~Pw6|(bgiZkgw3oS0gj@l+mljZjcmX(<k5Gh%0c5uqQH06@
z0i&1VQ-qZPqqlfegh&B@D-#B_4gdj1Tx<XzJvnLs5i1iC5orJ)yUzmw5i1i0wGIG_
z0f)o)01<%`ivb$?%>WU36CxwRGyoAS6HR~u03r#lp8ydp69%;o03v-z+5izR69%;o
z03wG-+5izS6G8w003wk{+5izT6Is^-03w@6+5izU6U5g803xM-NZJ4qGZUeyHUJ`V
zGOqv;G!spL0staDN!kDrH4_H44gexKGp_&<HWLQ54gexoN!kDrHxmZ64gex^k^}$|
zI1>i74gg2=T5JF!>w7-{5keCNwGIG2BF>;<01-kH2DJ_V002iOTWkPR_=yJr5mXb!
zrw#xK1Bk*m0001&zE*??e^e6&wGIFY1AxLe00000002i+TWkO#3+M;|5nK}nwGIG2
zOmuB<01;mk2DJ_VJ|fPbVgM0e69%;o00000M|4|k03z#qKL8P969%;o06rqlpke?K
zV-p6o4gdfEM}}K$03yzyVgM0j69%;o002jrTWkO#&Y)rd5oZ$y8MO`o07t4@Yycw8
zpke?KX%hyu4gdg0%3GJ7SA-q{>RXq-SA-n}M+#hQ0GHfXgd0UiDqL&;BI|oU01<r?
z2DJ_VJ|fPbVgM0+69%;o0000!BI|oU01<u@2DJ_VJ|fPbVgM0-69%;o00000!vUgG
z@c<DZmRN)#Js=YXwGIG1LN)+gyUzmw5g-!=wGIFmLpA_H39g?25g`)>wGIG`LpA^w
z2wH3a&1eY#00000U<m~P5hRyhScJL(NVgeTg!KafNVgANg!BOcVwVJ8ged`qmo8s~
zc>#jAs9%H~0Rd5$#$bd*0aLdNVT3>d0ce*@VuafPESD%_gqHy{m%w9$n*lMGTx5i3
z0W_DzWQ1D*F}ERQg!lmgEVmkGgku2#G?#v8gj)eDx7uigcL9IV5(c#n00FRXYyclU
zIcfkA(Gn67X#gL)&jSGw(Gmu=4giY*hr{;(5f2lK0UG+v01*lkA|N6K0TI#?a3KHy
zA_=aa01?v?2DJ_VB7I2O01?#^2DJ_VB8N!Y01?&_LI3~&B9Tbi01?*`6sQ6KBAZCs
z01?;{f2aZgBBg&w+5i#R5>nkY03vZRuK*F+5^x~^03tp~+5i#T5(c#n07qeOYycwb
zdp`gX<q`(94gfwP&Y)rd5#<sFwGIFP07r9gYycw8pke?K`Vt1U4gdg0lW%MQBI|oU
z01*ol2DJ_VJ|fPbVgL~f69%;o0000!BI|oU01*rm2DKUv06rqlpke?K3=;;m4gdfE
z0K)-R91Q^x#+PYpgdriu5(c#n0MlSL09?Dz0|61n5(c#n0EA&S0740_p8yfZm#b@p
z8v)9f$7_V20qeJCY=n^n0qeK<aD<Wq0r!`TbA;Og^p|LKglhrzm(+BGZvpq0M0JGk
z0r!_Qc7*={`Ij_zgogq5x4d_R0|5cjm%w<0yaC#meR+h$0q;)`umRVXLVAQ;0ou2p
zdV~uB0oS*ydxVw(0n)cZeuS$50ou3je}o4C0n)ejfP~)xe~S_ZwGIFQ#)oVGA3Zs0
z01=B45)o+tAG^;30TGK52DJ_Vivfqj_W%*K5{m&E`pp0lu@WK)uAcxAj1mU54gey3
zNZJ4qjuJuu001I~NZJ4qj}k%v001J9NZJ4qkP-&94gg1~him{M>w7-{5t<SPwGIG2
zBF>;<01=uJD+aX=0000!BI|oU01>hh2DJ_VJ|fPbVgM1c5(c#n00000!vWm{egF}G
zmw$tVAt8Vg2DJ_VDSb8oT)WQ$0TF-_2DJ_V$9*;cLJ6*)01<(g!GnYw0fd*`gM@Mc
zqL)yFgus7A5(c#n009AtYyclUIcfkAMG_JbX#gL)&jSGwMG^+J4giY*hr{;(5pNQU
z0UG+v01;~vA_=aa01-wK2DJ_VB7I2O01-$MLI3~&B8N!Y01-(NLI3~&B9Tbi01-+O
z2DJ_VN92iY03z#qKL8O_5(c#n06rqlpke?KR1zx&wGIFP06rq?dp`gXY!U{w4gfwP
z&Y)rd5o{6$wGIFP006@Qy9Go55jvN;g@hp?IuZu84gfrWHUM0^&jSGwIuZu84glJK
zHUL5iuAcxAJD2>0gc|`qml%eGasgbIi-v^2fBz8%wGIFQJdA7rA3Zs001^KY5)o+t
zAG^;30TKTZ2DJ_Vivfqj_W%(o5{m&E`pp0lB@!YDuAcxA01^hZ4gey3NZJ4q0un+1
z001I~NZJ4q0}?_2001J9NZJ4q1QG_d4gg0VjBEfR>w7-{5e^atwGIG2BF>;<01*xn
zD+aX=0000!BI|oU01+k<2DJ_VJ|fPbVgL~)5(c#n00000!vT@62mulFm-UB)AwBdF
z2DJ_VPlGl9T)WQ$0TJ{O2DJ_V?SnP|LJ6*)01@>O2DJ_VCWJNs7L$u?090ZL00000
z0B{5a01^0?Q;39e0T!3Wh=jmaz!3(u4gdioj%)xQJvnLs5x@}=5orJ)yUzmw5x@}!
zwGIG_0f)o)01@L6ivb$?%>WVL5h4k$p8yfT5eBsm03v-z+5i#55kdd}006@QwD~yz
z5w(}xiG(2`wGjrj4gg|?HUM0^&jSGwwGjrj4glANHUL5iuAcxAwwE4?gc||5mpY1s
zSOLwKlZu4o0r<Cgi-g7je+&@@wGIFQd6aAbA3Zs001*rk5)o+tAG^;30TB!l2DJ_V
zAI6e@01*rk2DJ_VAMzy801*rk2DJ_VAJ#p>01*rk2DJ_Vivfqj_W%*05sLvD`pp0l
zn-L;&i`@Vb4G{*l4gex;kJJDW4iN^m4gexE(BuFS4-p2n4gewvf3BYZ5fBjuwGIFx
zeMs5>5fTvwwGIFxhe+B05fc#xwGIFxkx1GA5fl*ywGIFxn@HLK5fu?a0000YrAXQU
z5f%|}Apig(aWby}5f>3~Apig(K1tdD5f~9ofC2zVn38M&A`9pU0TCz>2DJ_VK1_6N
zZ~zf25eBsm06rql1)yR85i6HLj)W6UFA)Z{4gfwP&Y)rd5ib!2wGIFP07uG_Yycw8
zpke?KG7$!~4gdg0_>yb@A`9pU0TDwH2DJ_VK1_6NZ~zfT5eBsm06rqlpke?KMwg+E
zgcC_h5eBsm06rqlpke?KN)ZON4gdfEM=FzS03yzyVgM0M5eBsm002i^lWYJ}_=yJr
z5m*t$rw#xK1DL`$0000008@Og`v4JGm;H`}8(3Zu2DJ_VK1_6NZ~zfu5eBsm06rql
zpke?KVG#zk4gdfE07t5mYycwbdp`gXWf2Co4gfwP&Y)rd5oHkuwGIFP07tx&Yycw8
zpke?KXb}dr4gdg0+>@8EkAxmX{F7_|BI|oU01=%L2DJ_VJ|fPbVgM1H5eBsm0000-
z9F%MTBI|oU01=)M2DJ_VJ|fPbVgM1I5eBsm0000-Jd|tzmkyAG8-GV!lxzSZ>w7-{
z5u*_XwGIG2BF>;<01=}R2DJ_V002HB>w7-{5u_0YwGIG2BF>;<01>1S2DJ_V0002P
z0R&+j01>+o2DJ_V(2@xN000006$AhP000000i4O1Yyez6IcfkAyATGo4gji*HUL~c
z7-IktyATGo4gh<NHh%zI1Oy)d5xWouwGIH<jWz(b0mGZ201*TcIFktg00000wgGlZ
zr~nZF5z><h000000JZ^nCrtnm{t$4J2><{9002U}&jSGwybuPp4gh$LHUL7#l79dZ
zy$}Yq4gkcCHUL8MB+&p7z7Phr4ge94HUL7_J;DGHzYqqs4kZ9yk2U~839g?25x@`z
zwGIHQk2U~apfLpivY2cDQ!D7101?EO5Rrrx0mYXlk%S--#}K-x4glkiHUL6%XHftV
z$d^}<gdYLwmv@nbv;pLoERuw@0p7Ral7yH60sohJl!W2|VV7N%gd~4M5C*jl00D-a
zYyclUIcfkALl6=XX#gL)&jSGwLl6eF4gep<l79dZLl6eF4geqWB+&p7Ll6eF4geq4
zJ;DGHLl6eF4gepf^xFUtLl6eF4giY*hr{;(5vmZ20UG+v01>1RA_=aa01-qG2DJ_V
zB7I2O01-wILI3~&B8PuS+5izp5C*jl03wk{+5izq5C*jl03w@6+5izr5C*jl03xMG
z+5izs5O5&?03vZRuK*EC5O5&?03tp~+5izu5KVvr03tavuK*EE5P{zg07uW7Yyct)
z=m-H3UJwSg4gfw(bZu||5n&JpwGIG2BF>;<01;sj2DJ_V004ggNAH?!03z#qKL8PB
z5C*jl06rqlpke?KWe^6n4gdfEM**8`03yzyVgM0n5C*jl002i5n`{6g&Y)rd5o!<y
zwGIFPM?srx03r+M2mujs5C*jl06t7~ZEye)bPxu$4gfwP&Y)rd5p)m+wGIFP002i}
zn`{6g>w7-{5qN(P2DJ_VJ|fPbVgM0%5C*jl0000-bDL}cBF>;<01<l-2DJ_V07r$J
zYycw8pke?KeGmq<4gdg0x0`GLQ}~Go0TF%>#itGc2?MCYHvj+t002{buloQIeh>z=
z4gd)Qpu#r*0000007vDUYyct)=m-H3gb)U`4gfw(bZugA01<}}2DJ_VJ|fPbVgM0`
z5C*jl00000M+Ka003z#qKL8Pn5C*jl06rqlpke?Kix3924gdfEM;DxI03yzyVgM12
z5C*jl002iRoNNFh&Y)rd5s(lDwGIFPM@gJ)0GAt>gd0UiX`E~TBI|oU01>7T2DJ_V
zJ|fPbVgM1Q5C*jl0000!BI|oU01>AU2DJ_VJ|fPbVgM1R5C*jl00000!vP%fi~tei
zmp_<<A%Ei!2DJ_V$CEYyTs|0M01@L42DJ_VpOiKLTm%Fk01@L42DJ_VLzXrGwgJPN
zp#Tv(5TBO`000000JZ^kN~i!4HxT5U2><{9006cDc_&Q(5j7BSoCyE`0000&yUzmw
z5#$dBwGIGWmo@-G#*%*k5#<jCwGIIJmo@-G@_!`J01@U72DJ_VmY6mGLe@RP01@X8
z2DJ_VGnqC3LZ<ZF01@a92DJ_V(3v&>LJ6*)01@dA2DJ_VZJIU!UeGZG0REtC08=aI
znE(;(4^`r700{%YyEgy;0000}vA#M05$+FkQE31Jk&u5SP%;GoLL%e~0TJ;Jx~L8S
z%?g?}077$TQ2-J0muQ)UBmo<jjhTcr0T!3SnS`_f5SMqFgtP$)w=A24l>q@fm%^Nc
zw*g9*dYy#80Zf-Wo`iz|OP9Z%gunq<w}797C;<U!x2d3n6aoQ!x4xo;B?1ADx8$UR
zP5}Xrmouh>j{%XF!={8r0Zf+)r-WMpOSfjHgbM)yO1E;Tgy{hROqX1$gj)eix45c=
z3jqO2x5TT2L<0dfmkF<g+W}0M<*$S|e}NAMwGIFQfVOM^A3Zs001<%?5)o+tAG^;3
z0TF=@2DJ_VANN)d01<%?2DJ_VA2Px;01<%?2DJ_Vivfqj_W%*n4~qdB`pp0l%nu?7
zuAcxAf)56@4gey3NZJ4qgbzXh001I~NZJ4qg%1X`4geyNNZJ4qh7Sg{4geyXe@NN@
z5r+>3wGIFxrAXQU5r_|PApig(aWby}5s42?fC2zVJGE>8A`9pU0TG%H2DJ_VK1_6N
zZ~zgV4+gak06rqlpke?Koeu`J4gdfE07qN3Yycwbdp`gXp$`VN4gfwP&Y)rd5upzT
zwGIFP07q@LYycw8pke?Kqz?wQe+~cuM}M_!03yzyVgM1Q4+gak002j^wQK-W_=yJr
z5vLEurw#xK1GvIB0000008@Og`v4KA4+gak00{%I!Z!c_00000N8hz<03r+M2mukT
z4+gak06t7~ZEye)unz{c4gfwP&Y)rd5wH&iwGIFP002k-wQK+)>w7-{e-X3~2DJ_V
zJ|fPbVgM1e4+gak0000-5w>gqBF>;<01>wj2DJ_V07oOXYycw8pke?Kxeo@l4gdg0
zL$+)HBI|oU01?d(2DJ_VJ|fPbVgM1%4+gak0000-W43GnBI|oU01?g)2DJ_VJ|fPb
zVgM1&4+gak0000!BI|oU5C9R+4+gak06rqlpke?K&zC8(gfam;mt3-hA%8m$2DJ_V
zc(gVETs|0M01-P62DJ_VD77{KTm%Fk01-P62DJ_Vqqa5xwgJPNp#Twi4`>Jp00000
z0JZ^kN~i!4bq@gz2><{9006cDc_&Q(5pxfq3<&@L0000&yUzmw5j+nDwGIHew>AJm
z_f`)85j_tEwGIG6xHbSnG8Mu!01-Y92DJ_V%eXcGLJ6*)01-cz5wnCC7ef!MNNE5G
z0|2}?0000008_EPIsg$wmq4?G6cI)bx~L8Spt&{xLUU(P01-!*WwV4I0b!SgvxKw(
zTemv2gna=4dY8Dhgp&b?mrS;V!~uzy5x0bc0hG6uw}dDG0j0OTxP*xV0j0M{yo4zN
z0l2qvzJyKz0l1gJzl4tgiI-8pgi`^Cx23>@;Q;}OmxscHQvry#+rory0s(ZFO2vf1
z0Xvrr#)LS3)(!@>4gdj(%xnN3JvnLs5!Mb85orJ)QjDJf5!Ma{wGIFu3d$b<5!Ma{
zwGIG_0f)o)01+w=ivb$?%>WT34<ZS!p8yfp4hFRj03v-z+5i#R4hFRj03wG-+5i#S
z4hFRj03wk{+5i#T4hFRj03w@6+5i#U4sam=03xM-NZJ4q-41Xe001I!GOqv;-VQ<l
z002iw%WME5>w7-{5$FyEwGIG2BF>;<01@a82DJ_V002jJ%WME53+M;|5$_HLwGIG2
zOmuB<01@&I2DJ_VJ|fPbVgM2H4hFRj00000N0!TM03z#qKL8Q-4hFRj06rqlpke?K
z_6`P5wGIFP07t0HYycw8pke?K`3?rP4gdg0y31?;BF>;<01^8R2DJ_V07vM{Yyct)
z=m-H30uKhY4gfw(bZu||5d{whwGIG2BF>;<01*Y3k;jA+J_!#7wGIG2BF>;<01*if
z2DJ_V002iC%xnN6&Y)rd5eyFowGIFPM=Z>203yzyVgL~i4+gak002i!%xnOc9>|0n
ze@AT0Yycwbdp`gXCJzR+4gfwP&Y)rd5hf1?wGIFP06rq?dp`gXCl3a-4gfwP&Y)rd
z5ho7@wGIFP005u@Hqo>I5t9xEwGIIAsR;l80000a2><{900001oXz8G09-veY5)<F
z4hFRj0O-&*09-y8V*nA84hFRj03gvee*jzr1RnqqlMV*84ge<7HUPE(!<(T15zY=I
zs|f%A0002C0d`8L01?U#%c=<g00000wgGu3O#l(d4pgcM00000076oXp8ye*4hFRj
z0C>|j0744N9{>@R4hFRj04mfr0740_p8ye-4hFRj0Nm6z0A8?Z1ptiBYyeX$1?ZUo
z5t^3CgcT;64ysCM00WVbe<d(#1pq=K<O=~2pANdH4gfsWHUL6%XHftVpqD$zgd_p2
zmte_+IRUemn#qK@0m+wC%7n=Q&zBy{go**&m#NEyzyaNtYRrVd0qM6Z&4d^M0s6N@
z&V-Ev0s6Pn(S$ky0sOZO(}Xbr0S&i5)P#ux0S&jw*MvF&0Tq`7*@Q9y-M2T{gbM)y
z+_yv8gs%Ys-M91GgbM)y+_(SSgiQhgDwj>)ggAd#4hFRj00DO7YyclUIcfkASPl{q
zX#gKoXjlLdSPllY4giY*hr{;(5rPhj0UG+v01<u;A_=aa01;UZ2DJ_VB7I2O01;ab
zLI3~&B8N!Y01;dca3KHyB9Tbi01;gd2DJ_VM_J@-03z#qKL8PD4hFRj06rqlpke?K
zXAUa{wGIFP06rq?dp`gXe+~w<4gfwP&Y)rd5q}N_wGIFP006@Q+U1Y{5lxpb;DjL|
zO%4XN4gml6HUL~yXjlLdO%4XN4gg;GHUL5iuAcxAPM2}ugc|`-myO_rZUJ1E{NRMd
z0cw}Q;e>MmYPU7wgbM)yZ<j&igw+9Wmlx!O?*VVO2jzqV0RddM3+99i0RddM7w3cv
z0Rd{aBj|(&0ReBfE9r!@0e=z>2DJ_V0nhJj03SU$Y5)-u4iXV*03TFnSO5_c4hFRj
z0E+>K!}kCYJPwNi8v4xu5jYMa39g?25fcsuwGIFxeMs5>5fu(X0000Yhe+B05f%<`
zApig(kx1GA5f=^ywGIGBvhQpFBI|oU01+V$2DJ_VJ|fPbVgL~#4l4$=4gdfEJ|gRT
zKL8Oq4hFRj06rqlpke?KISvN34gdfE0K)-6L7D&&2bYNJgdrgZ4hFRj0OSNW09;gP
zSO5_R4hFRj0QUtq0740_p8ydEm&)se8vzTK=IexN0T!20?1aDpB9{{FgyaGExA*OY
zQ2_xYmr3u0a{(l`z3+qv0RbtO#qflg0W+6h@q|197Pp-7gbM)y7PqMKgbM)yB)7El
zgqi_=&J6~&4gdis`fLCnJvnLs5zY+~5orL60f)o)01@^Livb$?%>WVd4I&Ayp8yfh
z4F<Ii03v-z+5i#J4H)|Z03wG-+5i#K4F<Ii07nY?Yycwbdp`gX+zke`4gfwP&Y)rd
z5!?+1wGIFP06rq?dp`gX^9=^I4gfwP&Y)rw01@*I2DJ_V0002P0hO#c01?BN;Pr$d
zJ;MzKwGIIE5jOx_F4$oJ5yK4zwGIF+5;p)s39g?25yTA!wGIGg5;p)AmicS|S<?pq
z00000;Mf5G5yzK4_Jmyl(U+X|gkS;QxBd2n!vX=`x54>@0RaKgx5N5`0RaKgx5fK}
zngM@`4F<Ii00B4xZ2%uVIcfkAiVYGGX#k4>hr{;(5wZ=70UG+v01>VYA_=aa01=A~
z2DJ_VB7I2O01=H1fP@18B8N!Y01=K22DJ_VM;iid03z#qKL8P!4F<Ii06rqlpke?K
zm<<NC4gdfEJ|gRTKL8Q04F<Ii06rqlpkf065wDlw{e)%#f0s!9gdrh+4F<Ii0GAp!
z09-EEVE_?-4F<Ii0MHsY0740_p8yenmx}&`8v%otrv8Lr0h*Tw|Ae=H8x01v4gdiM
z25kTzJvnLs5gQE>5orJ)xJOX{5gQE#wGIFuxt&h{5gQE#wGIFuTi43~5gQE#wGIFu
zeO%7~5gQE#wGIG_0f)o)01<8tivb$?%>WT-4I&Ayp8ydY4F<Ii03v-z+5izA4F<Ii
z03wG-+5izB4F<Ii03wlpNZJ4qAPok!4geyXNZJ4qAq{XL001JTNZJ4qA`NgM001I!
zGOqv;BMpJy4gexPN!kDrBn^Sz4gg2=1Z@B!3+M;|5i|`3wGIG2OmuB<01-D02DJ_V
zJ|fPbVgM004F<Ii00000M-&BZ03z#qKL8Os4F<Ii06rqlpke@j5jzbAwGIFP07oVT
zZ2%(9pke?KJ`D!74gdg0I0bD0BF>;<01-e92DJ_V07qs8Z2%$*=m-H3M-2wG4gfw(
zbZu||5lRgPwGIG2BF>;<01-+J2DJ_V0000-h6QZ^BI|oU01-|N2DJ_VJ|fPbVgM0N
z4F<Ii0000-m<4Sm03yzyVgM0Q4F<Ii002j-1#JK#&Y)rd5mOBYwGIFPN6H0l0GFHr
zg&Rdj>IH28BI|oU01;{p2DJ_VJ|fPbVgM0p4F<Ii0000!BI|oU01;~q2DJ_VJ|fPb
zVgM0q4F<Ii00000paO^nFaQzCm%ssqA%DsY2DJ_V>KiuzTs|0M01?Uz2DJ_VnjAL(
zTm%Fk01?Uz2DJ_VvmG}8wgJPNp#TvR4FTy1000000JZ^kN~i!44-L`n2><{9006cD
zc_&Q(5e*G+?Fj$?0000&xJOX{5z7n)wGIHI9yb6&xt&h{5zGt*wGIG5A2$F(TYuNf
z01?d$2DJ_V-yb&sLVaA%01?g%2DJ_VeIPdgLJ6*)01?j&2DJ_V8zDCUUT`l405uA2
z08=aInE(;f3_NXV00{#?yf**<0000}vA#M05!4J~QE31Jk&u5S;4cOMLL%e~0TI>=
zx~L8SXCXHLLUU(P01?*=#;6Ve{+HhZg&6_)m-zyPv;p&%t^<X%0qwVA1cj9W0Th?{
z1%;&nA(x*9g}?zKmtY5lzyU6|Aqa&S0Rca^Ithh|0|7s`#0-Tx0Rcg`{|$vP0Rd9C
zF%N|{0RdE(Xb^=$0acf)5QRnoBDeAog$n@zA-Db!g|7htBDby+g$n@zA-B2|g@ghD
z4wpX{g~0*gmjM`sH$IpQ2DJ_V0lge;03SU$Y5)<K3=$D(0E+>K!}kCYxeSW|8v4xu
z5w;8>39g?25t$4IwGIFP!vSGAt^g5@mv$J1Aw7)@2DJ_V;6671TrSvQ01=H02DJ_V
z89z4wLJ6*)01=K12DJ_VQa?8U7JwXW0C5!v000000FX@r01=Uw)EI@fe^d+xwGIFQ
zV;*e)A3Zs001;FS5)o+tivfqj_W%)m42uC8`pp0lcnl&5uAcxARSX8T4gey3NZJ4q
zR}6vQ4geyDNZJ4qSPTZW4gg0+9&G?3>w7-{5n>DmwGIG2BF>;<01;vg2DJ_V002HB
z>w7-{5qS&-wGIG2BF>;<695r;3<k9h00000!vWguApj9cmmC^}At6Z&2DJ_V!#_6w
zTrSvQ01-(H2DJ_V{69AULJ6*)01--;UK)iP0Zo^B8ii5;X_w6!g{*%i3<2C9Z2%uV
zIcfkAB@7Y~X#fD%0uLAZ01+Gv0rwwm03SU$Y5)-&3<k9h005r>+Lm?z5$X#%00001
zKp<@ZA3Zs001@g75)o+tivfqj_W%(M42uC8`pp0l3Jf9%uAcxA>kB#n001I=NZJ4q
z?F%ylHvj;i0X#hr0TF-P3pxM*00Dy_Z2%uVIcfkA+zSS^4geoM7-Ikt+zSS^4genn
z1Rnqq+zSS^4gewvuAcxA-3vMZ0001=0layk01>1M2DJ_V0Tdx^03ZC64gnFQ3kJ0g
z0E+>K!}kCY%nOSF8v4xu5y%T739g?25v2<TwGIFxeMs5>5vNuQLI3~&B8N!Y01>DQ
z2DJ_VNB1CY03z#qKL8Q33kJ0g06rqlpke?KvI_>a4gdfEJ|gRTKL8QQ3kJ0g06rql
zpke?K$qNRx4gdfE0K)-KHuV4znU~ERg&`rC3kJ0g07OAI09-EEVE_@C3kJ0g0DVC>
z0740_p8ye>mk%9<8v&h{D;<Sk0koHl9fc_YxtF#bg^hnp3kJ0g00D|4Z2%wqlnwzA
zOA7|I4giY*hr{;(5r+$l0UG+v01<)<BG$>+01-?Ja3KHyA_=aa01-_K2DJ_VB7I2O
z01;0M2DJ_VB8N!Y01;3N6sQ6KB9Tbi01;6Of2aZgBAZCs01;9PQr$HGBBe;$01;CQ
za3KHyB5{8*uK*EL3kJ0g07p9`Z2%(cdp`gXUJC}b4gfwP&Y)rd5nc-hwGIFP07p$D
zZ2%(9pke?Katj8v4gdg0Z6j>}BI|oU01<-=2DJ_VJ|fPbVgM0?3kJ0g0000!BI|oU
z01<=>2DJ_VJ|fPbVgM0@3kJ0g00000!vXy8$p8WoK$k-wg&{pa3kJ0g0F^>F09-EE
zVE_?83kJ0g0M0@;0740_p8yd-3kJ0g00%=i02YfPZ2;6!2LJ#7002+}0RRz1m!Thp
zjsau0Fd&7I1Oa2W!X$-~0s(KARVRhp0c@8sD1~bQZkL@Xg>M0GmklX}?*VU@{V9e2
z0dbf8DussuZ?}ajg#!TrOqYo)g}eb%mqsmx!~sy33oeCR0aLeVE`<vL0Z_MeFNKx@
z0Zg|GGKH%F0aLfPGld5M0Zg~QG=;PQfAR_jwGIFQx;t$EAN-UK0TJ>F2DJ_Vivfqj
z_W%(t3yT37`pp0lDGMT2(pvx#^9pbw001HhuAcxA^a=*G4gey3NZJ4q_6i2I4geyD
zNZJ4q_X-rK0stbBNZJ4q_zHih0stbLNZJ4q`3h3qH2@-|NZJ4q`U-F%001I!e=@HC
z5&H@TwGIGBYddWKBI|oU01*WX2DJ_VJ|fPbVgL~Z3kJ0g0000-eLHOcBF>;<01+7r
z2DJ_V07soWZ2%(cdp`gXDhmd+4gfwP&Y)rd5h@D?wGIFP06rq?dp`gXD+>m-4gfwP
z&Y)rd5i1J@wGIFP006@Q%4^O50ukqzb2o({J?9DrwGIGwTsHt*F4$oJ5$6g9wGIHU
zTsHth39g?25$FmAwGII2TsHs~yE<(EG)xBo00000F#G@j5$l)HH-(M?3b$Z5g^>gS
z3b*n+g^~gR7MF!Th1&rWmtjDKYXKFP%|L~30T!1%L51%D7MCqTh5rE<mn}nuhXEG1
zwL^si0Ri)uxkQD$0s5DAMTNux_m?|Hg<Jvpx0ptS3jqQ5x1>jfmI49uw>wIOs{sM}
zx93ZR2LS=|x9?1av;lva3I??f00HV#Z2%wqlnwzAnF<EA4giY*hr{;(5!DKd0UG+v
z01?m%A|{q901=uBa3KHyA_=aa01=xC2DJ_VB7I2O01=%E2DJ_VB8N!Y01=)F6sQ6K
zB9Tbi01=-Gf2aZgBAZCs01==HQr$HGBBe;$01=@Ia3KHyB5{8*uK*FE3I??f07sis
zZ2%(cdp`gXtO^FT4gfwP&Y)rd5v&RZwGIFP07tD;Z2%(9pke?KzzPPn4gdg0%~Ndv
zBI|oU01?p&2DJ_VJ|fPbVgM1*3I??f0000!BI|oU01?s(2DJ_VJ|fPbVgM1+3I??f
z00000!vQz~L;wO2j+diPg&{qT3I??f09ST509-EEVE_@13I??f0FZV!0740_p8ye$
z3I??f0LgYY02b>~Z2*);2LJ#7000p6000q^mjO_PjsdZ^kWht@1Oc(PB2|Tw0s+03
zwO57P0lJrwScPi=yqEo0g>M19mu*>v?*YA+U0Q|z0l$}BTZM-Ky|?9Cg#!TrnwRNZ
zg}ecwm!@5X!~vd{YhHz10in0ZUWE$*0iL(iUxk(e0h+gKVuh;#0im}SWQ7L-0h)&&
zWrc?yWrl|zWrv3!Wr&9#Wr>F$Ws0{SWsBhb0hzXAjQIuuPq+GMjD-UM__u&<jFJKY
zpSKfmjNB9fh_~>BjH&_w%9opmj6ng+m)(bqMghpTB#4Z20|B_Vn2U@P1p(8StdNX6
z0kxOukc^B0nU{8vj3@!sm!FZ0QUS@A@{x=<0kxMal8lT2nU}bdjOzjPx1^JdgaHBj
zw<(p3PXPgIw|SO~84Cdomy@B4(g6pTW}=Lc0hyQgqKu{izPFyEjQI`$f|r=GjMo8$
zw{WwJS^)vlx68DQECC(bRQtPh01bLCXaKQ51poj5000O900000000q|w-mOFt^olC
zm&3P=G6CGT_P2}z0Ra}5`M8XX0T{Q2xr_}20Uei=y^JCO5VyCzj12?<9=FB8jIsm)
zO}BW)j3xmAVwa!CjB^2@w-d;W(E<Ugm*vWgN&!WOFw2a$Fw2eEYylRR3OJ6y0T-9c
zIF7slBA19cj#mK~x8OOB3IhRjx9dEP0|5bbx9>fU2LS<gm-jx7bOCmkY(I`y0fx84
zKaK+d0f)E9K#m6i0f?8-L5_3*h_^*Tjv)a77?)>5j$Hv4x5Yz_3jqNZx6wq7mjMA7
zmuW_hT>%%j$3~6|0Ra|=(?^cC(?^fUDgncnoN|wO0j`%8bB{j()VEf1k4OOzyw`pR
zo)k-HNlQzHodblot#prP1Obzm6?u=X0gJc8d5?kt2=uUiN(2!DjkisEk4OOlke8Nx
zkCXwFw<~>*nF0=|?S9q-G6UxXN(2!Dh?mlTk0~O7?S8BT5dZ=N0t4rg0fGcF00IO8
z1DWl9JOmK{0t5mB=a(ITkGKJ|x7L7<F#$J&k^%Ao0~4MDZ7u^7AlL*c1CQ)}JOmK~
z^OFJA1Stca?S8rh5d-9s0iFZj@P5ckph-(h1E7~af{!l;%kX~k1DubSEQ60bm#>14
z5I4u{eh3l-5d*uE0ZIfZ1I6rqJOmK~y^{fS1IH}71G`g@1H~*n1G`f^1C{Z9E=!<E
zOG^Xlj+d^3k1qy+@qY3H=9epjk4!d?@qP#r1Q7%CjsZ#pDFco1emn#b1NDvpbOYo&
zx&xj>kOSg7JOiFYJOkbJelAO(NlQxuqL;0Mk1z+x^?vdLo{E<&gO59xAB2w)K>F=|
zJOmK~h>`&atOO|oiSK^o1Q7%IkO4dc%JP25OQ1<hO9Qoy0fGPm1OWzS00IO82AKpY
z1Bvf`@&pkB`Im5nk0S`S@_q<r1F?;lEQ60Tm(GNbr#7_iel7$N1I&>DN(3tdlkR@H
z0~b65K?Cssf&>u*xy%7d1G_}5195@_)I0JBAxk?;1S<o&m#c-3U=g{J0ayc~I|$?h
z!bD00D+9`x9EOi80kxM%hL0&Eq3?dk1VIB!SOgIRnUMkH1S<ob?tWMU31$Q=1EGId
z1Q7$_x0QyE-Yx;zw{*3SWdi}Hx7xUm_yGafx6`_hW&;7uw>rL$dISNqx3$8LKLZWl
z7V-pM1Nvzo1Q7$}mjT9)ApwcEBgT*Y6CRNi2rdI4N&{dn1S<oFY{&x_o&*sC;uDYr
zD+7mYmv5nvCJC7oSO^saUjvqxZQ74#6}%M41S<pFYghyk1EdrPN(3tdlXaIJqL3X*
zyrltx1S<ob;(ne45d-6$0fJkU1S<pd(SCFU5d(~_0fGc81O3r{JOnHQp?_Ee5d**!
zSOd0nkV~LROG^XE3xWfA2s{KU1Dcmm+mA>AnwOc|j~fN1be;pnm$}=ICjpC>;oFZO
zDWMct1S<o+Y*+*l1HuzZ1S<ozY*+&cSOhEsp?_Ee5d*ap2ucKA1K5`;+>f*#>J$hr
z10YHRU@inJ1Fmbx0~ekI5d*svkOV6Ou4|WXp^zpD@)TGI6$D=cy=#|l+K*-xl@-VY
zD+7aTSOgIR{S*jF1S<o&beA2XkR40Br2&EjD+8V4ex3vo1LK_mf?JdXD+Bb=esly8
z1B|W#f&?oA{n36r1S|uge^>+&1DO?A1Ko3wOQ1<hO9P?|f&+O7JOnEP!E2XM+mA;g
z^AuPE?3n=qF#!Y4=6>=6Y*+&n0wMtdBNV#>&2yduo(z||+m9v!vTK(iqmUmd`V?3M
zD+A?gSOgIRn-od}D+AnXSOW=I1S|uge^>+&1CA94N(5d5gO}*ukF*}c6$mZ^AW8#Z
zE(9wB)oREC7oG$W1C$kz1S<p8YL{=JkR}St6<7!r1YZN>YL{->k7gC_703iD1FC9R
z1Q7$)6$nZMD+A$kmmQ*z9ZS5W0fGc81D)c2o&*sC<DCJ5Ta*MV1N6~;bOaFtjIIHK
z1S<pm(SAGxECZo`SOgIR@)cMEgmaKfph-(h1N#hu19=EM1S<pRYL`*lk4GZR6<7n4
znE?VZ0RyM!e)0osSOXLSA^`#;6uSfTa-IYC441mwk0t`yYL_9SkRLJA6<7o-1B`1}
z1Q7%D6iNgu1A}W=0|{6JECZo`SOgIR=M@M_1YQHdYmgI51Q7$Gm#*QDE-Tb&2!aG`
z1CRm%1YQI3eRKp71H>4RTt)*9@&gSB5?usH1MO!#O9T-El9vbKj~xN3mnGtlKxNJq
z2)YC-1CC|L0~c5XK?CsskOUC}>KRxBD+7*Y@&pkB$HxJX2o?h&2=W6BW&@PdeypGb
zD+8!*2ucHocdP>mSOhEsp?{DB5d*f70ayep1EBAI2y_DpSOhEsp?}r{5d)H!%Hoe|
zIm2aw0|{6JECZo`kOUC}s2j+GSqT0FD+8`&2)Y9aSOhEsp?|sr5d)|j$OK*kq-F?M
zAv**q1MinF;*Tml-vC$yDFf~USOgIRloklO0~?S7Wjq8c1IT1b0~b65K?CsskOUC}
zuo_qdD+9=6@&pkB(;I>aA_OY~;+NFpk1<NTr2$w3D+8V4ex3vo1LK_mSYHGy1N6~;
zbOQ-k1Q7#+t^rB}D+B$}emn#$1EGId1Q7%B8(0J@1LS0o0|`n5ECZo`kOUC}*_Th`
zj~*KRWP$?;N(3wep?{DB5d-cU$b(r3{sb!n=$D7&j~*578}bBR1D|FHf)oQ3@)HOl
z1Q7$om%8MS9~Y-p0yO~>1Stdc06YX<1B`zN<OC4|pqKFEk0}_+seb}B0TTo%1ML7j
z1YQG_e+c9R5d*rHG3Adb0koG}<&Sd#)|b!ak0Ak;xAEnV93uhWmktS#Nj$e}2xbTq
zJM;4d5d+yS0xSUtSOW?2I|v~I8fKwO1StRl1Ofx6PD%t300IO81O71xo(LfbOaxy8
z?M|1W3Xm=r;xSkRDFf3`2rdJGepm!714^C*5d+4T*a?t$P}fih<O6|zSOhEsN>~ID
z1Ku(61YQH#Q1S>9JM;4d5d+OH0xSUtSOW?2I|v~I8fKwO1StRl1OfxEO-ck200IO8
z1NJfqo(LfbOaxy8^_QRukS`b7GFSvD1KUpsE(3vnSOhEsN}dD}1HPB$3Xpgr-%kkS
z1A%^61S|teSOgIR)-v)0UIXP%@*z7ttsevt1HzYp3y{DM<W2%K0U<j*tsevt1H&<w
zfeVlwmuw7>q7CCs0yF_jgGmGt1H>|y5Gjx%mpu)TZ40AK0y+T`1Q7#(IhPPAkRq4Y
z4UlaCjJHJ&kc|R=(Ut+^0}}`m1Stdf=zdBB5d-y>0dxZs@&j#j1StcH>3&!Q5d(;q
z0Rk%l<ykufDFfK=emn#b1FMh$SOh5p)$e{T1Q7rN1OfxtkO8^?0t5vH)&waA1E}wQ
z2v`IW1KE%PSOh5pr|*7d1Q7$-kO4ddDFdhPer5y_00IO90t4cgR1uI>5&G(Wo&$k?
zSOhEsN>~ID1HPBK5s+p9_m?3OkRAbwmp~GbT^;>a@&pk80t5mBgDIW_DFfA7)&vm)
zttC7JDFcz0x)P8(Gu$WE1StdlS7rne1A-`61StdNSXcvrepm!714>u~5d+>SSOh5p
z*jB6rECWiA1Q7$hmnajEF*2)GSOgIR=P6i&SqT0FDFfwL2)YA-eu4xn14>u~DFZgh
z1S|teSOgIR&?yKm1ZV^0w~G^ys{$p)8VEdG0~bmJDFgdtSOgIR>KX_j1StcSW%2|O
z1BDxq1YZN`ms=K)E<@%TSOh5p&u8QVfqqy7ECWiO1Q7$y8dzBa7YLpM5dtg$y8~BB
z1StdNXMzKPepm!714>u~5d-2H$b(r3{sdkFf|tw|kYoXjmoXQRAOV7xQWubW0)RG`
z&liv$7x+j#1Q7rN1OfwyHv%>R2q6ee1YZN=mjoD)FHh_?SOW<NAP7qcOal+(192_`
z4`v8lumng0^GZAf5d)1iN(dAPOY@cjEEEW!1KCmXOQ1<vO9QAbf&*y?SOX=j2$Zk{
zNCV4CJOmK~wYP#8kW>K#yhsR2m(gXF83PdmxtI1CkTC}AQ3!$qq_-^^kfQ<tx|h)$
zkRJi)m+>5sCIQ&D7#)y`0xg<1kO-aw3GzD#Ap;s_p-TiQ00IO81F1zy1Q7rN1Ofy6
zItZQ!AqY$aUjy%#&mWL#Vf~l^SO_o#DFdVHex3vo1K*SZ2y_G~1EK4FJOmK~-;@DZ
z14amR1StdT?0zl;5d)Ny0ayqW1StdA>wZcE5d*1|0jvXpEV=`LRFDIJEIb2%R6GNz
z@O~~!ph-(h1J;+CAdo)+{FmAwkR=_(I#>g-P696hOQ1<vO9PCM0fPGj{O^7OGywzB
zmmDFG83?xUegZfF1BjPBA&{5>@we6?kZ1$}gqJ@ikQf%RL`nn^00IO81N1uxkO(0N
zOaxy8_m^!ZkZL%ynE_Y`Fa#+B?dpDz1Q7$3l>rEJ1Stdj>wcaD5d(&l0Xze}?|ujp
zOQ1<hO9R4?0fOxWE(5oh4=0c)0lk+vCy*fplTDrjzL#DnkdX_JMuH$cgG&Sv1JRfI
zCy;Uh%$IE_kQf2ymxd^iNdc0V&?t}=0o|A9D3Blsf@<Uw1Q7%Bmk%kBkPE?50y+T`
z1Q7$~Ete1}kRq4NDUfXe#kV>tkWT>tf|rUbkaq~G$9_r#5d+tk2`rGrEYy&HkOUC}
zx$yw<2-h$OO9)E{O9Cqs0xJ^)DFd00e*!B31S|tef&>u*f|vO$kYyv*kbgV`5d+ij
z00>wETL_+yjsqwz16gzgDFf7ye~<(b1L2o}Eszrk)R2FW1Q7$Ymz*t-F&E8_e^>+&
z1GMn~kkte!1I><qSOgIRi<jgrkUR*~kbjT_5d-m;C@zq*2)y8atOOAQ#FyhPkT(|V
z=6+ZN5d)Q(0muX?1MB8~o&*sCm6tFtkWT@=mw+#j7blbO00>G1DFelmf7S#M1HJ43
zf&?i8fscP!1Q7%O@BjiN0R$-n!<W7<kT^1m@c;<O1StdVkbekD1Q7$0?*M`XDFfD#
ze^>+&1FG%-f&?i8*N=Z@1S|teSOgIR?UyJpkl!4DkADIz0R#~Pfbjs32xSB*1A&i!
z0xSUp5d;61BQcQEHId_fx&wiJtOP6rN>~ID1K{ugf`eHI{sbul){lPzCIJL214^C*
z5d+}x00=GwDFfA!e{`3=sgo8Hvwyk-UIUf`kOXW4#{vNa5d*H5R5Fk*0l>F|GLXjs
zf4DsYDggti-2ey@1K6T}tOKas00_DSDFgiEemn#b1B9Ca0)YVm<ynseDFgfDe#it7
z1C5&jE(9q9)a!m`1D|<<1K|dc1Q7$QlmS@uSJ#pQ8VHaB>V5(x0R$-n>*an}1Q7$3
zngIel0R$-n>*ao|1Q7$3ngP14Svv$NJ_F0_ex3vo1Gtj`f&?i8$n1VD1Q7$K!2ys1
zAqXx54Q2z8Qmh0i1DouAo&*sC=92-C1StcW?0$j-5d-Fz2Q`pD0qvJaHIOz4k(&Wn
z1StcNmxeWvTMy})0ayfI1L5U<tOOAQpO^48kdgxb<d=jtkfI3FJ)YPEDFgPGA3cy2
z74<y=Bmo2|1I<QC1Q7$XIc5YY1I?F7J&=J1h&+M>Ujy}*(LInFm-;u5d<WV+SOh5p
z+n0SfkXZr2m)AIuEeiJJegZTB1Q7%Cmkv3QYY2_yevkwa1BjP_Igl9v=$D&0kR21)
zJy-)2f~|vD0w)jzVFQCg2)dUMK#(B^`aQ@5VFQDg^Er@C3$xAv0xSX71YQH=w@f;a
z9|1=GJpv~I2s{JP-2ezK1FxcgtOO|o-a%Ld5d-s^0Rngd<ynseDFck;eyju$1Eib*
z2!aDy2y_Tbk^>qD5(5ed$OI_^wwJLxkSGX%oB>z_DFdRH;X9CG0gjhMJdiH|ke75k
zkfR8hK3LcUDFd*VA3cy26}vtHBmo2|1BXUR1Q7%IIc5YY1BaJMJ&=J1&OCwyUjwg~
z(LIosS^S&<SO_o#DFdYEex3vo1KyYc2y_G~1ES}CJOmK~-k1Sc1Stcv=za)h1Q7$1
zmjRFnFa#+B+UI^s1Q7$Mm;t)H1FGzP2rf&YNlQxu)sq2&?FbYD69|{FAdniD+CGpD
z0lb&#K9Ex#hdx*X?@9tN0ZX7sSxW<plmUYK1O4lM0xSUo(U)02kRdI%?0y0`0RxAW
z0jvWNf&*_#0}%ou0lNdeeh82VK1g;1DFcV(egZH71Q7$`m#06FC_0Jceh3l-5d)K)
z0Rkfdg9I!C`Qv^9F98H!1K|Kz1Stb-2ucJj14^C*5d*!O0rCUTqJM&&1S|vHmkB_S
zDNn#Y<bzoV{sdtIy+Hyf0R$-n%>Y;f5P}3O14>u~5d)k)$OI_^nM9rgfqtw6ECWhd
z1YZNn09XVO1HC=u1S|vMK?slp5d*S5f&?-H!?%GzkZuqwjFSNZB>@C71N7{E2s{HJ
zbOaFt>yiP21Tq7b?S2R@1Q7%3k^xu*G6VFN8c~oo0l$|~QIIVN<nVsF1Q7$*mxfW0
zE(zA`e!2sW1PFAO4pNXQmu_s62bb+pkO&EYlmS@R1S<p8m;O<ZO9+9K0ayep1G<+{
zQjp96yO$nQkU{~ymtj+oDhTb30ptWL1HzY+Q;<|d*^>bXbN~Vb0S0sgGX?|V>wcaD
z5dZ=N0|T0r0SIOQ0t5pC(jy3-00IO9H~<0!0|T;LJOn`l;q89p1Q7rN1Oo%0mn&3|
zQ~|n|fK-roKH}m42!aD02s{Kb1J#{>bOT;I10jM05d)9n0FVSS1M;1J@>-5r1Q7$O
z;s96zB^3h+2oeZPk_aILG6RN}7gmrh7TDYXo(Ne8Fuw#c1E8aS2oeMl1GC(hT2_!&
zNu;BH$OI7s+}r>NbOZ8!f&(20JOnZWv7Uc)1717>A%X-E1GC`(kOVRV&YpkrT8>!+
z5d+TQ09XPg6$1$f5(rF^2q6SA1LBt#R*)?f{M!JY2w4a)zXUP^{G)#e5(E(g*OysV
zkXH$Sq<_c+5d(nSm$_Du9T>--e@X-q00IO81K{8Qo&+)g0t5mB{+^faR**pz-{1gP
z1_%-c2tNY}2ol(m^8_*j(U(M5kS!LC+W?*jSqLuzEENL@2omO!^8_*jzL$qrkS-OV
z+W?*jSqL!c1Tq8Iq<;t!1Q7%0m&#X=yHJ^>e}V)N1Ekvk0wMu}SqT0E_IwD|104uF
z1Tq8Vp?`D(UOWRKf&>u*|J?wP1Tq7HqJQ#Qj#&f|1Df6dSOO&#0|^Kc2uzX)Ap|l5
zmZ_H*R*)<f$k+g$2w4a)zXUP^u%~|r5(E(gqu7^PR*+Uow5NZ_1Q7$w*Z>H01M+@?
z104uF1Tq7`qJMM)UOWRKf&>u*qul_I1Tq8OqJQ#Qj#&f|1HRn=SOO&#0|^Kc2uzX)
zAp|l5@~M{?R*)<f?AHLE2w4a)zXUP^jHrJI5(E(g$JduyR*+T+kf?vi1Q7%9*O$3g
zkR2G;qkl>S5dZ=N0t3+80G<Rg00IO81CFDY?pBaN7SG%OSOy3Z1_(a`2?!F{lJf*I
z1L3NdMOTn40sWT|TaXk2&Z?J)SCB0gkk<g72w4a)=>#$Z=%{}P5(E(g*OzHqkh>MB
zsegh55d)Oh00JTbgINgv1Tq7Xw>eyp76d!s)&L0P10D!G1Tq7prhjw;UOWRKf&>u*
zzu5qg1Tq7{rhoEGj#&f|1KZgESOO&!2uuVr1LCcJ2t1ankQkRxRgemofMJjb8iCXR
zo(Mk!2?!F{2!oOYG6U|be+VE15d-6w&R~!_R)eg62s{K41O3wg2y_Hq1B0o5kb^t(
zSp*RS`quym<O3cEJOnZWf~9|S1717>A%X-E1L4{LkOVRVprwEEOpaLu5d;0&09XPg
z6bMWNG6TS@e+WF5tB@F%P*so$mw;iA2pgl-0G<dx0|^Kc*a(A?1Tq86s(%O|1Q7#;
z)tAp;kU4+hs(%PP1Q7$D)c^=|1JJ<$kOV0Mwc>sVJOnZW_yFVt5d+hl0ayen1GVCQ
zSOgIR%bfv`1Tq7y<9-M(1StcEf1U&~1NZ<4kb?vf1Ld3nf&?-H_~L%710jOF1Q7$|
zoB_J!OOKN?11M$$G6U)2eu4xM1DBlvf&*AsX9yWf*aI2}SO5Y90R~tA0t5nP1_KJ@
z1Tq7wmvmx~E(x=j0SFQVG6S=ht74FB551QGSOhWy!{~lI1Q7#^mnLJ7B>}#-Ok<E2
z0Re=UWMq&e0jHOVWRNTml%4@t1G#?!Gy&KIa|7O&#AJ{)0gspZWROKQr09Nv1Q7%9
zmH`N?1Tq7V>3;GA5d((Z00Jff2s{Kb1D2hC2rdK>1M1=cSOX(i1Stdio0n~8kRAey
z<Cl$RkR1%#q<;t!S*}-0O9P~ry=9Oq0+gJW;AfB@7qr;`SO|mk1Tq85r+)}o1G9Sa
z1Q7#=mmp@4Dgx1=mqBQdECI)td1jC#2l1qT2y_FWm!xKpn=OE%f1U&p1D)Lf2p|LI
z00>G0^Lmg7JOnZWrlEfbE(8$++TH+I10z@jDFcg_Y-f-k0sEJXXOJEY>ZpGR5?QWS
zOG^Wkm$qk+C;_II-)E2@7NXSvSO|m!G6UYLe+XCuvwHFb5d-v>6=;wu0pyoJXpk)d
zx0h~ckR%6zsecG`1CW=QXpoK-kfwj+1Q7#`*#HP21LFV)N(5d5_qRN0kTU`SxtDNj
zkWv@7SV{yD00IO81N0|Y1S<dn1Ofx2Sag?>xsV$W^e3JM2oeSZhhhjk2ulO|mjrB(
zIzpjX2%ZB8SOhEsp?{DB5d*a-Sl0w!1C&_u1Q7%5C*;@!DFf<TN&~)oSOgIR*eD2|
z1StdRTS^2G1C}H_1StdRmw;@LObN*+SOhBro>=6U-)WN>4-o_KC|CmttOO|oy;lel
zmx0@l6PFNekOT(xCs+d+)|Y{|kQECADFf+N2)YCj1DKaOZIB)b`BnlY0R#~PhnHb(
zkQidnR{~1`1Q7$UC@urU00{B~DFcjG0!sk|5d-unE(1ykN(3nb!dC(%0R#~Pz9=pO
zp@snn5(Dk|e)0p%00JTb13?JL1StcjR{}2q1Q7$;D3Aj^0yqH#DFf+No&*sCmzUdZ
zkV_HsS6BoQ1JEc~gINgv1S<obmp5*ZHww8Z2p|M91Ep!KmpxpNMYp|fkV*ncyyXA_
zDglNBD+BVDe@X-q1EuKzJOnEP@|J&C1Q7%C=m3xeD+BhHfARzo1EuKzN(4az<ClLt
z1S<phepmz%1EuKzN(3tdnU_#;kZ4rAr2$w3D+8V4ex3vo1LK_mSX-0?D+Bb=eslu~
zSOgIRgRTKe1S<pm(SAGxECZo`SOgIR!s!551S<oSmw%802}%Sk1EGJA1Q7$P=>S*+
zD+7R+404bvHKOPM2rdH=0xSW$1B=A~@&qdb_R@ax1VIB!SOgIRi|7FI0~1(-SqT0F
zD+8F9e+c9Q2}+k<1dtXA5d*&I0LTPh1B<tda*$F10p*w8bC4eash9q9kX-@6mt1s^
zHv+C=mQs)um#K7+H5RBISO{4NF9IwT0|^Kc^pf)gG6UObm;b7e8JAF1kP;7r7@i1O
z2rvT=2!aGM1M8PAb&wzrf*x1{Ap$S~1S<o?W(ee$VcU@v1%d<-1FM&Jb&w|xuwqz-
z0~~?_V1fh@1E80zb&x_1t78IS0R%w<OM(Ou1B#dYb&zNSwrGNvFm;d~6u%zGgINgv
z1YQH^V*(=q69f?hg_oIjkc|Q4w<mXy8X*CUmqM73IRVU<e3+2>0f?7(nUFaF+?TPL
zkop12m#mtQIRT=#=9-WX0T84k0yF^wB*+6?E}jH31F@F{oRB6C@g%wk6bMWNUjzSI
z2ojfX+K(BRQk;-X7t<x41KWxLSOcZ?ex3v~1Cv^21Q7%Am$jUbClrih@=KseSxW=*
z9D)O92ucJp1Lc?SoRBa9^p_!>kYNXvV*)S%1MZiJoseA`fMK2l5d-`l@&qyigkhcp
z5d)MT2)YDb1G%=IkPQJa>?8s-0muXFpaJp(DFgkteo6!p1B1u`2$0MJAp#@;0}W;a
z{oa18pafq7x>*PkmvAPKX9txhSOh5ppqHzkkSqcGm)M_>Y60lCL!gii0UpjG0yF^w
zA;<$?E(9wB&s&}Y7oG$W1DYg|1S<p2mu{huCkxUfSO^pdOaxy8+?SxCkZ2V4B*+9S
z1EO141Q7$?BnV0bD+Ae=9HNjOOT484f&?oAo#K9;1Q7${odJSdlmsgS^wEBF1Q7#_
zt^tAsD+B$}emn#$1EGId1Q7%JBv=FdV~|UrNlQxugB*ebc?diND+AzLmr>h~M*-58
z&7zPS1?*#<1OJ!pqL4BH(U%~jkRU1ABv=G11BF~z1Q7%KBT57-1N>W90|{6JECZo`
zSOgIR?<5FH1YQG@w`rr04*@ljBnUhM9tdUxG6SJnbOT;k10jM05d*>{kOVRVzgqH4
zj#&f|1KuTA0woj(Oaw9m-(d(!maC8$mrzxZ5;ymb0iFai1G?;f2oeMl1KpDWSOXH)
z2wDhQ2w4ad1Tq7Y?0)hD5d+ne0fGcF1FP(QSOgIR)t3&XkU0m}UoHd@1Ll`brI1h$
zfM5tb1Q7#(AP95>R|BD!t)-CnI;|hN1S<oHUkE$|5d-%if&?oAiC+l11Q7%GAW8%)
z1E6161VIDw0J;P#1CM`#1Q7#=t^v9PD+A(}-lmWr6we@#1S<pcUUUR913ZEVJp>U0
z;Fl4nkW7E*UI>B&5d)YZN(3td-(FY)34#PH1EGId1Q7%GAz0T0D+A|V2#^F31DPR8
z1S<pHURVPOf&?rBp?_Ee5d-)kSl0wH1B_}2E(5cA$OI7s!5i|}0}TQq0SIRVD+AVE
z2#^F31F9iP1S<p0URVPOf&?rBp?_Ee5d(=LSl2`ZD+ASD2!aF=1FIoQ1S<o~URVPO
zf&?rBp?_Ee5d(@MSl0wH1Nv$51G9R_1Q7$%8_3uL4+0_q0|~4MJqTF@G6Rv9AgGWh
z0jrltsF0umyqD>ykR}0+mk_Ct-2v{G?x~P(0fm=ds*pEn^I!;q1Q7#`A4&u(1L|N{
z0||lzECZo`SOgIR>>pSNSqLu(y$B&$ky%+w2tNoRS3SFu^8_mc;9v-l1Q7$EA4&u(
z1J__!0||lzECZo`SOgIR{~uTgSqLuzEENL@2om&?^8_*jgqQxRkSqb(mn*A~8UvGN
zte2pykRS!f1Q7$nw{5GCFaZJdm!7PUCJUD!$b(r3{sdP8o43!bkOBh%mY346kZ1wd
zmpibKD;2#QkOV6L0t5mBtYdTp5dZ=N0t20wd$5ow0@-7iZpn~H0sWWmu#h<c#+N9u
zkVznr9asoi2rvWau>lA?1D(!(f&=5R0Xzep&VD=uD+9ilk+F~+6V@D910rSv6A0u3
zvwsNW1S<pWm%*`+PyvLO60(pv3F2e|H~|C^1FV-+vXD9*s$~df1Q7$z8VGa<F8~4r
z0R~tG1178lD+B*yf|s!&kr@pU1C$-egINgv1ZM-sx6!hY3<Cl5m)^CIYZv)dN(2!A
z0t5mBvnp5wD*yrn0t4h#beEC2kQ)%QDxL-i5(WdvUkE%1O9Q%>i?)zD0pgd&wvb7G
z^eR{b^`ZgRp9q};jNN`b*aRyB#8+4Yw0>9v%mF+E5d*v^SOhBrp;bx)qXJ3;-F-X*
z*8o@q5d(-SSOhBr-&M#130MRy1EGJA1Q7$TD_8>wtOP3qo>d4E0|{6JECZo`$OI7s
zv?^EwF@ghV2;>9h0CWSJe>?-_00=ynj<=8+0l=53w~#0S;+N95kT?m5R{}Tz1Q7$S
zmlL><YB8Z!SOW=I1S|ugf4T$_1NbU}gINgv1S<pHRRSde0|{6JECZo`kOUC}_$u-Q
zUIU|-uDFn10m+vQxsXy9r&LM=5dZ=N0t19Bo&+lZ0t5mBlT>t<k-3l*mx{TNA`pZu
z<OT>51_Sq92s{WA1B;iuxsV@!np6m81Q7$~D@p__1Cdl%0|{6JECZo`SOgIR#VlCY
z2%!iQOOZ=USqL!*AxS;Ek_Zz7D+B*i2%ZB8SOhEsp?{DB5d+sOSO{4NFa#?D;Z;}z
zw0>9v%mF+E5d)tpSOhBrzEesAqXJ3;-F-X**8o@q5d-ThSOhBr{8QEz0|{6JECZo`
zkOUC}k}X&R3AzL;1G|@Sx{x3Ng_n`KkkA9rRDzd)M3EdK5d+&S$b(r3{sb!n`%|6+
z30MRy1EGJq1Q7$<EXV|G1GcxKyO2N)B$F4q2y_Dr2rdJ4JOc_?0x%N;DrTZdOG}Xi
zD*yrn0t1j|ke5BMkSP^{8IS}k00IO81LtOR1Q7rN1Ofw{mqf{sC;^n0ZOM>H0m7HJ
z$&fi6if0Id1Q7%F7)k^y1A%8)0||lzECZo`SOgIR)t3#*kSPM~W|y%Yks1{d1DqLH
z10e!50Rs~V<OC}Nmum>*mmr3Z6$MxX5d+JYfy$694wh$FTmu|f17KJL5d*N7yULJI
z4w}<`2y_HN14~#05d*T98OxAw1G;B|m$4y{84VEw+Zo7%SqT0FYy+UTvCELu7Xj9{
zhU1V40Ri5ZkmQhh0quti<&d`v<&pRt8Tiiu0R))=1JSU4N(2!Dug?K=$OM@I1JSn}
z0+R3n3Hr_f2%ZEh1KO7r1d_@D(ud##l84{~lZW60l!xF1m51O2mWSX3mxtg4n1|p5
znTOy6nup*7n}^^8oQL29ormBAo`>KBpNHTCpoicDp@-lEqKDuFqle%Gq=(=HrH9}I
zrib7Jr-$GKsE6PLsfXYMs)yhNtB2qOtcTzPt%u+Qu7}_RuZQ3Su!rCTv4`LUvWMUV
zvxndWw1?mXwTIvYwuj&Zw};>axQE~bxrg8cx`*HdyNBQeyocZfy@%igzK7rhzlY!i
zz=z-j!H3`k!iV4l!-wDm#E0Mn#fRVo#)sep$A{nq$cNwr$%o(s%7@?t%ZK0u%!l9v
z&4=Iw&WGRx&xhay(1+jz(TCs!(ud##(}&;$)Q8{%)ra5&)`#E(*N5N)*oWW**@xf+
z+K1o-+lSx;+=t)<-G|@=-iP1>--qA?;D_J@;fLS^;)mb_<A>k`<cHt{<%i$|=7-<}
z=ZD|~=!f70>4)G1>WAP2>xbY3?1$h4?T6q5?uXz6?}y+7@Q2_8@rU39@`vCA^M~LB
z^oQUC^@rdD_J`mE_lMvF_=n&G`G?>H`iI~I`-k8J{D<HK{fFQL{)gZM|BK)R0gvDW
z0RfNT1Ofq%-~<B!kKhCZ0gvDW1p$xX1O@?*-~<N&kKhCd0gvDW2?3Ab1PTF<-~<Z+
zkKhCh0gvDW4FQkf1P%d@-~<l=kKhCl0gvDW5dn|j1QG#{-~<x^kKhCp0gvDW6#<Xn
z1Qr30-~<-|kKhCt0gvDW83B*r1R4R4-~<~1kKhCx0gvDW9RZKv1Rep8-~=B5kKhC#
z0gvDWApwuz1R?>C-~=N9kKhC(0gvDWB>|7%1SSEG-~=ZDkKhC-0gvDWDFKh*1S$cK
z-~=lHkKhC>0gvDWEdh_<1TF!O-~=xLkKhC_0gvDWF#(U@1Tq1S-~=-PkKhC}0gvDW
zH35&{1U3PW-~=}TkKhD20gvDWIRTI01Udna-~>AXkKhD60gvDWJpqs41U><e-~>Mb
zkKhDA0gvDWK>?581VRCi-~>YfkKhDE0gvDWMFEfC1V#am-~>kjkKhDI0gvDWNdb@G
z1WEyq-~>wnkKhDM0gvDWO#zSK1Wo~u-~>+rkKhDQ0gvDWQ2~$O1X2Ny-~>|vkKhDU
z0gvDWRRNFS1Xcl$-~?9zkKhDY0gvDWSpkpW1X=-)-~?L%kKhDc0gvDWT>+2a1YQA;
z-~?X*kKhDg0gvDWVF8ce1Y!Y?-~?j<kKhDk0gvDWWdV=i1ZDw`-~?v@kKhDo0gvDW
zX#tPm1Zn|~-~?*{kKhDs0gvDWZ2^zq1a1M3-~?|0kKhDw0gvDWaRHCu1abk7-~@94
zkKhD!0gvDWbpemy1a<+B-~@L8kKhD&0gvDWc>#~$1bP9F-~@XCkKhD+0gvDWeF2Z)
z1bzXJ-~@jGkKhD=0gvDWfdP-;1cCvN-~@vKkKhD^0gvDWg#nM?1cm{R-~@*OkKhD|
z0gvDWi2;w`1d0KV-~@{SkKhE10gvDWjRB9~1daiZ-~^8WkKhE50gvDWkpYk31d;)d
z-~^KakKhE90gvDWl>v|71eO7h-~^WekKhED0gvDWnE{XB1eyVl-~^iikKhEH0gvDW
zodJ*F1fBtp-~^umkKhEL0gvDWp#hKJ1fl_t-~^)qkKhEP0gvDWr2&uN1f~Ix-~^`u
zkKhET0gvDWsR57R1gZg#-~_7ykKhEX0gvDWtpShV1g-&(-~_J$kKhEb0gvDWu>p_Z
z1hN5--~_V)kKhEf0gvDWwE>Ud1hxT>-~_h;kKhEj0gvDWxdD&h1iAr_-~_t?kKhEn
z0gvDWy#bHl1ik@}-~_(`kKhEr0gvDW!2yrp1i}H2-~__~kKhEv0gvDW#Q~4t1jYf6
z-~`73kKhEz0gvDW$pMex1j+%A-~`J7kKhE%0gvDW%>j?#1kM4E-~`VBkKhE*0gvDW
z(E*R(1kwSI-~`hFkKhE<0gvDW)d7#-1l9qM-~`tJkKhE@0gvDW*#VE>1lj?Q-~`(N
zkKhE{0gvDW-2so_1l|FU-~`_RkKhF00gvDW;Q^1}1mXdY-~{6VkKhF40gvDW<pGc2
z1m*#c-~{IZkKhF80gvDW=>d=61nL2g-~{UdkKhFC0gvDW?E#PA1nvQk-~{ghkKhFG
z0gvDW@d1zE1o8oo-~{slkKhFK0gvDW^#PCI1oi=s-~{&pkKhFO0gvDW`2mmM1o{Dw
T-~{^tkKhFS0l(k`{Q)p~q9(H6

delta 118732
zcmV+W{{#S#{`C)#{`C*A*Z~xI%7R<(B{<3c*6n!-!fp{bbRGnQ>;Z@D0Re~X0Ro5Y
z0Ry+}0R;TG8LBcA000000Pw3f0000000;m80000000@@>y95_WFmx0E000000IW9v
z000002mk;800000V007!000002&^{%000002mk;800000fOHf9000005Ue)<00000
z2mk;800000Ft=g51Z0E((3i$P1yljlmm5F@MFG{9VL$~w0oJ#mKm|zw0oa%7K?O<y
z{FgOC1x5n?RF{211w{b>m%~B@NdW?v4MPRg0R)%gLj_U+370iQ1vLQ;muo}?Rsj!}
z!bAl-3+@5{0LKL_gfal2mk31#ISbJP0074Y>4P!=p_fcW1u_d%1polY1+0TI0HT+P
zMFm3vqnFJ^1v~+ymkLG&JOQPbM@9ua0j8IPMg=?pr<cJ-1waeY0{{TW1u23u0JxV7
zM+HC&)dK(k#|7SjG62JuS4Rav3)uqz0LKNBfieKjm!d}nKnvRg0074YNP#i{*_Z7{
z1waex0{{TW1^s|B0OXfANCiL(Ap`&b#|5&0G63(FgGdEH3nByn0LKMqfHDC3m(EB9
zN&y6y8A%070S%XANd-dz7?-F?1w#QNm+VOeLjf+AFiHhO0Xdg+N(Dm!M3=Zq1w#Q(
zm;6cvLjhWsKuZNV3xEIs0LKNfeKG)LmxN0NHvxc`yh{Z#3xEIs0LKM9eKG)Om-$Nt
zISY;e0074Y_<S+|ahE+z1vdeXmvu}9G7F9X0074YgM2aobeFPB1vv|w0002T1wVW;
z0DYJCOa(Uqo0l$41u_ep0002T1<HFe0DzZkO$9UosF$Ek1u_e$0002T1z3AB0Ew5~
zO$9&;cm)6e#{~_0G60g7C{6`B0eF{VP6as&cm)6e#|4ghG60pAr%nYx3yB2)0LKMG
zdNKf>m-0>pIsu87El&kG3yB2)0LKNyc`^W@mvK)8Knt1$0074Yd3iDbs+YV^1v&wm
zm-kNvISZNv0074Y`gk$`t(QAc1x^d71polY1)+E{0JfKuPz5>xsF%-B1u_e$1polY
z1u1wk0J)b8Q3Xy5x&;6L#|6-LG62GtWl;q>0lJr<Q3Wy!x&;6L#|2n-G62Pw-BATj
z3(N%o0LKOYb}|6Ymo`!bIswd=aZ&{`3(N%o0LKM~b}|6bm#$I;E(^2(0074YPIfW?
z2AAhj1#b(q0002T1=Mvi01dZcQw6{RDjZY*05RK0i3I|4G63n_H2?rejRg_X4gkjm
z_;NA;v;Y7A#|7$gG5`*jOI8I=0VS83Rs~7{Czs_`1x^7fmo!%eN&zgFe^&)g0WX)%
zR|QG|F_#%w1x^7pmuOf8OaV5RwO9pA0XUZeSp`f1I+ssb1xx`vmz-GzN&!BX=UD|x
z4nb4^07!`i^lmZ$=}rQdL01J6mr_~<A`5c>0074Ywr(;2E0=#-1v(3J0002T1#xaN
z04<lqS_Mf9bN~PV#|0g3G5|N17+VE83v>Vg0LKN`Z888kmt9*0MGJKR0074Yh;1?e
zLzk^v1xO2Z0002T1vPCl07#Zx1-A=z0002T1#@gN085v`Tm?G|bpQYW#|15HG5}4N
z2weqI3wQtk0LKN;Ycc>>mt|cANeg%Y0074Yd}}fQUAMVi1t9|qcmMzZ#|2PmG5}zg
z<zNMW3wZzl0LKN!XEFe4mv3POItzIK0074YfoC!RY?riQ1&j-N0002T1?pxp0Cty4
zVg)!0dH?_b#|5WmG5~m&j$#Fi3wi(m0LKLpW-<VKmn35aI173J0074Y&Sf$HeV1or
z1xX8f0002T1$t#N0DzafV+D%~dH?_b#|7kMG5~{@QDg->3wi(m0LKNMWHJDSmzQJ(
zMGJZW0074YPGm9wh?nPN1&RxL0002T1-WA~0E?G=Wd$?~dH?_b#|3(0G60R2yJZDZ
z3wi(m0LKL#V=@4dml$RRR110l0074YzG5-}m6vyB1w#vZ0002T1#n_A0GOA>W(8Rb
zdH?_b#{~>xG60;HDQ5+V3wr<n0LKM*VKM-wm%V2NHw$|J0074YHeoUVsF(j|1&IrL
z0002T1)^Xw0IQdiXazM3djJ3c#|2_wG61cY(`W@p3wr<n0LKLoU@`!(mnCThhzolF
z0074YeP1#FvzNGO1vU$N0002T1w3Ce0JWF-X$3<IdjJ3c#|7?QG61-jM`{Iy3wr<n
z0LKMhUNQi?m)B|qFAIAB0074YBwjKAy_XSd1yBon0002T1<7490Ku1RYXw#ddjJ3c
z#|3CzG62Pw&1(fi3wr<n0LKLyT`~a3mm6#aSqpmr0074Ywp=m*%$I*`1&j-P0002T
z1s_~80M?flZ3Q?Bd;kCd#|7M5G62|@S#1T53w!_o0LKMGTQUIKm-1}|H4A(I0074Y
z16wix-Ip?M1xgEi0002T1+H2$0N|I0ZUu-7d;kCd#|0Z&G63V38gB(Y3w!_o0LKN=
zSuz0Smt$`QNeg@c0074YepxaA=$E%|1&j-P0002T1?N~Y0PB}da0N9Bd;kCd#|5WY
zG63zDj&KE23w!_o0LKMMSTX?dm+Wu_R||Xq0074Y=T|ZS^_Na@1x5>e0002T1(sJb
z0Qi@maRpind;kCd#|1H0G64LS1abw73w;0p0LKNJRx$tvmzr_~I17CM0074YSXMFs
z2$$V*1&0fL0002T1;$k}01KCIa|JjHeE<Le#|48`G5`&ivU3GS3w;0p0LKL~RWbk&
zmjiSKg$sQE0074YqEs>f6PJ^81vMCb0002T1!7b(02NjM07!`iQ&chl>3SBI!F&Z1
zm-BQ5B@2B30074Y5>zq(7?&e;1&9lM0002T1%FdA02`OKbp<pFeE<Le#|1!BG5{Ty
z^>qbM3w;0p0LKO8QZfJ`mr!;ER119o0074YhEg&BC6}*u1w{*e0002T1vXML04SIL
zb_H4seE<Le#|6<*G5{=>WOoHn3xEIs0LKM(Q8EBFm%n!fItzdR0074YF;OxAIF|%?
z1yBo)0002T1=dh90792vcm+BOjsO4v#|4p4G5|!Eq<95T3!4A{0LKM3P%;2cm;ZPL
zIt!Zs0074Y@lP@UP?tn`1yBp90002T1(r`T09u!yc?CKPr~m)}#|2SOG5}nc<aq@`
z3*`X-0LKLcPci^zmoa(;Gz<6v0074Y#ZEE+bC+;>1%3-*1polY1w2kN0DzbIdId2H
zVFdsH#{~dRG5~^?HG2h33t<HS0LKNWO)>z5myLS`hznu`0074Y6-_b#kCz~P1uhF>
z1polY1>8(B0Fjqjd<94gVg&#I#|4Q@G60m9tb7HG3u6TU0LKOLOELhSmqUF8F$-e_
z0074Ywo5VqpqGDr1x*WM1polY1zt-s0Hc@JeFZ@aY6SoQ#|08gG61cYBYp)z3u*-b
z0LKN!N-_Ylmv4RrNegNP0074Ya!N7)w3owv1wIRE1polY1u0500JfJ5e+4)TY6SoQ
z#|7p|G61-jPk#lq3vmSi0LKL;NiqP%mk)piFbi=70074Y>qs&H$CpZg1vCqB1polY
z1*}Li0L+((fCWzraRmSX#|2VIG62q(<A4QN3wQ+p0LKOIM=}7|mqvjFSPOUs0074Y
zj7Ks6+Lx(;1xgEX1polY1vf`B0OXhcfdx1VaRmSX#|898G63e6L4pOe3xWjz0LKM2
zMlt~NmV*T{3xWjz0LKORMKS>Ompy|8ItzjY0074Yvqdrh`j>)(1xX8n1polY1z$xn
z0Q{HMg9TU%i3I=v#|8XEG5`jbIfMmK3yB2)0LKNOL^1#fmz9JCN(+Jo0074YOGGjN
z5|`?P1vm?W1polY1qehk02G%kg$1<>lLY_(#|213G5{f$?S%y}3zG!^0LKLlLoxs(
zmnen>Gz*gj0074Y&O$N(D3@o31y2i;1polY1$06(04kTkh6Pv)ngsv=#|05WG5|D}
zB!>l23z`K00LKNjK{5a~mw<-_N(++(0074YU_mkfK$q2r1vm?n1polY1sy>$0791-
zhy{5Is09E3#|4={G5}APoQMTF3#bJE0LKMXKr#SPm*I#7Pz$;R0074Y`9CrMTbDeE
z1v(451polY1++gh09}`Xi3Ly#%mn}d#|2qGG5}|n+=&G`3(N%o0LKLsKQaJmmm!J;
zN&#_~Y>EX$0du#piUkHF2F^(U0KU$by{-jD0nnEOuLVZ|(w9uH1waANw~em_C=D!l
z1polY1$7KE0K+)|0KUUH002md1z`*_09aZ`jSX{a4gg4vJwcfc0J?aW;k*U@8h-@<
z0KVKg0074YN((Xoe+2*l#|1nKG639{`Mm{45O@Uu0LKOD3Niq;IRF5@wwF-81yvb%
z1polY1%(PS0J1p%09i<h1$7EC0A9Ybx4ymwhY2i+1polY1?2!T04zEH0KP0b002md
z1<?R909aZ`jSb^p4gg4vJ?UQ#0J@2namWS!8juA50KP>!0074Yy8kf%kOcq$#|5na
zF#tuEiOB^=5Qzl<0LKMd|1kg`IsgE^AD7_C1yvb|1polY1@-<h02?|009i<h1?B!R
z0A9WumrlwB_bi$P0074Y<^3@Ll{x?bzLh!v07!`i(fu(1SXxMp4dYx607#8J2VD*T
zx|)}9%LV=#p#=Z{zOFg|0LKNo{4oHb1polY1+4rr0Irvb%mqgfngsv=#|2vaF#w1<
z006#-m*C6=RT-KE0074Y_4_dZggO8KSxAWm<@+%JUcQ94PR#|z1{E<$0074Y8uc*%
zG5`Po#{~@aF#s`_Dbxid3*|}x0LKO0^f3VCmsr#VToo}%0074Ydh{^>G5`Po#|3Qk
zF#s`_$<zfd0q2+a)CFV#>6cyA1xXydN&o=I1&H%80LlOW07!`qb@MR*#|3lqF#x=`
z!qo-(0|Ankz1anN0hE_G+67+*j7$IkzKoZh+65c|l9#&L1;PPbmwnp>dI4US_}c|v
z23AY}0KQh2UEBp60bG}Q+y%k`6qh~S1$qG%m$=;pUj`0J006!YmmS^(903%UI^G52
z0jHN9-vxmIsh6PN1y=^4O8@}Ap_l#N1snmVw;JFDf&>+qNdN%H1(M`301p5F0LKM{
z<S_u4m$Bpp{|0tB006#rm$u~vfd+Cp006#nw>{<s69EQ6IRF5@L6=tN1%U=VIRF5@
zJ-6NG1zH3w?gIb-#|87;F#x72006$GDgXdTi3Q)>F#uRvNR18gI}QLyjXg(i4gk9D
zw`1%DtqltX1ONcX1scyW0N=M8`UTno3kC!L0LKM2%rOAYmi+~83kn1P0LKNq%P{~c
zmwEjKG!g~`0074YeakTb+$sP7NQnh*mw^2R7Pr;?1wsK3?gIb-#|0D0F#wG!006#?
zmnZ)PDGLn*0074Y+Q~5hc$ZuM1#1fw1ONcX1!T!F0P~m8{{<@x3IqTE#|19QF#s`@
z2L>h!2?PKD#|8b!F#rvfI0hvP3IqTE#|6#EF#s5qXa-da3j_cF#|3T3F#u4N$Oa_~
z3j_cF#|1seF#ukb_y$c23IqTE#|7obF#tE0Pyq%t3kL)M0LKNT$1wovmyiJlB?}A$
z0074Yb;mIPYnQ+Q1|<s$1ONcX1x3d(0794W0R}Y-2Lu2B#{~t)F#!0NFaic03k?JS
z0LKO0#xVePmskP@C<_e)0074Ys>U$@dzXs>2CWMX1ONcX1@gr)0DhN40|qh+4+H=J
z#|5>;F#w8}fCC043l9VU0LKM|#W4Vwm#_l{atjm$0074Y2*oh~<(Dl41~3Z{1ONcX
z1<k}U0Hl{_1O_Dw5d;7L#|53lF#xZZm;?qT3lRhW0LKMw#4!N9m&gPLCkqk;0074Y
zJH#;n$(Q;B1~Lm21ONcX1^&Y^0MwT^1qO2q90ULW#|41HF#ye%wFL%03l#(a0LKM3
z!!ZC1m;VI@9}5)(0074Y48t)18kZ>s215%R1ONcX1-`;D0KAuX1_mJu7z6+S#|4tY
zF#sl)qXq^#3={+a0LKMQ!Z85;E0^g820sfJ1ONcX1p>k`04SF;2L@Xi?gIb-#|50h
zF#vrk003D?i3N<oF#uk^eV3;P1}X~}1ONcX1zEu{04kT<2L><;7z6+S#|0k2F#s}`
z7YGJ83mF6e0LKN}z%c-4w^|4WB>@=j0{{TW1*^X?0Cy??0AEOn1)IMy0KRvZmk9<t
z3mgOh0LKMbzcB#Ym)!{lPZ${l002cX(nyH~_P#LyNs9)@1@XQy0BM&*3I;t390ULW
z#|5drF#zV5j0y%H3mF6e0LKM?zA*r7m$nK9DFGRm;0gv70dJT03I<X|?gIb-#|7cN
zF#vHY002Q*NQni|y)gj3aVh`+L0akO;Q#+fjRn{o4ggGxY$Om!i)<tq#|5>$F#s9_
z0074Yr@b)%dzX$21||y{1ONcX1$MnL0DYIg3kF6D8Uz3U#|0(5F#vy;5DW%*G#Uf|
z0LKNQyfFZWEC2vV$3!F$00000>E-SJ|45Ampc@VV#|4MHF#r?<0074YdAu<I?U%d^
z29XOK1ONcX1>L(b0M(aR4F)C)7z6+S#|5mrF#tQ4i46uZ3m60d0LKMzyD<Pcm&FYR
zN(&kU0074Y8@n+8ke3<`1~o4l1ONcX1>3qY0FEpG07!%A0qNK4|Nlsh1>PAB0LKNz
zx-kGE1ONcX1-H5}027yf4hA0!Ap`&b#|4eLF#rgcs161%3m60d0LKMWx-kG(m*Nfv
zJQ5lN002md1qZq@0LKLdx-kH(moN_o@DPdv0074Y3b`==AtV3*z9E+|5C$6oinmM<
z2IvY4p#uN_#|8GSF#uJ!KN|+l0Sl!A0074Y9jq|`)wdWO208)@qyqo|#|7A_F#wd8
zT^|N86`=zF0LKNLsWAXVBme-uL?i$Jx}le%9|mg+p#uN_#|0g!F#uAx7$63H0uYJ=
z0074Ym8LNO2_ygjz6qD5A_gD}r2_x}#|34kF#y|_&>{v#3#9`90LKLsrZE7`mmwnt
zBnzYi0074Y<)twIoR?4|246Xf0{{TW1$(720RJNZ09i<h1#6`-0A9ZTBLD#DZ0!I4
zNR0(_1P%bl1zV*t0Ivf80LKMQr7-}zm*^t~Efu5#0074Y6{Rr%gCqa|zJnwH0J@}?
zDkKJLA*BNV0LKNDq%i=zBme+0?L{&FzPuy=07!`ie55e|x}}%4BnC$dqyqo|#|15<
zF#w{M2qgx93#tPE0LKNQqcH$Fmy;z1Di@~%0074YZKE*&^&|iQG3~zfBme-qr<c?v
z1|bWi0{{TW1vR5F0IilL1}zn$0{{TW1^uEi0AM5l0KQ-(006q8mqI26n+vA{0074Y
zO`<UX1DEI~1|<uk0{{TW1s$R>0B)BUCk7=Kr~?21#|7=7F#r!G001%Vz7Hh;0J^A`
zR3`=@3#9`90LKNjp)mmFmw+b*Eeogv0074YeW5V`9hbN#2BjCO0{{TW1<jx_04XH^
z05R>pDJ1{^x~i9SC<b&3uLA%8#|1*5F#w{M^C$*77pnsR0LKOWpD_SoB>(_1?Y?0p
z006qHmqaNBAq%Sm0074Y#h)<%Z<lZ>23`xW0{{TW1zevo0NR(?DF!SHssjK3#|0&y
zF#t%H5Gn>G3$6nI0LKOOo-qK0mq02891E%g0074Y&Ym#<Q<rBd1~n$H0{{TW1&*FE
z0FfmC07#2$Bp6JKd?XOQktF~CNQnh-o-qKrub09q24D-Z0{{TW1qhxo027xjD+V+s
zumb=9#|6oqF#yRW002mfY$O;=i+m&yzR4v307!`itDP|by0DjrD+VYFuLA%8#|3+x
zF#xcax+?}r8nFWa0LKL<oiP9cCIA3Pi)<tqz5*ry0J^c49V`Ye3$OzK0LKO2oG}3C
zmsTtWUl@u50074Yb(}E(`6B=TUr31sWt=epzWJBUECwbEvjYGC#|1Z>F#tT5{wxM2
z3$OzK0LKLeoG}3NmoF^_Vmpch0074YnVT^H^CJKNL0L$N1&Et50KW4h002Q+>4@V0
z|45AmkopY(#|3wrF#xdx0074YX`3+sA(zZ81|bWv0{{TW1wETF04$gIEe2f~iUR-u
z#|71zF#zo&002Q-NQni+nlS*r?U!yY1{Mpd0{{TW1*V!Y0793KE(R(KuLA%8#|3ek
zF#xKU#4ZLY3$OzK0LKMAnlS+4m-sFQDGRX!0074Y37Rng8@DVk1`Poa90dRX#|812
zF#xMI006$Lmq{=N8vz`*WiSTN3kx^}0074Y^^7q9cDF!01|<UvI0XOz#|6cSF#vd%
za6bku3pfP;0LKN5i7^0rm#9AmEekjW0074YRf#bGe3#%q2F(jM1polY1$~Gy0793z
zKn7I{HU$6x#|0gTF#sl)7(oUj3pfP;0LKOIhcN)4w?;t*4*?4{1polY1=WTz05q3i
zLIyYvI0XOzO^NQu1&@X?0Irv&LI!FJGz9<v#|0XOF#z+I8$$+y3p51)0LKN5g)so)
zm#9MqMGG_q0074YJ%up<0heGx1{9YJL<TGiI0XOz#|83)F#wL2LqrBD3p51)0LKNr
zgfRdNmtaB$6qk%d1}h6V1polY1#^Tk0GOA;L<UF@90dRX#|0pSF#wh|006$0w;DwT
z@&OAv1polY1s;Mi0Q#2~M+UDGIt2g#NQnhsfiVEb1zmwL0Q@$WF3=Acmw!A44wwH&
z21^+n1polY1>k@&0ERUH09i<h1<-&o0A9X^mt#l<B@;RY002md1)YE~0LKNJfH466
zHkX@7215%u1polY1x$c30LGW;NCqMcHw6Fy#|0mNF#t%H7D)!y3pxb=0LKMBe=z_X
zm-k5qG%`8`002md1^Rw50LKORelY+6Hvj<XSpNV2NR0*m-3<W81?GM+05k;v0LKN`
zelY<1mt9H*9}71H0074Yu6{89N|%R926_%U1poj=G5^N}CVnvh1Gf)L1}p&!It2g#
z#|7_vF#raaMob1;0XmnVOa>SO2sZ$iAV&rtm)lGRJsdg(002#i=SYbK3VbmDOp9bB
zP{#!Wd@%qDHvpF)M+O#`N=*jT3pxb=0LKLkdochlmnlvLL=HLy002cX_QwUfdNBYv
zmwQeIWehq60074YN_sH>U^ka=JO&h&PX<B`It2g#MKS-!1?71$0CAU4PX=fUHw6Fy
z#|3hEF#u+lKs*K&m)K7RNDDdz0074Y3VAUAOqVNA26GEK1polY1&nwx040~HPzFdV
zIt2g##|1ffF#uOL008MK|Ns9;jRip04FJalCwMUcIRyX!#|0U9F#yh&98m^g3poV<
z0LKNbcQF9WmxxgYiVHaf0074Y7<VxM>bD(I24xBh=mY=&#|5ZtF#wLYja&xy0xZx3
z0074Yr)e<&Kr;XUzCbeo07!`ilW8#kSXxMp4W-Bp07#8JLCOsPy3n_;Uj{@03+Ds?
z0LKM4XfXh8m;PV|dJE?S0074YvS%>>aF>H&2C@t11ONcX1@&ez0DqT2Vg^kM=mY=&
z#|52cF#wF0m|_N13-JU10LKMAW-$Qum-u1^wF~M50074YePuBKn3uR?1|<vX1ONcX
z1x;l!0H&AdV+J(~>jVG*#{~^#F#xldC}ajT3+n^`0LKN*WHA82muO@LH4E|t0074Y
zjbt$ZGMA`i1~m)p1ONcX1x;iz0M3`_WCmCZ@&o_?#|7$RF#t4|OJxQ!3+w~{0LKNb
zV=(~SmxyHsD+}@j0074YcVjUCIhVd=1}_Wm1ONcX1wCUi0O^<bWd=(N@&o_?#|7tN
zF#tW6PG$x)3-1H~0LKNWVle>gmyKoyV+-;G0074YI$|*ZM3?(!1~?1y1ONcX1@~bw
z06>>MX9hJ3@&o_?#|5`xF#t%HerE<W3-JU10LKM*VKD&xm%V2OUl7m)0074Y4q-6>
zDl-58zABe2Xa?yE_5=U`#|0H&F#wU5AZZ4J3-ts50LKM~Uoik+m#%3B+D7&S002mf
zY$P~EG5<)31(jYg0LKNBUNHcOGyniaG4@P}(?v1B=^OI@|45Amk-ZH7Nr~Ra1$ACA
z0PzF>0LKMqUNHa!m(FSiAPeyX0074YJYF#X2ABD21}qZy1ONa?i3I>&F#yK}{#`Ku
z$Co#21}O{m1ONcX1<GA90B)CSYX(jW_5=U`#|3d+F#vwI#A^oZ0Sop70074YdRs97
zfS0>%1}_r#1ONa?i3K}bF#yK}Ia@IR$(Q_X20;?{1ONa?i3R9dF#yK}=2|fT*_Tgl
z22Bh11ONcX1&&%V0MnPJZU!<E_XGd{NQnhZS}_2}1xi{m0M?i5ZU#3C_XGd{#{~#l
zF#z1REpG+~0Soa20074Y_E<3hB9}jK24xcV1ONa?i3NdJF#yK}e^@a9u9voO2A~V}
z1ONcX1=LqD0JgVbaRwj)8ukPL07#2$BsfJe_DG2ZoK-OZ#|4^IF#x8wn{x(D0W7uy
z0074YLR2vT%O(H-zRM;607!`iE>tl9SXxMp4M4LE07#8JX6Fq6y0*6zbp|CA4$}hw
z08NR{#|7>`F#w8}M~Mcg3(f-o0LKM8KQRCYxB7_&TLKHy0{{TW1=Tz;0CBfqj0P(L
z4$}hw08NR{#|4}^F#wR4nT`fm4$}hw08NSg#|1Y!F#wyF{*DGO3(f-o0LKOWIWYkC
zmpG3GRSV7o0074YojEZ81echP2B;9W0{{TW1?o640JA0l0KT)AP>=>u3(^As0LKN1
zI57Z5m#mNml^M1J0074Y={GR|sU`pbSxAWm**7r&UcRZ9T9F1D3(*4r0LKNjH!%P;
zmw=H59ShF`0074YjW;m>5|^lv1}F~G0{{R`iO<IcS2r;Lf|uTr1}_WG0{{TW1syjr
z03eqbk_NgwwgUhF#|2n6F#w|`003V|i3LbDF#x`!CIA3ISm_Yr|Nlsh1yGd@0LKM4
zHZcIy0{{TW1t~T$0JA3mmqCvP7MCQG1~3cM0{{TW1=ck&0KvCklLkEjM79F}0LKN6
zG%*03CIA3ISxAWmeKauuzMUoj06|*mMd1JcNR0)BlMMjJ1#C1i0M!Ek0LKMhG%*0q
zCjbCQi3LwIF_&SJ1{Rn0lm;d;)&l?l#{~{FF#y>o002md1^+WK0PC(ujRg^t4FF7u
z?Z*Z6Gcf?x0{{TW1?Mv{0Na;Nl?GA^)&l?l#|4TrF#z6|tCa?XJ+=b?0LKOUGBE&`
zCIA3ITS$oo>M}6^zLzEd09)y^-v9qdjRo_O4FF7u-NyyjGBE(w0{{TW1<5io0O6Nx
zmIh1>(gOeh#|3gSF#tp-0GC0J1{RmwmIg8l*aH9n#|0fSF#zqC7?%bd3)lky0LKOP
zF);x9mp_*VMGM&j0074YrZF)92bYhR24f4|0{{TW1voJ=07#epmj*2|+5-Rp#{~c}
zF#r-M002md1@SO30O_6l|Nlsh6#$S80LKOAFfjny0{{TW1=%n$03erKm<AjR+XDaq
z#|5%5F#s%=gO~<33(^As0LKM!FfjmOm&BL`X$##00074Y{Vy>9H<vh>20jbk0{{TW
z1+*_Q0B@IpnFbsS-U9#t#|4cqF#vs+sF?;F3*Z9)0LKMpFEId#m(Q6791Gn80074Y
zKQA!=VVCxq1}6*M0{{TW1r0AT09KbMng&n}-~#{v#|5%3F#wAw0GC0J1{9Z>ng*o@
z;sXEx#|7+{fSU#wmtC6%91P+E0074Yu`MwGt0<RXlLi@=o|^_73*iF*0LKMzEinL`
zm&Ka~L=52r0074YAuTZgnkbiGlLi^LE}RAw0Sn>-0074Y=qoV*x0g+w24xQ80{{R_
ziOa_Ybt^Fd?w7!w1|uBf0{{R_iF70oNQnhKD=`2}i*zI)#|1VkF#znh|D6V80Sn>-
z0074Y$0;!Yua|D02AnJA0{{TW1t2Lg08%Lc0O?Tb|Nlsh1t*9N0LKLmDKP-u0{{TW
z1pp~A08p1Tpaz2r;{yNy#|3gIF#z$G!=MHu0pyqGpavZQ0G9=!22Kv+0{{R_iOa_Y
z)F&|j^p|0w215?!0{{R_iOa_YgC{WnYnQU221pC$0{{TW1u!Qu0AQB|q6S(H<^uo#
zOo_|K1;{2b06~{+q6X9u<O2WzOpA0RAjbvdB{2X5mr<hzHw)wg0074Yq9ri^0+*Ac
z1|uBg0{{R_iF70oNQnh%B{2X@i*zI)#|2|0F#rsg(xV1|3+Mv?0LKN`BryQSmtCX=
zNekx#0074Ygd{Nls+X~(21*O(0{{TW1urBq0H~J+r3QTq=K}x$#|5V&F#w8}j->`!
z4(9^^08EL?#|1zmF#vg&^`!<x4(I~_08EL?#|7>pF#xrfN2Ufv3+Mv?0LKNMA~68Y
zmzSmnMGNNx0074YP9iY?l9%VE21N_!0{{TW1^yv10Hc>Trv^m}=K}x$#|5q-F#wpC
zho=Tk3*rL+0LKMYAu#~lm*A%cD+}TS0074YA0aUSz?T-N20;tr0{{TW1=Aog0KS)E
zs0LUI=mP)%#|3a8F#zV5#i#~K3+Mv?0LKLzATa>sml~-CLk{Qz002yh%f|)EA29&m
zmusm8NDJr#0074YcONkT)R(@g1{n+L0{{TW1ydg}00o!gsRk<x>jMA)#|0f9F#r*l
z7^((W3+4j=0LKNp9x(t~mwKuOIUM8z002yhbR-Z+i3M#QF#t@9bR;0h1!f*G05O-(
zss@x3<O2WzOo_`(i*zI)#|6I~F#r{pcB=+J6XXK`07Wt6OpA0RAjbu49WekCx5=vp
z#0m?70{{TW1)UNx06Mprwgznj3xWdx0LKL-4>15Amk_!JEenJL0074Y?G7;jeV0hO
z22~4$0{{TW1&t0d0DqULx&}B4gaZHo#|26bF#v&=>$(Oq3xWdx0LKLl4lw`>mngdi
zPz!_u0074YvJEi+gO`K51|<uE0{{TW1%eGR01lV4y9T-og988n#|6C%F#t)Ic)SKN
z3xopz0LKM^3^4$wm$bYFDGP!F0074YO$;#rFPG@N1~?9c0{{R`iS@??2MjR)wU;iv
z27VQU0{{R?iN{Ed1(9kE0LKM%3o!tdm%+UTAr*uJ002pe$4HF@TWSpe#|1tMF#w;J
z_q_%p3xWdx0LKLp3o!sFmk_!J6}LIQ2KE69h64Zq#|8TcF#rLVI=}|35{3f+07!`i
zKL{}Z#|1qIF#rRXaKi=|mk_}Q8w-U40074Y{|7Mu&6hU81}zeX0{{R>i3P$3F#yK}
zzy~n^1($Hc1{jwR!3GJJmca%l3xfjy0LKMx2QdIom&d^dy$gl|0074YrUo$pBA1WC
z1~N5<0{{R>i3MZ^F#yK}Vg@k)2qORh>7D-n|45AmX=M!nNr~FW1x*Gq0D}Vn0LKMC
z1~C9pm-fO2XcC44002md1;+(30LKNz1u*~$mvF-ds|$t$0074Y4+Sv*Q<o;h21pBr
z0{{TW1-t|?01cOU#0E_Zh64Zq#|2~rF#r#j(ZmKs3x)##0LKLt1Tg>+mmtLk=?jJf
z0074YAp<c0G?x^{20sgi0{{TW1=j*G06&*r#s)?Uh64Zq#|4H0F#vm)uf_&y4Tb{%
z05R>y1rGu-0B4sb#|Bypg988n#|5eZF#v6s5V{5xm!QW6I1+{f002md1x*1l0LKMO
z0Wknxm+Hp`D+`7L0074Y6#+2-ESDh21|$oG0{{TW1?B)T0M(aI$Oc;rg#!Qp#|3}@
zF#yw-wa5lW3xxv!0LKL{05JgZw+6`u5&<l90{{TW1rq--0OBD40KVcO002md1^)gp
z09aZ`jSUc94FE`uJ;`4U0J?OyM9K!90T6Tp0074YNc}JX&>;W-zR;KQ%LY{$bOQhY
z#|7j3FaX9O003D?i3QXAFaTb@#<yV122BGC$OQlZ#|3@)FaXY%xX%XI3&#Zj0LKNI
z_%Hymmz&TA!VAa+0074Y&G#?>z?W#z1|<u|1polY1)cXW0GOAU(FQFG$OQlZ#|35g
zFaXh)(9s4|3&#Zj0LKLe_b>pbmoL%=CJV>~0074Y)%Gv|-?w1W2Au&5#{~cY#|16*
zFaW8S2-F583&;fk0LKOW^e_P8w>Z=WMFk7C1polY1+DKe04bM<*#_7PwFLkG#|6sn
zFaQjfYuX0F3$_IS0LKOW?JxivmpI!7B@46#0074Y&FwG%^OtDb1}zJ=1polY1(oeE
z04$fF+Xhq%wFLkG#|1X+FaQFV|Jw#83$_IS0LKLd?Jxj3w=moWodFBA1polY1zqbf
z00fuV-3BEKwgmtH#|16xFaSNb2;K$_3=7`{0074Y72hxbRktAO2IT<@-~|8x#|0(b
zFaTti5bOrU3*ZF+0LKMY+%N!gm*DIM9Sh(E0074YE!;2wb(aY31|<vL1polY1^wGF
z06dpC?FLZ`-vs~w#|5F=FaS%pl<fw|0Sn&+0074Y&Dk&jO_ylz28RpZ1polY1v}X=
z08O|0?*={)ELZ~o0LKOL$S?pBAOHZq5+DEoNQnjB$S?p{T1brz@FxubNR2&CTnzxa
zShZva0|E<f0{{TW1;fTL0JxWO0tY4wZUX=S#|4tcFaV>sqXGxb0Sj&e0074Yy2LO5
zrI&jI2Vx6s0{{TW1x>^-07{qW1PAyFZUX=S#|28mFaW2Q>jej53v2@b0LKO0!Y}|i
zmskb|I}2_D0074Ym%=ar%9ox72S^KU0{{TW1w_Iy0NA(j1_wj|5Lg2M0LKOOz%T&(
z9{>Qp{Fg-t2S^KS0{{TW1)ab!09cop2nQt#Yy$uQ#|3S`FaTqh$Os2n3vB}c0LKLh
zz%T%Kw=4+<EddK|0{{TW1<}4R0K%7K3J2N?Yy$uQ#|7-YFaT1QNec%l3vdGf0LKNm
zy)XbFmwgKdSPO6i0074YRlP6(BbVR{2PX?~0{{TW1tYyM040|a3<pjUa037UNQni-
zyf6UA1;e~B04JAn3<q2pSOWk6#|2!xFaYu&003D?i3Lo&FaTb@@|Wif2XPo!0{{TW
z1=71P0PG(C0AEOn1;V>90KV*(a193~8gK&u07!}HNQ+z~7)Xf)h`TTV#|4JFFaS)K
zuMGz%3vdGf0LKMWyD$J?m*NcvXggQ~0074Y;JPpX<{tn6L0L$N1<$%L0KVoQ002Q+
z>F4qP|45Am*BK1}#|6H+FaT`>0074Yu(~h+c9(??2XPB;0{{TW1vI)a0QQ#w4+mQd
zZ36%R#|6&0FaVR6XAcKG8gK&u07#2mBp67E=}3tMdATqE#|3w}FaS80z7Gc|8gK&u
z07#2mBp67E=}3tMH@PqX#|1UHFaSfA5C<X)Zvy}T#{~$vFaY$IEf5Dt3vUAe0LKNj
zxG(?_w}21_wFwom0002T1@*2l0JjSO0KT^i006qNmr5E3Gz+l+0074YtgbKswYP~H
z2l*}wivR!s#|5Z=FaW!kjY$V+3yS~%0LKM2e=q>OmP!XbD~kXC0LKOGelP&63IG7U
ztO@`Cx{Ck+07!`q)P67k=~@5(|Gv8l006p+muN}{Cku-J0074Yn|?3=vbUN_2Z#g<
zm;e9(#{~#_FaQ*nEl&q%3zz@^0LKNDc`yJLm!MAvJu8?10074YO?fZ?1q%QGz6A>a
z0J@j}002md4K#T$0O?u(|Np)e3jhGRn3o7p2PX@d0002T1^aj~01LM|PzQ(v3#9-6
z0LKMrbT9yIm(5fMXbYtP0074Y^>Z))Z<jz-2R$pL0002T1+8;10A33K0KQ%e006qB
z0000;i4Bx<FaYUU|NsBKZ3_SZx}}$>RR<>vr2qf`#|2w+FaTti+EoYL3(Noj0LKM$
zaxeg8m%>&D%N6wi0074YopCS#oecm0zMTyK0J`;;qE`nw3-tg10LKMSaWDXzx8_#|
z90C;(0RRBU1t@PY03Z$k0KOm&006oWmls+GB^5DA0074Y@@_ByG5`Po#|7kWFaR-^
zQCbHa6%YXc0LKNrZZH5R4gdhYCk_Arx)7IwS_k+G5CH%H#|5x$FaRaDg<A*q0Tmws
z0074YuWK*>W)1)VzGe;p0J<NSkX;8I6(0cr0LKM=YcK$94gdhYZ4Lkcx*wOqT?bkV
z9{~UW#{~{+FaT?pCte5E6*~d|0LKL_YA^tV7XSdhgckq+x;vK=Uk5Y`I|2Xz#|7kR
zFaUy=QC|m~6)yw;0LKMAX)pk3EdT(%X)OQ%x-XXkU<Wh{F9ZMp#|89gFaT$lL0|`*
z6-NXB0LKMQXfOZ^E&u?&3oZZvx<{AtU<Wh{M+5)>#{~pvFaQa+F<}RM0TpNj0074Y
zeP%EKtu6onzO60*0J><G!eR$B3upuY0LKM1W-tJ&w*g}Z2SN+H0002T1tKIc0Qt8Q
zr3YF95OD$k0LKNrAus?p82|vjH<x{<2O9x#mzk#r!WDo50074Y&mb@WRT%&PzEv3j
z0J?yeaHt0>6@UT&0LKN7ATR({82|vjRv7>Qx`3Cks0SStfC2yj#|2#=FaTW{006#S
z82|vffS2N^2Z|7J0ssKV1->6J04f;(0KO`he5nUj8F2yt0LKMcA20wW82|uTNQnhW
zA20x3z9yIIsRt(ufC2yj#|0T5FaTS(9I6LV0uX}&0074Yy&NzAuNeRUzOR>ktOpwb
zgO{1C2f`JS0ssKV1<xBW0L>Wy0KUx`006p@mvF5ID;1Ig0074Yj~g%m&KUpzzRno{
z0J@Twu&oCj6_Nq~0LKMg8!!OX82|vj)foT)x{{aTtp|z_g8~2m#|6F`FaV+%006$C
zmwc`VRT+Z<0074YSsE|^o*4iDSxAWmNg6N!UcR1}>aGVT3z7l=0LKLx8ZZFUw;Znr
z*#RY$0ssKV1vMBj05Tc?0KPIB002md1tAzP09aZ`jSVw*3;?>7mm0ALxfQ1Z0074Y
zR~Ik<Q5paMzEK(g0J^7_=&=Vg6{i9K0LKLn7cc-)8UO&kQW^jNx~G>gvIjL4rvd-~
z#|6a}FaT2;006#I8UO&gr<Z!N2O|}y0ssKV1&tOk09qOV0KQrp006qDm$I@4s1TI`
z0074Y&=oKMBN_kzz9W}vvj<ffl>z_&#|3K@FaRGK003D?i3M8~FaTb@AD7s(2TmB3
z0ssKV1q2l^02vwp0AEOn1@sg!0KOTQKC}lZ3#S4A0LKNx6fgi-w{o-x!vQ6#0ssKV
z1@#gz0Jj<d0KT^x002md1>q7f09aZ`jScg03;?>Sms++5xfQ$u0074Y7ZNZ4)*1i+
zzSbH50J^-FD7ObQ6}$of0LKN-5ikJP8UO&k*BSr-y1bWgw+A&9yaE6K#|4EEFaX#Z
z006$&8UO&gyqCJS2O|}{0ssKV1x*n!0N)w_0KVTE006qYm-4p<s1T|G0074YkPt8c
zs2TtOzNnX~xCd1kssaE2#|0}8FaV?)003D?i3J-FFaTb@q?Z`E2TmBO0ssKV1;h_9
z0G=8E0AEOn1+)(^0KT4=ez^xJ3%mjV0LKM`4=@1Tx3akhTLC4%0ssKV1sx7B05KZ?
z0KPFB002md1qlu?09aZ`jSU-Z3;?>nmomEt$raB60074YFAXpNPa6OLzE2wf0J_hY
z5WEL970&_y0LKO83@`vt8vp>lP#XXMy3dzdyayu{&jJ7d#|5nnFaTE@006#M8vp>h
z&zF+C2Z|8C0ssKV1r7`_03;g#0KO!bEWHO+8NUJm0LKNX3orm68vp=VNQniR3orm)
zz95&Hy$2@?&jJ7d#|3E%FaTAz%)JLu0VNp&0074Y4GJ&-r5*qPzNH=j07!`i`3W!p
zSXxMp4GU=u0J<5MLca&e6(|D$0LKL%2`~V|9smHo!yW(tx+s?*zy~!IC<6ch#|700
zFaX3J006$k9smHkD3@x$2O||I0{{TW1)T^m0LvZ#0KUr}006oum!iN2iVztC0074Y
z{s%Asm>vKCzL=Lh!3R|t83O<S#|4=OFaVSu003D?i3N!VFaTb@l$WZ(2PX?C0{{TW
z1z86$0Li!9!3QY^3J^>H07!`iAOJ9zabgG+mn6vti2>!8uE_^D0RWfh$p<M2<4XVl
zG3~zNml(<i`z`nZ0074Y7yd5*E)@U(zAhC207!`i1O6`nSXxMp4HaDs07#8JYncoH
zy7&RNK+6Yh0TluQ0074Yef%!~NEHA8zDN}S0J;K~!psMY5cmNA0LKO0`!4_>6#xLf
zAeUXu2Qv!-0ssKV1)BRW08O`>%?FwW6#)SN0LKOO^e+J64FCYX;SB%)x&fC;)CW@w
z0RaF2#|4e_F96@SsMH530Shq#0074YS@SOdLbu%22h|D-;{pHx#|3%lF8~jhyx<2h
z3*!O+0LKMA=q~^YxA@=(_5l#m0ssKV1w7_20I3@Q0KTc0{^AE40n(Qm;|C)P;{pHx
z#|7=>F8~aeNaF`c3*!O+0LKNI<u3pnmz(1U_zU9#0074Ym*g)16qlal2R{qr0ssKV
z1x@5H02Y_%<Oe$o;Q{~v#{~!EF96?{F69S33*iC)0LKNt<1YZ<mv-d`JqzLj0074Y
zcjGSr@Rz>j2R#eo0ssKV1ux?-0P&Xw<_9ke;{pHx#|8D`F8~6UK;{RG5aj{@08EQ~
zBnZa^S>i7MIG5b!2P_cf0ssI^i+m&q#|0hYF90)_80QBq5aj{@08EQ~BnZa^;NdR-
zHJ4TA2U8Hz0ssKV1%lx(0Gt~D0KS};xaS8S5aj{@08EQ~BnZa^PT?;AMVII22Vxo0
z0ssKV1>xW?0GJy909i<h1<~Ly0A9YBmtyD#J`m*s002yjd?X0R1%==*06mwm=m$d-
z<pKZzG22Lq1u)<*0LKL`;4c6_mj~$wAq(XK0074Y0^lzIE0;6r2RjSr0ssKV1-#!c
z0Be_d=?6qH=K=r##|3TQF92v9001%VNQnhp-!A~`u}F;tZaxeE#|2g2F97EP0074Y
zN8c|1W|!{i2N?_J0ssKV1tZ@t0ArUD>IXCn=K=r##|7lxF92beQR)Xv3+Dm=0LKN6
z-Y)=Dm#69n9}DRM0074YWZo|Tk(bfx2TKd-0ssKV1rFXX0FRd^>jx_f=>h-%#|7Bk
zF94L6UF!!N3+Vy?0LKNf-7f%^mxSvFJqzyw0074YY27aXE|<*f2Pq5h0ssKV1vT9-
z03(*{2eu3E0ssKV1$5jm02`OV><3#5?*ae-#{~}DF8~pjC+!C=3-1B|0LKN@+b;kB
zmtgG&F%0hl0074YncFV_{v4N`?FSnR?*ae-#|3iRF97_v!|ewT0SoW~0074YSlTZD
zPM6*82WmC&0ssKV1>@N-06!f707!`i)Y&fp=?VV-|45Am<2VceO^N--1;^Ph0Pq3;
z0LKNn*)IS+mwfLBBn$2W0074YirFs!@RzIa2O$gW0ssKV1zy=N0O6O{?*}Oh>;eD)
z#|0|cF96n;3-AZo3+w^_0LKMJ*e?Lim+$ZgWee;A0074Y*4Hlp#Ft<32WkiG0ssKV
z1!0#D@&_N6@9_s24D12`0LKL)*DnCE9G9?;1{jwt@&`Bz>;eD)#|6gLF95HXZ}JB_
z3+Vy?0LKM@)-M30m$mW-EDPxZ0074YN!Bj_f|u;_2OkUP0ssKV1s~Qg0Ctxa^9N50
z;Q{~v#|6dJF97A2aPtQ}3*rI*0LKM=)h__{m$>r>F$?Je0074YLDeq+nYZ-w2L%WL
z$d~&22UG(7OqWpn2R;=90{{TW1)R(;0NWh^0KVHD006oKx1;<AECB(^m(cwOF#`Tf
zmjnI>GzP*<006$imp%RmNdW+thyDkCEcF5a0LKL-$}a%59RL8nwH*KeNQngz$}a#|
zT1brzBPk33NR2%KXAA(k^_MaK2N(g!mrDN!z7+%m0074YBFQfR)Exi-zSJE60J;Q~
z9SABF1Oos7#|7iaF96dW006$z9RL8j1eII}CI<LS006%Dm52y`5cL880LKL*$S(k>
z9RL8nsFxQ32vr&N0ssKV1-{2G0HhrN09i<h1+K?00A9YNmxTccCkq4v0074Ye#b8W
z*|)a=2m}lPG?%;$2u2ka0{{TW1t_~O08<_S0KQWm006ocmlqBQR{=JcZw?545C;PQ
z0LKMAx-S4N9smHoEtmZc2pa(hml+QT!W9++0074YPq{AuOCA6KzDphe0J;{J@DB(p
z6&3>k0LKLnxi0`r9smHoOdbFLx)zr(5C|t6E=>RczAjAw0J=Q@002md4aB%F0O@%B
z|Np);mvayZcn}8z0074YKe#UdARYh!z95(T5C~Nn2Lk{A#|7HAF8~}K003D?i3Q5H
zF92S?9G7Vk2qy~`0{{TW1)8@n08zJ_5ePE@6?p*w0LKMfwl4t75dZ+b%Mkzox_Ou5
z5(wi9c>w?b#|31yF96WD(Gv*J0Sndu0074YiL);NrnjsW2vGqgO#%P_#|162F91gv
z006#67ytlBi3J(5F929tNR16E5exvjO_w4U2+0*(0ssKV1wXMb0A&~e0KR1y006pN
zmKX>%6<h)U0LKOOurB~+7ytmiW*7hfx?Gn^7ziU3Tmk?9#|6ExF92;A006#i7ytme
zT$h3v2#OF*0ssKV1s<?306G`|0KPhx92p2z8BGEJ0LKNnuP*>L7ytlSNQnihuP*>z
zzBZSN83-o}Tmk?9#|3$>F92(|ycq~l0VQ7o0074Y9j-3`zZd`jzP}g%07!`i39c^y
zSXxMp4I2#%0J>k7G8+iV6>b6m0LKL{tuFxG7ytmi-53A>x^9;c90)ZPZUO)R#|7oA
zF96;c006$;7ytmeZkJjd2qP740ssKV1+A<v0Oc3}0KVlI006pfmy#R^iV$A{0074Y
z4y-Q#vKRmWzOt7r9SBt!UjhIC#|5dYF95C>003D?i3ORfF92S?u9un}2qz0}0ssKV
z1!=1<0OPmJ9SD610alkvCkP(_q?c+Z2yhEk1polY1yG?c0EM^ZCkP+`3sMCD0LKL-
zpf3PKw-6}^<pC@{1polY1vQ>80P;5g0KW1!002md1tFd<09aZ`jSVyX3jj!sJ()ZV
z0J=VxAu9+y3snUG0LKN)oi6~3muV{q><do?0074Y)toN?9+zM&2u%x61polY1$~?^
z04kSYEC?Ex(kuuo3s40B0LKL_oG$=4mk=ok8@D4Z2m}EcRs{e6G3`i+1^k*X07#2O
zBnZa^^_njLs+T}62r>&)1polY1-P0o0B)CkE(p*IRRsV5#|7P)F93U&ST6{23sMCD
z0LKMjnJ)lRm(?!_<_=W_002#i?8gOcm@fc*m&q^)=L=N@0074YdY3N%f0w&42vQDJ
z1poj|iT%d~8ka8sf|naI2tff>mt8Ul8v&Y^c`^uG8de1W07!`iP?j$MNQ*=y2t_gN
z#|1`~F94jE?=lEJ0alkTGYA&}o|i^52x1FX1polY1%j0?0HK$&GYC>2R|Nn7NQng?
zl`jBDi$o*{MKSG6iDV=o#|06UF95lhBs2(B3s(gI0LKNflrI3mw}dnZDFy*dx2`$}
zHUS2+PXGYEvzO{S2)731PXGYE<(H>C2xA83PXGYE=9dsX2u%<^1polY1<i{u0OL0R
z0KVgwZ9NE889oI70LKMri!T7*Hvj-xNQnhli!T6PzTcPIJqVB(J_P^(#|6KNF96v$
z003V|i3P8UF95#TmxDeCtP5EM0074Y`-v|A*Oxj!2tf;31polY1+R%O0M?g=KL|zv
z&X>(U2xJRc1polY1^I|C0N9s2KnN)dS_J?A#|6cRF97J5a6kww3t9yL0LKN6h%W%@
zm#07oH49n=0074YPKYl6`IqNF2=4(Qm)AiElncjC001%VMKR>Q$Cq|O2!H|Bm-j*l
zivc{BgF^^80oIqjLkK4UAD87r2ucfC1polY1^a|A0Ns~5L<rprIRgLy#|0CFF8~d<
zB1H%W3;~^&DNzVn0Svc;Q3&P*0dkjWR|t>+;kWx&2owPoM*{!=#|0a3F91#-006#D
z9{>QlN0%g72rm^!0{{TW1=nvc0979V0KQcp006p2muOiCGXag4oLLAd2yjgR05R>p
zaF@+l2rdENm;6}>n*scnm|6&j3r7P00LKLjZZ7~+w<=o*JqHPUP5=N%i3Pr9moZTY
z7Pp3C2wnvN$Cu@22vh<6mpW(&KNVF20074Yv0yI%n;!rGzMCHa0J>GTj%WyZ0RhRE
z2x<sf0sWU^Y6vw3!A<}GzQLEBY6wXJ|4x_aY6yB#P)+~<MKR<t?MR6QzFjW>>HGHo
z|7H&#ga7~l006$nP5=P9JpcdzOo`#?fc*deNR17$ybA!w1({th096A30LKN3T`vHX
z9{>Qplpg>9x>c97YX~L=_f7x+zW0~lYX~a{PEG&-G26aQmlJFVQ2}e0Wo!t+3snOE
z0LKLcTrU8bw=r!9A}b1PN&o;zi3QRyFPBh*1{Ak(m<T=s0ehFQnh0+Je3u}b2wnzK
zN&o=9QkR6A2x10eN&o=9Vz=y@2;>3_zytsQ#|2s{F8~Oa+nxwj3%~>b0LKOTDK7vE
zmph*bSQWqo0074Ymnkm*`Y`|izWOl$0J^}JsGkUhO27mF0LKOWC@%mLG5`R+6EXk*
zy1)bg07!`q<R~uy>AL>^{|_I400000z6de^0J^{g008Nv{{R1G4<Cd800000z6&w{
z0J^}Jcc2I-3%~>b0LKN4C@%mJx2d2ACjtw!1ONcX1zILA0IZkWq6k$Bv;+VE#|8Q&
zF95HXJEI6#6|@8Z0LKNFB`*M;F#rI*o-qIby0n+5qX>mcv;+VE#|8Z)F95eO006$X
zF#rI%v;+VENQn*PBrgEzy8i$F4<CR40002KtT6xpy0ioU0O_Ru|NmwWAA|q^0002K
zuQ31sy0n*fqzESqv;+VE#|4ZeF95Z-siX)T1Qn<R0074YWgssALNNdUzCtko0J^A`
z+NcO?ES&@Z0LKO8A1?q8F#rI*4>14$NQni}A1?q{T1brz<9-VONR2&d!3zMootJT`
z2;vo}1ONcX1=k)g0535B0KP9V006qEmuRX8B@3to0074YogOa$J(rlO2)zra1ONcX
z1=Srd06>>us|Z~Ss0082#|3R2F91-N$g2p23#bGD0LKO094`P<msqR_B^9Uy0074Y
ztsE}^TQL9tzFRQ>0J^A`lB@`j5S;`70LKLg94`O@F#rI*0+%wa2vr)L1ONcX1*01;
z0RAul09i<h1(O>u0A9ZSFqfgN2ulm71ONcX1wI=u09m*9tq2(c3ylN-0LKLy87~0%
zmm9DMRSS&-0074Yx)?72`<HvL2v`-31ONcX1y>j^0O~IQ0KV!k006p;m*}tvg-VSC
z0074Yy%#S4126ypz5_4-0J@C?002md4Wt(@0O`8^|NjpkfB*mh0KWJy006p;1ONc(
zr2hZ^W)B~P000000KWS#006p;m$$J9Cku@P0074YOcyTz0k`R~2qyvyfdl{m#|0V{
zF94jE8?*>j3xNax0LKNo6fXdumwU7bSQUW;0074YR}?P*jxPWJzK$;d0J?#f=(Grh
zN`V9b0LKNr6E6U#F8~0(r!N2ix`6}$07!`qq!TXy>AL>^{|_I400000zML-r0J?z$
z008Nv{{R1G4<Cd800000zMn4u0J?#fx3vf-3xNax0LKMP6E6Uzx9PPA90V101ONcX
z1tkwJ05UHC0KPIW006pmmm0bVYb<I60074Yr4BCu|1JOkzW*)&07!`ikq$2aSXxMp
z4Wn%f07#8Jp|A@8x@woPx(MPGbp!wa#|4)SF907e006!pF8~0#b(g5S2qg=31ONcX
z1zim<04<l;y9m7tbp!wa#|4!PF90x?pu7lO3v~nl0LKL_3@-pcmk7NGg$s2A0074Y
zoeM7jLzkGn2qhJD1ONcX1#Js2081|b0KQ8v006pmm(slmkPvDF0074Y$O<n2@-6@X
zzVerHz6ezrY6JiP#|2{wF97Z?003D?i3L*%F92S??w8=c2uurg1ONcX1^x*y07<tu
zzX*&B6*~j~0LKMq{Vo8jEdT(%t1SQkx;vNH%Lr2oI|Kj##|8iVE&#2UHp~bo3p)e=
z0LKN*{4M~rmuSogr3*U*0074Y9sDi;!<QJ%2rdgd1ONcX1?2lK0J67H%?N@56;lKN
z0LKMa`7QuWE&u?&O)dZcx>J|v&j?csQv?72#|7{BE&xxLM$iZ+3sVFD0LKNr_$~le
zmw3<!r3+I80074Y4frkqWtS+?2rdg#1ONcX1=RO0093bO(FoQD0cN*2*$65E0Xvst
z+Xy`YR+pjM2q_3TPXGWh+rBxs(%T4G1Pjar0074Y|L86NhqpH12!{a+%me@c#|3TY
zE&!F6$l(ac3(N!n0LKNI<}Lu3mz&}UA{ER80074YZRRcjTrvOvzFaZ@0J_YV(&7ke
z3(N!n0LKOH<t_kZmq+6WNIuO30074Yo8>M5t}*}szOFI=0J_Zt002md1%l-+0O=|J
z|Nlsh1(8Gx05R>y1$N~w0L=sd0LKMr<t_lSm(Ak{V++j$0074Y`s6MEu$MdJ2ucgh
z1ONcX1*POJ0J@iu<OpaB%>)1d#|1LvE&$1w1LX)I3(N!n0LKLe<SqbWmoMcAIt$DM
z0074Y!Q(Cff0uOS2uKUf1ONcX1#aUm0Jyiu<p}%(3z-A}0LKMq-!1@mx6bJZj{ysr
z1ONcX1<l<q0EL%m>j=mTnFIg;#|8S_E&zy^JM0J|6`2G80LKN*+%5o2FaQ9)OfUcd
zx|x@9><DTLnFIg;#|2K@E&x=Q=j;ef6`BM90LKOQ+b#f}FaQ9)o-hCax|)|r?FbwT
zngjp<#|6gQE&!*OZ|w*x3z`G~0LKN8+b#g0m!<6pDhrwf0074YTiY%GwwK!N2yF|Q
z1ONcX1>@Q-08*Dy?g%;ynFIg;#|56+E&y+rm+lBi3z`G~0LKMO+AaX4x9aW)HU$vn
z0RRBU1q#(J0A>^b0KR6IF!Tr;0p+($^aun43-JK}0LKN|(Jlaqw_EoJ0|6EF0RRBU
z1<cSc0NxY;0KVQ7006r6mvi_Cz!mWU0074Y_s=c>hZFz+zK0Y50J`y)Ncjj^74-oC
z0LKN1&n^JV6aWCe%M<_ry7iZ~`3NT!@&Nz<#|2H#E&!Dj006#~6aWCa@|W`Y2r~=v
z0RRBU1qIJ80Hc>M`Uo``^8o+=#|6dCE&#9;002yh>%OoQ006r4mwoyOATsj-0074Y
zjm|Cru@nFROo{8hu@nFRy7K`507!`qan3FP>C628{|_HP0KSP7006r2m(ThLXAtE9
z0074Y^vx~+RuljLzE+n;`v_?Z^#K3?#|3-ME&$n=y88&E3-bX00LKN)%q{@7w`u$c
z%m)<)1polY1@*)(06jGT0KPpn006oMmr4W)Y6}Jh0074YcEm0KJGZ|C2_^_EGXVeq
z#|1^aE&y^5006#n4*&p2i3KsdE&y0sNR16c8w&tPjXh=~3jn$^mk|vK`V~O|0074Y
zAG<C9kq-a>zL5_A0J=e!APxx~6+r<20LKOGx-I~c4*&qZk`Dj?x<Qvx4hcgGK>+{&
z#|4$TE&!mHpbiN=3qb(@0LKMQx-I~tm*@@&TNOb80074Y=eaHbr4IlAzNHTU0J=e!
zSPuzR3qb(@0LKM=xh?>+m$(lJ^$S4(0074YeYh?FwwJgN2}KYy0RRBU1ueKP0AmjT
z0KQ|F4G{@d88ZO@0LKN$w=Mu*4*&pJNQniww=Mu)zF(Jn5eZHhGXVeq#|2@xE&y2%
z003V|i3L%&E&#q+m*Ei!R|`P_0074Y^0qDjythLV2|58KJ^}y$#|5pmE&$dS006$$
z7XScAi3ORpE&y0sNR17v6bk^lK9`~s3CR^o0ssKV1;4Z|0Pz<90KV}T006p5mw*%r
zH5Ez%0074Yb+j%3@)rOAzVa6U0J=(-$`lEL5IzC`0LKO1vn~M17XSdh$(LLe2~`<B
z0ssKV1$nbB0L2#o09i<h1!=P`0A9Yum&z3hCksjf0074YIkPSR_qY5N35`h=rU3u|
z#|5EjE&!Pl006$35&!_Yrk9>y2_F@v0RRBU1#M|A0H6{80KT9S006qCm(pJeI18o$
z0074YAZacDp0^cX36ugAhyefq#|43AE&vS@006!X5&!_Yh?l-%2_F@R0RRBU1x;oy
z02C4c0KOCw006p(m-1o>I17jY0074Y0A?-#61O#D36ugAmjM6(#|2?yE&x*!006#I
z5&!_YmzUmU2_F@g0RRBU1ubJP09q0N0KQrh006p|ml9?PI185n0074Y;9@QSShrPX
z36lp4!~p;R#|1)LE&x)O^KA)&3&a5c0LKNjS}p)omw;{wy$i$v0074Y?O84WW0y#8
z2|WwM0RRBU1*KUo0A`nvZwZ|o!~p;R#|8aZE&xIk006#169526i3RCcE&#g3mq~C5
zHyy+Q0074Ys#q=nL=yl2NQniTSS|n$glPT$|Gq?*op1>u0aBO0a0w&}!~p;R#|1}N
zE&xfl?r;h91Qp@|0074YM^i2U2^0VTz6lfn0J`Fr_;m>#72*K^0LKLtQ!W4y6aWCe
z5EK9ay5g5Ab_wna;sF2v#|08nE&vX<BX<dp1q-zS0074YcuXz;^_RVT31|zo0RRBU
z1qDnl0QZ+LeF;4kwE+MC#|6DhE&$~c006$_5&!_YwU>f@2{t#i0RRBU1#(L+0Ot|_
z07!`qV@oao4}|*u|Np+{5&!_YwE+MC=?MS-|7H&#L;wK3^%4L8y0w?@eF-NEwE+MC
z#|0EiE&%MeA$|#Z1p!x=oP!Bz3)2Ar0LKL-MJ@ohmk@*rmj+Wy006#Im#u^eHUU+a
z<b(-V71IF#0LKOILoNWW6953dt`h(Ny3?0Xg$X+g(*Xbg#|585E&#BXmW2s11}aMc
z0KO`h&4mdl3)2Ar0LKM2LoNWumWByw0acf3h6yYS(*Xbg#|4x^E&#Q+p@s=k2Ni4q
z0074YH#;r>aS;FjzHt!%0J?0K2#yII6>I?j0LKLdJ1zin5dZ+bauEOkx@?y^jtNsD
zX#oHL#|51_E&w1A002cX<V7*|z910*07!`qgE}q%x@niMjtM3UX#oHL#|2e7E&wi<
z;EoAR7is|j0LKOWIW7P@5dZ)&?Y=t^006pbmqd>VAq#2&0074Y#W^kjNtbYs2}2lb
z0RRBU1$Q|v09FwI07Wt6zE%+c0J>|J%Z~|?3u^%Y0LKN*I4%HWmuQd)B^7J|0074Y
zoj5K4a}fXlzH<=(0J?0KqL2wA6>I?j0LKMoI4%Hm5dZ+bbP)gmx@?!)kO@H*Z2<rP
z#{~~KE&ziO006#&5dZ+XZI>{S2^|$}0RRBU1>H9;0E7_$0KS9~006pemtv6#BNc4{
z0074Yr8h1Bg%JP%zJ(D00J?3Lnvn@36>R|k0LKMwH!c8%5dZ+bh7kY&x^0)zkqIsf
zZ2<rP#|0`kE&z*{3z7*z3v2-Z0LKO1HZA~pw^x!0UjZyx0RRBU1#~qo0R0dE0KWYY
z002md1z<HU09aZ`jSX^?3IIrrJ)qkP0J>SX;*$xd0T5XM0074Yp))Q3><|C|zU-Ho
zl?hcDSpfh5#|1nyE&%8d003D?i3Kb(E&yJ>=$8kU2~HSU0RRBU1=TVx0OAk;0AEOn
z1;sKh0KVdvZI%g88Cd}U0LKMwGA;nz5C8x{SxAWmTrw^IzTCIgmI?9&EJOhS0LKMl
zEG_^v5C8zaG!Os)NQnhbEG__8T1brzV2cU>NR2(rlL`R3M7Q*u2|oc9RsjG2#|8H)
zE&x>!006#K5C8zWR+mVg2^|$y0RRBU1;r{Z09FtH0KQfb006pHmwKKFLkm^`0074Y
zZ7MDRWS7XE2}KK50RRBU1sy6b0B5%tp9xn15JUk00LKNpC@ugV5C8za9+!Wh2_6+z
z0RRBU1&t^!0BH~a0KRDu006pHm$IM<BMVjm0074YRVXe1cbDLx2^|Yo0RRBU1uZBp
z0D6}Qp$SVBRsjG2#|76XE&zNG006#x5C8zWR+ng@2_*|w0RRBU1)V1@0E(BGp$VM}
zRsjG2#|8B!E&z>}K%xmN6;=TN0LKNtCN2Pv5C8zaj}QO=x>lEfq6t+QL;(N*#|2d;
zE&vn|003D?i3LR_E&yJ>6qoLz2~HS90RRBU1@0v-01glU0AEOn1>Pkt0KN{FRHF${
z8AJg90LKM}B`yF65C8x{SxAWmbtNtUz6h7SqX}^fRsjG2#|8W(E&!CbIiv}81`r|v
z0074Ydmb(T#|{7hzQ>opstFqbBDdSB3GxILFaZDn#|2#(E&%5a006${4gdhUFqh)5
z2}27o0RRBU1qB%{0QHwJun9d2FaZDn#|6C@E&%wKc(4gY6)*t+0LKMw7%l+%4gdhY
z`VIg9x-gg0un9X0FaZDn#|0f2E&vFZ7_kW@3oroy0LKOG7cKw`mq@V*B^59M0074Y
zy%#P36Au6Yz7r1s0J<=jg0Tr~5F!Bp0LKM97cKy{4gdhYwwL~~3GE9o0RRBU1wR%p
z01>zLvI$BD5CsAN0LKOD4lV$t6#xLfrI%2;2^#?gw`{rz+yV;_0ssKV1>y@X0LZsf
zy$N6e6%Yad0LKM)3N8S~6#xLf#uWenx)7Jhz6rJ!6aoMM#|5kjE&v@C006!n761Uc
z6ql2~2{IKC0ssKV1!V~?0Mr!#0KU`}006oXm)gGxPZbmb0074Y0SPVu3KjqWz6urq
z0J;>HKEMed3lRbU0LKN*2rdBPmuSEVH5d{C0074YjR-CP?-c+5Oo{8h?-c+5x)PVQ
zzzHri5&{4K#|1|SE&%Wq002yh>%Q<6006oY0ssI=i47|VE&%C!{Qv)E4<AGT0KUi-
z006oWmlweaix33@0074YhzBkJlobE~zLb};!3p3C6aoMM#|4-ME&v#pox%x93lahV
z0LKMG1}*^gxAMXX!T~H70ssKV1$G230CpAt0KRq>002md1!4p)09aZ`jSX~T3IIrr
zJt%An0J;~q;KT|30~IF%0074YOaLwb02crNz5o{h0J<la^T-L%6(a%w0LKMq|1AKQ
z761Ugmlgm3x+9m^$q7IUBLV;b#|0AqEdZI9BgzSZ6(<4!0LKM}{w)CQ761Ug?G^w4
zx+j;p$_XSFBmw{c#|2IPEdZq!001%DzNHob0J<cX^vVe#3nT&n0LKLt{w)Brmmtdt
zH5DZS0074Y)%`61z!m@izQ7g$0J<fYYRd^D6(s@y0LKNL{Vf2&761Ug!4?1jx+RyQ
z%LyY3B?15d#|35mEdbA#(8~!O3nc;o0LKMA{Vf2}m-x#Gju00D0074Yr2H)aW)=Vd
zzGjz}%n3*fCjtNf#|2FMEdcwM>C6e!3nc;o0LKMx`z-+2x5v#1?jJ1K0002T1$Emk
z01OQP0KN<j002md1!3DQ09aZ`jSX`;3IIrrJ+Xrd0J_<?;q3`50Tt;00074Y9oj7b
zEDZnvzAOy@0J`ayBJT+^3+Vs=0LKN@*)0G%mtgM+Jqzgo0074YjoB>#J(sBO2{jez
z0002T1x?v406z@?0KPvB006q_m-6okTMOv`0074Y-PkPvO_x~k38f3^0002T1ufVu
z08p0*@d+~(=>Px##|7`#EdWvt006#H4FCYT>6cLP2`d%p0002T1+UjF08<SB0KQWV
z006q_myq!ZMG)Bl0074YRo5*5`wRd8zWbNu@d;HK*#H0l#|7}#Edcin003D?i3Q-+
zEdXA=_m@-h2~HT<0002T1&P)z0Pzd}0AEOn1$ovj0KV~;x$+568QA~;0LKL()-3?*
z3;+N@SxAWm5Y{aKzU!AF^9foD=>Px##|5s{EdW@Dhw}-Chw}=Dhw}@Ehw}`Fhw}}G
zhw~1Hhw~4Ihw~7Jhw~AKhw~D*hw~GTfR}xB6nG3`W??chUtcz1Ibt@KA#@a^3}R+s
zGBICYHeoqpHkZ+K6s(v2G86`f`*jqD`*jtE`*jw#`*j!XH7Tkx6aWAK000Pd6aWAK
z001y_6aWAK003Zg6aWAK004k=6aWAK001zD5v&)75v&-85v&=95v&@A5v&`B5v&}C
z5v(1D5v(4E5v(7F5v(AG5v(DH5v(GI5v(JJ5v(MK5v(PL5v(SM5v(VN5v(YO5v(bP
z5v(eQ5v(hR5v(kS5v(nT5v(qU5v(tV5v(wW5v(zX5v($Y5v((Z5v(+a5v(<b5v(?c
z5v(_d5v(|e5v)0f5v)3g5v)6h5v)9i5v)Cj5v)Fk5v)Il5v)Lm5v)On5v)Ro5v)Up
z5v)Xq5v)ar5v)ds5v)gt5v)ju5v)mv5v)pw5v)sx5v)vy5v)zO5v)#$Vt;Q4chC+1
zv;)dQ5C9R22s!`&0Fnjo5;p*h2nhqgy*B^=0000}#c2cq5swH4wGIFY0|?1C00000
z08=Its{j#@2nMwd00{#C$u|H10000YThZD85s?T6wGIGNM0>UX5t0Z7wGIFY1Ng`{
z0000003s6GVgM172nMwd03Ra!&sqQxln4g34gezMgWCWRl?Vp44gezZ?Kc1smX|mW
zPa*-Ams<}{EEkvv%M&*M2?J(N6aWAK002|w->(1>nU|gqPb?Rj2o4lC00{$OPZR(E
z0000}4k5q*5u2Cc4^I_YoCppSHvkC(TTc`K00000Q>U-M01=%C4iq;42?JP96aWAK
z002{3>Tv)Oo(K*UHvkC(RZkQE00000Q@}Om01=-E4iq;42?J736aWAK002{~VRrx#
zpqEb&PZcep2r>mX00{#~PZR(E0000}Zak_05uykc1vdZ*13pg_0000008?lIvj7pJ
zmyr-p6<MSR6$Li{2?I7y6aWAK002|^qJaPrr3jV>HvkC(F;5f#00000Q?_V001>7L
zi3K+R2?Hok6aWAK002|`NrM0prwEn@HvkC(98VMg00000Q_kF301>E{{}4|VEvX2W
z2R8r-0}@XZ0000008?hA$N&+l2$lyo00{#PPZR(E0000}6LwDk5v!L$5l<CatO%9|
zHvkC(15XqH00000Q(SjX01>SSmIpTg2?PF46aWAK002{VoreGst_YR~HvkC(`A!r7
z00000QyZik01>YUi3K+R2?Ov>6aWAK002`EphEx=u$Qk9PZce(2$lyo00{%*P80wD
z0000}l=Kb&5wZxD2R8r-1Kv&)0000008?6>X8;kim+=u#6<M?hi3K+R2?Nkh6aWAK
z002{k7wiBLwFuA$HvkC(#!eIf00000Q#*Gr01>tb&<8gF2?M}R6aWAK002|f+N1yx
zw+PS&HvkC(x=s`T00000Q@pnK01>#CUlLChEx8EL2R8r-1FlXK0000008@)M)Bq8>
z2+#*N00{%AP80wD0000}!{&Sd5xbY65>FLbya><-HvkC(oK6$~00000Q}0Pv01>?i
z&<8gF2?Lf+6aWAK002|9P1gVsz6j6<HvkC(kWLf;00000Q>S}&01>|k&<8gF2?L5w
z6aWAK002{4_k;iuz?TmbPZce}2+#*N00{$rP80wD0000}c|JS<5yA-a1UCQ)19wgo
z0000008>VwCjb${mr)Z>6)nUF@&q>k2?J?P6aWAK002|Y-MIh}#R&2QHvkC(VonqQ
z00000Q_U#@0TITRlM_!Y0mqlX6HgU?$OwrAHvkC(Ku#0@00000Q-5ah01?Ru@&q>k
z2?H}u6aWAK002{y9De{2$_N7(HvkC(C{7dr00000Q{^#@01?Xwi3K+R2?H2T6aWAK
z002{z4p;yY%n0%XHvkC(3r-XO00000Q=~o)0TImz0~j{|2?GF56aWAK002`Q{af4s
z5zYuQ1vdZ*1Nco80000008<Z+F#r+Imv9tM6@SnOG6gpP2?OX&6aWAK002|hYkvR{
z(FigHHvkC(;7t?&00000Q<ljL0TI#&G6gpP2?N+o6aWAK002{5&vyV3(+DyJHvkC(
z&`lHo00000Qzm=401?y(G6gpP2?NMY6aWAK002|s66XLB)d*|_HvkC(xJ?uQ00000
zQyi7FMgS4k2#EzZ00{%AO%wnC0000}Z@kk05!aU>6;Bm^*a(RQHvkC(h)om#00000
zQ}&dE01?>;PX#vs2?Kad6aWAK002`F2XX)r+6ajSHvkC(XiXFV00000Q$dGn01?{=
zi3K+R2?JP76aWAK002{#?gjx7+z5#UHvkC(NKF&~00000Q`U@501@2?i3K+R2?IDy
z6aWAK002`Q5g8o-5#9)i1vdZ*11L=t0000008<jmn*b5tm(UeY6@TCei3K+R2?GdC
z6aWAK002|CQ!W4z;RsI!HvkC(_)HW400000Q{z+*01@H{c?UND2?OX%6aWAK002{l
zF(d#H;|O^NHvkC(*h~}v00000Q@caG01@N}c?UND2?NMX6aWAK002|}*x&#W<p_BP
zHvkC(xJ(oP00000Qyf=As{j$^2zduL00{%AOcVeB0000}s#aG35$BhH7Ecvf=m>cS
zHvkC(h)fg!00000Q)ER)01@d3c?UND2?Kac6aWAK002`8KMMg7>IiuUHvkC(XiO9U
z00000Q?XX;01@j5c?UND2?JP66aWAK002|!oH+my?3eBqPb?Sh2wodE00{#)OcVeB
z0000}B<@oH5$=~U7f%&g?+9KSHvkC(7)%rZ00000Q;>k>01@yA0~j{|2?GdB6aWAK
z002{gr$zt~@dyJLHvkC(_)8Q300000Q;{3H01@&C0~j{|2?OX$6aWAK002{AX#@cg
z^Ov6&PZcfn2m=^500{%gOB4VA0000}<z-?35%mZI7&ib31Gq~R0000008@K8kpL0)
zm*E#r6<PNP0~j{|2?Ll*6aWAK002|ttqcJX_y}wTHvkC(h)Waz00000Q#Rnt01^2J
zYy~#}2?Kab6aWAK002`6zBvF9`Uq?VHvkC(XiF3T00000Q|Rqj01^9_PZ&=XfBXn+
z1vdZ*14v600000008?<b5dabW2y6v600{#)OB4VA0000}9S9@<5&j5l1vdZ*11L)r
z0000008?WL!2l8e2y6v600{#aOB4VA0000}aNNcK5daBn1vdZ*0|-kL0000008=;h
z82}Lh32X&700{&5N)!M90000}90Vdo01*NSYy~#}2?OX#6aWAK002`dUO)g51DF38
zPZcc$2~Pz#00{%gN)!M90000}-MFOy5d{fP1vdZ*1Gq{Q0000008`9t<p2=|mq8g%
z6<G%fPX#vs2?Ll)6aWAK002|WK#u?s2nkOGHvkC(h)NUy00000Q>L4_01*iZPX#vs
z2?Kaa6aWAK002{x4rKrl3JFgIHvkC(Xi5|S00000Q~XpU01*q9uNhAjEer`>8#e$6
z14v2~0000008_bC2LKTb39|<`00{#yN)!M90000}g*+et5e}E}8BY}#4+&lyHvkC(
z2uc(H00000Q*>;{01*(E9U4z87ZC}K9yb691K>#%0000008<s4?f?-ImtY!C6)h79
zTn9G*2?M}M6aWAK002`_3*rC~6bW1hHvkC(ut^jE00000Q@lpV01*|Jq8d*u0T!3f
z8c!817YPy{HvkC(fJqbp00000QvwN(01+4o5+64J2?KCR6aWAK002|y4q^Zi8J7_o
zPZdKN2@)SS00{$7NfZD80000}S7%WG5gQ2-A2$FA13*a>0000008`3ZJ^&FM2@)SS
z00{#yNfZD80000}S*4Z$5gnIl8&52M9tqhWHvkC(@JJK@00000Q$lyC01+Pv?FTmi
z2?Nkb6aWAK002{HUwHr#APKVvHvkC(ut*dD00000Q(EdE01+Vxvj;Z-2?LNw6aWAK
z002`9gjN6%A_<KiHvkC(a7YvY00000Q}m_%01+bzjUG1u2?J0_6aWAK005Df6dX?+
z0VJ0(98VQlB?;LdHvkC(5J(gN00000Q-Lx%01+k$*&jCm2?Ov)6aWAK002|jcY**B
zCkfdfHvkC(&_@&i00000Q-D&I01+q&*&jCm2?MZ46aWAK002`Q-NFD7DVLudPb?QI
z2@fGR00{$dM-%`60000}VQ-ND5i6JC98W9(ESC=*PZcdK2~r_900{#SM-%`60000}
z2|*D75iSW*AvXXC1Mo%^0000008?Q!1pyH+mr)&06#+1pXB|&07cmLfAvXXC1CT}(
z0000008@#OIsg$em#Q64ELk%N8X`9U2?J0@6aWAK002`3W&r>ZGzkm}HvkC(Fh&#r
z00000Q~U>?01-6_3<@^@2?G#D6aWAK002{2ZGZp~HVN+uHvkC(@I@2=00000Q;)Uo
z01-ErFCI@7EjS6%2{!-<1F%ID0000008>IQ9RLwI2^u0d00{$-MHB!40000}2=smc
z5jvM~9#1R*JC~jwPb>jEm&+bcECD^2_#RIz0X~-{A5Sa+KbKM;Pb>jImwz8mEdfE7
ztshS;0YaDBA5Sa+Lzf32Pb>jMmoy+xECEHAVIWT|0Y;aOAWtj-N0+-GPb>jQm*^l*
zECETE6(LV70ZNxbAx|s;OP6mUPb?Qq3Ew3*00{%2JQM%`0000}2hvdh5lxq}Ax|s;
zPM6;yPZcds2?n(e00{%|$Tt800000}Du*io5l{&RwGIFY1L(*%0000008<gdTmTVK
zmmwlg6<JaV2DJ_V2?N;3Hvj+t002`NMnV7)QwavO4gd)Q(8xCc00000Q?Fq=01;FP
z2DJ_V2?NN;Hvj+t002{rR2KjdRS5>Q4gd)Qz{ocM00000Q}dn!01;M~k0MVMEmsK!
zwGIFY1F*<90000008`_azyJ|g2?n(e00{%A$Tt800000}6bQ8d5m}efB2N`rS_uZV
z4gd)Qn8-H(00000Q(V?P01;aW2DJ_V2?LPGHvj+t002|;iw*!0TnPrX4gd)Qh{!hp
z00000Q=AOe01;gY2DJ_V2?K!0Hvj+t002{EN-F>nUY9>3PZcd+2?n(e00{$d$Tt80
z0000}D%Rux5nu@hwGIFY18B%M0000008{R-v;Yxdmw_Ws6<J~l2DJ_V2?JQjHvj+t
z002{!q5}XCV+jVe4gd)QP{=m`00000Q<EBB01;#f2DJ_V2?I#THvj+t002|Ry`=yV
zWeEng4gd)QK*%=$00000QylmC01;-F?;}qYEoTV^wGIFY12D)p0000008<{#F#r*0
z2?n(e00{#q$Tt800000}9H_|v5own(Bu^DtY6%9l4gd)Q7|1sO00000Qvlo`01;~m
z2DJ_V2?G$wHvj+t002{k4>kZ1YzYRn4gd)Q2*@`800000Q#CM*01<5o2DJ_V2?GGg
zHvj+t002|`Te|=eZkL}VPZce12?n(e00{%|$2R}~0000}fX0Ua5pW3xwGIFY1L(&$
z0000008<SZc>ob{m*FH&6<Kl#2DJ_V2?N;2Hvj+t002{fDw_Zia|s5u4gd)Q(8o6b
z00000Q?uZt01<Qv2DJ_V2?NN-Hvj+t002|(qXhvGbqNNw4gd)Qz{fWL00000Q@I#a
z01<YVPbE(kEq4h9wGIFY1F*+80000008>O7CIAt52?n(e00{%A$2R}~0000}RXx)H
z5qX!9B~KMudI<)#4gd)Qn8!B&00000Q-n1F0TFu%2DJ_V2?LPFHvj+t002{t+DZTs
zd<h1%4gd)Qh{rbo00000Q<^mh01<r&2DJ_V2?Kz~Hvj+t002`3mM8!bewY6xPZceH
z2?n(e00{$d$2R}~0000}EGP5;5r7E>wGIFY18B!L0000008{NaG5`^Qmq8{^6<LA_
z2DJ_V2?JQiHvj+t002{*4xj)Lg9!$;4gd)QP{%g_00000Qw_(u01<=<2DJ_V2?I#S
zHvj+t002{loCp9Bg$V|=4gd)QK*u)#00000QwiJI01<|luO?3wEr$sPwGIFY12D%o
z0000008`41zyJ}52?n(e00{#q$2R}~0000}Iw5cX5s8=aCQlVviU|g_4gd)Q7{@mN
z00000Q;;A|01=A`2DJ_V2?G$vHvj+t002|REKUFsj0py{4gd)Q2*)=700000Q}HxV
z01=G|2DJ_V2?GGfHvj+t002{VZ=V1Wj+b92PZceX2?n(e00{%|#y0=}0000}jtjE@
z5s(Q6wGIFY1L(##0000008=7ENB|L$m!T(56<LxA2DJ_V2?N;1Hvj+t002{knI`}d
zlL-d34gd)Q(8f0a00000Q@ab001=c42DJ_V2?NN+Hvj+t002`nZPx%1l?ev54gd)Q
zz{WQK00000Q#L1W01=j#4=7I+Etd%fwGIFY1F*(70000008_RI^8gW;2?n(e00{%A
z#y0=}0000}3sgk_5t)}!C{Gnxnh6HA4gd)Qn8r5%00000Q+0Vb01=xB2DJ_V2?LPE
zHvj+t002`qKllI<oCyZC4gd)Qh{iVn00000Q?85y01=%D2DJ_V2?Kz}Hvj+t002{k
z^^pJ(o|nHUPZcen2?n(e00{$d#y0=}0000}BFseq5ugbMwGIFY18BxK0000008{2i
z4FD0LmjNkH6<MMQ2DJ_V2?JQhHvj+t002|<gqHvjqX`DJ4gd)QP{ua^00000Q$>~k
z01>1K2DJ_V2?I#RHvj+t002|go(TaFr3nVL4gd)QK*l!!00000Q?!by01>8_Zz)d|
zEvE?vwGIFY12D!n0000008`q&od6N22?n(e00{#q#y0=}0000}+%a_k5viB4DNhwy
zstE?Q4gd)Q7{)gM00000Qx6q<01>MR2DJ_V2?G$uHvj+t002`7etZBCtO*9S4gd)Q
z2*x)600000Q#%u(01>ST2DJ_V2?GGeHvj+t002`$CvN}|u9qJwPZce%2?n(e00{%|
z#Ww%|0000}f)=R&5wHmcwGIFY1L(y!0000008?zhF#r*<mtiVT6<M+g2DJ_V2?N;0
zHvj+t002`K7zzLpvk3;Z4gd)Q(8V_Z00000Q}N<?01>na2DJ_V2?NN*Hvj+t002|i
zoIL;$wFw5b4gd)Qz{NKJ00000Q{x5(0TH&B&niz9Ew>2<wGIFY1F*$60000008`!z
zy#Nuo2?n(e00{%A#Ww%|0000}z_<nh5xJKUD^C?!x(Nog4gd)Qn8h~$00000Qx2IT
z01>+h2DJ_V2?LPDHvj+t002|=BjErMya@)i4gd)Qh{ZPm00000Q-9=l01>?j2DJ_V
z2?Kz|Hvj+t002{t<aYoOzL$S1PZce{2?n(e00{$d#Ww%|0000}7jOUp5x@xswGIFY
z18BuJ0000008^B*vj7pnm%%Gf6<NXw2DJ_V2?JQgHvj+t002|HTLu9U!wCkp4gd)Q
zP{lU@00000Qy)M)01?Cq2DJ_V2?I#QHvj+t002|QA}#<C#R&$r4gd)QK*cuz00000
zQ>dZc01?KQFDy?LEyoE4wGIFY12Dxm0000008{elmH-jR2?n(e00{#q#Ww%|0000}
zpSK4A5y_WvEKe0#$_WOw4gd)Q7{xaL00000Q#M$801?Xx2DJ_V2?G$tHvj+t002{!
zLJ9y8%n1gy4gd)Q2*o!500000Qv%=M01?dz2DJ_V2?GGdHvj+t002|DBE$d@&X?aT
zPZcfC2?n(e00{%|#5Vu{0000}&#3_c5zq++wGIFY1L(vz0000008{-TA^;K5mmw`r
z6<N{=2DJ_V2?N-~Hvj+t002{X=YIeZ(+LK(4gd)Q(8M<Y00000Q{kCK01?y)2DJ_V
z2?NN)Hvj+t002|Lx0nDC)d>c*4gd)Qz{EEI00000Qvn8_01?)gk1bCXE!PPKwGIFY
z1F*z50000008_Jnc>odE2?n(e00{%A#5Vu{0000}nD77q5!si~El(9$+6e}=4gd)Q
zn8Y^#00000Q$r#;01?{>2DJ_V2?LPCHvj+t002`$`~?6J+zAG?4gd)Qh{QJl00000
zQ&Ob@01@2@2DJ_V2?Kz{Hvj+t002|3<*)z|-j_cvPZcfS2?n(e00{$d#5Vu{0000}
zlM2!R5#R|1wGIFY18BrI0000008=LY@&FOxmw_%%6<Oj52DJ_V2?JQfHvj+t002{H
zl?wn7;|T_}4gd)QP{cO?00000Q;uD301@N~2DJ_V2?I#PHvj+t002|;{_+43<p~D0
z4gd)QK*Toy00000QxQid01@Vw?=DXjE$0aawGIFY12Dul0000008_!;5djhC2?n(e
z00{#q#5Vu{0000}Rk~{c5$TsPFHaR&>Inw54gd)Q7{oUK00000Q>BRY01@j62DJ_V
z2?G$sHvj+t002`NMk@dj><I?74gd)Q2*fu400000Q^ocF01@p82DJ_V2?GGcHvj+t
z002{3OqKu<?w6l0PZcfi2?n(e00{%|!#4l`0000}i^Pop5%38HwGIFY1L(sy00000
z08`Eu009y4m*Fo@6<P8L2DJ_V2?N-}Hvj+t002{&%hCW5^9csE4gd)Q(8D(X00000
zQ&AQ;01@;F2DJ_V2?NN(Hvj+t002|c#l`>;^$7;G4gd)Qz{58H00000Q;7MK01@_=
zPcTmvE%ylqwGIFY1F*w40000008@081_2TH2?n(e00{%A!#4l`0000}Jzzxu5&4&q
zFi#a)`UwWL4gd)Qn8P;!00000Q<iNQ01^8M2DJ_V2?LPBHvj+t002`~Pdfk+{0RoN
z4gd)Qh{HDk00000Q%yR~01^EO2DJ_V2?Kz`Hvj+t002{b9WDS7{t1>bPZTZx2?n(e
z00{$d!#4l`0000}bo_$=5daDXwGIFY18BoH0000008@X2mjDp~mq9U46<Gob2DJ_V
z2?JQeHvj+t002{G91sB!0}2MU4gd)QP{TI>00000Q_W|)01*TV2DJ_V2?I#OHvj+t
z002`KBD??*1queW4gd)QK*Kix00000Q>@7i0TEZ1uQ5**Em#T$wGIFY12Drk00000
z08{mczW@<g3I??f00{#q!#4l`0000}xTkLb5n7k=F;5j)TM7oX4gd)Q7{fOJ00000
zQ~&Up01;dY2DJ_V2?G$rHvj+t002`p#a;jrT?z)Z4gd)Q2*Wo300000Qxu&f01;ja
z2DJ_V2?GGbHvj+t002|QYSaJ`UzcApPZcd-3I??f00{%|!Z!c_0000}-azyK5n&1j
zwGIFY1L(px0000008{Qj^Z*fJm!UFG6@Oz22DJ_V2?N-|Hvj+t001IlLi7L;WC{kg
z4gex@Li7L;WeNth4gey8Li7L;W(o$i4geyOLi7L;X9@<j4gezlSMdN5XbJ|k4gewv
zL-YU<X$l6l4gew<L-YU<Y6=Fm4gez2eenPhYYGOn4gexne(?YiYzhXo4gex<S%2vO
z5p4<vwGIFxh+p&o5pD_wwGIFxe{}Q!5pN0xwGIFxu$%D!5pW6ywGIFxH+J*@5pfCz
zwGIFxM|Si85poI!wGIFxS9bIO5pxO#wGIFx2nDAA5p)U$wGIFxHBrX^5p@a%wGIFx
zeNo2%5q1g&wGIFx7YL^S5qAm(wPOwdQ|r>_01<c!2DJ_V2?Nl=Hvj+t002|`;O77l
zc?t%#4gd)Q$ig=O00000Qx9*)01<i$2DJ_V2?M~wHvj+t001J&Z^r-;dkO}%4geyx
z;^zPnd<q7&4gezLa>oD>eU}$BPaYSF3Trqw00{%2hc^HK0000}sgRxk5sQ~hG*2iO
z;|i)cHvkC(Aci*p00000Q(?C+01@Pui!@It7d#8;I5z+Z1Av4#0000008>~GbN~@O
zm&!CxCl{3qC^<I(2?KzGHvj+t002`-Z^Qr*mX`=MPbUFJmoPO?C;^I>Sv5~60mqku
zHBTr3e3z>=PbUGum(w**C;<eQ{54M~0ZNx8Hcu!4ke5z2PbdL6mv=T#IsxF9vNlgB
z0l1gjHcux3vX=ulPbdNLmn=6=C>J*piaR#|2?M}uHvj+t002`(hWr2#eV1!DPbdMy
zmy|b8C;=Upy*E!v0gsmmI8Sf^?U#8tPbdL1m!ddNC;<eQ%Q#Oc0dAM{I8P`6Lzf*n
zPbUE;mqj^GCjn!ZZ8=XU0jHOiIZr477?-~}PbdM9m*_c9C;<qU5jsyO0f3h}I!`D8
z_?KfkPbdL>my9}3C;<YOwK`8I0eY9-I!`D8hL;CBPbdQZ9hWgXPbUF(msvYcC;{%5
zf;&$r0Wz1XJ5MM9c$d^WPbdMtm;F0WC;_CGCOl6l0m7G0JWnVAY?pXEPbdLJm!UjQ
zCl^R9|3EhY2?Ky2Hvj+t002`m*O>qjjF;a$PbU}ZEj>Xu00{%29yb600000}Z|A-M
z5iXY<Jx?bAjF&|{PbdM<mu@{zCjoGmmOW1=0nC@aJx?cDC@>sCHvkC(Kp8gx00000
zQy$7601@CYqd_+S2?GEaHvj+t002`r2gm>sfHBNLHvkC(Ko>Uv00000QwKyo01@0V
z^g%ZO2?Kx@Hvj+t002_}OrHP|IF~*?PbdNNmuEgtCjlInk3LT*0brN7K2InCxR>ER
zPbU|XIX^=;00{$N4mSV*0000}2xtTW5tNr7KTj1c{5d~EHvkC(zzsJ500000QyKsw
z01^E;fX5C12?HRvHvj+t002{p?kWHgWtU+;PZa@Xmv=u;C;`rwpg&I)0neAVKTjwD
zbC=&gPZa@lm-RnSC>Iz!Q$#la2?HPrHvj+t002|${R;sR8J9FbPqP=7P##4$00{#?
zt~USx0000}*(?J95!;vkKu;$DAeSaVPbU|^QG`V|00{$tt2Y1u0000}+YJc;5yzKg
zK~E<E<(G^>Pcs2vm%u?!CjoVr=s`~>0fLthLQfR|QHLf%PlqN#P=_W$QHLf%Qimo&
zQ->x(REH))Rfi@*R);1+SBEA-ScfJ;S%)S<T8Ab=TZbk>T!$t?U56$@UWX<^Uxy|_
zV236`VTUF{VuvO|V}~X}WQQg~Wrrq0W``z1XNM+2Xon_3X@@34YKJC5YlkL6Y=<U7
zZHFd8Zigm9Z-*vAaEB&BaknNya=i%+_h&i)0018}&jSGw_lNLHa)<CsbBFLtbcgUu
zb%*dvc8BmwcZcvxc!%&yd57>zdWY~!dx!8#e24H$eTVQ%ez)*Te-6EW^=AgP4gdpX
z4QK!#?%zKE5%p(600000p8@%p4*?PKX9l$n00V*zXaFN(01@(M<p2Nx0G|QCT2TNI
zv}d}g4gdq;4QK!(VgM1eX9l$n03z<+KL8Q6XS%2k08DrQ5w>RrwGIGB(hX<;A~-IK
z01?4w;SFd2B2+SS01?7sXU3=w002HB&Y)rd5$|UPwGIFP022sI#0~(T0bn?m01<;{
z2DJ_V111h=03&z+5rbz2wGIGGZ2%F4XH2~g03s!J*#HrTXV3%y08Daaa{v*DX9l$n
z006@QQ8Zlu5pS2@34kFHZ)XOz4gl%kE&yE3bUOeMZ<i4YfD{3DmnI5;ApwAwND6>&
ze`9Al00003C=X}=BX|H2V`m1n4gey-F24W~WM?`60001=0sbr701-)N&;$Sg18EOv
z03&Yz5lLqTwGIFu2Z~|<5lLt1whjP3B8kIC01;YeWdQ&HOmYAbTW2}|00000p8+5X
zLjV!`W;y@>00D*%XaFO001^9Uh7V`}mo^K44lC<sIsgCw0mTn!03v!>_y7^?W{v{@
z03taZZ2%GMW{v{@06rp~$qWGz@Rx53fFOV6W;y@>00RIJXaFN~F#r+eX3zux03&lU
z01@S8&;$Sg0G|OkYqJ0m&1O0P0009k5NH4+WpZ)=5zS@>wGIFuk-U5W5zS@>wGIFP
zp8@_Fv;YyfW;y@>00DLoXaFCPynFx=xMl{m4geq3u95%|xMl{m4ggGP01>%nx~P8+
z08DNG5xHi%s15)=OmqMd!e$1w4gdfEp8-^{WB?JcW;y@>00X%YXaFN+a&iC>ux19e
z4gen(odW?8ux19e4geqV5q1C(ux19e4ggGYWpe-#v1U2|0001=0peLG01=92IsgCw
z0|F6f03%@l5sGGa&<+42VgM0}W_N$k4ggGTa&G_;i)IG34gfw(X#f$AX1b^j08DNG
z5szlNs15)C0G|Q;X2<{$f@V4Z0003g5oiD-VE_?=W_QpI03%`m5rSrS&<+3qp8=tK
z`v4J2WI_M{00SfxXaFO401--L2DJ_VivhB@ga8rzWVQiVYT5u10A*RV1pt2l0000Y
zfj&k65lduE0000YAfo*M5lduE0000Ym(%tD5lm#t0000-Q4(kXA~-IK01;he;SFd2
zN3;=W03vN9hyW2_WI)6Y06roBFlqo1U}QkV4gdfEN7NB$03vN9hyW2}WHJB%06roB
zFlqo1WMnb`00000NAwYB03v^FB!~bJZDcY4002HB05EC*5p85L0000007n!OXaFK@
zB!~bJc4RUD002HB05EC*5q4xU0000007o<uXaFK@B!~bJjASwZ002HB05EC*5sYLq
z0000006rpZB!~bJk7TQ&4gfwP05EC*5szf6q7DE6002jy6KDV;I4*yS01=dA;SFd2
zM}iV)03vN9hyW3nWKqNp06roBFlqo1mt;}I4gdfEN1_sF03vN9hyW3oWI)6Y06roB
zFlqo1m}Ef24gdfEN5T?l03vN9hyW3sWaR(=06roBFlqo1oMhzy00000N8%D_03vN9
zhyW3#WHJB%06roBFlv7Q5v61@0000007n87XaFK@B!~bJu4Lr^002HB05EC*5w2wA
z0000007oJdXaFK@B!~bJw`4K^002HB05EC*5w~PA0000007pU-XaFK@B!~bJz+~kB
z002HB05EC*5x`{S0000007qgIXaFK@B!~bJ$z(DB002HB05E@Q01?S#G5`Po002jV
z6KDV;Z6t^Q5z=Jk0000!A^<RI01?t;<p2Nx002HBZ6t^Q5!GZxq7DE)A^<RI01?$>
zMWPM>0000-`x9sYOm<;x01@6~LI3~&Om+Yf-(&{04gfw(Z*FA(5$I$(0000Y{!|D6
z5$I$(0000YFH3)301@eA`PL2qOmby(01@hBIsgCwJ|fPbVgM2CWCpbk00000K1_CD
zYyc7UWI_M{08DgkZ~zhZWCpbk06rqlpke?K`D6yQ4gdfE0G|PUpq>B`xnc&j4gdp<
z6lee+Z6t^Q5xHVI0000Yfj&k65xQc^0000YAfo*M5xRe3$^ZZWBA3(l01>-l$^ZZW
zK1^?JWdIS-Vmbf-03rcZ2mle#Vmbf-03t6-UH}o%V)@n%0001=0ir$=01=vE2DJ_V
z1NIbX03U56hyW3qVhZRE03v}tMgS3;Vi)KR03slw{Qwc0Vi)KR03w&u_5cx_V#)vj
z06t7_Ze=O}5wT)A0000Y0aXY95wT)A0000YFH2qk5wc?W)(!vw0Hy_Oq&ol+W|yxJ
zfE|BkV$cKt0L|ep09>X5XaEsrVmbf-0MOzt09*udqyQ0TVmbf-0OI2=09<t94gnEo
zV$cKt0KDWb0M-IxVPpUiequtnkg@;~XkvB{HvqroE&xJ26_5ZCX=3FYXaIobE&xIR
zm4pBhYGO3c4gdw`E&yJT2LJ%h6=(oL>w6hL01<y;2DJ_VhvzN;C6EUI07A~7VgM0;
zmt+xuHUWT_n-PE@0f3jh5r829ZkOH>fMo$}mqZePUje_Dq!NI+0ok`+6M$F&0dTjs
z6o9e;0du!%7JxYce==fr5H|n;yBuf$AA{Mq01-1{&;$SgAEp9m01-1{IsgCw9|Uov
z01-4|IsgCwA9Ugl0TDD}&;$Sg)&gQ-WB?IiVj{SZvH%e^Vs;QW03sP=_y7?$Vg|Ji
z03v|dw*V10Vg|Ji07q;bXaFM4pke?KUSbBd4gdg0ejI23e<IGHVgM0eVg|Ji002jx
z9B2R{>w7-{5n^HnwGIG2BF>;<01;wh2DJ_V002HB>w7-{5o2NowGIG2BF>;<01;zi
z2DJ_V0002j0&*bZ01*OW0WuwE03S0N009vKVmbf-03V6aYyc4gVs;QW08DUia{v(o
zVmbf-08D8Be-Q&>IsgCwB9@d~01*UYTM{<_M<X3*08DgkZ~ze&Vs;QW06rqlpke?K
z7-9ys4gdfEJ|afBz5o#%Vmbf-0001=0e;YI01@<Ib`UoJ0cjm*03S0N009y7VLAW+
z03wOdYyc7UVRjHV08DUia{v+cVLAW+006@QGw@FU0ukSrA{l@pf8SwS5;p(=Ra|>Y
zg8&iVVLAW+03`M<09-Q~009x-VLAW+0E6}}0780e+5i#YVLAW+0Oj^B0AX!uWB?K1
zVLAW+05JD10AXcrWB?K1VLAW+0HXIU005r>hRH1e5w>AE00001p&n=e9{`nv01>uf
zG|vtIAEp9m01>ufOgaDn03!3Io&XWIVFtAh03zrr1OO4ZVFtAh03!O(!T=GtVFtAh
z03tk^;s6o2VFtAh03xcA{QwcVVZFr;07rx#XaH0Ai3b4@!<P>lfE9nkVFtAh00{%=
zzBd2>00000J|fwN0ss-oVFtAh06rqlpke?K%wYz#4gdfE0HOmBp#lIAlVLgl0004>
z0gqq801-N22DJ_V0Ye{X03Y2E&;SuSVFtAh03XC1TL2L{VFtAh03&O601-N22DJ_V
zB8kIC01-Q32DJ_VBHMqzC;$;WVM8Bi06ro!F%bX}K4H434gezXvvB|sKVb&74geyY
z^bP<KKw$>84gg0pA7}s~&Y)rd5l3MLwGIFPJ|fPbVgM0KVFtAh0000069~!?Hvpdj
zh$!X&5jSB5wGIFSe;;T7AKekq01-D~2DJ_VBVlr901-D~2DKIr03&O601-D~2DJ_V
z0H+0un9u+b0hc%%fFXYYVFtAh008(d09>5(4ge7WVFtAh0NVI20BvD%X8;iaVFtAh
z04(}00Bvh`01*LU2DJ_Ve)}!}VP*gk0%0m4XaKGJE&xKro1_2{17T>#4giP!E&xJQ
zGIIbC1Yx?U4ggsGE&yI21^@sxAZP$0iNi+#5gTC!wGIG6{uu-j01+Fv+#7%l0Ri`y
z=^TI{0r!^*9e^}{=3oZ34gdkiA!q;}-4W0L5$0e9wGIFyZ(?cy5$0e9wGIFuhLc4A
z5$0e9wGIGGZ2%GIU_&2h005r>ER1ph5z$~e00001=ptwUBW(Z?(O^R#XaFDH-gE#F
z(O`GC4ge!za%TV$(O^*l002h~B4_|2X~E0@5z}Bg0001g07oApXaFK<!OQ>=)L=RQ
z0000-F(POHB5A?Q01?$-IsgCw07pY2XaFK<!OQ>=)?hjS0000-RU&8rB5A?Q01?+<
zIsgCw07qvcXaFK<!OQ>=*kC#U0000-c_L^4B5A?Q01??>IsgCw07r`=XaFK<!OQ>=
z+F&{W0001GN1Y;Q03vC@%m5MFU^)N*002j?B4_|2X~E0@5!_%p00000N5LX!03vC@
z%m5MHU^)N*002HBR)*F95#M0F#SQ>WX#f%6U^)N*06rpV!OQ>=;$S)e000000JQ`z
z$a(+~%$K1afExkKm%Sc<8v)Ok*&cv#I>lfCt0QOtBW(Z?#b84pXaE490%BIl01>TV
zIsgCw0Hy^100000)Bykh000000i4ArXaH?(01>U1gdcz;0k4;-AAleMub0XnfF=R3
zm+>EfAOWzK5+Hyk0kM}kAb=nNv6osPfF=R5mw_OFAOW(Mq9A}K0kfCMAb=nNvzO{1
zfF=R7mk}X=AOW<OG9iE_0kxM|A%GwOwU>G!fF=R9m!TnmAOW_Q!Xbbr0k@awA%GwO
zx0ebcfF=RBmoXxMAOX0SQX+sR0lAlXB7h(PxtE$EfF=RDm%$={AOX6U;v#?|3%p>x
z#SQ>rX#f$umk%R=BmuvdG$Vi@0l$}3BY-G>i(m$|4gdjkC};p5hLc4A5sP4V&<+42
zZEOG$i(uTx4gepR87%-2i(onc001AeLuLRGi(m$|4gen_8+ZT_j9><}4geo#_uBvw
zj9><}4geoW(+U6)j9><}4geoa%bfrbj9><}4ggGVZ~zgFU_&2h07p|OXaFLkjQIdD
z5tLvCwGIFPM`b8z03xG|`2Z1?U<S1g002HBqm20g5ujiOwGIFP0K)+R0zv>0dY3IE
zfFBWhU_&2h0N?{J09?c!TL2Mymt7=)9|2aEeI$S&Ay!~RA7}tl1TO$w0r%nn5msPP
z0RRA=1TO$WiNi+#5m%SKB!C`&MqoMs0004$DQEyAZ*Tw+Mqoo9XaFB=B!~bJMqmcD
z4genk_u>E%Mqp6^002yMZEye)M_>lE4gg1YDQEy9qm20g5l>(SwGIFPJ|gRTKL8O@
zU<S1g06rqlpke?KQeXzP4gdfE0K)+)d>a4}IA8{~4ggS~1poj5002psTP1)JAvj<|
zA7}s#1up<x0r%nn5jbE`0RRAh1up<XiNi+#5jmHRC4eP=C}27O00045Drf*BZ*Tw+
zC}2Y$XaFB=B!~bJC}0M)4genk_u>E%C}2?m002yMZEye)DPRV*4gfwP>w7-{5j0>1
zwGIG2BF>;<01-4`2DJ_V0002P0Uh&J01+Bs2DJ_Vpr8c+00000NdN!<000000i3`p
zXaH?*Z~ze@8el^oXaG_MF92Kt_u>E%8emZY006`WF91S`!$$xS8<$fifF)lFU^)N*
z00AB=XaFN`Z~zesU_&2h03U56hyW1^U<S1g03QMO;s6l}U{L`808DgkZ~zetU<S1g
z06rq?dp`gX7GMUo4gfwP&Y)rd5f)$uwGIFP005u@W6Ayi5%ibWCV(Ix^j||CXaJW7
zF93l8iNi+#5%pg>00000VRB`201@_=7AJrt0S1>lCx9XW?U!CBfFUsLUqc^g00C9S
z0Ri{o01@q9Q2_t|0aQYX!$$xS?_UPB4giY?F8~0;0k)1801@YxuP1;ZAm?90A7}sp
zRmA}T_u>E%=U-6)000400K);f)dBz!;g|3yfFXb3Uqc^g00C9S0Ri{o01@F|Q2_t|
z0aO5=0c+{+01?e!IsgCw0oE*N03&a301?e!Lmy}WA8jOv01?e!2DJ_V9|8B`01?e!
zQ2_t|OmuB<01?k$2DJ_VJ|gRTKL8QkUk0@f06rqlpke?K-CqW^4gdfE0K)-icDeu&
z!e5t=D1aXk!e2ulXaK<oF92Kt_u>E%!k54(fF1$3m)|IWAb+@DLmy}W0ae8TA{%%B
z5x8Gb0RR92R8ySv4ge9lUk0@f00B?{p8<9Cp#Tx2UpfE)00Gi1XaFN`Z~zgcUqc^g
z03U56hyW3!Uk0@f03QMO;s6n(Ur_-708DgkZ~zgeUk0@f06rq?dp`gXv|k3b4gfwP
z&Y)rd5wu?hwG9pc0002P0byNq01=*-iz$F1f1Y1MA7}spRmA}T_u>E%o?lS`00040
z0G|QjEh_*KieEYa0004TE@%KFZ*Tw+ieEz?XaFB=B!~bJieCn`4genk_u>E%ieFIy
z002yMZEye)j9&(|4gfwP>w7-{5t?5HwGIG2BF>;<01=vB2DJ_V0002P0l(Bs01<m%
zmn$lO9}#<BLmy}WMhPzfTmkpu01<naTPlDa0dJRjDu5tHZ(l<nXaLIzF92L^B!~bJ
zZ(jzr4gePlF92Kt_u>E%Z(mUX002!2F92b5ZEye)abE_t4gi@7F90O~0000&>w7-{
z5qFp1Du5J!cV7mz4gduUF8}}l0G|P};8Oq*TVDpX4gdjqFK7TGZ*Tw+TVF#TXaFA~
z8+ZT_TVGKD0001=0g5;n01+}@chC+10r4+r03U{vMF0^pUw6<B0E+=EnKb|rPhZvo
zVqs(e5mH|wwrZOI5i?%~wGIFxZ>#SB5j0<S&<+59N7OH903yzyVgM0PUk0@f002HB
z>w7-{5mR3VwGIG2BF>;<01;DP2DJ_V0001=0qF8f01*&h2DJ_V0m?9F03TBZQ2-GT
zUk0@f03RiG*#Hp`Uk0@f03TOVcK{I(Uk0@f03Y6-a{v($Uk0@f03&T{cK{I)Uk0@f
z03WV@8pr?<5??w10019J22lVJ5??w1001J1!$$xS6JG|k4gez2gBJi16JG|k4gezT
zD<}XF6ki6l4gg1+FlYcG>w7-{5g}g&wGIG2BF>;<01+Wy2DJ_V002j=FlYcG&Y)rd
z5i4H?wGIFPJ|gRTKL8OeUk0@f06rqlpke@j5iMT^wGIFP005r>opj*<5$|3GwGIFQ
zr7>s#A5#WV01@wA2DJ_VA6HX%01@wA2DJ_VA0>9#01@wA2DJ_VAG1Sd01@zB2DJ_V
zBW-JU01@zB2DJ_VAHWa{0TJ+C2DJ_VBV=!501@zB2DJ_V)&gQ-WB?HiUrcUu01@$j
zUIw)e07sTFXaFJv?SlXj^IitE4gfwPv342&5%pdMwGIGBbTMcEBF>;<01@|I2DJ_V
z07r%~XaFM4pke?K`(6gM4gdf?BF>;<01^CN2DJ_V0000!BF>;<01*ve2DJ_V005r>
zqVEC$5y)N!wGIFQ05WI*A5#WV01?Q4UU$$A03TOVcK{K{UIw)e08D8B5y@V<s15)k
ztqWZM5y@V<s15)=BA@@v01?YxchC+1OmuB<01?by2DJ_V005r>Prrfy5xQOmwGIFQ
zH!^4dA7}U501>)g2DJ_VA0>9#01>)g2DJ_VB5fpy01>-h2DJ_V0G|O@C~p9N5qVw)
zwGIFQ?=olrA0>9#01<g!2DJ_VAFJ0`01<g!2DJ_VA1j4@01<g!IsgCw)&gQ-WB?Jc
zULtDcA^;J3UIw)e03v7i+W--JUIw)e03wj;1pyI!UIw)e03x{a1pyI#UIw)e03v&*
zkpK~XUIw)e07ud?XaFMXdp`g>5wu<gwGIG2BF>;<01>oa2DJ_V002HB>w7-{5w%_h
zwGIG2BF>;<01>rb2DJ_V0002P0iCGY01-Zys4jpZe?DFYwGIF%3oig%SCrKN5k6kf
z1ONbG3oig%=}yZ45k6kf1ONc73oig%R5Eh_5kFo+0002_3oig%i-=SJ5kFq15H|ot
z3@-rI0%Bog01<LtLW#pi01-i62DJ_VdkilCQ=9iJ01-l7#Sdrz1CfA!LQ?cA01-o8
zx~L8Smq;&w9|3chXfJ>u0du#BFMtXGe*j)O000012sUT{A3dJ#01*IQ2DJ_Vivg=A
z=>QQbUW)-Y1S0?uE?!J>Wpn@$0bV))001IPH?sf{0$v8S4gexnI0pd{16~HT4geyo
zJ`Mp91YQQU4geyA+=~Da1zrZV4gexzjP3vt23`iW4gey+Q!W4z2VMrX4gexyOr@3p
z5eQxewGIGBi#2EfBI|oU01-1@2DJ_VJ|fPbVgL~{UIw)e0000-t2Jl<BI|oU01-4^
z2DJ_VJ|fPbVgL~|UIw)e0000-%a<N8fFA+tmp(Co9sxd=VKIOk2>`<ZX%?XX5%8D6
zF@PaT@LdMA4gfL@F92L6cG&<C@LdMA4ghNnF92b3Wpn@$@m)Fq007_(F91S#qNxB8
z@?8eC4ge|+F92R(2LJ%KHfR7s>w7-{5&f4eGJqff{g+KLfM@~dm$ov1AuZ=!2DJ_V
z-wrPTT%It801@Y12DJ_VXAds`7G*eS07#<+000000AL0H01@k#1T%oG0n?YVGk|M<
z%3TJv4gdqsIA{PPZ(?cy5z1W#wGIFuo-l_15z1W#wGIFxml8n$5zAc$wGIGGbZ`I>
z%w2cT4gfw(W&jb|UCO);06t7|Wpe-#++7B>4gfwP&Y)rd5#3z|wGIFP0001=0hgoK
z01>}k2DJ_V0r@y+03&p801>}kchC-h03V((hX4`3T?Vxd005r>zo&cv5%F9)00003
zJ2_|oBX44A01@$A2DJ_VAD%FW01@$A2DJ_VA8jOv01@$A2DJ_VOmuJn5%OGj&<+3q
zp8-NGDgY66Tn4oc009*`XaFN`a4`T8bzBCu4ge!>a54Z9bzBCu4gen>sEYu95p`VU
z0000Vo#CJW5p`TT0000V)J;wR5p`TT0001s0Rie$01?_;NAfvn08C;45rtgs0000?
zVE_?^T;%`&08DrQ5r$ml0000?Y;0lx5r|w(q7DE|Y;0iw5s6$yq7DE|Y;1S{5s6$y
zq7DEeI4+9-5sO^m4QK!&R5Eja01=E_#;6Ve06t7&01@L{?EnA(Okn^K<XkEM001J1
z!$$xS<y<NN00000p8=w|4gnD_Tsi;%00Y!IXaFC{JyHM>FkA+;4gepN7dQYBFkHH*
z4geqX*Sr7`FkI=j4genl0T2KYF<iaH4geqe(82%_F<iaH4genwc`yKf5iwl7#SQ=;
zBy)ZM5iwjk0000Ypbkd>5i(o`wGIGBvpQ%1OlxI%01-7@2DJ_VB5fpy01-7@2DJ_V
zJ|fkwk^m7tTn4oc0000-#X4vJOmJ^=01;DMx~L8S06t7`Z*u?<SzNlP4gdfEp8?TL
zD*zD#Tn4oc00TceXaFC7&Mv+H5d&NXwGIFuD|3DT5d&O00000VDORrl5d&NXwGIFu
zk{37t5d>W6whjOvB6EHK5d>U&sty1?A{g7101+Wv2DJ_VJ|fPbVgL~;Tn4oc00000
zp8-nGg8&idTRH#$00DeFXaFD8u95%|=UWD~4geoVNq_(m=Ucsh#SQ=;_Yv0s5$9XI
z#SQ=;gYKyS5$Ibw0000V&aRRG5$Ibw0001s0b(&#01@|F)&gQ-WB?KRTOttY)c_Ic
zTe_$Y03t|9fB+HdTL!fb03!Gi*8mafTL!fb03sOMl>iayTL!fb03zn-KL8QyTfMvv
z07vaRXaFM4pke@j5$;<CwGIFPNB=u$03yzyVgM2ITL!fb002iHJZJzS>w7-{5&K&P
zwGIG2BF>;<01^9J2DJ_V002inJZJzS>w7-{5&T;QwGIG2BF>;<01^CK2DJ_V002i{
zJZJzS>w7-{5&c^RwGIG2BF>;<01^FL2DJ_V002HB>w7<c01^IM2DJ_VJ|fPbVgM2T
zTL!fb00000p8-(1Pyi9nTRH#$00ZSbXaFD8u95%|&|3zz4geoVNq_(m&|AI54geqb
z5!V0_&|AI54gep{u95%|&|5kH001ILNq_(m(OU+!4gezf5!V0_(OU+!4gfwP&Y)rd
z5!zb@wGIG(007qlD7U8o5yV>o3O#55BXj@}#9Icn4genz>D2%c#9O+k4gdhC1&Uk}
z01>}i1D^p9Kneg6vs*d<0003^J!k+Q3{!#t5wlwcwGIFuv<55y5wlyms15)=BF>;<
z01>rY2DJ_V007qlegYQ&5v^MTXgz2EAG8K601>T!Te_$Y007qld2<Q@5vE%MgFR>f
z9}wx)01>8Jx~L8S0G|QOXmJ1$mRkn34gdlBJ!k+UZ(?cy5tds9wGIFuDORrl5tds9
zwGIH*0%Bog01=^EBC$|B01=m42DJ_VB8kIC01=m42DJ_VBC0LU01=p5IsgCwN9;Xl
z03yzRpke?Kn_C984gdf?BF>;<01=>D2DJ_V005r>fx+Sc5jI;o000015I<-DBX44A
z01-A@2DJ_VA0>9#01-A@2DJ_VAF)t801-A@y~Pdyivf6J9RLxBTOxg;sQ?i-TL!fb
z08Dgn01-J`chC+1B9{_D01-M{2DJ_VOlAOo5j$JTybb^&Ji&7S5j<PH#SQ=>-kx&+
z5j<N4wGIFx>XHQj5j|Tv0000-ojzy)BF>;<01;eU2DJ_V07tJrXaFM4pke?KXIlof
z4gdg0&pv1XBI|oU01;?g2DJ_VJ|fPbVgM0nTL!fb0000-;XY^pBF>;<01<0j2DJ`3
z002kxK4<_U&Y)rd5p!DxwGIFPJ|YX~2muj_TL!fb06rqlpke?Kj9Uh^4gdfE0K)<J
zKXCvN6PL+4fFU{+TQtuO0G|>s078+xd;k#^TL!fb0O%4g0A4^E003V=XaGW`pGN=@
z8CwRm4gek#F91S6^9%tI8J8qGfFuDQmq|N-H31-(e>;F60U(#4JAg7GAzKEu4gk#)
zF90ic&jkPg0000N000000000&&Y)rd5h0iKJAf_$7?&eFfI(IMS_ZWa00YZFXaFDG
z5zqh;{#pjL4gdh30sCXQ01@+A2DJ_V0{}s203&Z=Y5)=QS_ZWa03&N<c>odfS_ZWa
z08DKF5%gMEy$%2X!vP4m6#x<8m$^KEAy?sA2DJ_Ve-tkOToCEi01@F@2DJ_Vc@-}J
zLUMRT01@I^2DJ_V6c#T4BA+@z01@L_x~L8SVQl~r<XTd_4gfb6F8~&T4QK!~zy$yR
z0002s5&!@Z>6bA*fLZ~=myA7t{eQGt2DJ_V18hTR03&Yz5wuzcwGIFyX#f$lTDqtX
z03TrwWB?JgS~>s#03S|Q&Hxd$S~>s#03Y=i9{>@yS~>s#07o}NXaG!Z01>%bx~L8S
zK1^}|5xZIjwGIFP07p|pXaG!Z01?4jx~L8SK1^}|5yM&rwGIFP06t7@0Dlq4T2j3Z
z06t7?01?Ytx~L8S0001=0WJUu01=Q{2DJ_V1H(gT03&Yz5s+F2wGIFyX#f$BTDqtX
z03S|Q&HxdSS~>s#03Y=i9{>@OS~>s#03yae3IP$6TDqtX06t7|01=s52DJ_V006@Q
zCYGH55s6v`wGIF<s09E30GEe9fD|@~S_ZWa0QeX$0Bvdj5s6v`wGIG887}}~asUyE
zS_ZWa0HYZ%0A4T$0025fXaGXapke?KjhE3sfIWX6S^-f-XaFCElSKd#9$E&q4geo*
zB!~bJ9$E&q4ge!`VgL~yS_ZWa03QSFd;k$1S_ZWa0M-IxVPpUif?6WmK7arbA6f>r
z4gg2!L}&makF_NL5i(i^wGIGB%S31ZBKteK01-A?IsgCw06rqb9a{hqL0Sg24gfwP
z&Y%Qh01-@=kwAbGe_mP!wGIFPM>Iuf03raDga8qFS~SlN03vXNQveZqS_ZWa06rq?
zdp`gXep&{#4gfwP&Y)rd5q??*wGIFP002HB>w7-{5rbL=wGIG2BF>;<01<;)2DJ_V
z0002j0`r%^01@9=1CK>$03QIAga8rWSv1cM03U{vMF0`se_6f74geo*B!~bJ-&wuI
z4ge!`VgM1}S-r&$007ql*8~^<5z|=%O-5({9{`nv01?w!G|vtIABK}f01?w!2DJ_V
zA8jOv01?w!2DJ_VBXnW_5z|=)wGIFx16*$a5!6`*wGIFxwF2b;5!6`*wGIFxaD-C;
z5!6`*wGIGBe-K7!03z#qKL8QiSq8NZ06rqlpke?K+gS#+4gdfEM=(Zc03z#qKL8Qj
zSq8NZ06rqlpke?K+*t;-4gdfEJ|gRTKL8QkSq8NZ06rqlpke?K-B|{;4gdfE0G|PP
z5m*2buvrGR4gdj=MrZ&ZC3e{W5wKYXwGIFu{)q<xSrM>VkVa?#AA+y@01>cRy~Pdy
zB8kIC01>rW2DJ_V022tsrw#z00T(M401<du2DJ_V1HndU03Y2E&;Sv5Sq8NZ03%^?
zX8;j+Sq8NZ006@Q=#I+(5pI`yLVzJHZdnGk4gmZaF92;}a%TV$ZdnGk4giE2F90Hm
z!$$xSZ&?Pl4gf;L9a{hqaF@tJfFuERm+eA;ECF<v8bg5mP+D0AwGIFStw?A9AKekq
z01;YQ2DJ_VBVlr901;YQ2DJ_VBAoON01;bR2DJ_VB8kIC01;bR2DJ_VB6S2{01;eS
zXvYo!0K);CJ;eYKJeOuffFVjeSq8NZ0J<D609@&%2>}s2Sq8NZ02v)G09=~*DgY5Y
zSq8NZ0I?k}0E+=EnKb|rR9QmA9a{hqJy`~|4gj7WF91S`!$$xSJ(tl$fFuDxm-a+}
zJ{?9`s!C`81Cg+Q7d}mB0L8=w000000FV;^01-!*QAL1p0aBO4MSwK{Q<wQgfO~%_
zSq8NZ00Rn3XaFDG5zqh;DOm=!4ge!za%TV$DOm=!4ge!-cK{J7Sq8NZ03wOQM*tBj
zSq8NZ03!8!1OX8%Sz5IY005r>G2A-<5zkl#wGIFQ3ruJLAKekq01?kv2DJ_VAH*G7
z01?kvy~PdyA1#bi01?kvx~L8SA5VXK0{{`vSO&EY0E+=EnKb|rAX!XfZvYX{SauLM
z03t}!3IGw&SO&EY03z%?7yuE{SO&EY03sqAcmNU8SO&EY03sv2J^&HbSO&EY03xlP
ze*h8JSO&EY08DED5!YC~#SQ>WWB?J^SiQv#08DOV01??(x~L8SOm1rc5!zB%x~L8S
zB8kIC01?|*2DJ_VNA62#08DUia{v+bSh}bV08D8B5%pNQs15)CJ|gRTKL8OQSq8NZ
z06rqlpke?KA6W*q4gdfE0K)-)HoyQ8vX{h1fFOUeSauLM0D&Mc09?c!TL2NVSiQv#
z0N@}m0Bvqz01>iSx~L8Sdm%3XTu9Rj01>lT2DJ_V@F6b%VP*gkv{>aEXaG+lF91RS
zm4pBhwOBOI4gi=UF91Sw!{z`HwphKz4gldIF92a_01>xXx~L8SJR>gvLW#pi01>!Y
z2DO)CNPr;$%9n*mfN_7DSO&EY00X5>XaFC(*9HI)n^*?54gep-9a{hqn^?WY4geo5
zj8gy+n^?N24gezE5zqh;oLDL$XaFL_o1_2{omgnc4geyY^bP<Ko>&I84geyU87%-2
zpIABo0001=0clyU01-4;IsgCw0gg^+03Qkj9{>?GSO&EY03Uxd=$QZ!Hdu~MXaFCH
zugU-sH&_O>4gepqzB&LAI9R>K4geo-XHftVIas==4geo#j|>12I#|#I004^t@Q(!n
z5sX-i0kMrU01=2-ive?0ssIs%Sk?k!VPpUil~_z`WqAM*JXi*`4gex;B!~bJJXi*`
z4ggGWZ*u?<Jy?Ibs15)kC3e{W5k6RsPG|rkxpm+G5kFXtPG|r}aZYFeOk-_u01;wX
zIsgCw06rq;wU_`Aa9EB`XaGJ;V{LE%5pq~M00000000vRy~Pdy*8(`NC;$;ASOL3E
zXaFD9InV$RCs@z~0018<L`DD+C|Cxy4gdh30m6eJ01<!wS2_Ry00A6NXaFA=IVb=T
z|5paJ4geo#j|>1209eoj0019pn<xMg0a!W!004^t@Q(!n5gAyE0kMrU01*^eOlxI%
z01*RN2DJ_VOmJ^=01*UOx~L8S0M`Qf4Ceq5<W~WIPiO!i);Z7s5#?9V1ONaZWV%oQ
z5$0Dq004ggA9s8Y01@X`x~L8S9~Ar!01@a{x~L8SA4fkH01@d|x~L8SBCrN501@j~
zx~L8SBK-}v01@n0&;$Sg0G|O`x55As$yWxo4gdq@PiO!iz8X&e5z1H41ONaZ2Z~|<
z5z1G(s15)hdTZJM5z1G(s15)h=+j*Q5z1G(s16_iAMk`q01?Yq&;$SgA9`&E01?Yq
z&;$SgAL<(F01?cWkxPIrf4f%(wGIFQF;Hj#A0>9#01>-a2DJ_VB8kIC01>=b2DJ_V
z0G|QDN6P>au2%-N4gdpZP-p-nZ(?cy5w2GTwGIFuo-l_15w2GTwGIGGbZ`I>uUB`_
z4gdh30Y89K01>lPIsgCw0eD<!03YpUk^m92R0g#U0E+>K!}kCYe~(v-0UG+v01<&#
zB7I2O01>xT2DJ_VB8N!Y01>!U2DJ_VB9Tbi01>%V2DJ_VBAZCs01>)W2DJ_VBBe;$
z01>-X2DJ_VN99mx03!M8Gyo9|RR*;V002k#P-p-m&Y)rd5iV5*wGIFPM+s4A03yzy
zVgM03RR*;V002iDe^F=vBF>;<01--62DJ_V07orRXaFM4pke?KPE`iA4gdg0KT&7^
zBF>;<01;1B2DJ_V07p?#XaFM4pke?KUsVRR4gdg0V^L@TBF>;<01;tT2DJ_V07rFE
zXaFM4pke?KZB+)f4gdg0r%`ACQ}~Go0TFgp#itGc2?J2TQ8xep0000}e6RZe5q4Du
zwGIFY14zI(00000002k1QD^`n&Y)rd5qebywGIFPN9$2&08{vh2LTa#RmG<c00{#?
zz&8K@0000}e6RZe5qp=6O@JGJe^myx4gdg04^n6VBF>;<01<*!2DJ_V07pSmXaH0A
zi3b4@gH^?+4gd)QFu*qe00000Q+%)c01<;#2DJ_V2?Hp=Hvj+t0000-Q&MOEBF>;<
z01<~(2DJ_V07qp~XaFM4pke?Kid6=+4gdg0mr`f|Q}~Go0TGK;#itH`00{#iz&8K@
z0000}e6RZe5sOs@wGIFY0~o+J00000002j+QfL4o&Y)rd5sy^{wGIFPN4rvJ03yzy
zVgM14RR*;V002kLQfL4o&Y)rd5tCI0wGIFPN8eIt03yzyVgM18RR*;V002kvQfL4o
z&Y)rd5tmg4wGIFPM*~w;XaFM4pke?KnpFn14gdg06;o&cBF>;<01=y22DJ_V07oZN
zXaFM4pke?KomB?44gdg0SyN~LQ}~Go0THHE#itGc2?G$oHvj+t002{buloQIrkDCo
zfEy{QRR*;V002jwQ)mEF_=yJr5vx_jrw#xK0|3A`0000008@Og`v4KEmp)H`8-K4=
z2DJ_V07t=7XaFM4pke?KvQ-AP4gdg0^HXR5Q}~Go0THuR#itGc2?OxIHvj+t002{b
zuloQIvsDJQ4gd)Q=)X4r0000007nH>XaFM4pke?Kw^atU4gdg07gT5fBF>;<01>)X
z2DJ_V07pqwXaH0Ai3b4@yH&-f4iW$f1K__m0000008@Og`v4KUm+DV|8-Kr52DJ_V
z07q?9XaFM4pke?Kz*Ppd4gdg0e^h7yBF>;<01?Ag2DJ_V07sEjXaFM4pke?K#8n2h
z4gdg0qf}@BBF>;<01?Mk2DJ_V07tb{XaFM4pke?K%2fun4gdg0$5dzlBF>;<01?Yo
z2DJ_V07uzWXaFM4pke?K&1_W$wGIFPM+;SG08{vh2LTb*RmG<c00{%ozc&B?0000}
ze6RZe5!O`(wGIFY1IWKO00000002iFRcHVr&Y)rd5!zJ-wGIFPM^9B~08{vh2LTb=
zRmG<c00{%Yzc&B?0000}e6RZe5!;vWP=FgH-&F>+4gdg0b5&>nBF>;<01@I<2DJ_V
z07s=&XaH0Ai3b4@<Cic|fE8EcRR*;V00{%Azc&B?00000N4HgI03yzyVgM26RR*;V
z002kHRcHVr&Y)rd5$aV2wGIFPNBdQ108{vh2LTc5RmG<c00{%2zc&B?0000}e6RZe
z5$l(nQGgqN?^Oo14gdg0A694pBF>;<01@z22DJ_V07o%aXaFM4pke?K^Hm154gdg0
zLsn=2BF>;<01@<62DJ_V07q3;XaFM4pke?K_f-b94gdg0XI5wcBF>;<01^6C2DJ_V
z07rRNXaFM4pke?K`&9<D4gdg0i&kg=BF>;<01^EtRR*;V002k5R%ifI_=yJr5e8Pp
zrw#xK1CYNr0000008@Og`v4IJmtInU8-EH`2DJ_V07n5=XaH0Ai3b4@3s%La4gd)Q
zfWJ2Y00000Q+%)c01*pT2DJ_V2?Kb)Hvj+t0000-6IW;eBF>;<01*#X2DJ_V07oTP
zXaFM4pke?K5>^Ja4gdg0S665NQ}~Go0TB~c#itGc2?KDyHvj+t002{buloQI6INCR
zwGIFY18BcD00000002j6S7-nt&Y)rd5f@elwGIFPM|)Rj03yzyVgL~uRtB{W002j=
zS7-oJ_=yJr5gS&;rw#xK17N>50000008@Og`v4IemrhfF8-E{G2DJ_V07ubRXaFM4
zpke?KAXWyo4gdg0<5y?^BF>;<01+cr2DJ_V07vy#XaFM4pke?KBvuBs4gdg02UutT
zBF>;<01+ov2DJ_V07n^EXaFM4pke?KDpm%y4gdg0D_Cd%BF>;<01+!z2DJ_V07pGo
zXaFM4pke?KEq_)9wGIFPM^9L203yzyVgM02RtB{W002i}SZDwu&Y)rd5k*!8wGIFP
zM{`(c03yzyVgM0QRtB{W002jYSZDwu&Y)rd5nWaWwGIFPN4HpL08{vh2LTa)R>h|d
z00{$7zc&B?0000}e6RZe5r9?(wGIFY14zF&0000007n2v$yjIrBF>;<01<>%2DJ_V
z07u(cXaFM4pke?KhE@i(4gdg04OwUaQ}~Go0TGW@#itGc2?IdCHvj+t002{buloQI
zkC(1gfE$04RtB{W002iZS!e(v&Y)rd5tUX3wGIFPM`KxN08{vh2LTbDR>h|d00{#y
zzc&B?0000}e6RZe5uR2CwGIFY11P^Y00000002jIS!e(v&Y)rd5u#QGwGIFPM~7Kx
z03yzyVgM1ORtB{W002jsS!e(v&Y)rd5xG_dwGMv(07t7?XaFM4pke?Kx>g3Y4gdg0
zy;*1gBF>;<01>-Z2DJ_V07uVRXaFM4pke?KyjBLa4gdg00a|DPQ}~Go0TIPk#itGc
z2?HR%Hvj+t002{buloQI#a0Hj4gd)Q7{50F0000007nyAXaFM4pke?K+g1j(4gdg0
zC0b`_03yzyVgM26RtB{W002ifT4(?w&Y)rd5$IM1wGIFPM@d>}03yzyVgM28RtB{W
z002i@T4(?w&Y)rd5$aY3wGIFPM~zx&08{vh2LTcDR>h|d00{#Szc&B?0000}e6RZe
z5%ZTWR)8CS30DTS4gdg0v07*VBF>;<01*{e2DJ_V07t`GXaFM4pke?K7FPzf4gdg0
z)mmr(BF>;<01+2g2DJ_V07vIqXaFM4pke?K7*__h4gdg08Cz%oQ}~Go0TCfr#itGc
z2?GGXHvj+t002{buloQIAy)>q4gd)Q_`Wv)0001&$X0+2e>YbKwGIFPM?G6;03yzy
zVgM0CR|d5X002i%TWA0x&Y)rd5kyx8wGIFPM`2rN03yzyVgM0ER|d5X002jGTWA0x
z&Y)rd5k^-AwGIFPM}=Ex03yzyVgM0RR|d5X002jqTWA0x&Y)rd5nfjYwGIFPN2yzA
z03yzyVgM0pTUQ3P4gdg0yIW`gBF>;<01<Ro2DJ_V07u(fXaFMXdp`gXf>#E$4gfwP
z&Y)rd5rS6+wGIFP07v^<XaFMXdp`gXgI5N%4gfwP&Y)rd5rbC-wGIFP07n~Kmv2{q
z9s)aDmycI~9sygIu2+B_e?B6F5fA_oi&qA<4gfwP&Y)rd5sOy_wGIFP005r>p%?uC
z5t~#7wGIFQ{#<APA4vk101=y12DJ_VBV=V}01=y1{#<AP)&gQ-WB?JZR3c~h+W--q
zR0g#U03yC-O#l&`R0g#U03yH;3;_|GR0g#U07v0mXaFM4pke?Kf233fwGIFPJ|gRT
zKL8P~R0g#U06rqlpke?Ku2crK4gdfE022rw$qoRY0XaD?01=E-IsgCw0eM|$03S&L
zmH-iqR0g#U03W_)O#l&$R0g#U03Vfa000q;RL}$f03U!xWB?J2RL}$f03QWeLI4qr
zR5}0v03vN9hyW3dT~r3O4geyF!$$xSj#N4T002HB&Y)rd5tCE~wGIFP0G|PDTNnTl
zbyPY400055U1$JP?7t=e5q?yz&JF+x16aT}0000003u|mQ~(iwRIbhr005~4+5+|f
z5pb7hS%5!(Q&c(t0004^0}G(Z01-=6IsgCw0iOY_UxNS<7gRa`0004~UT6S|0UG+v
z01-u0B7I2O01+8f2DJ_VM?+p{03yzyVgL~)R0g#U002i-UT6R!&Y)rd5i3*%wGIFP
zM`vDW03yzyVgL~^R0g#U002jMUT6R!&Y)rd5j|9Y2DJ_V07r{nXaFM4pke?KLR1E|
z4gdf?BI|oU01-x12DJ_VJ|fPbVgM0FR0g#U00000p8@RQ+yD^*R5}0v00G2aXaI`=
z8v4xu5fW4Yq5}%_0{{{FQ#t?u00E){zMkL!5%N<y00001q60F90Ra)}Q#t?u00Ex?
zB(bJ{01??!IsgCw0SaGe0E+<{`pp0l-ctad0WsH`01>THIsgCw0VZE)0E+<{`pp0l
z)KdVT0Z3^C0TG{4IsgCw0YYDB0E+<{`pp0lsZs!+0lGY(01=Z>IsgCw0Wx4{03&Sx
z5tC5{wGIH*0%Bog01>fKOlbfSlu<eW002yXa%FS?5tUIo0000?a{v*RQDp%D03v1m
zpa2n<Q3kaR03v&*kpK~xQ3kaR03slE*#HrmQ3kaR03s!J*#HroQ3kaR03y5A&j1mf
zQ3kaR07undXaFM4pke?KsZj>C4gdg0^<QWJBI|oU01>iL2DJ_VJ|fPbVgM1cQ3kao
z4gdfEM-^ab03z#qKL8Q4Q3kaR06rqlpke?Kvrz`M4gdfEK9{RofE#}Rp8=UGJ^&F@
zQ91ws009$WXaFB&{h$C5Q&9%B4ge!|01;DB2DJ_VOkr$n01;JD2DJ_VBHo^J01;ME
z2DJ_VA|-a&01;ME2DJ_VB5fpy01;ME2DJ_VA~o7N01;PFIsgCwOmJ^=01;SGIsgCw
zOl5L%01;SGIsgCwN5FqzXaFM4pke?KYEcHY4gdg0(qL!+BF>;<01<&v2DJ_V07v9t
zXaFM4pke?Ki%|x(4gdg01Yu|ZBI|oU01=E)2DJ_VJ|fPbVgM10Q3kaR0000!BF>;<
z01=Q;2DJ_V005r>rHU8;5hPJM00001b75!zBX0l^BvA&n4gfwMcSZXE5hPItwGIFu
zC3e{W5hPJ%0RRBj0%Bog01-}6Q$UlO01+iob75!z2?M~ZHvj+t001J*bUOeML6;s~
zfFA*dmp)y98v%xwT3vu7AdOH4wGIFRkw8=ZyXpWDjZg-)4gdm?NZ|oxbYcJzjhCKX
zfEodim$_YlAORhh+FgLBTz60gwGIFvqz3>10000+0RR9100001oPlL%0Bv*t5qD5`
z&<+4`BrgDMVE_?#PzJRQ0MjHd0Bvgk5qD4qwGIF^B`*M~1!7@j01<&uucQY600000
zVQ&Btc$eK?fFuEemjYjaECGL)FJFK>0pXWyUx4QU-<MQifaQM(P}qh800GNpXaFCM
z-@yP82v7#K4giY*hr{;(5pht90UG+v01<3ZBCX=V01*jL*oFfDBB?OL01*mM2DJ_V
zA_=aa01*pN*oFfDB7I2O01*vP2DJ_VB8N!Y01*yQx~L8SB9Tbi01*#R9kmVsBAZCs
z01*&S2DJ_VBBg&w+5iy|P=JI307oNcXaFLSynFx=IZy_*4gdg0QD$fWOmuB<01-k^
z2DJ_VJ|gRTKL8OzPzJRQ06rqlpke?KLQn>^4gdfE07qkHXaFM4pke?KS5OAE4gdg0
zgJx&|BI|oU01<6a2DJ_VJ|fPbVgM0sPzJRQ0000-qh^0-03z#qKL8PKPzJRQ06rql
zpke?KZcqlb4gdfEN5f`l03z#qKL8POPzJRQ06rqlpke?Ka!>}f4gdfEJ_P5$X8-^I
zp8?4)qW}?^Pk@9200CxbXaFCM-@yP8nNJ3_4giY*hr{;(5&2Jx0U4P!01?+uivd|N
zRR9svPm6y68v4xu5%W(X{d<1^5t>g1wGIFxWai@l5t~m2wGIFx39g?25u8tegaZH~
zeMs5>5uQ&5wGIFxhe+B05uZ;6wGIFxkx1GA5ui^7wGIFxn@HLK5ur~VwGIFxrAXQU
z5u#5*0000YaWby}5u;B+0000-muF}IB3UDf01<z#PX@IP03uW}a{v*qPr9fM002kH
zXJ`N-5b4zb5ynrts15)}x@TwrBF>;<01?nn2DJ_V06rqlpke?K(@zGq4gdfEN9kv1
z03z#qKL8Q*PX@IP06rqlpke?K^iKx04gdfEM+s<X03z#qKL8Q+PX@IP06rqlpke?K
z^-l=~wGIFP07ofkmwjV^9z;oKXaFMXdp`gX`cDS64gfwP&Y)rd5&BOCwGIFP06rq?
zdp`gX`%eb74gfwP&Y)rd5&KUDwGIFP006@Qo?-p~5h9l#WPl-mB2Naj4gldOF95ay
zhr{;(5sOcL*$4mt0001s0UG+v01<{yQ?25`01+ckS=R#q1Cg+PLa8vr01+fl2DJ_V
zRw*w4LJ6*)01+imS=R#qO)4({LVZZu01+oo2DJ_Vu_`YBLWfA&01+rpx~L8SsVgr4
zLXk+?01+uq9kmWo0Gli?079Ec+5izMPX@IP068r$08^z%+5izNPfdUV00WWGeqOL^
z0|1|CXaGWyynFx=Q%?r94gfSRF8~$*5NH5v)d&Cp0002sYXblgRhJNDfLj4vmuzK#
zAOT#Ljb(r!0bG}^Wq>yUahK+0fHeV!mmFq*AOVM$J!XI=4v0?%wGIG2BF>;<01=3n
zab|!b0gRWFW`H08jF+`$fGq(fm*ZxDTmdVWKxcrL0V9{#XMkJ*X16tHfIa~MW|wqn
zfMNk>x7TTa5dr~cx9Mwud<y|*mu7c>SpjF4z;}RM0V|gtcz|62Be#5bfcOCcE4Oue
zfH(nv<W2^)4gdk@fM@_8t>VD|5#&x2&o%&y0f)o)01*^Vivb$?%>WS(Pa;6t!2l8E
zPC@_x03r#lp8yf&P6o9O03v-z+5i#gP6o9O03wG-+5i#hPC@_x03wk{+5i#iPC@_x
z03w@6+5i#jPC@_x03xMG+5i#kP6o9O07sdBfM@_B>w7-{5%x|7wGIG2BF>;<01@_1
z2DJ_V002j;fM@_B&Y)rd5dcpHwGIFPN6Ua{03z#qKL8OBPX@IP06rqlpke?K5KjiR
z4gdfEJ|gRTKL8OCPX@IP06rqlpke?K5l;rS4gdfE0G|Q2`}+V9u1*HE4gdk$foK4K
zAFblS01>WEWa|z9ivfqj_W%*zPKyB=`pp0l*-j!r+Q9%3uTDY$001HhuAcxAuucZG
z4gey3NZJ4qvQ7rI4geyDNZJ4qvra+)001J9NZJ4qv`#_*001JJNZJ4qwN63+001JT
zNZJ4qwoV4M4gg1tfoK3C>w7-{5x`D=2DJ_VJ|fPbVgM1qP6o9O0000-o`GlpBF>;<
z01?Vg2DJ_V07t%oXaFMXdp`gX+D-<w4gfwP&Y)rd5!y}$wGIFP06rq?dp`gX+fD|x
z4gfwP&Y)rd5!+4%wGIFP005r>erN{)5jRfo!UX^U8iQy6ACBL_01-D%2DJ`<0E+>K
z!}kCYpiYYc8v4xu5t~jTt>VD|5jalp!UX^#sW8I;5jjo<wGIFx39g?25jsxr!UX^#
zeMs5>5j;)?wGIFxhe+B05j{@2s15)kkx1GA5k5{GwGIFxn@HLK5kF1_wGIFxrAXQU
z5kOA%O#uK$eS&BJB9Xj&01;?^P6o9O002j;f@lCtbZu||5phliwGIG2BI|oU01<Ic
z2DJ_VJ|fPbVgM0wP6o9O00000N4<h*03yzyVgM0_P6o9O002kbf@lCD>w7-{5u8p2
zwGIG2BF>;<01=!{2DJ_V002k*f@lCD>w7-{5uHv3wGIG2BF>;<01=&kP6o9O0000!
zBI|oU01=^12DJ_VJ|fPbVgM1LP6o9O00000p8=j1lK>IOO->2}00CKqXaFCM-@yP8
z$V~>d4giY*hr{;(5iw4S0UG+v01+%sBCX=V01?SeP6`76BB?OL01?Vf2DJ_VA_=aa
z01?YgP6`76B7I2O01?f9O$N0N03wG-+5i#GO}eNK03wk{+5i#HO&zrk03w@6+5i#I
zO$N0N03xMG+5i#JO&I$E07txoXaFLSynFx=`Ar734gdg0>Vs$iOmuB<01*OC2DJ_V
zJ|gRTKL8N|P6o9O06rqlpke?K0!{|C4gdfE07v|TXaFM4pke@j5f@GdwGIFPM;wG`
z03z#qKL8OeP6o9O06rqlpke?KElvit4gdfEM?8dR03z#qKL8OfP6o9O06rqlpke?K
zE=~ru4gdfEJ|gRTKL8OjP6o9O06rqlpke?KGEN4y4gdfE0G|O~zv%!GSxp%G0ssM{
zg=hdDj^Dum5m`-t2DJ_Vivfqj_W%*XO^X2<`pp0ly-gym;=uqBT1^=H0stbZFv9>5
zTTKSF4gewvuAcxATum7J0sta?NZJ4qUQGtI4geyDNZJ4qUroBG4geyNNZJ4qU`-vh
z4geyXNZJ4qVNC|L4geyhNZJ4qVogE-002k(glGUFk-U6=01=8!2DJ_V07of>XaG!f
zZEye)lT8M-4gfwP>w7-{5tB^@wGIG2BF>;<01=Z-2DJ_V0000-JB4TfBF>;<01>E7
z2DJ_V07qMeXaFMXdp`gXzD)+T4gfwP&Y)rd5xz|ZwGIFP07rX;XaFMXdp`gXzfA_U
z4gfwP&Y)s{01>}U2DJ_V002jug=hdG>w7-{5yMRewGIG2BF>;<01?AY2DJ_V002G&
z;nQaT005r>9)YX?5gSbgwGIFQmWF5mAFblS01+EaPntFWivfqj_W%(~O^X2<`pp0l
zMol6>+Q9%398E$1001HhuAcxA9Zd$c4gey3NZJ5@5g$zkwGIFxhe+B05g<)M0000Y
zkx1GA5g|=N0000Yn@HLK5h6`O0000YrAXQU5hG0owGIGBM}}wsBI|oU01+)s2DJ_V
zJ|fPbVgL~>O$N0N0000-S%zo;BF>;<01-D$2DJ_V07rR-XaFMXdp`gXM@<H`4gfwP
z&Y)s{01-z`2DJ_V002HB>w7-{5lBr2wGIG2BF>;<01-${2DJ_V0001=0VBOD01@R(
z2DJ_V0f~ob03WU5!2l8EOhyhj0E+>K!}kCY6-|o)8v4xu5fDuxK-$3o5#~%n0000Y
z39g?25$8+>wGIFxeMs5>5$Q|@wGIFxhe+Cg01@g;LI3~&B9Tbi01@j<LI3~&BAZCs
z01@m=LI3~&BBe;$01@p>2DJ_VM>>aS03z#qKL8Q;Oa`?M06rqlpke?K_e=)04gdfE
zM@)xk03yzyVgL~VO$N0N002j9hiCvI>w7-{5fM!WwGIG2BF>;<01**Q2DJ_V002IJ
zBI|oU01*;R2DJ_VJ|fPbVgL~mO$N0N00000p8*?Wc>ocwOa`?M00Df6XaFCr;=uqB
zuS|b$5CDq-hr{;(5#LOU0UG+v01?_uB0$=~01>cELI3~&A_=aa01>fF2DJ_VB7I2O
z01>lH2DJ_VB8N!Y01>oILI3~&B9Tac+5i!?OhNzv03w@6+5i!@OhNzv03xMG+5i!^
zOa`?M07or|XaFMXdp`gX!Au6V4gfwP&Y)rd5y4CbwGIFP07pNFXaFM4pke?K%S;Bf
z4gdg0Ux;V`BI|oU01?|v2DJ_VJ|fPbVgM1_Oa`?M0000!BI|oU01@0w2DJ_u06rql
zpke?K+)M_w4gdfE0K);lozDOfd6zYWfFXZ*Of{x90L(-$0JZ^#!}kCYsZ99U1poj5
z004^t8v4xu5u{8)K-$3o5qeBQ0000hMK1t>0tv3401<mk2DJ_V07898+5i!KOa`?M
z0GvfH03wG-+5i!LOhNzv078*S+5i!MOhNzv03Sv#079Ec+5i!NOhNzv0BA-p0759G
zNZJ4qflLOq4gkeQF91hViD&>K>w7-{5sOR)wGIG2BF>;<01=Cq{e*xb0hN~^g@7p*
zrA!934gi!#F90PlF#rHU&Y)rd5v7-3g@8%{rkARPfFJ>;m&=8KF#(R3{e^%o0g{(2
zhJYpkl$TS6fQbP^m)(YdA%8<m#+fz%xk)bowgHF3_W%)dOt{+x000000E+<{`pp0l
zZcIWz+Q9%3L`*^e000<DF93l839g?25k*V}wGIFPLVZZu01-z_2DJ_ViApa3B8N!Y
z01-$`LI3~&LXk+?01-({LI3~&3`;KnLYqk101-+|LI3~&RZA}bLMWw3+5iztOa`?M
z0J2Lj07v_aXaFMXdp`gXRZIr84gfwP&Y)rd5mlF^hkzmhUzf;-fGHMlOa`?M0D(*|
z03{$Y002VHpke?KZ<hy%fJ!28Oa`?M04Gf^04omM1poj5001%o00000002VHpke?K
zaF=t4fGq)7m!XJ&Dgj)V%!q(b0bZ9JiGX|o%$J#ofFXa(O9r(L0Buh%0JZ^#!}kCY
zHB6Am2LJ#7004^t8v4xu5im?st>VD|5zR~XO#uJ{k)VG<sW8I;5zb2nwGIH{PcHyM
z39g?25zkBZO#uMLP%i*NeMs5>5z$KqwGIF(Q7-^Ohe+B05z<S#s15)RQZE2Pkx1GA
z5z|W@wGK4^sZuWhLYqk101?zn2DJ_VE>kZ6LZwLB01?$oDgXcg0aPylUU0Yp02z;H
z078+xd;k&smu`xH8v_4Jmy3#kS^)-^=8AwI0S1=`i+~^j2A3#{fHwgfmtc#4H32b~
zmWzNO0Wp`ii-0B$GE4@w4gfwP&Y)rd5i*zSi+~~lHkS^JfFJ=jmoJQfEdkG$U5tQW
z0nL}bjDTbT5tk#4fJy<)r*4ga0s+mpZ;ybf0RfJe8Igb?e~wE}nl=E5S}y>$0f)o)
z01>`RS>FW!00000ivb$?%>WU(OF}@}!2l7DOF{qu0O(pT0D%GtuAcxAkV^)&4gdf`
zeMs5>5t2&=wGIGmTQ2}2he+B05tB<o0000&kx1GA5tK_p0002*TQ2}Yn@HLK5tU0q
z0000uTrU7ZD5Xf+01=i;2DJ_Vs9Y}qM>m#e03z#qKL8P+O9r(L06rqlpke?KpqJs1
zfFc2^mjsf4DHgg*2DJ_VcU><4CD1bf07A~7VgM1kmqU_(N&&l<j*@^N0lSy4l7KM*
zqnF{5fGPo}mkg7DCIP9JGn0Upe;7*!wGIFQHkoJuA0Oet01+5VH48TYivfqj_W%)p
zON#*-`pp0ldP^dX-@yP88A}GW4geyTIza#t8cQ~rXaFL?8)^U%8%v5f2LK`ouAcxA
z97_hZ4gey3NZJ4q9!mzb4geyDNZJ4qA4@s_001J9NZJ4qAWH_d4gg1|f0$?hBF>;<
z01;SA2DJ_V07to)XaFM4pke?KSxW}B4gdg0*_dbmBI|oU01<3U2DJ_VJ|fPbVgM0r
zO9r(L0000-`Iu+`BI|oU01<mj2DJ_VJ|fPbVgM0)O9r(L0000-8JTDRBI|oU01<pk
z2DJ_VJ|fPbVgM0*O9r(L7ytkOJ|gRTKL8PcO9r(L06rqlpke?KfS2EtfE59%m-Uo@
zA%Ch$E7~;xoL?^hivfqj_W%(QON#*-`pp0l3`;_e-@yP8t4ap74ge-#F90H!Iza#t
ztV$1_XaGXN8)^U%txC~=0sy^WF91RbuAcxAu1W^A4ghyyF91S)NZJ4quu2BC4gmaN
zF91S^NZJ4qu}V4s005L?F91T3NZJ4qvX_;WfFuFxm%5dJF#+qB?3I8aI{r!qwGIG2
zBF>;<01^I52DJ_V003T)AOHZdnrHw*>w7-{5e-WQwGIHXV=n+DkRSj6Le8LK01*wB
zN0xv#0S=dbmVh7u4ws*nfHnaVm)4emAORAW_m+S$0mzprmw+Jw(w9q@fI9)@mxq^t
zl>srA9hiVI0q2)bn1Gc5F_-0-fH48|mlv6Uk^#Jzs+oWz0Xdh>nSddGIZ6h#4glC@
zF95ayhr{;(5t>Slm<Rv>0001s0UG+v01=c*LJZ;K01-M$2DJ_V0cbA(Q}*NI01-P%
zO@IOb1ChXeLjBr801-S&2DJ_V0BJ7(LdhMn01-V(2DJ_V9BMBBLJ6*)01-Y)O@IOb
z|7tG)LVZZu01-e+2DJ`<0F7%e08@uZ+5izjO1h{H00WVzeL|5)+5izkO1h{H0H16x
z079Ec+5izlN;&`l00C_;079im+5izmN(QwK08MQ#077vxuK*E6N(QwK0Kjc803tp~
z+5izoN;&`l08=?LuK*E8N^l_n00WWmeN$IS+5izqN^l_n00WU5kbXyApJ)Ig>w7-{
z5oSsTwGIG2BF>;<01;-F2%CT+0dJQpn}8qzZ<kD)fFJ>Hmu;JXHvw^%rJH~t0dbec
zn}8qzahK<tfHwhummi#fH35~ERh)nz0hO0_oPZ`SmP!V-4gfwP&Y)rd5td2@wGIFP
z07oF9XaFMXdp`gXmr4e;4gfwP&Y)rd5to<BoPZ($o0sjJfE)pvmkOPLG69^IJ)M9c
z0i2g!oq#q0otKrJfFJ>#m$sdNEdf54<eh+*0Xvs`o`9DDN0$_zfXD$zm))O$$N^-R
zrl5dX0cy6PfJ^}ad$&!ZfRqCPd$;1HfCm8qe7EYRfQbVEe7AI}fC~WueV2f%fZhRp
zmtU-a)d79CHLZYW0|9-Pov?sP0e!dVuz)uL0e!b1vw#Z$0e!bDw15i%0e!bPwSWr&
z0e!bbwtzDN0e!c4xPS`*0Z6xmxqu4+0Y|rxx_}D-0XvtRyMSE*h?n}ifJ^~Mw@18y
z3jqN~w^O}<W&r^^x4OQ7cmV<XmovbCA%FWx#+fz%MVc=FwgHF3_W%(qN_gD`00000
z0E+<{`pp0lCQ3p;+Q9%3{7FIp005+#F93l839g?25&cO9wGIFPLVZZu01^L52DJ_V
z6`L;rB8N!Y01*I6LI3~&LXk+?01*L7LI3~&n42#ELYqk101*O8LI3~&;hQf2LMWw3
z+5iy)N(QwK06v^A07p^9XaFMXdp`gX4N3;J4gfwP&Y)rd5e=97z<?qF7ndHvfGHLy
zN(QwK01cfl03{GJ002VHpke?KCzoBpfJ!1LN(QwK0JEJh04vVj1poj5001%o00000
z002VHpke?KD3{5>fGq(Km-oSdDghLiBEo<!0T!20!hn1MzL(CzfFXarNvx$d01}@s
z0JZ^#!}kCY?MaKT1poj5004^t8v4xu5$8!l39g?25x+?WwGIG>pDzGHeMs5>5y43Y
zwGIH%pDzGHhe+B05yD9ZwGIGQpf3PIkx1GA5yMFawGIFRp)UYO!pCR;BI|oU01?ee
z2DJ_VJ|fPbVgM1%Nd_vl4gdfEUN8y(02at-07C0~KL8QvNd~nJ0Cb@*03|RA002VH
zpke?K=$DwofJ!3iNd~nJ0Q8|R04rjz1poj5001cf00000002VHpke?K>6Zk=fF=&=
zNd~nJ06rqlpke?K>X$ggfE@wOmsZ4pE&<S&g~WhX0f3j>#DF1xfJp|m4gjg5F95ay
zhr{;(5wS_o;{^Z!0002C0UG+v01>W9wB!W<00000Q%0({01<&nx~L8S1CgMAB9e|C
z01<*o&;$SgLJ6*)01<;pa3KHyN24zQB7I2O01<^r&;$Sg7Qpal0OH~W000000DwmT
z01<;pkIHBO1kUi6e8qs)0hpIl#(;zYqL<jlfYSk@x2(s2i30(mw>Haw9RdNEmsHJw
zJ%1xf2DJ_V176Z-03RRW!2l5>Ni^jS0E+>K!}kCYL`jPQ8v4xu5kN^I39g?25hO_l
zwGIFxeMs5>5hh6nwGIGBLDFadBI|oU01+@r2DJ_VJ|fPbVgL~^Nd~nJ0000!BI|oU
z01-h+2DJ_VJ|fPbVgM0ANd~nJ00000pML>L2WS8hB|HYT4gdjJ)o1`8{FDv>5hXkZ
zwGIFuVb^E?5hXkZwGIG_0f)o)01=ivivb$?%>WUPJR&YoY5)-?JO;H603!X`LI4pb
zJO;H603yj9vH%e%JO;H603r#lp8ydlJO;H603v-z+5izNJO;H603wG-+5izOJb${V
z4geyNNZJ4qEj+rY4geyXNZJ4qE<8E_001JTNZJ4qFFXdd4gex?GOqv;Fgyme4gexP
zN!kDrF+4f|001I6Gp_&<GCT&g4gg0=(`W!9>w7-{5l=h@wGIG2BF>;<01;0-2DJ_V
z002jL(`W!pbZu||5m-D1wGIG2A`|O-KL8O}JO;H606rqlpke?KSeGHsfD@Mn&wwN$
z*wbhLQ}~Go0TE<8#itGc2?L10Hvj+t002{buloQIWS4o*fE#~mJO;H606t7~ZEye)
zZ9E3G4gfwP&Y)rd5p6sMwGIFP002iO)Mx-A>w7-{5pg^QwGIG2BF>;<01<IK2DJ_V
z002ig)Mx-A&Y)rd5p+BTwGIFPM@rOa03yzyVgM0#JO;H6002i^)Mx-A&Y)rd5qLZX
zwGIFPM|{+103uxLdp`gXfjkDa4gfwP&Y)rd5rI4gwGIFP07snEXaFMXdp`gXkUR#p
z4gfwP&Y)rd5s*9vwGIFP07tykXaFMXdp`gXkvs;q4gfwP&Y)rd5s^FwwGIFP07u-H
zFVKJ=0sNOw(10I*9Mxz5BI|oU01=lw2DJ_VJ|fPbVgM1BJO;H60000-Jk@9bBI|oU
z01=ox2DJ_VJ|fPbVgM1CJO;H60000!BI|oU01=ry2DJ_VJ|fPbVgM1DJO;H600000
zpaO3MJ^&FKJO;H60FdSe0000009^|J0000000EqF<7fawT>O*{0TCKJ2DJ_VxU4S#
zTw&K}01+BI2DJ_V@vJWZLJ6*)01+EJ2DJ_V9j-3`7GBb50I}u<00000005u>01+OS
zaM6IO0Zo@1(twZwBA2YvfKLIKxBSw8xB~%Gx3$)QcmV-vw=UR#M*;zGmv-8KDFJ|&
zpxS`f0fV=A+kmkHe}Fp%wGIFQ*yd;eAN-UK0TF;Z2DJ_VACQL601<#Y2DJ_Vivfqj
z_W%(EJc|Ju`pp0l{W~H){9^zSfjb7Z4gezk+Cl&kf;$Ga4gey_9kKusgF6Pb4gewv
zuAcxAggXYc4gey3NZJ4qhC2qe4geyDNZJ4qhda8c4geyNe@NN@5r{jws15)kn@HLK
z5s5oG0000YrAXQU5sEtowGIFxaWby}5sNzpwGIFxK1tdD5sW)J0000YIWw;S5sf<r
zwGIFxS4r9c5so_swGIGBY2;`CBI|oU01>M@2DJ_VJ|fPbVgM1VI|j850000-m*i*w
zOmuB<01>o1BL=k&06rq?dp`gXv^xg14gfwP&Y)rd5wtr7wGIFP002kDmmJ`LEr0Xm
zXaFJ$=m-H3#ybYJ4gfw(bZu||5y?9SwGIG2BF>;<01?SM2DJ_V0000-6Xj?CBI|oU
z01?eQ2DJ_VJ|fPbVgM1%I|j850000-CFN)UBF>;<01?nT2DJ_V07p~hXaFJ$=m-H3
z*E<Ha4gfw(bZu||5!yQjwGIG2B7e@HVgM1^I|j8500000M|0(903z#qKL8QlI|j85
z06rqlpke?K-a7`h4gdfEM}_5R03yzyVgM20I|j85002jq<!AsR&Y)rd5#u`swGIFP
zN2%p#03yzyVgM23I|j85002k3<!AsR&Y)rd5#~DvwGIFPN89CS03z#qKUx40?>h#y
z4gfwP&Y)rd5$`(&wGIFP07v`fXaFMXdp`gX{yPS>4gfwP&Y)rd5&k;{wGIFP07o0<
zXaFMXdp`gX|2qb?4gfwP&Y)rd5&t^|wGIFP07pCKmyh9q9sygIuHk?m0ehFu;ea1N
zo91W$BI|oU01*Z}2DJ_VJ|fPbVgL~aJO;H60000-yXI&BBI|oU01*c~2DJ_VJ|fPb
zVgL~bJO;H60000!moegi8wmiQ0$!PX01<VUk>Y?MJ#{+<wGIHO)-M2DkcQ6y5p_ET
zwGIH{)-M1;39g?25q3KUwGIF`*e?JUapPzJk<11F00000ppXCn5qX#I;()gSsh6kY
zfRF(qmpJ5rPXU;>h~$8F1OdRe|LA}m0Rhps9O{6w1b^E)2DJ_V0e$vp03ZC64gnF{
zItH~403R{8?En$mItH~40E+>K!}kCYU^|Ne8v4xu5m-ASKKx?<5!^ZkwGIFx{n|nR
z5#2flwGIFx$sMu)5#BlmwGIFx39g?25#KrnwGIFxeMs5>5#c%pwGIFxhe+B05#l<!
zs15)kk$*_q01@Llx~L8SBAZCs01@OmIsgCwBBe;$01@Rn2DJ_VB5^XW01@Uo2DJ_V
zB0fpl01@XpIsgCwA~`d!01@aq2DJ_VB3DV;01@dr2DJ_VM-KF803z#qKL8O1I|j85
z06rqlpke?K20I3|4gdfEM?Cas08DgkZ~zexJ0k|Q4gfwP>w7-{5f3{CwGIG2BF>;<
z01*#62DJ_V0000-YL~(9fGs7M^k@Jg3+M;|5g|JUwGIG2OmuB<01+fR2DJ_VJ|fPb
zVgL~&mqGA=6HO>P2DJ_VJ|fPbVgL~+I|j850000-%JgUeBF>;<01+!Y2DJ_V07v-r
zXaFJ$=m-H3GCKyf4gfw(bZu||5j8sowGIG2BF>;<01-8pq40nce>pn_wGIG2BF>;<
z01-Jm2DJ_V002iS^=JSh&Y)rd5j;BvwGIFPM?Cdt03yzyVgM07I|j85002i$^=JSh
z&Y)rd5kETywGIFPM_~1603yzyVgM0AI|j85002jT^=JSh>w7-{5lTA-wGIG2BF>;<
z01-+%2DJ_V002jzfAwepBI|oU01;U`2DJ_VJ|fPbVgM0YI|j850000-!1ZVVBI|oU
z01;X{2DJ_VJ|fPbVgM0ZI|j850000-;Pq$#BI|oU01;a|2DJ_VJ|fPbVgM0aI|j85
z0000-0QP7ABI|oU01;d}2DJ_VJ|fPbVgM0bI|j850000-e<1c~03z#qKL8P3I|j85
z06rqlpke?KT{{M~4gdfEM?m&y03z#qKL8P7I|j8506rqlpke?KVLJx34gdfEM_~47
z03z#qKL8P8I|j8506rqlpke?KVmk)44gdfEJ|gRTKL8P9I|j8506rqlpke?KV><@5
z4gdfE0H6Y11rp!@5zv=5@_->d&^iXS4gf>=F92LIx9tEC&^iXS4gh@kF91RbuAcxA
z(K-gT4gk^lF8~%6^JoBJ)&>9o0000Wl>h({)R&a<fVTkzmrV13kO3l>-SdD?0hqTf
z^ni8*0UNiZ_<$S%0W7z{`GB$nR-!rvwGIFQ5CdrdAN-UK0TH4)2DJ_VA3Zs001=`(
z2DJ_VAMtEA01=`(2DJ_Vivfqj_W%*TI*S1s`pp0lyE-BXuAcxAqdEq)4gdh40y$Hn
z01;=G$^L*Le`h)dwGIH?8884`J{V&F5obCEwGIH08ZZD{1Oy)d5obCEwGIIA8ZZF1
z0mGZ201=uxsmcWa00000wgGlZr~na`I$_HN000000JZ^nCrtnmlsW**1poj5002Tg
zIcfkAXgUV94geV(FaSdFY&QTAX*veA4gjGXFaUu91_`d801;}JZ~uT5JZ?Hv2WbEa
z1HiU70000008_EPIsg%GI;u)&00WV*e<k1-002TF<O=~2aXPxF4gf11FaSbxXHftV
za+kpWfFA*bx8VPPWB~!2mqh`AQ~|D+ngM}a2(CH>0Z{`Ap_l#vfhPf#mnH&%!2z0=
z=mLQ@S2;QcwGIFQy9a3iAN-UK0TDSm2DJ_VA3Zs001-Jl2DJ_VAMtEA01-Jl2DJ_V
zivfqj_W%)9I*S1s`pp0lQaU0DuAcxAIywfm4gdh40<EqX01^H<mvsYyAb<Wj2DJ_V
zogpv)Ts|0M01^H<2DJ_VQ6ew^Tm%Fk01^H<2DJ_Vup%%3wgJPNp#TvvI;qD600000
z0JZ^kN~i!4EjnSy1poj5006cDc_&Q(5i2?X$OQlZ0000&JvnLs5&t;`wGIH;BQO9$
z@oYB$5db;{wGIGbBrpJh0vrjhp8yd7ItH~4003Tq761S&3TXgSE9jX35e1hp1c4PM
z20E%rXaEC|uzw}c761T3BIFAJ5ePcEs15+@BrpI%b7xTi5eb)T1c4s`8@G!Dfn)&z
zGMDfLfm8uSmqiAFTnI%v1p!e53Xqrn0f85nzy^U30W6o%27$o=iI<ZHfh1PYIR>>3
z00C4CX#gMmlnwzA&^ZRR4geoLIcfkA&^ZRR4geqVY&QTA&^ZRR4giY*hr{;(5$rjO
z0UG+v01@dqA_=aa01?qS2DJ_V0H6YG#u5M#m6r_&fgyjDIR>>309`3C09-y8V*nAA
zIR>>301+xM09*tF9{>@RIR>>30B|ZW0JZ_co1p*^$T=Cj1poj5006cDc1oxK5yUys
zyafON0002C0eL4)01?7DaJ&To00000LOnTZ01=is2DJ_VnJX{=Lh)=j01=lt2DJ_V
zAuKQefdYRCuAcxAm^lWu4gdgN;1B=+#SUozQ!D7101=!yR0nAQ2?Id4Hvj+t002|5
zzB&LAojIyXXaEC|uzw{m5dZ)}BIFAJ5uZ7_s15+DEHD5<b7xTi5uiE7s15*XEieEY
z98YKfU%mwZ00000IRF3v000005wbZE4ru@ZT2Hqu34vSz0m+w%3V~Ds+?U=8fm{gO
zIRybx0}6ka{sDm&msksd5dp)OY72qE0eF`z41pwnXE_G74gdiX5orJ)JvnLs5ob9P
z5orJ)b_%lq5ob9DwGIG_0UG+v01<#WA_=aa01;?82DJ_VB7I2O01;|A2DJ_VNA(bC
z03yzyVgM0!IR>>3002HB>w7-{5rH`dwGIG2BF>;<01<&X2DJ_V0000J2=os(0K);I
z8_NI!5nPwi41pm%Tsa1{4gei7FaTV33bOzaTsa1{4gh~KFaSacuAcxAT{#A|4gldX
zFaQ^r5NQAk<plr$0001xQvd)FV3#iqfk**zw|EVKUjctHIR>>300E5?X#gKRIcfkA
zF*y<uX#k4>8v4xu5lT5C39g?25i&UjwGIFxeMs5>5i~glwGIGBa1&_&BF>;<01-er
z2DJ_V06rq?dp`gXOF0I$4gfwP&Y)rd5lcA+wGIFP006@QG4lce5hghXwGIHF<^=!%
z0001z0RWd)4}lRqCOHPR4gec8FaTUG*kJ$>COHPR4gh^LFaSacuAcxACpiYS4glXY
zFaQ@H6KMdd<^=!%0000GQ~&@GDwm`Wfj<E~x8x6jUjcvrHwLv100GDwX#gKRIcfkA
z|2Gm5X#gLbWv~Dd|2GD;4geo#Q?&pQ|2GD;4genmdIJFw|2GD;4giY*hr{;(5fV9z
z0UG+v01*i}BJROF01*H<2DJ_VBK57F01*K=2DJ_VB0$=~01*N>2DJ_VA~uJ101*Q?
z2DJ_VA{~E4&;StxI0m&203x6aXaErfI0m&203xO&3jh%YI4S@D03r#lp8ydDI0m&2
z03v-z+5iy=I0m&203wG-+5iy>I0m&203wk{+5iy?I0m&203w@6+5iy@IJ&3~03xMG
z+5iy^I0m&203vZRuK*DaI0m&203tp~+5iy`I0k>U4gexKGp_&<5I8yj001IaN!kDr
z5ja8s001I$k^}$|5;#Ht001H<HVOa{6FAt00{|j2lmq|~6gVmX002k*7HI$?3+M;|
z5imFgwGIG2OmuB<01-1d2DJ_VJ|fPbVgL~{I0m&200000M;#Yw03z#qKL8OoI0m&2
z06u>r&Y)rd5jQvnwGIFP07owuX#gV5pke?KIyeTk4gdg0K^JKNBF>;<01-Pl2DJ_V
z07r8dX#i9Bi3b4@L^#E#4gd)QxWG3600000Q+%)c01-qu2DJ_V2?MaeHvj+t0000-
zp%-ZYA`9pU0TD_#2DJ_VK1_6NZ~zfaI0k>U4gfwP&Y)rd5luJ-wGIFP002k97ij<@
z>w7-{5m7h>wGIG2BF>;<01;6*2DJ_V002kR7ij<@&Y)rd5mYz^wGIFPM*$dV08Dgk
zZ~zf%I0m&206rq?dp`gXYB&b94gfwP&Y)rd5o$OFwGIFP002iX7-;}ZbZu||5p91s
z2DJ_VJ|gRTKL8PJI0m&206rqlpke?KZ8!$C4gdfE07pR>X#gV5pke?Ka5x6F4gdg0
za~Np=Q}~Go0TF#T#itGc2?MCWHvj+t002{buloQIeK-cS4gd)Qpujf(0000007r!w
zX#gV5pke?Kfj9=W4gdg0ml$aPBF=xHVgM0?I0m&2002j+7-;|^&Y)rd5rjAfwGIFP
zN4pql03yzyVgM0_I0m&2002kn7-;}ZbZu||5s)|rwGIG2BI|oU01=Ql2DJ_VJ|fPb
zVgM14I0m&200000M;jSw08{vh2LTbBIK`(900{$_z&8K@0000}e6RZe5uBI26oDIm
zpg0D#4gdg0KN)ENBF>;<01=`%2DJ_V07p?7X#gV5pke?Kqc{e&4gdg0V;N}xBF>;<
z01>4)2DJ_V07sD-X#h-gZEye)uQ&#^4gfwP>w7-{5wAD~wGIG2BF>;<01>Y^2DJ_V
z0000-qZw%cBF>;<01?7C2DJ_V07tcd8EF6_&Y)rd5zIISwGIFPN5>gy03yzyVgM1%
zI0m&2002kz8EF6_3+M;|5%4$$wGIG2OmuB<01@*z2DJ_VJ|fPbVgM2II0m&200000
zM->`r03z#qKL8Q;I0m&206rqlpke?K_c#W%4gdfEM<*I-03yzyVgM2NI0m(84gdg0
zIT~pIBF>;<01^8*2DJ_V07pw2X#gV5pke?K{5S@+4gdg0YZ_?)BI|oU01*l~2DJ_V
zJ|fPbVgL~eIR>>30000-iyCPFBI|oU01*p02DJ_VJ|fPbVgL~fIR>>30000-tC#*2
zfgb_OmmU^@9|7x^J{Ex=0t*|LUlxHLK`R?+03z#qKL8OEIR>>306rqlpke?K6FCO8
z4gdfEM@t)N03z#qKL8OFIR>>306rqlpke?K6gdX94gdfEM{Ae97J(lDi<jOOfgb^@
zm;M%k9sxd=As2xg2>_r13ajw|5xJLv7l9#vxi<#24gj|{FaTUW7-Iktxi<#24ghC2
zFaTTx1Rnqqxi<#24gk_PFaWjz!<(T15%)LJA_xEg0002C0d`8L01@*y5i1A)00000
zwgGu3O#l(`H?S%Q00000079H)umBOdHwLv101-Mc077R|wEz*jHwLv10IoVP073&U
zdIJFwyf+554ggC#FaSacuAcxAy*CE64glyoFaTZvd;$O&A87znE9jX35y6*F7=aZg
z!Z*H1X#fL}pnoNRd;$PMBIFAJ5yUsTs15)$JTL%4b7xTi5yh8^7=a%F-<PTwfwTeH
zw_h27eE|XZx7`|noC5)FmwFz7oB@ru7axJ90Re@#z950P0Rg19cOrou0Ra=Yl_P;h
z0s#}Z+$Dil0|D5VGbw?60ST9}DS>DK3zr%yfn)(pmwzgOo&n{T9xH+40rs~7EP+e`
z0p^!aErH_!_O~-Gf$sqUvX>(;fj|MVmt-)3Q~}PHxiEpQ0oA8yF@XXBO_ymifm{JQ
zx5hJpP5}W4w-q&kR007Bw`w<m3IPGww{<vyBmn{1x0*SDN&*4ex8*y5Ndf`dw=X_{
z0RaIMw=q9~I0OOmmuf?S!2v#(EJT4^f4DUUwGIFQVo_-TA3Zs001>z~5)o+tAO6O5
z01>z~2DJ_VAG^;30TH-02DJ_VA0Sd>01>z~2DJ_VAK^Y*01>z~2DJ_Vivfqj_W%*4
zH;Vy!7?J=HIW~&{8v4xu5tlb25!v7X5xF%6wGIFxg0@ot5xO-7wGIFx39g?2e-XPi
z2DJ_VB7I2O01>@4IsgCwB8N!Y01>`5IsgCwB9Tbi01>}62DJ_VBAZCs01?172DJ_V
zBBe;$01?482DJ_VB5^XW01?79IsgCwB0fpl01?AA2DJ_VA~`d!01?DBx~L8SB3DV;
z01?GC9kmVsB6E@i01?JD2DJ_Ve<CL~3IGwuH3qc~03tG!1OO4pH3qc~03t6x5C9R$
zH3qc~03yhg1OO4rH3qc~03y?r1OO4sH3qc~03zO$1OO4tH3qc~07pwoX#gS%=m-H3
z=`{wm4gfw(bZu||5$rVvwGIG2BF>;<01@mp2DJ_V0000-c}i&jB3V>Qe*h8kHNHq`
z06rq?dp`gX^fd;x4gfwP&Y)rd5%e_%wGIFP002j)N@)NhSyV~@5&AX0NNE5*BI|oU
z01^B(2DJ_VJ|fPbVgM2RH3qc~00000N4ZLA03yzyVgL~iHU_m0002kJN@)Nh&Y)rd
z5fC;8wGIFPN9#&y08Dgke{cX15jF<34gfwP&Y)rd5fL^9wGIFP07n^1X#gS%=m-H3
zEH(zU4gfw(bZu||5id3dwGIG2BF>;<01+=X2DJ_V0000-IZJ5(BI|oU01-1b2DJ_V
zJ|fPbVgL~{HU_m00000-OG{}0BF>;<01-Ae2DJ_V07qR*X#gV5f1qLj5jQplwGIFP
zM{i4M03yzyVgM0LHU_m0002jUOKAWik-U5W5o0z6wGIFPN3Tn108DgkZ~zf(HU_m0
z06rq?dp`gXY&Hh94gfwP&Y)rd5o|UFwGIFP002k9OKAWi&Y)rd5q>rXwGIFPNAF8%
z03r+M2muj<HU_m0e*ivAbZu||5r;MgwGIG2BF>;<01<~a2DJ_V0000-K}=}?B3V>Q
z01=HgVo_-TM<7gT03z#qKL8PrHU_m006rqlpke?Kk2VIi4gdfEM>0%l03yzyVgM15
zHU_m0002HB&Y)rd5t23rwGIFP07sKdX#gTwR7wC5mo{Qie`x?ma!hFeBI|oU01=rs
z2DJ_VJ|fPbVgM1DHU_m00000-giL7wBF>;<01=xu2DJ_V06rqlpke?KoHhov4gdfE
zN2N?@03yzyVgM1aHU_m0002j~Olbfj&Y)rd5wbQ0wGIFPN6AcS03yzyVgM1dHU_m0
z002kZOlbfjf6ky{01>@52DJ_V07vaiX#gV5pke?K&^89O4gdg0|4eBBB9Xj&01@Xl
z2DJ_V07orNX#h-gZEye)@HPgu4gfwP>w7-{5%4w!wGIG2BF>;<01@yu2DJ_V0000-
zKTT->BF>;<01*H;2DJ_V07p?xX#gV5pke?K1vdt@e+~cuM}JLe03r+M2muihHwLv1
z06t7~ZEye)6gLL74gfwP&Y)rd5fnEDwGIFP002kPO=$okSyV~@5gIqXNNE5^uuW+I
zBI|oU01+HF2DJ_VJ|fPbVgL~wHwLv10000-!cA!aBF>;<01+NH2DJ_V06rqlpke?K
zA2$ZIe+~cu07oNEX#gTwR7wC5BsacDX#ht8PH6xl>w7-{5hgbVwGIG2BF>;<01+lP
z2DJ_V002i6PH6xl&Y)rd5hynXwGIFPJ|fPbVgL~-HwLv10000-HBM;&BF>;<01-Mj
z2DJ_V07pkoX#gV5pke?KJU0fl4gdg0bxvsje<BO$2muj7HwLv106t7~ZEye)MK=bu
z4gfwP&Y)rd5k)r!wGIFP002kXPH6xlSyV~@5lc68QE322rcP-9BI|oU01-_$2DJ_V
zJ|fPbVgM0MHwLv10000-xK3#RBF>;<01;0&2DJ_V07uGBX#gV5pke?KP&Wp(4gdf?
ze<IGHVgM0QHwLv10000-Jx^%>B3V>Q01;L<bWv#lM+#4A03z#qKL8O}HwLv106rql
zpke?KST_c>4gdfEM;uRS03yzyVgM0ZHwLv1002iWPiX)m&Y)rd5nDF~wGIFPJ|fPb
zVgM0bHwLv10000-Pfuw8BF>;<01<OHe+IP<002i}PiX)m&Y)rd5p_2PwGIFPM{`eU
z03yzyVgM0#HwLv1002jYPiX)m&Y)rd5r8)awGIFPN2O0`03z#qKL8PlHwLv106rql
zpke?Ki8ltd4gdfEN5xNR03z#qKL8P!HwLv106rqlpke?Km^TKs4gdfEN99jx9{?ij
zdp`gXnKuTt4gfwP&Y)rd5t%mzwGIFP07nH-msv}J9s(s$mw8Kp9sxy{m`i~l0cDrC
zOMxE&g_qb%fgb^-m-tJ89|6Ue7)*g50p*uCOo1N)1yPq-Oo1LiB~fVrBI|oU01>7)
z2DJ_VJ|fPbVgM1QHwLv10000-MNw%0BI|oU01>A*2DJ_VJ|fPbVgM1RHwLv10000!
zm%B`X8<!GIfh+-Gmp@H`A%9^t2DJ_VpQkVYTs|0M01;s|2DJ_VsH`vmTm%Fk01;s|
z2DJ_V6Rt1-wgJPNp#TxEHN7wf000000JZ^kN~i!4tTk0I1^@s6006cDc_&Q(5vn!d
zF9rYr0000&{>F6x5n?q4wGIGPuP^{YyUzmw5o0w5wGIFqurL5ZAW~9f01;$02DJ_V
z;jl0OLg7AJ01;(12DJ_Vr?D^qfdUDxp8yeNH3qc~003SfbN~R1RA~THE9jX35o$F&
zZD{}r1E9Az0000008_EPIsg%CmoZL(6cKMVx~L8SZL%-`LUU(P01<GPR!)H=0gIP=
zPJuZAk(aejfiwYzx9U!TX#oMNmsC)J%mJ~tAyI+I0Rj8B>QaHz0s&>0zEy!}0hqTZ
zR)Ks10erWlSb@?30qK`_T7hT*CAZ*OfrJAASGQ$efoci?9GB2&fi{0AH3qc~00G}^
zX#gKRIcfkAC^ZrhX#k4>hr{;(5l}UY0UG+v01-?zA_=aa01+uQ2DJ_VB7I2O01+!S
za3KHyB8N!Y01+%T2DJ_VN5XAs03z#qKL8OnH3qc~06rqlpke?KHZ=yd4gdfEJ|gRT
zKL8O;H3qc~06rqlpkfmM5luA)wGIFP006@Qk?VQ@5gnJ9X@Maj9W@5E4glo`F#udH
z*kJ$>9W@5E4gekqF#tjduAcxA9+%o_fg1rKm-cCaTmdVWQ)+?00XVl5Yk`Ra0XVmw
zZh;E{0X>(hZ-LbTJ(qxRf$sr5w{UTR0|5amw{&uW3jqNuw}5kj3jqN=w~TawhXH?0
zGX}K|00D4<X#gKRIcfkAO*0Y^X#gL)&jSGwO*00y4geojcd`HxO*00y4genpKfM4E
zO*00y4geo1CL#b4O*00y4giY*hr{;(5d$@g0qEhl01>w|ivg88QveaNG>ZWm`pp0l
z`ZOZ6aq|EXPBR9z4gezCz~TTAPcwf8wGIFxV?yWv5l}M*wGIFx;eMh35m7S+wGIFx
z_f`)85mGY-wGIFxikdM15mPe;wGIFx@+8p!5mYk<wGIFx#*%*k5mhq=wGIFx!<9k+
z5mqw>wGIFx39g?25mz$?wGIFxeMs5>5m_?^wGIFxhe+B05n3|_wGIFxkw|~q01;a=
z2DJ_VBAZCs01;d>x~L8SBBe;$01;g?9kmVsB5^XW01;j@2DJ_VB0fpl01;m^2DJ_V
zA~`d!01;p_x~L8SB3DV;01;s`IsgCwB6E@i01;v{IsgCwA}2Nq01;y|2DJ_VA~KW&
z01;#}9kmVsM+baq08{vh2LXQ(ZZpNF4gd)Q$iO!M00000Q+%)c01<982DJ_V2?M~u
zHvj+t0000-8GLB~BF>;<01<LC2DJ_V07plBX#gS%=m-H3cQXdH4gfw(bZu||5qdKQ
zwGIG2BF>;<01<jK2DJ_V0000-XMAY@BI|oU01<vO2DJ_VJ|fPbVgP>;elrHO4gdfE
zM|pf{03yzyVgM0=GX}K|002jed}#n8&Y)rd5sxzlwGIFPN4b1y03r+M2mukEGX}K|
z06t7~ZEye)p)&@x4gfwP&Y)rd5uq~%wGIFP002kXd}#n8>w7-{5v4N*wGIG2BF>;<
z01>4#2DJ_V002kpd})6GBF>;<01>D&2DJ_V07n^pX#gS%=m-H3xibc}4gfw(bZu||
z5xg@7wGIG2BF>;<01>=12DJ_V0000-IelpWBI|oU01?152DJ_VJ|fPbVgM1qGX}K|
z0000-OMPhoBF>;<01?A82DJ_V07rR!X#gS%=m-H3)H4RP4gh~XOmuB<01?+S2DJ_V
zJ|fPbVgM1>GX}K|00000N11(T03z#qKL8QiGX}K|06rqlpke?K+cO5W4gdfEN2`5l
z03yzyVgM1|GX}K|002k5eQ5w9&Y)rd5$rPtwGIFPN6&p}03yzyVgM2UGX}K|002kf
zeQ5w9&Y)rd5eR=Y2DJ_V07vtEX#gUTynFx=A2bHF4gdg0A%1B9OmuB<01+rO2DJ_V
zJ|gRTKL8OZGzPT}06rqlpke?KC^QDO4gdfE07o-^X#gV5pke?KLNo@o4gdg0MSf`j
zBF>;<01-ws2DJ_V07qATX#gV5pke?KN;C$w4gdg0X?}lc03yzyVgM0KGzPT}002jO
zerW(A&Y)rd5ll1&wGIFPM~!}I03yzyVgM0MGzPT}002jyerW(A&Y)rd5mz(@wGIFP
zN6mg|03r+M2mujhGzPT}06t7~ZEye)X*3444gfwP&Y)rd5ot6AwGIFP002krerW(A
z>w7-{5p91o2DJ_VJ|fPbVgM0sGzPT}0000-|9)uzBF>;<01<FB2DJ_V07oBxX#gVY
zdp`gXf;0xT4gfwP&Y)rd5rQ-ZwGIFP07pxIX#gTwBZ>eKj5G$d4gexlGIIbCjWoKb
z4gfwP&Y)rd5u-E)wGIFP07s91X#gS+>D2%cr8IxKs15)}ZhvV2BF>;<01>D(2DJ_V
z07rm-X#gV5pke?Kurvm>4gdf?BF>;<01>k^2DJ_V002k5e`x?rbZu||5xF!5wGIG2
zBI|oU01>$~2DJ_VJ|fPbVgM1jGzPT}00000N9%uS08DgkZ~zgzGzPT}06rq?dp`gX
zyEK<tdw~@K8Gx5@dx0!}D}ZSLBF>;<01@6a2DJ_V07pH5X#gV5pke?K-!ulb4gdg0
zPk?CvBF>;<01@Cc2DJ_V07qefX#gV5pke?K;WP%d4gdg0bAV|8BF>;<01@Lf2DJ_V
z07sL6X#gVYdp`gX>@)_o4gfwP&Y)rd5$rSuwGIFP07tWcX#gTT>w7-{5&JX-wGIG2
zBF>;<01^8%2DJ_V002kRfN20C>w7-{5&Sd;wGIG2BF>;<01^B&2DJ_V002kxmqvVn
z9|9ABmuGx|9s)Cgmxp|T9syIArhI`PL34p=03z#qKL8N~H3qc~06rqlpke?K1T_Y=
z4gdfEN0WhR03z#qKL8O0H3qc~06rqlpke?K1vLh>4gdfEN3)j(eSseV)0ZZFfgb_$
zmqvYo9|9ACmuG!}9s)Chmxq0U9syIArhS1P0X~<>eSsSZ0H6Y1bszu{^p`1qfgykN
zG6uB{0H_!-09-y8V*nBKG6uB{00|s109*tF9{>^bG6uB{0E8Ve0JZ_co1p*^MKkSE
z1^@s6006cDc1oxK5kWJbUj_gG0002C0eL4)01-bkAYTRm00000Lc7lc0TJ~w2DJ_V
zbRIDPLRELN01@^w2DJ_VCm%5YLIx#2y#NvSG6uB{0NNih075AyA^;KiG6uB{0F59q
z0740_p8yg0m+pRn7(D$lJZ)(J2?OA_Hvj+t002|5zB&LA{xV`wX#fL}kbfoMdH?`I
zBIFAJ5d$;2s15*(Au#|#b7xTi5d@b)e}N<c8JA~&fjI#mm!W@wIRPb?+kb)00Y<l}
zfPvWs0U?)zhJnKYIk!iLfinUDfR}rTfsg?rmkf%5PXU;hT#A8+0ZW(Vih+;;m$!C{
zfn)&zK)2hBfvo}oke7*%fj|L|m&%WUcLA=KLXd&30kyaBkb(OP0YI1SoPp~B(3jbr
zfh2!KG6uB{00BOxX#gKRIcfkAL^2W)X#gKCjy?bpL^1}o4giY*hr{;(5#ch60UG+v
z01??TivhGK009x9GK&G4h0OpFo-!i8!6X0?MKT7p4gey?J;DGHMluGq4gewvuAcxA
zM=}Pr4gey3NZJ4qNiqht4geyDNZJ4qN-}>2wGIFxkx1GA5lb?<s15)kn@HLK5lk{2
zwGIFxrAXQU5lu1%wGIFxaWby}5l%7&wGIFxK1tdD5l=D(wGIFxIWw;S5l}K&*8>0|
zS4r9c5m7QtfC2zVHKS<&B9Xj&01<pL2DJ_V07qk^X#h-gZEye)g)#=U4gfwP>w6_X
z01<^U2DJ_VJ|fPbVgM0^G6uB{00000N0p;#08{vh2LTa>mmr^k6@P~^2DJ_V2?OxJ
zHvj+t0000-!=q^cA`9pU0TGQd2DJ_VK1_6NZ~zgIG6uB{06rqlpke?KkTM3f4gdfE
z07v7aX#gVYdp`gXlrjdj4gfwP&Y)rd5tK3pwGIFP07vzsX#gV5pke?Kmof&m4gdg0
z2c&5LBF>;<01=ro27k2<002iBq-g*m&Y)rd5t}jwwGIFPM>(Wv03z#qKL8P&G6uB{
z06rqlpke?KoH7Qr4gdfEM_Ht403z#qKL8P*G6uB{06rqlpke?KpE3ru4gdfEM|q@a
z03z#qKL8P}G6uB{06rqlpke?Ktuh9+4gdfEN2{c108{vh2N?kot}?}^4gd)Q=)gAs
z00000Q+%)c01>X2OrU`q7qBwLrw#xK1K7Yf0000008@Og`v4KJmwupu8-KGh2DJ_V
z07o6AX#gS%=m-H3z%mB44gfw(bZu||5yLVDwGIG2BF>;<01?A72DJ_V0000-J*8;?
zBI|oU01?MB2DJ_VJ|fPbVgM1xG6uB{0000-Po-%9BF>;<01?PC2DJ_V07qe^X#gV5
zpke?K%Q6PF4gdg0fu(5xB7f_9KL8QhG6uB{06rqlpke?K+A;>U4gdfEN1>%@03z#q
zKL8QiG6uB{06rqlpke?K+cE~V4gdfEN5Q3O03z#qKL8QjG6uB{06rqlpke?K+%g8W
z4gdfEN8zPu03z#qKL8QkG6uB{06rqlpke?K-7*HX4gdfEM**g30AM2Pdp`gX-ZBQY
z4gfwP&Y)rd5#BNewGIFP07oIFX#gVYdp`gX;xY!c4gfwP&Y)rd5#lliwGIFP06rq?
zdp`gX<1z-d4gfwP&Y)rd5#urjwGIFP006@Q%asKI5jmIlp@AVGIWh*d4glqIF#udH
zjy?bpIWh*d4gei=F#tjduAcxAI+r=3fg1rmmsO&H(g9JID5HU`0e-jEqk)zI0hqUK
zrGcga0Z^9|r-AMPvbO`Mfyx2_wU^GSfpP)0w?V6coB{z+m)@;`T>((HJg$K<0s&FC
zZ?J(N0)OZ+2DJ_V0erY=03SU$Y5)=FF%l7J03X&p!T=HIF$T2`0E+>K!}kCYC^Cxy
z8v4xu5h5}o39g?25$Q1owGIFxeMs5>5$iDqwGIFxhe+B05$rJrwGIFxkx1GA5$!Ps
zwGIFxn@HLK5$-WffC2y_rAXQU5$`c70000-dqB5o03z#qKL8Q=F$T2`06rqlpke?K
z`7s8y4gdfEN3FMM08{vh2LTcKF~z4200{#K!8ZT^0000}e6RZe5&D--vw<6b05S%(
z4gfw(bZu||5d$&?wGIG2BF>;<01*Q+2DJ_V0000-`?qNTBI|oU01*c=2DJ_VJ|fPb
zVgL~bG6uB{0000-4Y+9lBF>;<01*l@2DJ_V07oCVX#gV5pke?K3^E3_4gdg0Ke%ZC
zBI|oU01+cH2DJ_VJ|fPbVgLXIfck$CO(QY}wGIFP07qZAX#gVYdp`gXBr*oI4gfwP
z&Y)rd5hOAOwGIFP06rq?dp`gXB{BxJ4gfwP&Y)rd5hXGPwGIFP006@Qni3rV5#5(H
zw1FW#-7yBW4gl_vF#ufFJ;DGH-7yBW4gi^wF#tjduAcxA-Z2KX4gg1!F#s0VwrK$L
zoCp8_0001B3IhNU;g^%NfyV*vw_CM=KLG&?x0|+sP5}W6m+H5HlL79xez<`N0Rir}
zh`E7P0e_4!2DJ_V0m{H>03SU$Y5)<8F%l7J03X&p!T=GBF$T2`0E+>K!}kCY%rT1r
z8v4xu5ymkh39g?25sfhhwGIFxeMs5>5sxtjwGIFxhe+B05s)zkwGIFxkx1GA5s@(l
zwGIFxn@HLK5t1=YfC2y_rAXQU5tA`00000-$w0qp03z#qKL8P(F$T2`06rqlpke?K
zoiPTr4gdfENBh5N08{vh2LTbDF~z4200{#a!8ZT^0000}e6RZe5uTTvyMY^jq%j7y
z4gfw(bZu||5vMT*wGIG2BF>;<01>A#2DJ_V0000-Nx*3UBI|oU01>M(2DJ_VJ|fPb
zVgM1VF$T2`0000-Tfk`mBF>;<01>V+2DJ_V07q@WX#gV5pke?KurUU;4gdg0jlgLD
zBI|oU01?MA2DJ_VJ|fPbVgM0M$1w)A4gdfEN3FnV03z#qKL8QPF$T2`06rqlpke?K
z$T0@B4gdfEJ|gRTKL8QQF$T2`06rqlpke?K$uS1C4gdfE0K)-aXjlLdftQ24fgwGC
zF$T2`05hC109@8R!T=G0F$T2`0N$K20740_p8yepF$T2`0F0e602U>`X#n(~2mk;8
z006KI0{{_)m*KpD#{rVJsl9<e0RgYK>ArzZ0RgX<IKY9E0g|`Oz<~(?0g|`Y!GTo)
ze;hFewGIFQ)X8Z8A3Zs001+H95)o+tAKEL*01+H92DJ_VAElXM01+H92DJ_Vivfqj
z_W%)aF^d5j`pp0lYB3@SuAcxA9We&A4gey3NZJ4qA29~C4geyDNZJ4qATb8D4geyN
zNZJ4qAu$HE4geyXNZJ4qA~8*X0stbVe@NN@5hF28fC2y_aWby}5hO7x0000-FUV;C
zBI|oU01+-R2DJ_VJ|fPbVgL~?F$T2`0000-VaRC!Q}~Go0TC}T#itGc2?IF6Hvj+t
z002{buloQIFEIwS4gd)QFu^wf0000007s9=X#gS%=m-H3H8BRY4gfw(bZu||XAw9t
z2DJ_VJ|fPbVgM01F$T2`00000N3Y0f03z#qKL8OtF$T2`06rqlpke?KJTV5f4gdfE
zN5RNx03yzyVgM08F$T2`002kR$Y}r~&Y)rd5kWBqwGIFPM+M1g08{vh2LTa6mypDP
z6@Nl82DJ_V2?HR(Hvj+t0000-Gs$THA`9pU0TD<s2DJ_VK1_6NZ~zfYF$T2`06rql
zpke?KOECtu4gdfE07p~FX#gVYdp`gXPca6y4gfwP&Y)rd5l=A&wGIFP07qrXX#gV5
zpke?KQZWX#4gdg0cgbl0BF>;<01;F%21T_F002jq$!P#0>w7-{5o<99wGIG2BF>;<
z01<032DJ_V002j~$!P#0>w7-{5o|FAwGIG2BF>;<01<342DJ_V002IhEyaNw2>`<Z
zbvs4?5#g7O#epGz;V=fZ4gh$jF#udX7-Ikt;V=fZ4gg7~F#ucy1Rnqq;V=fZ4ge6S
zF#xs!!<(T15fm{+r3e530002C0d`8L01*%|VW|iJ00000wgGu3O#l%NF#xFu00000
z07BX;$^a4KFb1^_01&A$079jiVgM22Fb1^_0Gp{X073~guAcxA<S+)c4ge;qF#uiw
z4+8*5%xM5qE9jX35$G^f2WbEa0}!}30000008_EPIsg&rmtn?%6cOt%x~L8Sa;h-^
zLUU(P01@n$hQ@&(0{$?!r^bO~0Ra`449J0s0V0=|$bpvuBeyTffo=f-K)0UCfldJd
zK)3D7fnosxQ@2pffldJdQ<tC4fs+9vmov|STLB`skI#W~0RbbI1JQw70V20w(Sf@G
z0S}iO(}BSOotNFyfjECOFb1^_009!*X#gKRIcfkAG%ykoX#gMLexd*oG%yCW4giY*
zhr{;(5za7+0UG+v01?J8ivhGK009w^FpB}2h0OpFjxZt!C;$KvH82LX4geyWFS-B`
zHZTUY4gewvuAcxAH!udZ4gey3NZJ4qIWPvb4geyDNZJ4qIxv3*wGIFxkx1GA5j!xt
zs15)kn@HLK5j-#*wGIFxrAXQU5j`*lwGIFxaWby}5k4>mwGIFxK1tdD5kD{nwGIFx
zIWw;S5kN3m*8>0|S4r9c5kW9bfC2zVP1$JxB9Xj&01<332DJ_V07rY-X#h-gZEye)
zbub3C4gfwP>w7yt01<UC2DJ_VJ|fPbVgM0!Fb1^_00000N3Gdu08{vh2LTaxFvX`1
z00{$F!8ZT^0000}e6RZe5qFnP)qxv-eJ}>K4gfw(bZu||5r8lTwGIG2BF>;<01<#N
z2DJ_V0000-``Kv#BI|oU01<>R2DJ_VJ|fPbVgM0@Fb1^_0000-4cch{BF>;<01<~U
z2DJ_V07oC%X#gV5pke?Ki7*DW4gdg0G1_SWBF>;<01=BY2DJ_V07p^UX#gUB>w7-{
z5sWYfwGIG2BF>;<01=EZ2DJ_V002jE+GzkH>w7-{5sxqiwGIG2BF>;<01=Nc2DJ_V
z002jk+GzkH>w7-{5uGpwwGIG2BF>;<01=%q2DJ_V002kB+Gzk&_=yJr5uPx`rw#xK
z14zL)0000008@Og`v4K1Fb1`M4gd)QK*2Wv0000007voKX#gS%=m-H3tuO|)4gfw(
zbZu||5wI`@wGIG2BF>;<01>b-2DJ_V0000-5!-11BI|oU01>n>2DJ_VJ|fPbVgM1e
zFb1^_0000-Bim^JBF>;<01>q?2DJ_V07o_3X#gV5pke?KxG)B_4gdf`M^)Qt03z#q
zKL8QOFb1^_06rqlpke?K$1n!94gdfEM|In203z#qKL8QPFb1^_06rqlpke?K$S?-A
z4gdfEN0pc8)`1@ZwU-Fjfgb_YmnhePA4K)rX#gVYdp`gX&oBnH4gfwP&Y)rd5zjCN
zwGIFP06rq?dp`gX&@cwI4gfwP&Y)rd5zsIOwGIFP006@Q;@9Q?5h<6;*MT7+DKG}L
z4gf;AF#ufQexd*oDKG}L4gh+&F#tjduAcxADwhq|fg1rWmnqnR(g8u2{Mdo50dBXd
z*@2b;0f@Ij+kvJ50YI1N+<~V7ptq3SftLaSqL(P(fpP(&x0~RBoB{zsmpJ2rT>(J1
zmg9jk0s%p{$>o6{0)N;q2DJ_V0oLtl03SU$Y5)<~FA@=H03R~KGyoCUF9x*^0E+>K
z!}kCY7%+<g8v4xu5fU&W39g?25!o*WwGIFxeMs5>5!){YwGIFxhe+B05!^2ZwGIFx
zkx1GA5#28awGIFxn@HLK5#BFNfC2y_rAXQU5#KK=0000-)qm`103z#qKL8QwF9x*^
z06rqlpke?K=`RMg4gdfEM+fa`08{vh2LTc4FU6-000{$V!8ZT^0000}e6RZe5$Z1n
zwGIFY17N{700000002id?P&la3+M;|5%4btwGIG2OmuB<01@*q2DJ_VJ|fPbVgM2I
zF9x*^00000M}JlAX#gVYdp`gX_b&#u4gfwP&Y)rd5%(_!wGIFP07qx-X#gV5pke?K
z`Y#5x4gdg0dF^QcBF>;<01^Bz2DJ_V07seaX#gVYdp`gX6EFs~4gfwP&Y)rd5fd;5
zwGIFP07tp)X#gVYdp`gX6fg$04gfwP&Y)rd5fm^6wFM3U002IhFX@3B2>`<ZTP$e+
z5zUv6>4703%`XPE4ggiwF#udL!ZZL8%`XPE4gdhxF#tjduAcxA&X>{Yfg1tQm+k3+
z#{u5Awd#RC0Rj8B_3MF70Rj7$MD2l-0p7RR?STma0p7Rb?txVSe|#?nwGIFQB=%_l
zA3Zs001<pI5)o+tA2Px;01<pI2DJ_Vivfqj_W%*RFN*;h`pp0lwl5+HuAcxAeJ=*J
z4gey3NZJ4qe=i2L4geyDNZJ4qfG-BM4geyNNZJ4qfiDKN4geyXNZJ4qf-g;g0stbV
zNZJ4qgD)xo002iLD)ngqBI|oU01=HZ2DJ_VJ|fPbVgM11F9x*^0000-RrP5AQ}~Go
z0TGUu;P8PJe~vE(wGIFY18~7N00000002jW^=SYi3+M;|5tJ_mwGIG2OmuB<01=lj
z2DJ_VJ|fPbVgM1BF9x*^00000N2B#=03z#qKL8P%F9x*^06rqlpke?Kn=b~n4gdfE
zN452703yzyVgM1IF9x*^002kF^=SYi&Y)rd5uh&yMYRq907vKbX#gVYdp`gXw=V{@
z4gfwP&Y)rd5w|Y}wGIFP07nP*X#gVYdp`gXxGx5^4gfwP&Y)rd5x6e~wGIFP06v#}
z@qrr&0K);58uI`VahKikfgvGrF9x*^0Gi-209-P{GyoBCF9x*^07c<30740_p8yeZ
zmml(h8v%8fJMw|Y0fM&!^MOAB0iU-+^np$R0iTzW^?{QCg1092fe8Ttg10dDfmH#2
z3@-+?4gdiz|7idpJvnLs5ezR95orJ){%i995ezQ|wGIFuz4HG65ezQ|wGIG_0f)o)
z01;p>ivb$?%>WTvFCq!9p8ydJF9x*^03v-z+5iy`F9x*^03wG-+5iy{F9x*^03wk{
z+5iy|F9x*^03w@6+5iy}FHL{~03xMWNZJ4q6E97G0sta$GOqv;6fY_O002jc{b>Lq
z>w7-{5gsoFwGIG2BF>;<01+N92DJ_V002k3{b>MG_=yJr5g#wbrw#xK1DL@#00000
z08@Og`v4Iimt6UQ8-FD)2DJ_VK1_6NZ~ze~F9x*^06rqlpke?KC@%)J4gdfE07nV_
zX#gVYdp`gXEH4JN4gfwP&Y)rd5iBnTwGIFP07o1CX#gV5pke?KFE0kQ4gdg0E&gc$
zBF>;<01+`S2DJ_V07qZ`X#i9Bi3b4@GB3ra4gd)Qh`~1i015yA08@Og`v4I#mimDk
ze>g7&wGIG2OmuB<01-Pc2DJ_VJ|fPbVgM04F9x*^00000N3H&803z#qKL8OwF9x*^
z06rqlpke?KKQ9Kg4gdfEN5B4Q03yzyVgM0BF9x*^002kP{%HUr&Y)rd5kxNrwGIFP
zNAdn?03z#qKL8P1F9x*^06rqlpke?KO<OMpwGIFP07nu3X#gVYdp`gXTrUQ-4gfwP
z&Y)rd5nL|@wGIFP06rq?dp`gXT`vZ;4gfwP&Y)rd5nV3^wGIFP006@Qh?X({5z&|D
z`hg*T(JltH4glZkF#udX7-Ikt(JltH4gj$0F#ucy1Rnqq(JltH4gh!TF#xs!!<(T1
z5d<%Vwg><K0002C0d`8L01*H$p}GhF00000wgGu3O#l)8E<m~n0000007Cw2^8gXj
zE(Wy@0C(*%07AX;{{RuwE(Wy@00QnY073}`uAcxA)R&C=ffzj4E>s6;00{#yxHkX*
z0000}vA#M05!o)PN@xHBk)VGi&=&&$LL%e~0TJ6Sx~L8S+3qm_LUU(P01@1m;QN6e
z0q(Z^fn@;!1(#<1fr<eVm+}39mjM&EiT;6Z0Rb?#`2T@U0Rb?#M*)Ij0Rcm|sse&e
z0Rcmo`2&KJ0TY*v1cF-u61V9Df^z`@6PH~Ef?EL+x48y_y8#0KE|+!)g1`Z4mp2K5
zIDeQf2DJ_V0Z0yN03SU$Y5)<KE)o%G0E+>K!}kCYz%Gjc8v4xu5xg!U39g?25t%Lq
zwGIFxeMs5>5t}Y50000Yhe+B05u7dtwGIGBD-LP^BI|oU01>7x2DJ_VJ|fPbVgM1Q
zE(Wy@0000!BI|oU01>?|2DJ_VJ|fPbViN!ny)Fi|4gdfE0K)+svu^+qjh6umf+0PP
zE(Wy@0IB^k09-EEVE_@0E(Wy@0OI{I0740_p8ye#E(Wy@02}@>02a3mY5;*^2LJ#7
z0000H0RRz^mtP8k)qgrJ2DJ_V0ZkEV03SU$Y5);BE)o%G03Ra=PXG}*E(Wy@0E+>K
z!}kCYd@hRt8v4xu5p^yi39g?25j!phwGIFxeMs5>5j`#jwGIFxhe+B05k4*kwGIFx
zkx1GA5kD?M0000Yn@HLK5kM|YfC2y_rAXQU5kW2nwGIFxaZEC=01-ki2DJ_VM=TI(
z03z#qKL8O-E(Wy@06rqlpke?KOfCks4gdfEM_v$W08{vh2LTaHF2$z~00{%A!8ZT^
z0000}e6RZe5lxr>3xXSeTrLK+4gfw(bZu||5nnC_wGIG2BF>;<01;m<2DJ_V0000-
ztPpAdBI|oU01;y@2DJ_VJ|fPbVgM0iE(Wy@0000-z7T2vBF>;<01;#^2DJ_V07uXe
zY5*e6pke?KXD$Y{4gdg0@DOSMBI|oU01<XB2DJ_VJ|fPbVgM0;b}k0B4gdfEM-UNe
z03z#qKL8PTE(Wy@06rqlpke?KcP<9C4gdfEM=%j;03z#qKL8PUE(Wy@06rqlpke?K
zcrFID4gdfEJ|gRTKL8PVE(Wy@06rqlpke?Kc`gRE4gdfE0K)-`d^P|PFD?eP4gk>2
z2mk;8007Pd0001&7!86CJufZ>wGIIK{xJYtBM46b5ic$VwGIFy05SkV39g?25il+W
zwGII105Sj;d=F{><jn{G00000pko675i^%$4T8`CK$kTRf|~(Rx6uxQP5}W?w;&LL
zHvs`amsk;k=K(;sKoWvr0e|)_2DJ_V0eKi|03SU$Y5)=TEfNuF0E+>K!}kCY9xjUk
z8v4xu5gINc39g?25%(<ywGIFxeMs5>5&11b0000Yhe+B05&A6#wGIGBTo`HqBI|oU
z01*T(2DJ_VJ|fPbVgL~YE(Wy@0000!BI|oU01+E52DJ_VJ|fPbVt)V;8!iU54gdfE
z0K)-WFr)wx>n#Sg4gj!V2LJ#7006ZB000000004;(-~?2Ts=8z01@jg2DJ_V845B0
zTrSvQ01@jg2DJ_VQVKEvLJ6*)01@mh2DJ_ViwZIT7U&mh0I^^P0000008kGB01@vk
zz8Pu&0s0qe0K*D00GFN=f>Hqsm+=&Wntzxr2DJ_V0i+vh03SU$Y5)<KEfNuF03Ra=
zPXG~^Ee5p?0E+>K!}kCY+AWI#8v4xu5z{Rq39g?25t%IpwGIFxeMs5>5t}UrwGIFx
zhe+B05u7aswGIFxkx1GA5uGhU0000Yn@HLK5uPngfC2y_rAXQU5uYsvwGIFxaW*oq
z01==q2DJ_VM}rz_03z#qKL8P`Ee5p?06rqlpke?Ksx1b!4gdfEN3|Mi08{vh2LTbQ
zmp~SR6@RNO2DJ_V2?MafHvj+t0000-;~HuJA`9pU0TH?_2DJ_VK1_6NZ~zg#Ee5p?
z06rqlpke?Ky)6c{4gdfE07nBGY5*eZdp`gX!7T>04gfwP&Y)rd5y346wGIFP07n%Y
zY5*e6pke?K!Yu~14gdg0CmU)2BF>;<01?G421d0G002iv8)^U|>w7-{5!5XPwGIG2
zBF>;<01?zJ2DJ_V002j48)^U|>w7-{5!EdQwGIG2BF>;<01?$K2DJ_V002jam)aJB
z9sxd={T6~72>`<Z5A03=5sjB!7lI)njV%VX4ghKkG5}m72u}bJjV%VX4gix3G5|se
zuAcxAj+dVof*S#mm%A5&&;g#8j2MEO0k5|x8G=p$0k5}r8iF?g0iKty8-nKnp0}4A
zf?xrER4oRz4gdkrB5D91JvnLs5mYS_5orL60f)o)01<pGivb$?%>WU2Eg}i7p8ye6
zEe5p?03v-z+5iz(EkXbQ03wG-+5iz)Ee5p?07tVTY5*eZdp`gXVl4)>4gfwP&Y)rd
z5n?R{wGIFP06rq?dp`gXc`XLD4gfwP&Y)rw01<gD2DJ_V0002P0jSz!01-)-i5`L>
zAxSL;wGIG@7BT=_F4$oJ5lJluwGIHn7BT=r39g?25lWZK9)cSIO_%2$f?xq@mq;Ii
znt!t@2DJ_V0bwm_03SU$Y5)<lDiRTC03WzVQ2-IMDh9O<03W%XPXH0KDh9O<03Sjl
zhyW3@Dh9O<03Wlg)Bq8)Dh9O<03QeE4*(IfDh9O<03Q@|H~<l|Dh9O<0E+>K!}kCY
zH7$z)8v4xu5iBhtHvtO)5wt1>wGIFxqkrM!01>q+2DJ_VA{^Sm01>t-2DJ_VA_=aa
z01>w;2DJ_VB7I2O01>$=2DJ_VB8N!Y01>(>2DJ_VB9Tbi01>+?2DJ_VBAZCs01><@
zIsgCwBBe;$01>?^IsgCwB5^XW01>__2DJ_VB0fpl01>|`x~L8SA~`d!01?0{9e=eB
z03ugO+5i#3D!Ql+03vge1OO4jDjl^B03s(g3IGwqDh9O<03tG!1OO4lDh9O<07sA|
zY5*b&=m-H3)G7wG4gfw(bZu||5!WgPwGIG2BF>;<01?+J2DJ_V0000-uqA2$BI|oU
z01?|N2DJ_VJ|fPbVgM1_Dh9O<0Dk}gN5Um)03yzyVgM1|Dh9O<002kSC29a7&Y)rd
z5#TBYwGIFPM*=2l03r+M2mulBDh9O<06t7~ZEye)^C||l4gfwP&Y)rd5%VerwGIFP
z002iKCTaj8>w7-{5%($vwGIG2BF>;<01@{p2DJ_V002icCTaj8&Y)rd5r6tB2DJ_V
z07ph9Y5*e6pke?K`zi*t4gdg0SSD%!BF>;<01*!>2DJ_V07r%<Y5*b&=m-H3A1em6
z4gfw(bZu||5h5!FwGIG2BF>;<01+Z92DJ_V0000-v?gi*B3V>Q01+rFzDQ{RJ|gRT
zKL8ObD+aX=06rqlpke?KDt{{mwGIFP002kgCTaj8SyV~@5il#hNNE5*BI|oU01+}P
z2DJ_VJ|fPbVgL~`D+aX=00000NAxCY03yzyVgM0DD+aX=002h@Cu#s9&Y)rd5k@Np
zwGIFPM;Iq+03yzyVgM0GD+aX=002iuCu#s93+M;|5mqY(wGIG2On-E3Z~zfmD+aX=
z06rqlpke?KSt|y$4gdfE07qsgY5*eZdp`gXT`LB)4gfwP&Y)rd5nU?=wGIFP07rNy
zY5*e6pke?KU@Hc-4gdg0iYICSBF>;<01<5~2DJ_V07slBY5*e6pke?KdMgIC4gdg0
zt|w{$BF>;<01<sF27k2<002k8Cu#s9&Y)rd5r8WOwGIFPN75&103yzyVgM0=D+aX=
z002h-C~5#A3+M;|5s)hewGIG2OmuB<01=Zb2DJ_VJ|fPbVgM17D+aX=00000M<6I_
z03z#qKL8PzD+aX=06rqlpke?Kmn#Of4gdfEM=~gC03yzyVt)V;nkxpi4gdg0L?~(i
zBF>;<01=xj2DJ_V07q6RY5*e6pke?KrYi=u4gdg0geYnNA`9pU0THz;2DJ_VK1_6N
zZ~zgwD+aX=06rqlpke?KxGM&=4gdfE07s-KY5*eZdp`gXyekH^4gfwP&Y)rd5xgq~
zwGIFP07tecYFGdw&Y)rd5x*-2wGIFPN608@03yzyVgM1qD+aX=002kYC~5#A&Y)rd
z5z8wEwGIFPM+hlu03r+M2muk>D+aX=06t7~ZEye)-YW*R4gfwP&Y)rd5#E=vB!Uxv
z<SV{NX#hSV>w7-{5#}ofwGIG2BF>;<01@UZ2DJ_V0000-WGQL@B3V>Q01@mfzDQ{R
zJ|gRTKL8Q#D+aX=06rqlpke?K?kfhh4gdfE07rHyY5*e6pke?K04xTz4gdg0h$(6S
zBF>;<01*N#2DJ_V07sfBY5*e6pke?n5d$m+wGIFPN60B^03r+M2muiiEC#g>06t7~
zZEye)6)Xm|4gfwP&Y)rd5fztlC4v)w87v014gfwP&Y)rd5g9B7wGIFP07v>MY5*e6
zpke?K94rR44gdg03@T~>BF>;<01+K52DJ_V07o7wY5*e6pke?KG%N<S4gdg0Fe+*Q
zBF>;<01-Vb2DJ_V07pV9Y5*dUynFx=R4fLy4gdg0a4Kp5OmuB<01;g*2DJ`<06rq?
zdp`gXT`UH*4gfwP&Y)rd5nU_>wGIFP002jVDrx{C&Y)rd5o;_4wGIFPN0cgR03yzy
zVgM0tEC#g>002j(Drx{C&Y)rd5pgUAwGIFPN4P3#03yzyVgM0xEC#g>002kIDrx{C
z&Y)rd5s54YwGIFPN8BoE03yzRpke?Kk}L+b4gdg0?kZ{kB9Xj&01>Gy2DJ_V07o1v
zY5+`hZEye)vMdI*4gfwP>w7-{5wa`>wGIG2BF>;<01>h*2DJ_V0000-N-JsrA`9pU
z0TIJ22DJ_VK1_6NZ~zg;EC#g>06rqlpke?K#w-T44gdfE07rN$Y5*c2SyV~@5zH*U
zNNE5*BI|oU01?hC2DJ_VJ|fPbVgM1&mmw#D6Mxh!zDQ{RJ|gRTKL8QdEC#g>06rql
zpke?K)+`3K4gdfE07tkhY5*e6pke?K=qv`c4gdg0$}4IBBF>;<01@ge2DJ_V07u*_
zY5*e6pke?K>nsMf4gdg0?kj2lBF>;<01@>p2DJ_V07n2UY5*e6pke?K|11Wz4gdg0
z5`Qde03yzyVgL~eEe5p?002iMENTEE&Y)rd5fv>4wGIFPM?@@Y03z#qKL8OPEe5p?
z06rqlpke?K9xVp74gdfEM`SE&03z#qKL8OeEe5p?06rqlpke?KEiDGM4gdfEM}#bD
z03z#qKL8OfEe5p?06rqlpke?KE-ePN4g~-J07s;k_a}lM0mPRVD1sjW<d-)nf*%3|
zEtgm*f*wF5EouNF>w7-{5jHIbwGIG2BF>;<01-AV2DJ_V002isEouNF>w7-{5jQOc
zwGIG2BF>;<01-DW2DJ_V002Ihx+sDh2>_r148HsT5k{99DS{z?Mk)ri4glI0G5}mY
z7-IktMk)ri4gek*G5}lz1RnqqMk)ri4ge7wG61##!<(T15vwZ9>;(V-0002C0d`8L
z01>AuQ|tu*00000wgGu3O#l(4D&6Y^0000007AG&Q2-G~Dh9O<0DK%W07AK)PXG}}
zDh9O<06`rx0761WBZvSINh$`l4gdxoG5|uetkeJzN-74m4gkv@G5|sc=MMl8ODYDn
z4gioJG5|spbT|MJOezMo4gggkG5~=B39g?25lxmVf)zYbDveHR00{$NxHkX*0000}
zvA#M05mG9jPHF%Hk$`_Cz-0gcLL%e~0TEUzx~L8S8X+<OLUU(P01;Q0Q7VEY0fLuy
zDuOuyh?lS`f;0hrm*pygv;lXQnk#~|0dcoaEP|N<0j8JfErQGethcr<g53cD-?x-7
zf_DM|F}Dmef@lT-?YDI~g4zNB)wfVQf;RyHtG9?ff`9=5R<{X1f)@(`?w27-f=GYZ
zCI+<*00B^1Y5*TSIcfkA*d`JYX#gMA0sa6H*d_+G4gepN_TvB%*d_+G4geoQBZvSI
z*d_+G4gepstkeJz*d_+G4genq=MMl8*d_+G4gen%bT|MJ*d_+G4giY*hr{;(5icr>
z0UG+v01+lCBD8Vy01??H2DJ_VA~$~l3jq<@CI+<*03!MTlmHRiCI+<*03y<1ng9{p
zCI+<*03xH|;{XxeCI+<*03saP!2l87CI+<*03r#lp8yfxCI+<*03v-z+5i#ZCI+<*
z03wG-+5i#aCI+<*03wk{+5i#bCI+<*03w@6+5i#cCOQBB03xMG+5i#dCc1y94gex?
zGOqv;<|ZAr4gexPN!kDr=OzZV4gexKGp_&<=q5S<001IaN!kDr=_UrX4gex^k^}$|
z>L$9V4gex2HVOa{>n0tw4gexDlmq|~>?Q`a4gg0xP-*}o3+M;|5&R|wwGIG2OmuB<
z01^Kt2DJ_VJ|fPbVgM2UCI)}C4gdfE07qL;Y5*eZdp`gX11AQx4gfwP&Y)rd5d$X%
zwGIFP07q?5Y5*e6pke?K1}6r!4gdg0e^6=wBF>;<01*f$2DJ_V07tD*Y5*b&=m-H3
z7$*j`4gfw(bZu||5gR84wGIG2BF>;<01+D}2DJ_V0000-%}{CpBI|#9KL8OQCkC|+
z06rqlpke?KA14O24gdfEN8eCt03yzyVgL~$CkC|+002kvP-*}o&Y)rd5hEuCwGIFP
zM*~r603yzyVgM00CkC|+002i7QEC7pk-U5W5nd+-wGIFPM?+C+08DgkZ~zf!CkC|+
z06rq?dp`gXXD0@=4gh~XBF>;<01;;=2DJ_V0000-aZzdjA`9pU0TFm72DJ_VK1_6N
zZ~zf|CkC|+06rqlpke?KdnX394gdfE07suuY5*cxR7wC5fhWF5X#hSV>w7-{5rZcN
zwGIG2BF>;<01<;H2DJ_V0000-%~5IqB3V>Q01=5NzDQ{RJ|chXdp`gXizf!P4gfwP
z&Y)rd5sN1VwGIFP002kdQEC7p&Y)rd5uGOnwGIFPNAXc=03yzyVgM1JCkC|+002h=
zQfdGq&Y)rd5uhgqwGIFPM-@_P03yzyVgM1UCkC|+002irQfdGq3+M;|5xFM@wGIG2
zOmuB<01><=2DN_<06rqlpke?Kye9^=4gdfE07qj|Y5*eZdp`gXz$XT^4gfwP&Y)rd
z5x^$~wGIFP07rFFY5*e6pke?K!zTu{4gdg0hf-<)BF>;<01?n92DJ_V07scpY5*e6
zpke?K-6saM4gdg0t5RwJBF>;<01@CP2DJ_V07t!2Y5;#C&Y)rd5#uKYwGIFPN6%7f
z03yzyVgM23CkC|+002k*QfdGq3+M;|5%VVowGIG2OmuB<01@^l2DJ_VJ|fPbVgM2L
zCkC|+00000M;%jY03z#qKL8Q>CkC|+06rqlpke?K`X>gp4gdfEM=w)q03yzyVgM2S
zCkC|+004hSK~rh~BF>;<01^Ht2DJ_V07p|(Y5*e6pke?K2`C1&4gdg0fm3P#A`9pU
z0TCD|2DJ_VK1_6NZ~ze-C<e6-06rqlpke?K8z=^~4gdfE07s!yY5*eZdp`gXA1DU3
z4gfwP&Y)rd5g#Z9wGIFP07tV^Y5*e6pke?KA}D_bwGIFPN5xZW03yzyVgL~%C<e6-
z002kVQ)&Pr&Y)rd5iTeOwGIFPM+H=B03r+M2muj3C<e6-06t7~ZEye)K_~{b4gfwP
z&Y)rd5kV*hwGIFP002ibRB8YsSyV~@5l1M#NNE5*BI|oU01-(j2DJ_VJ|fPbVgM0I
zC<cGE4gdfE07qd|Y5*cxR7wC5Pbj`fX#hSV>w7-{5m6`xwGIG2BF>;<01;6r2DJ_V
z0000-b5v>oBF>;<01;y-2DJ_V07r#XY5*e6pke?KWhe%<4gdg0msDy1BF>;<01;*=
z2DJ_V07u1CY5*b&=m-H3btne44gfw(bZvid01<d72DJ_VJ|fPbVgM0%C<e6-00000
zN99y%03z#qKL8PYC<e6-06rqlpke?Kd?*IB4gdfENB2}}03yzyVgM0;C<e6-002h`
zRcZht&Y)rd5r8NLwGIFPM;ldY03yzyVgM1jC<e6-002iVRcZht&Y)rd5yB`2wGMv(
z07pMnY5*dUynFx=*(e6J4gdg0ZB=RjOmuB<01@IS2DJ_VJ|gRTKL8QpC<e6-06rql
zpke?K;wT2S4gdfE07rjSY5*e6pke?K@F)hg4gdg0kyUB{BF>;<01@*j2DJ_V07s)$
zY5*e6pke?K_9zCm4gdg0wN+{WBF=xHVgM2MC<e6-002kFRcZht&Y)rd5e_K^wGIFP
zN7+?s03yzyVgL~rDF(F;002kpRcZhtk-U5W5iThPwGIFPM;TUX08DgkZ~zfEDF(F;
z06rq?dp`gXHz@|S4gfwP&Y)rd5jQCYwGIFP002ivR%!qu3+M;|5lATpwGMv(K1_6N
zZ~zfYDF(F;06rqlpke?KODP7m4gdfE07rFJY5*cxR7wC5Q7OJiX#hSV>w7-{5mPA!
zwGIG2BF>;<01;Cu2DJ_V0000-qgHAFB3V>Q01;U!zDQ{RJ|gRTKL8P1DF(F;06rql
zpke?KTPX&$4gdfE07tb}Y5;#C&Y)rd5p5|3wGIFPN5@ub03yzyVgM0uDF(F;002kX
zR%!qu&Y)rd5pXF6wGIFPM+aAG03r+M2muj*DF(F;06t7~ZEye)gDD2J4gfwP&Y)rd
z5rZiPwGIFP002idS84zvSyV~@5s4|jNNE5*BI|oU01=BR2DJ_VJ|ch4pke?Kizx=R
z4gdfE07qk2Y5*cxR7wC5ktx1NX#hSV>w7-{5tAtfwGIG2BF>;<01=ZZ2DJ_V0000-
zbysQtBF>;<01>4r2DJ_V07r*cY5*e6pke?Krzr-t4gdg0nOAB6BF>;<01>Mx2DJ_V
z07t7=Y5*e6pke?KtSNs6wGIFPN7+|u03r+M2mukiDF(F;06t7~ZEye)zbOW_4gfwP
z&Y)rd5x*%0wGIFP002h^SZV+wSyV~@5ydIKNNE5*BI|oU01?M22DJ_VJ|fPbVgM1x
zDF(F;00000M>SY#03umbN&pefDZWT)06rq?dp`gX&nX7A4gh~XBF>;<01?kA2DJ_V
z0000-M_6hABF>;<01@FS2DJ_V07qF^Y5*e6pke?K<0%HU4gdg0YglRkBF>;<01@XY
z2DJ_V07rdTY5*e6pke?K=qU!Z4gdg0k63B|BF>;<01@&j2DJ_V07s!%Y5*e6pke?K
z{3!;t4gdg0vsiy>03yzyVgL~bDh9O<002kRSZV+w>w7-{5fCZ{wGIG2BF>;<01*%>
z2DJ_V002kxSZV+w>w7-{5hp4JwGIG2BF>;<01+oD2DJ_V002i5S!w_x>w7-{5hyAK
zwGIG2BF>;<01+rE2DJ_V002ibS!w_x>w7-{5h*GLwGNk*RDux!Q<tVxf*%2Mm&R0r
z9|4n>=2U_oL9<zE03z#qKL8OhDh9O<06rqlpke?KFe(PM4gdfEN7Gqq03z#qKL8Oi
zDh9O<06rqlpke?KF)9YN4gdfENAs6PRe~P^6Iz#NRe~M@Gg_C2Re~J>K9{Lgf*T0{
zpaPQ>WB?IumkCyaA%ATq2DJ_VzJ4+QTs|0M01<5_2DJ_Vkclz?Tm%Fk01<5_2DJ_V
zB#SZtwgJPNp#TxkCV_PZ000000JZ^kN~i!4%q9eN1^@s6006cDc_&Q(5y~czbOrzb
z0000&)&c$i5pE_1wGIG}j4}X1l=kBQ5pO02wGIGOjWPg2LPjHq01<E|2DJ_V8ICdl
zLbI&Y01<H}2DJ_V-i|T=LI>v$01<K~2DJ_VqmMEGLKJj301<O02DJ_VXpk}hfdUDx
zp8yebm+w}B6+Cz*jZSI+2?L<GHvj+t002|5zB&LAc_yDuY5)U~fPW=mf&c(QBIFAJ
z5qu`Ps15)vkum^6b7xTi5q+0HSArw~u9sz3f;0iCmzr0Cv;m`+PgsJq0iBlwS%S0y
zm$$iDg0TSs%a?0gg31BWw>DgY_yGY2w=Z6T(E$NqmjPgcXaS11YG8u#1_4mFTxWvH
z0s&gLCu)LG0Rb?#d252~0RfV?W^IBe0s+pqj&Fju0Rh{$Msb3e4FR^7-+qE5e=Z~j
zwGIFQ7K~~DA3Zs001+-E5)o+t9|$*C01+-E2DJ_VAMvoV01+-E2DJ_VA3`ID01+-E
z2DJ_VAG55~01+-E2DJ_V9|z|T01+-E2DJ_V9~5*r01+-E2DJ_Vivfqj_W%)CCW`?Y
z`pp0lQ6?gz;o|@iFC+%F4gew?f7-zS5ileMwGIFx`T&#w5iukNwGIFx(qWna5i%qO
zwGIFx39g?25i=wPwGIFxeMs5>5j7+RwGIFxhe+B05jG?SwGIFxkx1GA5jP|TwGIFx
zn@HLK5jZ3|0000YrAXQU5ji9}0000YaWby}5jrFWwGIGBb%bgFA`9pUe*qCnBnGt(
z06t7~ZEye)O(X`j4gfwP&Y)rd5ltipwGIFP002joglYgH>w7-{5m6)twGIG2BF>;<
z01;6n2DJ_V002j)glYgH&Y)rd5mY1wwGIFPN4bP*03yzyVgM0VBnGt(002klglYgH
z3+M;|5oRO?wGIG2OmuB<e*h6_BnGt(06rqlpke?KX(R@<4gdfE07nOfY5*eZdp`gX
zZ6pS@4gfwP&Y)rd5p5&}wGIFP07n^xY5*e6pke?Ka3lt`4gdg0D}`zRBF>;<01<H{
z2DJ_V07qGcY5*b&=m-H3iX;ZL4gfw(bZu||5sf4UwGIG2BF>;<e*h7UBnGt(00000
zM~8)K03umbN&pd)B)&*#06rq?dp`gXl_UnW4gfwP&Y)rd5tSqcwGIFP002j|g=zpI
zSyV~@5t}5wNNE5*BI|oU01=%e2DJ_VJ|fPbVgM1HBnGt(00000N5_R~03yzyVgM1Z
zBnGt(002kXg=zpIf6ky{01>ey2DJ_V07nOgY5*b&=m-H3!z2c^4gfw(bZu||5ym72
zwGIG2BF>;<01?I{2DJ_V0000-HHK;cB3V>Q01?b2zDQ{RJ|gRTKL8QVBnGt(06rql
zpke?K&Ljr44gdfE07qknY5*cxR7wC5)Fi%0X#hSV>w7-{e-YLs2DJ_VJ|fPbVgM1=
zBnGt(00000M|Fm303yzyVgM27BnGt(002jahH3yJ&Y)rd5$YrcwGIFPN417(03r+M
z2mulMBnGt(06t7~ZEye){v-yq4gfwP&Y)rd5&k3wwGIFP002kThH3yJ>w7-{5dtL!
zwGIG2BF>;<e*h5zB?h$)0000-=Z0zkBF>;<01*Wx2DJ_V07v<TY5*e6pke?K1|<fy
z4gdg0Cx>bPA`9pU0TC4?2DJ_VK1_6NZ~ze)B?h$)06rqlpke?K7$pX^4gdfE07plM
zY5*eZdp`gX93=*|4gfwP&Y)rd5ga83wGIFP07qGee`)|C&Y)rd5g#Q6wGIFPM{9>_
z03yzyVgL~!B?h$)002jshiU*K3+M;|5ilhNwGIG2OmuB<01-1K2DJ_VJ|fPbVgL~{
zB?h$)00000N5_Y103umbN&pc#CB8^$06rq?dp`gXJ0%9S4gfwP&Y)rd5j!OYwGIFP
z002kze}`%SB3V>Q01-hYzDQ{RJ|gRTKL8O!B?h$)06rqlpke?KLnQ{a4gdfE07nOi
zY5*e6pke?KRV4<s4gdg08Hj2CBF>;<01;Ou2DJ_V07ol`Y5*e6pke?KSS1Fv4gdg0
zS%_)?A`9pU0TE{<2DJ_VK1_6NZ~zf%B?h$)e*iur&Y)rd5o#p{wGIFP002jMh-v^L
z>w7-{5pE?0wGIG2BF>;<01<8_2DJ_V002jeh-v^L&Y)rd5pg93wGIFPN1cdj03yzy
zVgM0xB?h$)002kJh-v^L3+M;|5tt<gwGIG2OmuB<01=xd2DJ_VJ|fPbVgM1FB?h$)
ze*gdg07v<VY5*cxR7wC5p(VaZX#hSV>w7-{5u+srwGIG2BF>;<01=}l2DJ_V0000-
zCy8nRB3V>Q01>GrzDQ{RJ|gRTKL8P{B?h$)06rqlpke?Kt0e}t4gdfE07p5AY5*e6
zpke?Ky(I><4gdg0ONnX#BF>;<01>|>e+IP<002i_iE02M&Y)rd5x^w|wGIFPM~jJS
z03r+M2muk$B?h$)06t7~ZEye)(<KJA4gfwP&Y)rd5z{3GwGIFP002k1iE02MSyV~@
z5!ofaNNE5*BI|oU01?|I2DJ_VJ|fPbVgM1_B?h$)00000N9T!Z03umbN&peze<i+1
zX#hSV>w7-{5#uEWwGIG2BF>;<01@LQ2DJ_V0000-`H5-(BF>;<01@>i2DJ_V07nao
zY5*e6pke?K_az3k4gdg0D~f6WOmuB<01@~l2DJ_VJ|fPbVgM2NB?h$)0000-S&C`^
zA`9pU0TBr%2DJ_VK1_6NZ~zeue<lXC4gfwP&Y)rd5ey~<wGIFP002jaifRBNSyV~@
z5fUc8NNE5*BI|oU01*@>2DJ_VJ|fPbVgL~oCI+<*00000N41J-03umbN&pcWCca2%
z06rq?dp`gX93}>}4gfwP&Y)rd5gaB4wGIFP002kFifRBN&Y)rd5iTYMf3*$(07u!1
zY5*e6pke?KFeV1I4gdg0`HE@)OmuB<01+`J2DJ_VJ|fPbVgL~_CI+<*0000-8H;KF
zBI|oU01-PT2DJ_VJ|fPbVgM04CI+<*0000-Ig4rlBI|oU01;9q2DJ_VJ|fPbVgM0R
zCI+<*0000-S&M1_BI|oU82}MeCI+<*06rqlpke?KQzizr4gdfEM|qdbhJqgfnV0N_
zf*(P-i)sKO>w7-{5m_b%wGIG2BF>;<01;Ux2DJ_V002kXi)sKO>w7-{5n3h&wGIG2
zBF>;<01;Xy2DJ_V002k%mrRF(9sxd=Z-;^#2>_r1uS);{5yO|yhk_x0!y^W@4ggyc
zGXPvZ7-Ikt!y^W@4gj|lGXPu!1Rnqq!y^W@4giN1GXS;$!<(T15hf(D5C#AM0002C
z0d`8L01+Z2JP-x|00000wgGu3O#l%fB*+g200000073{iSO5{kBL=k&07Muw07CJw
zvH%gqBL=k&05KUe07624BZvSI#v=x`4gek+GXO%ftkeJz$0G){4gd=qGXO#d=MMl8
z$Rh@|4gmNYGXO#qbT|MJ$s-1}4glpGGXQ}C39g?25y~S5wGIFPUNCC_0BMnG08=aI
znE(;aBaKdK00{%&xHkX*0000}vA#M05zix^PHF%Hk$`_C;A;RF074?<3jq<*Bf6*#
z0MZ>Z077$TQ2-Ismj;P~Bmo4MD~W<M0{<hIVTppX0r{5|ih{HO^Owqsg0uncw||R*
zu>k=gmobfk$^j?0{EdRJ0RdIFt&f5>0s)-2<dK4@0Rh#wj+27*0|7#}gqMQ)0s*SG
zf0}{`0s-Q;iJXE00s$MhjGltk2LXzgVy1#50q>WGrh*}V?;#QqX#h5GGXPwp;o|@i
z?;!@Y4gl|PGXPu)${zp`?;!@Y4gkb-GXPws^xFUt?;!@Y4gmaiGXS;$hr{;(5w9a;
zxCQ_K0001s0UG+v01=@hLR4s201@yZ2DJ_VG<P!qLJ`^E01@#a2DJ_V@pv-;LV~ta
z01@&b2DJ`<0L^<d0740_p8yf_AqKS$0Kt7T07898+5i#tAqKS$07rf^078dI+5i#u
zAqKS$02hNZ078*S+5i#vAqKS$0MUmt079Ec+5i#wAqKS$07{E9079im+5i#xAvyp6
z056a;077vxuK*GHAqKS$08x@N075=V+5i#zA-br44gma^GXO$4Gp_&<{2?8+4gh(Z
zGXO$YN!kDr{UHXm4gmU{GXO$!k^}$|{vign4ghJOGXO#-HVOa{{~-po4gl_?GXO#|
zlmq|~03rsp4gi0qGXO#_KM(*B0U`#q4ge6SGXO%!lmq|~0wM;r4gdkFGXQpLtOfu8
z0001&C8&ZD0Tq`xsDdB?7?)Y7f*=7HmwKpzD>fb?zDQ{R1CfA#M;@(e03z#qKL8OR
zA_lb%06rqlpke?KAR-2}4gdfEM=-5w03yzyVgL~$m)od<H~}Y@6sdwF0V$U`se&K@
zDVJKQf;IswmyxN0ECDE&zNvz40X&y2s)98EK9^&vf;9nCmzk=9DIIDe2DJ_V@vSoe
z761@v0N$?#0000008oqo01<4L->QOJ0dto-tAZc_bC+DJf*=8Nmwl^(Hvxs0wX1?P
z0gIRGtAZc_j+YFqf*=8omo2P<D*=?3S*(I20hX73tb!l`mY1Kbf;It|m)5L;B>|b2
z{H%gJ0iKsRt%4*0pqE~)f*=8)mw&B-HUXlSwylCC0i&1Pt%526pO*)&f^Gq~mv*j#
zH37MotgeDJ0lJssu7V{2#Fq%Kf;9oxmp-q8DIM-22DJ_Vkg_uX761@v0EM&$00000
z0HBQk01@z)gs*~IBKIN&wGIF>voioI@~{Q~000006#xJL00000LhE}!01@|>0kDD~
z0r!_8u!1)M2bWr~f;9mQmyoc6DJvKw2DJ_VR<tt!CE$(#0AX}(Z~ze-BL=k&0F1OV
z043m#002VHpke?K8<*s;f-3<bmk_an6f+|u2DJ_VJ|fPbVgL~%BL=k&0000-nzd>G
zBF>;<01+i42DJ_V07tB~Y5*e6pke?KCYM*Sf<OT)mzc4FBmpg#y|IEI0WFu_v4SQ6
zFP8<ff+Ybkmn*V@E&(i;TC#$20YR6|vVu1OLze}!f;9n2mpik9AOTF5T(g280Zf;D
zvw}BTQX`*EY5)U~fPY6VxM~0*>w7-{5mX}vwGIG2BF>;<01;Fp2DJ_V002inxM~0*
z&Y)rd5mqAxwGIFPM^U(H03yzyVgM0WBL=k&002j0xM~0*&Y)rd5m=WFw1PPSU6)9-
zf+PW7mu<9yAOT;Oj<kX{0b!TLw1Oo8VwdZ*f+YcCml3ssECF7ZKDB~!0ezRTwSqSR
zf0yaCf;9nvmmju*H6)572DJ_V1iCW-07uTcY5*eZdp`gXlOqPT4gfwP&Y)rd5tEmA
zwt^x7qL-Vtf*=8+m%O%u8Udr1*S3Nn0i&1rwt_YRq?a$Zf*=8;mr%EYCIO|FcejEd
z0i~Ciw}LhSrkBsRf*=8=m+-fOHUX!XC%A$j0jHNpxPmwVsF#7bf*=8?m!i0WCIP9J
z$+&_b0jZbjxPmw$sv`!q4gj>gGXN{*w*~+J0000U000000000&&Y)rd5vrF&xq>(W
ztCxMbf*=8_m!7$TI03Ae*SUfq0j!t!xq>zUu$M2of*=8~mr%NbI03PjiMoOy0kM~=
zx`H+VvX|kyf*=91mjb(jEdhhKFuQ_M0|BL%gT8_xf2AP?wGIF=zcT<_J{V&F5v3sp
zwGIFkz%u|`1Oy)d5v3spwGIIJz%u~00mGZ201@XQWXc8r00000wgGlZr~nb;AvDSc
z000000JZ^nCrtnm;UU1u1^@s6002Ux;o|@irXdEk4gll9GXO#g${zp`ry&Nl4gfF0
zGXO%SF7(>~5vU;swGIG$!ZQFu39g?25vd^twGIHt!ZQF~P>=uslf`NPQ!D7101>U1
zQNMx}0j`&4zk(nUu_3ys4gdhdGXO$!XHftVvX_{@f+PXUm%YD&Gy%r9^1p&;0RiHd
zTfu_c0d$u)!h$@1A0P&`4gdk(%xVB1JvnLs5g#BD5orJ)QjDJf5g#B1wGIFu3d$b<
z5g#B1wGIFuM=F8<5g#B1wGIFudRGtt5g#B1wGIFuru5qY5g#B1wGIG_0f)o)01=TP
zivb$?%>WUGAtF?0SO5_qAO^J#03s3D-~bUJAO^J#03w2awo?ERA|M8}4gewvuAcxA
zBOnH~4gey3NZJ4qB_Ia14geyDNZJ4qCLjj24geyNNZJ4qCm;s34geyXNZJ4qC?Glj
z001JTNZJ4qDIf;54gex?GOqv;Dj>S34gexPN!kDrD<B=U4gexKGp_&<EFcE84gexo
zN!kDrEg%MewGIFxbCLuA5iTGGwGIFxCpHQI5icMHwGIGB3CL;yA`9pU0TDnT2DJ_V
zK1_6NZ~zfQAO^J#06rqlpke?KLm&pV4gdfE07o~-Y5*cxR7wC5Ng%#RX#hSV>w7-{
z5lbKjwGIG2BF>;<01-<d2DJ_V0000-Wyop(B3V>_N&pd2AihXx06rq?dp`gXQy>Pl
z4gfwP&Y)rd5mO)rwGIFP002jK$Z7y0&Y)rd5oI6-wGIFPM~TR403yzyVgM0mAO^J#
z002ju$Z7y0k-U5W5s)ATwGIFPN6E-)08DgkZ~zgRAO^J#06rq?dp`gXnIHzW4gfwP
z&Y)s{01=rW2DJ_V0000-_sD7hA`9pU0THes2DJ_VK1_6NZ~zgpAO^J#06rqlpke?K
zu^<Mu4gdfE07ps5Y5*cxR7wC5w;*CsX#htk$!Y*1>w7-{5xF1+wGIG2BF>;<01>$$
z2DJ_V002ii$!Y*1&Y)rd5xXD;wGIFPJ|fP4pke?KydVa(4gdfEN1MrN03umbN&peT
zAYxHz07rVsY5*eZdp`gX!ypE=4gfwP&Y)rd5yKz`wGIFP07s0;Y5*e6pke?K#UKW?
z4gdf?BF>;<01?I@2DJ_V002j=$!Y*1&Y)rd5!xUIwGIFPN59Ex03yzyVgM1`AO^L6
z4gdg0(aCB6BF>;<01@3F2DJ_V07v7=Y5*e6pke?K=O6~P4gdg0^~q`gB9Xj&01*x$
z2DJ_V07oOrY5+`hZEye)7a<0<4gfwP>w7-{5f>o_wGIG2BF>;<01+1<2DJ_V0000-
zQOarnA`9pU0TC`C2DJ_VK1_6NZ~zg1F(C%E4gfwP&Y)rd5iubKwGIFP002j|%4z^2
zSyV~@5jP>9PHF&0g34+DBI|oU01-JM2DJ_VJ|fPbVgM02AqKS$0000-l*(!VBF>;<
z01-PO2DJ_V07s_EY5*e6pke?KJRt_P4gdf?BF>;<01-VQ2DJ_V002iB%W43BB3V>Q
z01-kVpH6B3N9M|E03z#qKL8O#AqKS$06rqlpke?KL?H&X4gdfENBGKW03yzyVgM0F
zAqKS$002h{%W423&Y)rd5l0~gwGIFPJ|fPbVgM0HAqKS$0000-E6ZvCBF>;<01;v#
z2DJ_V07pH`Y5*e6pke?KWFZEBwGIFPM^DRY03yzyVgM0kAqKS$002i}%W423&Y)rd
z5pN*|wGIFPM}f;~03z#qKL8PUAqKS$06rqlpke?Kcp(P04gdfEN1@AV03z#qKL8Pj
zAqKS$06rqlpke?Kham>F4gdfEN5RW#03z#qKL8PkAqKS$06rqlpke@j5r`oMwGIFP
z07v1=Y5*eZdp`gXi6I8H4gfwP&Y)rd5s4uNwGIFP07n7LY5*eZdp`gXiXjHI4gfwP
z&Y)rd5sD!OwGIFP07oIrY5*eZdp`gXiy;QJ4gfwP&Y)rd5sM)PwGIFP07pU0Y5*eZ
zdp`gXj3EZK4gfwP&Y)s{01=EK2DJ_V002i}%xVB4>w7-{5se`RwGIG2BF>;<01=HL
z2DJ_V002jU%xVB4>w7-{5so1SwGIG2BF>;<01=KM2DJ_V002j!%xVB4>w7-{5t1PW
zwGIG2BF>;<01=WQ2DJ_V002k9%xVB4>w7-{5tAVXwGIG2BF>;NVgM17AqKS$0000!
zBI|oU01=cS2DJ_VJ|fPbVgM18AqKS$00000paPNB9RLx(mzT(bA%DLg2DJ_VJjOEs
zTs|0M01>|*2DJ_V*T*veTm%Fk01>|*2DJ_VLdi1#wgJPNp#TvVARGAx000000JZ^k
zN~i!46ChLo2LJ#7006cDc_&Q(5fLEZ{{{d60000&QjDJf5x^e?wGIH*$}<2$3d$b<
z5y2k@wGIHA%QFB%M}I1U01?6;2DJ_VV9YZBLV8ya01?9<2DJ_VCCxJcLZ<ZF01?C=
z2DJ_V>di9%fdUDxp8yfX9|pA!003Uls{jC0(P{uwE9jX35y>A_;%Wd113<Yq00000
z08_EPIsg&MA9PV^00WVLe<g6N002TF<O=~2%^$j`4gjytGY9}eb7xTi5zd#s$$}&S
z_m|_zf;0j0ml(=|v;pjw&B}tb0p_=Y%Yu~w0T{O!&4Q@`0aCZZ&Vo4s0g{*e&w^+H
z!nbA6g6IJOqqk1df<*xV5SNtGf@lFlxBb(C4+#Mhmkr&5+W}aY>D_`he}o?fwGIFQ
z5aVh9A3Zs001<>A5)o+t9~s;#01<>A2DJ_Vivfqj_W%*DABzDR`pp0lsvjZ=uAcxA
zg&zjB4gey3NZJ4qhad351pp$4NZJ4qh#&C61pp$ENZJ4qi5~{F4gg2<;%WdQ>w7-{
z5tAPVwGIG2BF>;<01=ZPD+aX=0000!BI|oU01>Mn2DJ_VJ|fPbVgM1V9|pA!00000
z!vWYbP5=>km&x9OAt8Am2DJ_V3Jo*>Tp8Rd01<f~2DJ_VY7I01LJ6*)01<kZ3*Ukp
z0ezPz--2!dl$Vy@f`I`zml)uJXaPExf#8B<0YI1L;DTlWYL`vng2MrOw-MrkKmq}K
zmr~?{{Q-QpQ00QN0s(xN1n7cF0erVl=z;+O0erVn>4Mk+0YH~3>w^9PIJYV6f|&vV
zK)1v0g1`X*hnIx#f@lGVm+kO^XaSVBR`G(H1Ob${^7Vq10s)q{i1>m<2LYD1&i{g&
z1Ob+}Y6OFs0s)q{1P6na0s)(snhAq+0r{5?3WF{I{g*xpgDL_2muU)vcmczg;tGQ*
z0m-)w3xjn50{>38gbagP0Rf1&-wlJA0s)7&c@Tqg1Ajgr2DJ_V0o)jC03SU$Y5);F
z9}*F103R9LDgY5a9|pA!0E+>K!}kCYXdjCK8v4xu5o8}C39g?25kDUWwGIFxeMs5>
z5kVhL3IhNlhe+B05kenM3IhNlkx1GA5knsawGIGBzZh!(BI|oU01-_e2DJ_VJ|fPb
zVgM0MA1em64gdfEJ|gRTKL8PB9|pA!06rqlpke?KWgiB$4gdfE0H6W|u~GmLGnbVV
zgCRXL9|pA!0N7zP09+Z|DgY5P9|pA!061ba0740_p8ydw9|pA!0NP?S02VVCYXBii
z2mk;8003~00ss*=m-7^ZZUIi0Vikjd0m+x-6@zF2%a=_SgJc2Gm$DXvW&!(`85e_6
z0SuRH7lU~Lo|onqgGT|Kw=Nii9svRPmrWUiUjg>Ftr>&J0Ri^6b{m5!0RfDcpd5q8
z0a}-69fM>6T$jxqgXIBQx3eCDQUU>5w+11D>Hz@^xA7u_Kmq{_mog=T{Q(WPFeZbu
z0s#$|<S2tl0S&h<DT4t40S&h>Dudnu0nwKaEQ9<3$+rwGgP8&W(YK~AgQNigL6=K0
zgJ=Olm$Wg1XaP>Q9WsNP1OZOBxi*890s&CBPdS4{2LVvGl|X}=1OZUDF-3!!0s&CB
z%SVHm0s%s|WlDpY0s%p{|4f681AqD+2DJ_V0cBNd03SU$Y5)=X9ug600E+>K!}kCY
zA|HzZ8v4xu5gs2R39g?25&IqnwGIFxeMs5>5&a&4-wpsGhe+B05&j+qwGIGBNL6b9
zBI|oU01*fu2DJ_VJ|fPbVgL~c9|pA!0000!BI|oU01+P_2DJ_VJ|fPbViN!nA0Gy_
z4gdfE0K)+*TbTe6@0T7?gCQaB9tO1z0Ij_=09-EEVE_^D9tO1z0Oq|k0740_p8yf?
zmtawY8v*l|d{Ki^0S}kYQG=|1vK|Jt4gdj!S8D(tJvnLs5wac<5orJ)qv7KK5wacz
zwGIFu9NNJE5waczwGIG_0f)o)01@6Eivb$?%>WVF9wG^@p8yfF9tO1z03v-z+5i!?
z9zp;B03wG-+5i!@9zp;B03wk{+5i!^9x4C;03w@6+5i!_9tO1z07qv-S8D(w>w7-{
z5yBn@wGIG2BF>;<01?6-2DJ_V002HB>w7-{5!)UHwGIG2BF>;<01?|B2DJ_V0002P
z0a9AQ01<bWJyU}re|H`RwGIF-zcc_`J{V&F5qBO2wGIIPzcc_`1Oy)d5qBO2wGIHr
zz%&510mGZ201>JlZ@vcr00000wgGlZr~nbB9yz}U000000JZ^nCrtnmq#oG52LJ#7
z002Ux;o|@icpe6|4gki%Gyp;z+Q9%3c^(F}4gfpCGys7De+jOi01<i~2DJ_V0A3)H
z0RZ+{YXDO#=$QZ!ejZc@X#fcWaJe@C00000Q?b4}01<y4s!C`81Cg+QC18>P074?<
z3jq;<9=fOw0I<R|077$TQ2-Hw9>%B+09?Z~02&-mXaGyW2LJ#70021v00000000q`
z9zj`a00CN0w_H?%T>$~Bx4Kn>76AdJm(f;(=6^aK2DJ_V0l-{q03SU$Y5);B9ug60
z03V~_;{Xvl9tO1z03RIM!2l6D9tO1z0E+>K!}kCYW*&<H8v4xu5n>)939g?25j!3R
zwGIFxeMs5>5j`G40000Yhe+B05k4M50000Ykx1GA5kDSI3IhNln@HLK5kMXWwGIGB
zqeNV503z#qKL8O*9tO1z06rqlpke?KN*)Ha4gdfEJ|gRTKL8P99tO1z06rqlpke?K
zV;%;z4gdfE0K)<Fig5rD{~eclSc4#c{~ZRk4ggNaGyq&a7-Ikt{~ZRk4giP9Gyq%#
z1Rnqq{~ZRk4ggNcGyt{%!<(T15i%Zg3kU!J0002C0d`8L01+-82@ePW00000wgGu3
zO#l%r9;glo00000079eT;{Xu=9tO1z08Yv@074wv!2l5f9tO1z0I$k4073}`uAcxA
z0+$|HgBV-}9#jWu00{$-xi<g+0000}vA#M05eFWsN@xHBk+6RyKnel?LL%e~0TBrv
zx~L8ST+1{7LUU(P01*lv#;6Ve2+TA98XQk(016QZ000000672v0000001+Iwl39ag
z0Rc0Y_*#R60Y8_0TZ3o;PPgV;gNp<KPPcMkgM|VCQ<wN+gLnZjmvLi*E&()`pksq3
z0XDbCV}on~0YA4iWrJ1&0WFt{XoJB42bWE0gE)V}9R{@y00E9}YXBcTIcfkA!W|M3
zX#gLi;o|@i!W{;+4gen<+Q9%3!W{;+4giY*hr{;(5$+v}0UG+v01@gPA_=aa01?9-
z2DJ_VB7I2O01?F<LI3~&B8N!Y01?I=LI3~&B9Tbi01?L>K0pTmBAZCs01?O?2DJ_V
zM{z@LYXBnadp`gX(j5l14gfwP&Y)rd5z-w7wGIFP06rq?dp`gX>m3HQ4gfwP&Y)rd
z5$hcWwGIFP006@QGBLXV5r>yYYJ(wvhaCpB4gl=mGyq&a7-IkthaCpB4ge+LGyq%#
z1RnqqhaCpB4gl=oGyt{%!<(T15xO17`3C?10002C0d`8L01>twA^!&e00000wgGu3
zO#l(J9l-tv00000079eT;{Xwe9R{@y0PNy4074wv!2l779R{@y0Ab@a073}`uAcxA
zikIwagBU!H9aINt00{%Ixi<g+0000}vA#M05sw|JN@xHBk)VGi0Pq0-LL%e~0TGcM
zx~L8SBIGmxLUU(P01=XxKx=~^0i?HJYlCG00lSyRY=eXW$CpcOgJ=QPx3q18iv$7I
zw?J}(g#rQFm&SC1bOHL8J9UF90{%>wXLW;j0mPTzb%QDa$d?LsgIod1w`X>PUjYLC
zOqa-agDC+3xAJ#`TLA&bw@!J3R|5gHm!^D!!2$J`Xnlh?f3_P2wGIFQv5IQ|A3Zs0
z01>tu5)o+tAEV*p01>tu2DJ_V9~|1j01>tu2DJ_Vivfqj_W%)b9g6`4*2@49q8y6>
z${J<>5uY530g?O!0TFZ@ivd*KegF}19E$-O`pp0lXdNPi76<_mw;R}o0{|kXBnto$
zxEm?}001JHf93B05xE-%wGIFx-Lr835xN@&wGIFx+myWk5xX0@s15)khKEW35xg6^
zs15)k19}4i5xpA*wGIFxK-$3o5xyI`s15)knlQrv5x*ONgaZH~udN^e5x^UOgaZH~
zoIrX25y2Y<wGIFx*mtM^5yBe=wGIFx39g?25yKk>f3*$(B7I2O01?F-LI3~&B8N!Y
z01?I;LI3~&B9Tbi01?L<*oFfDBAZCs01?O=DgXcgBBe;$01?R>2DJ_VB5^XW01?U?
zLx%$ZB0fpl01?X@fP@18A~`d!01?a^2DJ_VB3DV;01?d_2DJ_VB6E@i01?g`x~L8S
zA}2Nqe*h8B8wRxw03tG!1OO4x8#(|003t6x5C9R;8wRxw03yhg1OO4z8#(|003y?r
z1OO4!8wRxw03zO$1OO4#8@i|t03zv>1OO4$8@i|t07uS)YXBn7pke?K4IBow4gdg0
z0EBA*Q}~Go0TC1&#itGc2?GeiHvj+t002{b1+V)65fqo9e}fxM7#s$*4gdg0B!p`K
zBF>;<01+J=2DJ_V07o{2YXBn7pke?KHXH`E4gdg0XoPD3Q}~Go0TDeM#itGc2?O}S
zHvj+t002{buloQIJ(mf9gBxi<90s)x002jfglhmI&Y)rd5k?#awGIFPN4kV-08Dgk
zZ~zfs90s)x06rq?dp`gXUmOOt4gfwP&Y)rd5nmhzwGIFP002kqglhm(_=yJr5n&v~
zrw#xK1L(mw0000008@Og`v4JPmyLjf8-HaS2DJ_V07npoYXBn7pke?KXB-B#4gdg0
zB86)JBF>;<01;^%2DJ_V07o>1YXBn7pke?KZX5=+4gdg0RE28*BI|oU01<B-2DJ_V
zJ|fPbVgM0u90s)x0000-bcJgGBI|oU01<K=2DJ_VJ|fPbVgM0x90s)x0000-hDwEN
z03yzyVgM0$90s)x002j{g=+vzbZu||5se%MwGIG2BI|oU01=HG2DJ_VJ|fPbVgM11
z90s)x00000N9KiV08{vh2LTa}mvMoE6@QN$2DJ_V2?Nl<Hvj+t0000-_=Rf#BF>;<
z01=ZM2DJ_V07nXjYXBn7pke?Kl^h1O4gdg09ENKEBF>;<01=lQ2DJ_V07ou{YXBn7
zpke?KoE!$V4gdg0PKIj$BI|oU01=%W2DJ_VJ|fPbVgM1H90s)x0000-ZiZ_BB7f_9
zKL8P+90s)x06rqlpke?Kpd1FZ4gdfEM}UTF03yzyVgM1P90s)x002j_hHC&*_=yJr
z5w9G@rw#xK1IWQQ0000008@Og`v4KI90s)x00{%Y!8ZT^00000N8pBQ03r+M2mukg
z90s)x06t7~ZEye)y&ML$4gfwP&VQg{01>?$2DJ_V0000-GKXsbB3V>Q01?9+zDQ{R
zM-hi>03z#qKL8QM90s)x06rqlpke?K#T*8;4gdfEM<a)803yzyVgM1x90s)x002HB
z&Y)rd5y%_{wGIFP07ry}YXBlyR7wC5%^bc+X#huKhid>L>w7-{5zib3wSNu(J|fPb
zVgM1(90s)x0000-b%$#JBF>;<01?p~2DJ_V06rqlpke?K(i{f04gdfEN0x_c03yzy
zVgM2290s)x002j*hid>L&Y)rd5#<~PwGIFPN7jdH03r+M2muiX9R{@y06t7~ZEye)
z3LOTu4gfwP&Y)rd5egj!wSNu(0000-D2QtSB3V>Q01*%!zDQ{RM+b;&03z#qKL8OD
z9R{@y06rqlpke?K5*-G$4gdfEM;VA~03yzyVgL~o9R{@y002HB&Y)rd5fvQ<wGIFP
z07rU=YXBlyR7wC58Xdk!X#huAh-&~M>w7-{5gZ)`wGIG2BF>;<0Dlo29R{@y0000-
zYlv$ABF>;<01+M?2DJ_V06rqlpke?K9~}m@4gdfEM~sMT03yzyVgL~^9R{@y002jx
zh-&~M&Y)rd5i%VHwGIFPN3e)%03yzyVgL~{9R{@y002kch-&~M3+M;|5k(yawGIG2
zOmuB<01-$X2DJ_VK7S(4pke?KNF4^X4gdfE07w3aYXBnadp`gXOdSTb4gfwP&Y)rd
z5lkHhwGIFP07npsYXBn7pke?KO&tcc4gdg0B8h7NBF>;<01;3f2DJ_V07o>5YXBn7
zpke?KQ5^=g4gdg0REcW<BI|oU01;Xp2DJ_VJ|fPbVgM0Z9Y_YX4gdfEM|6p603z#q
zKL8PF9R{@y06rqlpke?KX&nZ&4gdfEN0f<c03z#qKL8PG9R{@y06rqlpke?KY8?i(
z4gdfEN3@p*g@Ydf)R!lPgC7C(mq&$z9|9DLmuQ899zisUYXBnadp`gXavcV>4gfwP
z&Y)rd5po>{wGIFP07q1cYXBnadp`gXa~%e?4gfwP&Y)rd5px{|wGIFP07rC}$AyC*
z0hE{Lg@YaeK9>rHgBu9|paKXh2LKU!muiNCA%A-t2DJ_V+5R*DTs|0M01<l|2DJ_V
zqX0DkTm%Fk01<l|2DJ_Vya6=;wgJPNp#TxC8<E!s000000JZ^kN~i!4sv8R92LJ#7
z006cDc_&Q(5vUue;s*c#0000&qv7KK5quj4wGIFd12q6b9NNJE5q%p5wGIHQ12q6b
z2?eg701<wd4~K&oT!I@^2WbEa1JJoQ0000008_EPIsg%a8>&iZ00WVre<g6<0RTcG
z<O=~2g&Vr44giz{H2^|$XHftVh8xDH4ggaHH2@kMPiO!G=m!7*0000v000000000H
zn74z6gJl5$uea%lgM9%3x0h#%gJ=QAm(YrXXaVB4Ig5kC0s-Q;{f&bG0Rab>|Bi#x
z0S}kckApP<9G3}@gNp$$mz9u%(*ZcQX_1410Rd>Y?2?0n0s)q{aFv5-0s+pq*_VT?
z0RbDghM9vR1p&vmsh)!Y0RgwStDl2W0s+Uj{h@;a0RgwS|DuD)1p%p-$f$$C0n(R>
zse?FwJ{tzL4gdkmu4@1vJvnLs5k4Cd5orJ)qv7KK5k4CRwGIFu9NNJE5k4CRwGIG_
z0f)o)01;{%ivb$?%>WT*8zKp=p8yd*8wRxw03v-z+5izj8$tj803wG-+5izk8$tj8
z03wk{+5izl8!7+*03w@6+5izm8wRxw07tGvu4@1y>w7-{5l$NhwGIG2BF>;<01-|b
z2DJ_V002HB>w7-{5oa3)wGIG2BF>;<01;;!2DJ_V0002P0f&Y001*S1gsOufe*+r^
zwGIGkS2X}!J{V&F5d#|rwGIF~STz7#1Oy)d5d#|rwGIFRSv3H*0mGZ201-ADU%CeX
z00000wgGlZr~nZ%8!5X7000000JZ^nCrtnmFdN9a2LJ#7002Ux;o|@i1RDmm4gdmL
zH2^{!+Q9%31sevn4ghyrH2{GE90{(U01*Zo2DJ_V0A65^0RTO*YXDO#=$QZ!3YS2u
zgB2zV8>&iZ00WV*e<h%g0RTcG<O=~24I8?s4gl<0H2^|$XHftV4wrkYgC7APx0|bj
zWdQ*<w*jq#76Ab<ml&>t=6}Q*2DJ_V0Uxw$03SU$Y5)<$8WIs{03V~_;{XxF8V0ov
z03RIM!2l7&8V0ov03Y{O4*(Iw8V0ov0E+>K!}kCY@*0Z)8v4xu5$+ly39g?25yct?
zwGIFxeMs5>5yu)r0000Yhe+B05y%=s0000Ykx1GA5y=`t0000Yn_EcQ01?U>LI3~&
zBBe;$01?X?2DJ_VM*_5K03z#qKL8Qc8V0ov06rqlpke?K)fxu14gdfEJ|gRTKL8Q$
z8V0ov06rqlpke?K?-~ZR4gdfE0K)-s2)F<dewWy<gCT!@8V0ov0E%BV09-y8V*n9;
z8V0ov07zgp09*tF9{>@48V0ov01sg`0JZ_co1p*^yc#gI2LJ#7006cDc1oxK5x5#r
zwg&(J0002C0eL4)01>tt@U;g300000LZjj101<y02DJ_V{9!c!LLAz`01<#12DJ_V
zabh(9LicA@4*(H?8V0ov0On#f0D%GtuAcxAf*J<34gdgNkd6TWZntXyQ!D7101<{7
ztVn492?HRyHvj+t002|5zB&LAhZ?>}X#fL}uzw}sjsXBdBIFAJ5s4bQs15*HV>JLm
zb7xTi5sH`Ju!A1~q?iA&gR}vjx3sZ?eE|Wzmol@1jsf47!Lx(v0k@Z|w1eh<KN<$L
z4gdkOx@!O*JvnLs5kDFd5orJ)qv7KK5kDFRwGIFu9NNJE5kDFRwGIFu_f`)85kDFR
zwGIG_0f)o)01<5(ivb$?%>WT;8X^g<p8yd+8V0ov03v-z+5izk8bSa703wG-+5izl
z8bSa703wk{+5izm8bSa703w@PNZJ4qMH)f?001JTNZJ4qMj8gS4gg1%x@!O;>w7-{
z5l|WiwGIG2BF>;<01;3c2DJ_V002HB>w7-{5o#I+wGIG2BF>;<01;{$2DJ_V0002P
z0o89Y01^3@Y_@|TfB6{(wGIFjX*B>`J{V&F5&0PgwGIH%X*B>`1Oy)d5&0PgwGIH8
zYBd110mGZ201-DD@UI5|00000wgGlZr~nZ&8WFJv000000JZ^nCrtnmF&eP22LJ#7
z002Ux;o|@i`WXhb4giL0H2^{!+Q9%3`xyqc4gmXWH2^~Qe^w6w5&RhjwGIGqY&8IZ
z0tv3401^Ee2DJ_V0A9e20RRHNYXDO#=$QZ!0UE4GX#fcWK)N>o00000Q?b4}01*Nj
zzDQ{R1Cg+QB@m7Q074?<3jq-X8oH<s0OV{n077$TQ2-GI8pfy&099=@02&-mXaIS$
z2LJ#70021v2mk;80000HAD5H2gR}t|w@A2yeE|VDm&Un+jsaSiRl0-f0Wz0ByMyL`
zycq_y4gdi!!D|2?JvnLs5xf}^5orJ)qv7KK5xf}&wGIFu9NNJE5xf}&wGIG_0f)o)
z01@aJivb$?%>WVP86pX;p8yfP83wfu03v-z+5i#18A1R603wG-+5i#28A1R603wk{
z+5i#385sKl03w@6+5i#483wfu07nu-!D|2_>w7-{5zH9|wGIG2BF>;<01?a?2DJ_V
z002HB>w7-{5#<>MwGIG2BF>;<01@RG2DJ_V0002P0TC?)01<(g=)8j=e}NeWwGIHD
zay0;4J{V&F5rG*7wGIGpb2R{51Oy)d5rG*7wGIFcbTt6B0mGZ201>nqBdZ4h00000
zwgGlZr~nbL8Hub1000000JZ^nCrtnmt{Et-2LJ#7002Ux;o|@if*A(24gd{xH2^{!
z+Q9%3gBb?34ghX-H2^{h1+Jd}5rmhGy@MD$h#6D|X#fcWV7fN|00000Q?b4}01=59
zs!C`81Cg+QC4h_p074?<3jq;}8M>$r02_8S077$TQ2-H)m*KsG9|4}X0ltG}0Rgp_
zXTO7l0l}B-zk_H2(6?2<gTn#=(6<@Gg8=~n!M7X4g8=~n!M7d7gVO;4ub1)0gTVoV
zm$S!%IDbSL2DJ_V0rbpk03SU$Y5);L84?j`03V~_;{Xvv83wfu03RIM!2l6N83wfu
z0E+>K!}kCYa2bmM8v4xu5o{SE39g?25k(mWwGIFxeMs5>5l0z90000Yhe+B05l9(A
z0000Ykx1GA5lI;s`vL$Wn@HLK5lR^bwGIGB*F(%}03z#qKL8O_83wfu06rqlpke?K
zR2c@f4gdfEJ|gRTKL8PJ83wfu06rqlpke?KZ5al&4gdfE0K);&&u9P<374(NgCTzj
z83wfu00M$F09-y8V*n8e83wfu0Mmjs09*tF9{>>v83wfu0C9sg0JZ_co1p*^JQ*XV
z2LJ#7006cDc1oxK5jYu%rw0H40002C0eL4)01-ABD5nPi00000LZjj101*lq2DJ_V
zafCGhLLAz`01*or2DJ_V(u6etLJ2gkp8ydI83wfu05gR(0A9d~0RU*vYXDO#=$QZ!
z5E)blX#fcWfVwvT00000Q?b4}01*+Ffysju5fd4@s15*sg*5;|b7xTi5fqoE$%7vO
zCb!1PgJl5$J(n!YgM<M|m$A!(XaQKa8q9;k0s&aJ-p+#o0Rc(3;Ln2r0Rc(3;?RTB
z0RcCcw$g*a0dbdj(}OsF%oql>4gdkU*lPeEJvnLs5zH785orJ)qv7KK5zH6{wGIFu
z9NNJE5zH6{wGIG_0f)o)01@~Yivb$?%>WVf7$OO-p8yff7zVWt03v-z+5i#H7(xI5
z03wG-+5i#I7(xI503wk{+5i#J7%Bh&03w@6+5i#K7zVWt07smE*lPeH>w7-{5!@IC
zwGIG2BF>;<01@062DJ_V002HB>w7-{5%m}bwGIG2BF>;<01@>V2DJ_V0002P0fIY4
z01=TG2DJ_V5E=*o00000^8)|?000000h~15YXDq5IcfkAkr)QG4ghA6H2_>b7-Ikt
zkr)QG4gfcjH2_?H1Oy)d5s?@MwGIIOk~ILf0mGZ201?C(Um6Gi00000wgGlZr~nbb
z7%3YF000000JZ^nCrtnmz8J_F2mk;8002Ux;o|@ik{AZH4gmg>H2^{!+Q9%3lNbiI
z4ghnMH2{GE39g?25tJAPwGIFPUNC9`04v*T08=aInE(+ym>5(CX#fcWpt?5z00000
zQ?b4}01=rOs!C`81Cg+QC2(p2074?<3jq<E7`mtq0O*u8077$TQ2-H~muc059|5kn
ziPeK;0RhFg@79AB0Rg|42H1n<e^eL-wGIFQ_1$X#A3Zs001;Fe5)o+tAEV*p01;Fe
z2DJ_V9~|1j01;Fe2DJ_Vivfqj_W%)q7>fZK`pp0ld>A4LuAcxARTu`f4gey3NZJ4q
zR~SM7001I~NZJ4qSQtV8001J9NZJ4qSr{q+001JJNZJ4qS{Mek4gg2kL)~itBI|oU
z01;#u2DJ_VJ|fPbVgM0j7zVWt0000!BI|oU01<r{2DJ_VJ|fPbVgM0+7zVWt00000
z!vRK|LjVyOm#*1^A%7Vd2DJ_VhM6@0Ts|0M01+7&2DJ_VSDG~dTm%Fk01+7&2DJ_V
z9-B1)wgJPNp#Tv~7*7@m000000JZ^kN~i!4NEjIx2mk;8006cDc_&Q(5k?re76<?U
z0000&qv7KK5gHf<wGIFtoHYPK9NNJE5gQl=wGIH2oHYP}0x}7%p8ydY7zVWt003Tq
zXaWFd;A;R=E9jX35g-^;2WbEa1HigB0000008_EPIsg$Nmw?)X6cHmBx~L8S2%R+m
zLUU(P01+gYrP_lZ0XDbA+Jj{Q0Zq3n+=CVY0Y{fY-Gk<T;}-_C4gdk&;%fjOJvnLs
z5#tvU5orJ)qv7KK5#tvIwGIFu9NNJE5#tvIwGIG_0f)o)01*fnivb$?%>WSv7$OO-
zp8yf$7Y4Ns03v-z+5i#e7eW9403wG-+5i#f7eW94006@QPb>%k5vUgiwGIG)5(xkR
z000290{{S*r{04Qf2bD*wGIHLp)~+pJ{V&F5vUgiwGIGxqBQ_q1Oy)d5vUgiwGIG2
zqcs4w0mGZ201?|4&k_j$00000wgGlZr~nby7nu_Y000000JZ^nCrtnm)fYGu2><{9
z002Ux;o|@isTT&d4gfx+H2^{!+Q9%3suu>e4gj;HH2{GER|&4401>Mf2DJ_V0A6qn
z1OQLvYXDO#=$QZ!uNPDYX#fcW;JP;e00000Q?b4}01>bks!C`81Cg+QC9n+y074?<
z3jq<b7rLkp04Sw3077$TQ2-IMmyO?p9|6U;t>1%X0Rh~X6XApR0oIob;)CXYZx;r&
z4gdkl=W75TJvnLs5pNd~5orL60f)o)01=lLivb$?%>WUT7a|F+p8yeX7Y4Ns03v-z
z+5i!97b*Y%03wG-+5i!A7Y4Ns07tFoYXBnadp`gXeHR9`4gfwP&Y)rd5q%d1wGIFP
z06rq?dp`gXlotlI4gfwP&Y)rw01=cI2DJ_V0002P0Uq;x01;%DgX4oCJ!BUKwGIFs
zsWkvxF4$oJ5o8wzwGIGQsWkvX39g?25oH$!wGIG}sWku=Ip=Et(Ow4t00000&<+6r
z5onj;<Ac>cG8YE54gdi=>1zNVJvnLs5i%DN5orL60f)o)01;Ccivb$?%>WTl7a|F+
zp8ydv7Y4Ns006@QV#JLA5hs`Q<bxqSCl>~_4ghYdH2_>L*kJ$>Cl>~_4gjaBH2^{h
zuAcxAC>I8`4glb)H2@a==xYF>)dv6o0001B+yMX)E0<E`gSLNz76!Er00BhpYXBcT
zIcfkAgBB7IX#gLytkeJzgBAw04genq=MMl8gBAw04giY*hr{;(5fvAU0eTpc01*Ng
zivb$?%>WS(7a}YgD*zFM76!Er03r#lp8yes76!Er03v-z+5i!U76!Er03wG-+5i!V
z7CHa`03wk{+5lM*i55Bl001JJNZJ4qiWUa74geyhNZJ4qixvj84gex?GOqv;j23Vq
z001IBN!kDrjTUer002in>uUg0_=yJr5tkOlrw#xK11Q2b0000008@Og`v4J_m*D1u
z8-JP>2DJ_V07rl8YXBk(=m-H3s}=^e4gfw(bZu||5v~>nwGIG2BF>;<01>Vh2DJ_V
z0000-pX+M?BI|oU01>hl2DJ_VJ|fPbVgM1c76!Er0000-vFmF9BF>;<01>qo2DJ_V
z07u{JYXBk(=m-H3#})>)4gfw(bZu||5oXF32DJ_VJ|fPbVgM1!76!Er00000NB`?<
z03z#qKL8QV76!Er06rqlpke?K&K3r>4gdfEM-l9603yzyVgM1*76!Er002iL>}vob
z&Y)rd5z`h1wGIFPM^)@=08{vh2LTc7m*D4v6@Tp(2DJ_V2?G$qHvj+t0000-gY0Vn
zA`9pU0TJ^S2DJ_VK1_6NZ~zhZ76!Er06rqlpke?K_7(=U4gdfE07s+jYXBnadp`gX
z`W6PY4gfwP&Y)rd5&9MewGIFP07td#YXBn7pke?K{T2qb4gdg0$LwnWBF>;<01^Kd
z21m6H002kl>}vob>w7-{5fB#!wGIG2BF>;<01*%u2DJ_V002h^?P~xc>w7-{5fK*#
zwGIG2BF>;<01*)v2DJ_V002iP?U#G#gB>M4BI|oU01*}!2DJ_VJ|fPbVgL~q7Y4Ns
z00000!vXgkEdUWjm;LC2A%8>`2DJ_VQL8loTs|0M01-qM2DJ_VDXcXBTm%Fk01-qM
z2DJ_VWvw*;wgJPNp#Twk7AVCD000000JZ^kN~i!4cNWmZ2><{9006cDc_&Q(5p@<)
z#0dZZ0000&v#itr5k(dTwGIG|t~CHc2j>p}5k?jUwGIG5uQdQd2{f*s01-zP2DJ_V
z^shAlUXZ>80448h08=aInE(+>7E}jm00{#CyEgy;0000}vA#M05lolS>4OvzPZqkU
z4ggfJH2^|$XHftVP?z@UgCqfAmmBJX(E)as?dpTr0eqLn>w~HRjF)ljgTMiew=(U6
z3jqO{w?6KJbpZjux3uqr<pBZGx0>;TIspOIm(}uv*#XtJuk(Wo0Rh#wyYz!i0RfG-
z3HE~v0RfD+759S&0|AY<9r}X{0RfD+Df@#C0|6J8IR1k;f3g(@wGIFQIt6S1A3Zs0
z01>hk5)o+t9~5*r01>hk2DJ_VAMrXY01>hk2DJ_Vivfqj_W%(x7K;IT7?J=H9~O%N
z8v4xu5iAxWN0Ma#5wjHrwGIFx39g?25wsNswGIFxeMs5>5w;ZuwGIFxhe+B05w{gO
z0000Ykx1GAe-XG9IsgCwBAZCs01>$r2DJ_VBBe;$01>(s2DJ_VB5^XW01>+ta3KHy
zB0fpl01><uf!_`QN2dd908{vh2LTbr6~(6x00{#~!Z!c_0000}e6RZe5ylk;wGIFY
z13<zz00000002k118e{y&Y)rd5y=$>wGIFPN9O}<e*hv2=m-H3+7$-14gfw(bZu||
z5#1FAwGIG2BF>;<01@342DJ_V0000-2Lx;YBI|oU01@F82DJ_VJ|fPbVgM206$Z5q
z0000-83b$qBF>;<01@OB2DJ_V07pv%YyeaEi3b4@3Kqqu4gd)QIKnpo00000Q+%)c
z01*lnm+k+98h;QL2DJ_VK1_6NZ~ze#76!Er06rqlpke?K6BY)w4gdfE07sbwYycwb
zdp`gX7ZwJ!4gfwP&Y)rd5f>H)wGIFP07t6?Yycw8pke?K8Wsk%4gdg0y##CkBF>;<
z01+G(2DJ_V07u;fYycwbdp`gXEfxl~4gfwP&Y)rd5i2bg2DJ_V002k*1Z)5z>w7-{
z5iS-6wGIG2BF>;<01+-02DJ_V002iF1(k7x9VI>@>w7-{5i=GBwGIG2BF>;<01-15
z2DJ_V0002P0X1tN01<DM^MoOPZxsf$4gf*QH2_>b7-IktZxsf$4gf*RH2_=$1Rnqq
zZxsf$4gg2YH2}5&!<(T15vmoax(NUP0002C0d`8L01>7YT)GJW00000wgGu3O#l(3
z73H}J00000074XWH~<lF6$Z5q0C3GU07CIPEC3O46$Z5q03*&d073~KuAcxAauo)(
z4gl89H2_|~js*Zc2W$XSE9jX35q6ic0fZGMcNMBiXaEC|kbfmWj|BiiBIFAJ5qlN7
zs15)&&oux-b7xTi5qy{I0fZz0iI))qggF6^mrDYKxdEk@1p|c10jrnK1B9vpySI7-
zgbx7$$d`%*gqH!-w;~3FC;<WQw@U|vi30)ex77-S2LS=_m)#45KmqcX9t?!p0rI!_
z41@~-0rIy54unks0lT+S4}{YK0lT**5`;|x0R@+I6ofc`I~4}C4gdj38EgO_JvnLs
z5jzzU5orJ)|Ku6~5jzzIwGIG_0f)o)01;jlivb$?%>WTw6(R|)p8yd&6$Z5q03v-z
z+5izg74}U5006@QN(1}=5d{?nwGIGa$_D@d0001+0{{R300001oCYOq09-veY5)-h
z6$Z5q02S?jH2_>b7-Ikt1r-Lh4geMIH2_=$1Rnqq1r-Lh4gksSH2}5&!<(T15i=F2
z%Lf1e0002C0d`8L01+=0FUto200000wgGu3O#l%s6|Kq#0000007C!d8UPUn6$Z5q
z0Hg0U0740_p8ydD6$Z5q02uH!0AA3z0RZY6YyeXRE9jX35et{F6@(Qg3>8ns4gdp@
zuzw{Wxd8w|BIFAJ5f2r*s15*l@HGHJb7xTi5fGQ_6@(-K8<!9kgrNa0m$DXw)&Vq^
zi5G-m0Z6y%7lc3o0ZO+k8HBe20ZO;y8-xJ?0X~=I9EAJ<O1I)2gaH8oKDXo^grNZf
z_yL!PAcRSOa})-(4gdj~Cu{&8JvnLs5pxt05orJ)LL-O(5pxs<wGIFu|Ku6~5pxs<
zwGIG_0f)o)01@*PivfBVk^m9l6pH~G`pp0l?Gz$68K(ddbQA`)4gewvuAcxAbrc4*
z4gey3NZJ4qcN7M-4geyDNZJ4qcoaGS001J9NZJ5@5qT6k0000Yn@HLK5qcB`wGIFx
zrAXQU5qlH{wGIFxaWby}5quPIApig(K1tdD5q%W)O#uK$2PSL)Q}~Go0TG83#itGc
z2?J=tHvj+t002{buloQIhZF|24gd)QV8S;50000007n@nYycw8pke?KiWCO54gdg0
zM<#5403r+M2mukB6b7{p06t7~ZEye)o)iYP4gfwP&Y)rd5uOwVwGIFP002j4CTsvA
z>w7-{5uy|ZwGIG2BF>;<01=`T2DJ_V002jMCTsvA&Y)rd5v3FcwGIFPN2?}m08{vh
z2LTbw6vd|w00{$F!Z!c_0000}e6RZe5z7>pO(KLEf6){MwGIG2OmuB<01?y_2DJ_V
zJ|fPbVgM1;6b7{p00000NBJgf03z#qKL8Qf6b7{p06rqlpke?K*c1k}4gdfEM++xx
z03yzyVgM1_6b7{p002iFCu{&B&Y)rd5#1C9wGIFPM?EKO03z#qKL8Q#6b7{p06rql
zpke?KD((~pwGIFP07qRXYycwbdp`gX?-T~L4gfwP&Y)rd5$_ZRwGIFP07re7&?1B$
zB|akSdp`gX^b`iQ4gfwP&Y)rd5%d%WwGIFP006@Q1QMtK5j2-mBZMJ;G!zE44gkmd
zH2_>b7-IktG!zE44ghWaH2_=$1RnqqG!zE44gjhCH2}5&!<(T15o;8P^aua|0002C
z0d`8L01;;tJoE?v00000wgGu3O#l&P6u$Ea000000760|hyW2a6b7{p0Mh?807C!d
z8UPVC6b7{p0EGZH073~KuAcxAHxvf74gfd-HUM4_gaZJ3DQo~!E9jX35j&R=B!m?v
zJQS)*XaEC|kbfnBgaZIVBIFAJ5kC~Vs15*_0X6_ab7xTi5kQwrB!naZQI~NfgwX+J
zmqR6l*a2*p943US0erW|CWH$C0g0E>Cxn*)mA7jsge(C8u(y~ggbM)yvA3iugfIaC
zv$xeNgtr0#vzLu7gxLYJw`VVe3jqPMw{tLr0RaKCw{<au2LS=Iw|z2%M*#tSm&G%L
z`~iHo!Zd_I0Rert1U7``0Reos>o<f80Rert_c(-m0s-Zhb324MfB6#zwGIFQoIq><
zA3Zs001^2U5)o+tAG^;30TKBV2DJ_Vivfqj_W%(i6pH~G`pp0l9~2@9uAcxA`V$7V
z4gey3NZJ4q{1ZX|001I~NZJ4q{S!_K0{|kCNZJ4q{u2hZ4gg1gKx_aa>w7-{5eO6p
zwGIG2BF>;<01*fjD+aX=0000!BI|oU01+S*2DJ_VJ|fPbVgL~!6b7{p00000!vXH>
z009y1msLE3AwBLB2DJ_V10gm5T)WQ$0TJ#K2DJ_VDIzuiLJ6*)01@vK2DJ_V-6A#s
z7V|%B0G)FQ000000Pqk601@(+vpj@q0ST8LJ%r=|_O}B*gi!$j{g+igglGW^x4u7w
ziv$4+w@O5Wi2?!rx70?2NdteI69%;o00A~oYyclUIcfkAn-dZdX#gL)&jSGwn-d1L
z4giY*hr{;(5#1Au0UG+v01?&`A_=aa01=!M2DJ_VB7I2O01=)O2DJ_VB8N!Y01=-P
z2DJ_VB9Tbi01==QLI3~&BAZCs01=@R1~>-*BBe;$01=`S2DJ_VB5{8*uK*FF69%;o
z07n&1Yycwbdp`gXtrG^d4gfwP&Y)rd5v>yjwGIFP07pkpYyeaEi3b4@t`o(l4gd)Q
zc)~XT00000Q+%)c01>Ve2DJ_V2?KD#Hvj+t0000-bx&*nA`9pU0TI0u2DJ_VK1_6N
zZ~zg&69%;o06rqlpkjXj5x^4$wGIFP002joPiz1p>w7-{5yTS)wGIG2BF>;<01?C!
z2DJ_V002j)Piz1p&Y)rd5ycY*wGIFPN4ZaI03yzyVgM1y69%;o002kXPiz1p>w7-{
z5!Vw2wGIG2BF>;<01?*{2DJ_V002k%Piz1p>w7-{5!e$3wGI;iJ|fPbVgM1?69%;o
z0000-8BmvPON1Q(K9`S6gd3OUON6)qke8oKgdri269%;o01GlU09?Dz0|60`69%;o
z0QE990740_p8ye&m*Gr=8v&G;{Y-?=0il=9O@x~PvbSnZgiZkgvX{9}gj@l)mljZj
zcmXz-k5Gh%0b;ioQH06@0il=TQ-qZPp|^Ncgh&B@DH8^@4gdj1Tx<XzJvnLs5h)WA
z5orJ)yUzmw5h)V}wGIG_0f)o)01<x^ivb$?%>WU16CxwRGyoAQ6HR~u03r#lp8ydn
z69%;o03v-z+5izP69%;o03wG-+5izQ6G8w003wk{+5izR6Is^-03w@6+5izS6U5g8
z03xM-NZJ4qF%zMwHUJ`VGOqv;G80XJ0staDN!kDrGZO~24gexKGp_&<G!q834gexo
zN!kDrH4_H44gex^k^}$|HWLQ54gg2=T5JF!>w7-{5kM0LwGIG2BF>;<01-eF2DJ_V
z002iOTWkPR_=yJr5mFPyrw#xK1Bk*m0001&zE*??e^L_$wGIFY1AxLe00000002i+
zTWkO#3+M;|5n2-lwGIG2OmuB<01;gi2DJ_VJ|fPbVgM0c69%;o00000M|4|k03z#q
zKL8P769%;o06rqlpke?KVG{<m4gdfEM}}K$03yzyVgM0h69%;o002jrTWkO#&Y)rd
z5oHqwS+x!T07t4@Yycw8pke?KXA=gs4gdg0%3EvzBI|oU01<c-2DJ_VJ|fPbVgM0%
z69%;o0000->RW68BI|oU01<f;2DJ_VJ|fPbVgM0&69%;o0000-3S5_7SA-q{DqNR-
zSA-oUJ|gRTKL8PY69%;o06rqlpke?Kd=mz>4gdfE0K);IQ}F;19+p^yAw3=w2DJ_V
zJwi4BT)WQ$0TCV(2DJ_V7eh7xLJ6*)01+P(2DJ_Vj6*g676@8w0L^F#000000AL9P
z01+aWURZ>>0Y<kOS%mci0Y<kEU4--j0brK|UxX<EgO@H}gn0pgx2Ru)9RUGPm&Ra(
zL;+E^3t@yn0Rd*0Ok#xF0V<a$V}zFhGMB((gqr~`mt16oXaO>p#bktA0WY^9WrX+v
z0V=l|XM|$`0Wz0<XoOn<D!1Bbgm(dd&k_c;4gdkLZ)^Y`JvnLs5zi755orJ)yUzmw
z5zi6^wGIG_0f)o)01*univb$?%>WSy6CxlY1pyJz5^x~^03r#lp8yfj5(c#n03v-z
z+5i#L5(c#n03wG-+5i#M5<&m~03wk{+5i#N5)`Nc03w@6+5i#O5`U-y03xM-NZJ4q
z*Ai0QH2@-UGOqv;*b;Cd001IBN!kDr*%Ah|4gg1CZ)^Y}>w7-{5#tgDwGIG2BF>;<
z01@L72DJ_V002jGZ)^Y}&Y)rd5%>}YwGIFPN0V=C03z#qKL8O469%;o06rqlpke?K
z2@?jj4gdfEJ|gRTKL8O569%;!4gfwP&Y)rd5egFqwGIFP006@QRvZlh5yY2iYlI;o
z#1aOz4gk|&HUM0^&jSGw#1aOz4giE<HUL5iuAcxA#h0sVgc||Km&a>_p8@H&XKaL#
z1Oe%{`EZ1i0s-}xjB|wB0rHn<bcAaG^q16hgl_@$mqc}h?*a9fG<Jml0r!_QcZ7!l
z^|!otgaZKq(3ik?guDURmwkDJ!~xZpLVAQ;0ob>ndV~uB0oAvwdxVw(0noQXeuS$5
z0ob?he}o4C0noShfP~)xe~A(XwGIFQ#)oVGA3Zs001=520nAS`uo4kz03W;00|60<
z5(c#n0E+>K!}kCYvl5E|8v4xu5w8*=39g?25sDH9wGIFxeMs5>5sVT-0000Yhe+B0
z5seZ;0000Ykx1GA5snfDwGIGBsfTO;BI|oU01=oH2DJ_VJ|fPbVgM1C5-SF^4gdfE
zJ|gRTKL8Q15(c#n06rqlpke?Kuo4Ef4gdfE0K);@1bzS!ewTlPgdrh*5(c#n04aSo
z09?Dz0|60!5(c#n0LOhc0740_p8yemm%)RC8v%ls-GhX30ic&qgoMC<LlOqH4gdiG
zifjNMJvnLs5knFZ5orJ)yUzmw5knFNwGIG_0f)o)01<5xivb$?%>WT;5+Vt%p8yd=
z5(c#n03v-z+5izo5<&m~03wG-+5izp5<&m~03wk{+5izq5(c#n07vABYycwbdp`gX
zQW6HW4gfwP&Y)rd5mFK>2DJ_V002HB>w7-{5o!_!wGIG2BF>;<01;{u2DJ_V0002P
z0lNi601-Hsx`l)xAvh8SwGIG0fHnYJyUzmw5jYYCwGIH<fHnX^39g?25jmIqg@hXc
zJeL@TgmM8|my3pkz<>P_2DJ_V0X&Rs03SU$Y5)=a5fTw;03W;00|62J5eBsm0E+>K
z!}kCYClZSR8v4xu5hD^J39g?25&jVdwGIFxeMs5>5dacG0000Yhe+B05djiH0000Y
zkx1GA5dsnhwGIGBAB=1OBI|oU01*rl2DJ_VJ|fPbVgL~g5-SF^4gdfEJ|gRTKL8OV
z5(c#n06rqlpke?KBoYR-4gdfE0K);1uLuDV@|X38gdsih5eBsm08fK909?Dz0|628
z5eBsm0PTY|0740_p8yf_5eBsm049Vs02Y&rYyebZ2><{9003|V1ppEDms5y@asd>V
z#)yQ#R=yDiwGIFQBaUnUA3Zs001>_s5)o+tAG^;30TI3t2DJ_Vivfqj_W%*$5sLvD
z`pp0l-Vq`RuAcxAzYzwt4gey3NZJ4q!4X0L0002P0kru!01>m7+=+xCA+r$%wGIGc
zhBg3PyUzmw5wj5nwGIH+hBg2~39g?25ww>cii8^hx0gDKgjfN~my?Qw<N@}#c#DL_
z0e=b+2DJ_V0eO^c03SU$Y5)-m5fTw;03W;00|5~V5eBsm03XJZe*h5*5eBsm03Y%s
z(Et$&5eBsm03X&p!T=Ep5eBsm0E+>K!}kCYpAm}z8v4xu5t$JpbBo;o5epFpwGIFx
zZI9Fd5eyLqwGIFxGtlG!5e*RrwGIFx34gAi01*xm2DJ_VB7I2O01*%o2DJ_VB8N!Y
z01*)p2DJ_VB9Tbi01*-q2DJ_VBAZCs01*=rLI3~&BBe;$01*@sa3KHyB5^XW01*`t
za3KHyB0fpl01*}uO@IObN0^dq03r+M2mui$5eBsm06t7~ZEye)DG>&>4gfwP&IO=i
z01+vdL5_qIO)U`ywGIG2BF>;<01+(_2DJ_V002kIl57AX&Y)rd5ik)3wGIFPNBEL#
z03r+M2muj65eBsm06t7~ZEye)L=gtH4gfwP&Y)rd5k!}vj)W6QND&6L4gfwP&Y)rd
z5l9gRwGIFP07oj5Yycw8pke?KOA!XO4gdg0T$5}7Q}~Go0TEUa#itGc2?LnIHvj+t
z002{buloQIR+s&bgd2Zc5eBsm06t7~ZEye)Ul9hi4gfwP&Y)rd5nmAowGIFP002j-
zlWYJY>w7-{5n~YswGIG2BF>;<01;ym2DJ_V002k4lWYJY&Y)rd5oQqvwGIFPN8FQa
z03z#qKL8P$5eBsm06rqlpke?Knh^%I4gdfENBom)03z#qKL8mKn-K=J4gfwP&Y)rd
z5t|VPwGIFP07o2@mvE1S9)CQPYycwbdp`gXpb-YO4gfwP&Y)rd5ugzUwGIFP07qPu
zYycwbdp`gXp%DhP4gfwP&Y)rd5up(VwGIFP06rq?dp`gXq7eqQ4gfwP&Y)rd5uy<W
zwGIFP006@Q1YsNi5xEcswGIH#k_i9+0000L1ONa400001oXMGN0DoLPIcfkAxex}m
z4gji*HUL~c7-Iktxex}m4gh<NHUL}%1Rnqqxex}m4glJXHUPE(!<(T15dsl7lL-I-
z0002C0d`8L01^HW(vt}Q00000wgGu3O#l)65O9+T0000007ARZ0|61b5C*jl0C<iz
z07Ax+e*h7?5C*jl07S%&HUL8MB+&p7ybuPp4ge94HUL7_J;DGHy$}Yq4gg+{HUL5i
zuAcxAz7Phr4gjl<HUM6rF$Dm!m}~%3E9jX35yF=ck%Sci!<QzJgdh>c5W1)i0OOA~
z077$TQ2-IfmsgR59|7o>caem&0pgb|l7zGY+_&M9gqQ&V{g--_gyI2NmtB>FB!58=
z2DJ_V0fwAx03SU$Y5);I5E2n-03W;00|6015C*jl03XJZe*h6d5C*jl03Y%s(Et%a
z5C*jl03X&p!T=FL5C*jl03W9G+W-+k5C*jl0E+>K!}kCYs1S<*8v4xu5uy+x39g?2
z5ke3KwGIFxeMs5>5kwF|0000Yhkr=g01-tH2DJ_VB9Tbi01-wI2DJ_VBAZCs01-zJ
z2DJ_VBBe;$01-$Ka3KHyB5^XW01-(La3KHyB0fpl01-+MO@IObA~`d!01-<Nf!_`Q
zN6(sU03r+M2mujX5C*jl06t7~ZEye)Ul0bh4gfwP&Y)rd5nm7nwGIFP0Dk~S@0x4?
zBI|oU01;yl2DJ_VJ|fPbVgM0i5C*jl0000-0h??9BF>;<01;*o2DJ_V07ny>Yycw8
zpke?KXb=Xq4gdg0L7Qv<A`9pU0TFKy2DJ_VK1_6NZ~zf<5C*jl06rqlpke?Kau5c!
z4gdfE07qe)Yycwbdp`gXc7G5CwGIG2BF>;<01<W&2DJ_V002jGn`{6g&Y)rd5qS^>
zwGIFPM}?bg03yzyVgM0)5C*jl002j~n`{76_=yJr5quEErw#xK1E|6`0000008@Og
z`v4Jq5C*jl00{%2!Z!c_00000N9CJr03r+M2muj-5C*jl06t7~ZGUh85rq&2wGIG2
zBF>;<01<@{2DJ_V0000-1)OXEBI|oU01=502DJ_VJ|fPbVgM0|5C*jl0000-7o2PW
zBF>;<01=E32DJ_V07ogDYycw8pke?Kjt~a54gdg0Nt|o|BI|oU01=}Q2DJ_VJ|fPb
zVgM1N5C*jl0000-X&;<y03z#qKL8P=5C*jl06rqlpke?Kq!0$R4gdfEK9}E@gc}I}
z!vP%fi~tegmp_<<A%Ecy2DJ_V$CEYyTs|0M01@F22DJ_VpOiKLTm%Fk01@F22DJ_V
zLzXrGwgJPNp#Tv%5TBO`000000JZ^kN~i!4H4x;S2><{9006cDc_&Q(5i<~QoCyE`
z0000&yUzmw5#kR9wGIGWmo@-G#*%*k5#tXAwGIIJmo@-G@_!`J01@O52DJ_VmY6mG
zLe@RP01@R62DJ_VGnqC3LZ<ZF01@U72DJ_V(3v&>LJ6*)01@X82DJ_VZJIU!UeGZG
z0REtC08=aInE(;%4^`r700{%YyEgy;0000}vA#M05$q3iQE31Jk&u5SP%;GoLL%e~
z0TJ&Hx~L8S%?g?}077$TQ2-I}muQ)UBmo(hjhTcr0Th?QnS`_f4wrYDgtP$&w=A24
zl>q@dm%^Ncw*g3(dYy#80ZNxUo`iz|NteH#gunq-w}797C;<Uyx2d3n6aoQyx4xo;
zB?1ABx8$URP5}Xpmouh>j{%RD!={8r0ZNw&r-WMpNw;RFgbM)yNVjsRgy{hRN|#)!
zgj)egx45c=3jqO0x5TT2L<0ddmkF<g+W|zE<*$S|e}4}KwGIFQfVOM^A3Zs001<x=
z5)o+tAG^;30TF)>2DJ_VANN)d01<x=2DJ_VA2Px;01<x=2DJ_Vivfqj_W%*l4~qdB
z`pp0l$`2w5uAcxAfDZ<>4gey3NZJ4qf)7Ff001I~NZJ4qgAWF^4geyNNZJ4qgbxO_
z4geyXe@NN@5rq#1wGIFxrAXQU5rz+NApig(aWby}5r+>=fC2zVJGE>8A`9pU0TGxF
z2DJ_VK1_6NZ~zgT4+gak06rqlpke?Kn-2!H4gdfE07qN3Yycwbdp`gXpAQDL4gfwP
z&Y)rd5uXnRwGIFP07q@LYycw8pke?Kq7MeOe+~cuM}M_!03yzyVgM1O4+gak002j^
zwQK-W_=yJr5v32srw#xK1GvIB0000008@Og`v4K84+gak00{%I!Z!c_00000N8hz<
z03r+M2mukR4+gak06t7~ZEye)t`7#a4gfwP&Y)rd5v~sgwGIFP002k-wQK+)>w7-{
ze-W||2DJ_VJ|fPbVgM1c4+gak0000-5w>gqBF>;<01>qh2DJ_V07oOXYycw8pke?K
zw+{xj4gdg0L$+)HBI|oU01?X%2DJ_VJ|fPbVgM1#4+gak0000-W43GnBI|oU01?a&
z2DJ_VJ|fPbVgM1$4+gak0000!BI|oUm))_17nd-ygew6#mt3-hA%8g!2DJ_Vc(gVE
zTs|0M01-J42DJ_VD77{KTm%Fk01-J42DJ_Vqqa5xwgJPNp#Twg4`>Jp000000JZ^k
zN~i!4a}NOx2><{9006cDc_&Q(5pfTo3<&@L0000&yUzmw5jqbBwGIHew>AJm_f`)8
z5jzhCwGIG6xHbSnG8Mu!01-S72DJ_V%eXcGLJ6*)01-Wx5wnCC7eNoKNNE5G0|2}?
z0000008_EPIsg$umq4?G6cI!Zx~L8Spt&{xLUU(P01-u(WwV4I0biGevxKw(S+_d0
zgna=4c$c`fgp&b=mrS;V!~utw5x0bc0g|_sw}dDG0i(CRxP*xV0i(A_yo4zN0k*et
zzJyKz0k)UHzl4tghnG>ngi`^Ax23>@;Q;}MmxscHQvrsz+rory0s(TDO2vf10W+5j
z#)LS3)D8x<4gdj(%xnN3JvnLs5!4P65orJ)QjDJf5!4O_wGIFu3d$b<5!4O_wGIG_
z0f)o)01+q;ivb$?%>WT14<ZS!p8yfn4hFRj03v-z+5i#P4hFRj03wG-+5i#Q4hFRj
z03wk{+5i#R4hFRj03w@6+5i#S4sam=03xM-NZJ4q+YWFc001I!GOqv;+zvtj002iw
z%WME5>w7-{5#|mCwGIG2BF>;<01@U62DJ_V002jJ%WME53+M;|5$z5JwGIG2OmuB<
z01@yG2DJ_VJ|fPbVgM2F4hFRj00000N0!TM03z#qKL8Q*4hFRj06rqlpke?K^bQ73
zwGIFP07t0HYycw8pke?K_YMZN4gdg0y31?;BF>;<01^2P2DJ_V07vM{Yyct)=m-H3
z01pPW4gfw(bZu||5d#kfwGIG2BF>;<01*S1k;jA+e+Lf+wGIG2BF>;<01*cd2DJ_V
z002iC%xnN6&Y)rd5eg3mwGIFPM=Z>203yzyVgL~g4+gak002i!%xnN6>w7-{5hD)<
zwGIG2BF>;<01+b(2DJ_V002j9%xnN6>w7-{5hM==wGIG2BF>;<01+e)2DJ_V002HB
z>w7-{e-R}Q2DJ_VJ|fPbVgL~(4+gak00000paM40v;Yy24hFRj0Pv{^0000003-<j
z0000000ErM<7@z2JvnLs5s?lCwGII2&^7>EJ{V&F5s?lCwGIFv(KY~F1Oy)d5s?lC
zwGIF#(l!9L0mGZ201?a%B&!Jk00000wgGlZUZ?;O$PUY@2><{9006cDc_&Q(5ylQw
zstEu90000&QjDJf5t0rDwGIGy(>4G?3d$b<5t9xEwGIF()HVP@39g?25tI%FwGIH>
z)HVQKuxSMVjLvKTQ!D7101=p$$%GXqnGUK-XaEC|kbfmGY6Sp7BIFAJ5uFaYs15);
z)iwY^b7xTi5uTSj$%G^UtCwKOggF7Rmzv3hxdF$QRLX?O0nL{l%Y=#n+Lx)zgunsY
zmuk#}zyasCE6s!$0Ri~8M9zeb0|EHA($R!E0Rj5A4by}%0RaoQKh%VY0|5)S$=8HB
z0Ra=21lfc#0o%7X*@O!L0ou1i+JvtG0o%9p+k^`N0ou3!+=NX60V0=8--I}SRt^TW
z4gdjm<ZJ*RJvnLs5mpWo5orJ)RA^WL5mpWcwGIG_0f)o)01<!=ivb$?%>WU64k8Jz
zp8ye84hFRj03v-z+5iz*4nhC`03wG-+5iz+4sam=03wk{+5iz-4hFRj07qHmYycwb
zdp`gXWex_l4gfwP&Y)rd5oHc52DJ_V002HB>w7-{5q%B@wGIG2BF>;<01<r-2DJ_V
z0002P0ovt|01-=<FW`hBAxjPhwGIIP_cj1rRA^WL5lapRwGIGY_%;AS39g?25lokH
z;Dj3iPnV70gl+*^m;B&_!~tlR!Qq5+0cf{1;)DwU0d1E-<Al`#ZI>71gzo`uw+H2f
z0|5bAw+rTk3jqOIw-@Jx3jqOWw<G9;2LS<Xw=3y{vjKk)4hFRj00GbMYyclUIcfkA
z5DpR%X#gKoXjlLd5Do^l4giY*hr{;(5jqZw0UG+v01-A0A_=aa01*)m2DJ_VB7I2O
z01*=oLI3~&B8N!Y01*@pa3KHyB9Tbi01*`q2DJ_VN3!p103z#qKL8OQ4hFRj06rql
zpke?K9}X)9wGIFP06rq?dp`gXHx3514gfwP&Y)rd5jPG7wGIFP006@QK|z`T5e1is
z>x3a81r7$a4gll?Hvn8zXjlLd1r7$a4gmKBHvmEjuAcxA2A9g~gc|_~m*(q)Y5^3N
zQS5}k0U(zW?S$k3_P6)#gi!$jB9}?;gmVERx4rL#2LS;mm&NdenE^4EU-5)I0Tj2K
z@q`Nj0Tj2W@`MWk0V21w^MslKf6NUAwGIFQC;DsvA3Zs001?a$5)o+tivfqj_W%*}
z4T}L9`pp0l@C_mfuAcxA%?$>%4gey3NZJ4q&kY#+0stb1NZJ4q&<zH)4gg0A`fLCq
z>w7-{5!wv~wGIG2BF>;<01?^^2DJ_V002HB>w7-{5%CQMwGIG2BF>;<695tM4F<Ii
z00000!vU47H~<mBm*Dk;Aw9tj2DJ_V^bt1zTrSvQ01?3r2DJ_VEfO~XLJ6*)01?6s
z2DJ_VW)e357MA&J09n%q000000N~gG01?HPKK6uN0neA5_Jm*o+_(Mqgu?;>+_%B`
zgaH8o&$q++gaH8o&$q?<gqi_=hz$m{4gdi-0&M^vJvnLs5r_>E5orL60f)o)01>ba
zivb$?%>WUs4I&Ayp8yew4F<Ii03v-z+5i!Y4S<9L03wG-+5i!Z4F<Ii07n}FZ2%(c
zdp`gXmJJ5A4gfwP&Y)rd5ta=GwGIFP06rq?dp`gXtqlgX4gfwP&Y)rg01>U1;r)bW
z0ezQ9{)8bReGLY+4gi-LHvn8N*kJ$>eGLY+4gk;^HvmEjuAcxAewT{=gc|{Ym!|%N
zU;&ty2mge(e;Ex1wGIFQ2L^2bA3Zs001+7t5)o+tAGk+R01+7t2DJ_VAGw`R01+7t
z2DJ_VA6wVU01+7t2DJ_VAAMZU01+7t2DJ_Vivfqj_W%)W4T}L9`pp0lW(^_<uAcxA
z8Vv@u4gey3NZJ4q91RAw4geyDNZJ4q9SsJx4geyNe@NN@5grW&wGIFxn@HLK5g!e3
zApig(rAXQU5g-k4Apig(aWby}5g`qM-wpsGK1tdD5h4wN-wpsr^aO1HA`9pU0TD6{
z2DJ_VK1_6NZ~zfC4F<Ii06rqlpke?KH4O%}4gdfE07n!BZ2%(cdp`gXISmH24gfwP
z&Y)rde-Swi2DJ_V002iO1#JK#&Y)rd5j+hBwGIFPM>qv-03yzyVgM074F<Ii002j3
z1#JK#3+M;|5k(CKwGIG2OmuB<01-$H2DJ_VJ|fPbVgM0H4F<Ii00000M}`G$03z#q
zKL8O-4F<Ii06rqlpke?KObrIL4gdfEN0<d|VE`h|pke?KPYnjO4gdg0ss(KTBF>;<
z01;6Q2DJ_V07uFNZ2%(cdp`gXXAK6m4gfwP&Y)rd5oZkswGIFP07vQtZ2%(cdp`gX
zXblFn4gfwP&Y)rd5oiqtwGIFP06v#s0fie00H6Yh1~32-$d|wYg&}{)3<k9h0O}hz
z09-y8V*nAz3<k9h0Gb>(09*tF9{>@^3<k9h0J9x80JZ_co1p*^5e)(92><{9006cD
zc1oxK5e*H|?Fj$?0002C0eL4)01*ofaP0{I00000Lbyj!01?Ry2DJ_VqaHT^Lb;t!
z01?Uz2DJ_VK_52&LR){=%K#C}3<k9h0N)=s0789S&j1n33<k9h0DT}g0740_p8yff
z3<k9h02?7U0A6q}1^_h*Z2(g%=$QZ!(F{CoX#fcWK)g2q00000Q?b4}01?s*Vo_-T
z1Cfw_CEzax074?<3jq<-47#We0B0dL077$TQ2-Iu492Jq0RETX0)-g?_m}wsg|q?j
zm#zbav;ph4V+4hj0Ra-1`2~fg0Uwv228F-@AeUeVg}?zUw;>3H7y$u2w>k-hi30&W
zx5NyEIspMcxBm@=F#!Qkw=oZeHUR-rmuL`$LIG2ks}O}o0U)>X5QPf?0Ux*i5rwY-
z0U)=o6NL)_0Ux)z6orHW0SuQv7lpwA*_Qzrg*QHy3<k9h00F%mZ2%uVIcfkAmJAXR
zX#k4>hr{;(5w{GB0UG+v01>ndA_=aa01=lA2DJ_V0K)-cIj#T^i<fp7g&{qQ3<k9h
z0N_4109-EEVE_?}3<k9h02x0w0740_p8yez3<k9h08&3U02Y88Z2)l<2mk;80059p
z0ss+@m(&=AwtrF#2DJ_V0b?F*03SU$Y5);Z3=$D(0E+>K!}kCYdJKyJ8v4xu5q1nB
z39g?25mO8XwGIFxeMs5>5mgL<-wpsGhe+B05mpQawGIGBMjmYdBI|oU01;pe2DJ_V
zJ|fPbVgM0f3<k9h0000!BI|oU01<Z#2DJ_VJ|fPbViN!ncMJx#4gdfE0K);=?I8dW
zN0%HLg&`qF3<k9h0K-2w09-EEVE_?F3<k9h0Q^5U0740_p8yd^mtGo$8v#p~dK!gN
z0cV%Z8ilNXBMbrDA8i01JvnLs5hDx|5orJb*8&e0`T!9c3<38aZ2%uVIcfkA8Vm-t
z4gdh30os;!01@a5IsgCw0YD&a03SU$Y5)=F3lb4&0E+>K!}kCY3k-_^8v4xu5eN(-
z39g?25$Ov$0000YeMs5>5$g*x12+Hwp8-5Q5CIW?+6y`W0004lAZ-92JvnLs5!wp|
zwGIFuJ{V&F5!wp|wGIFu1Oy)d5!wp|wGIFx39g?25!(wo00000p8>pip#Tx03kJ0g
z009&sZ2%wqlnwzAq6-GK4giY*hr{;(5y}gT0UG+v01?IuA_=aa01=}L2DJ_VB7I2O
z01>5D3qk+@03wG-+5i!z3kJ0g07v&AZ2%(cdp`gXunPvY4gfwP&Y)rd5wHsewGIFP
z06rq?dp`gX#|s9v4gfwP&Y)rd5yuM#wGIFP006@QPd4=c5to<E9EBkvmkS294gf?!
zHvn8N*kJ$>mkS294gh^YHvmEjuAcxAn3oS7g&P5zmn$8GU;(n1j2(q30k@a79fgg5
zNec$G4gdj)BW(a5{FDv>5lIUMwGIG_0f)o)01<@?ivb$?%>WUA3nJFZ*Z>hq3veL-
z03r#lp8yd{3kJ0g03v-z+5izv3kJ0g03wG-+5izw3lyjV03wk{+5izx3xB8r03w@6
z+5izy3sT)R03xMG+5izz3veL-03vaJGOqv;QVRyP4gg0xBW(a8>w7-{5nKxfwGIG2
zBF>;<01;dZ2DJ_V002i#BW(a8&Y)rd5pW9zwGIFPM{Of*03z#qKL8Pd3kJ0g06rql
zpke?KfeQw;4gdfEJ|gRTKL8Pe3kJ0g06rqlpke?Kf(r(<4gdfE0K)<N@W}uI5k8kg
zAB7=3J_`o54gi%xHvn8N*kJ$>J_`o54gk(VHvmEjuAcxAKMMx64gd#3HvkrkB5eTF
zQ3n730000`0|5XLLYJW*g^mGXw=f`ukpux@x56Zak^%v3msKZ)+W~5qF(`#=0c@9@
zD1~nUZI=xxh3^4vm;EV){{e58{VIir0d2R1D}@6A0ZNyNEQP!QQI|$7g~S0)mkTb1
zTmezHXfA~d0Rc|8bT5UL0s%_53o?bP0Rd6Bw=;zY0Rc+4zchuk0e|oc2DJ_V0lGVF
z03ZC64gnGH3I??f0E+>K!}kCYEDMVP8v4xu5hn{GR?=Gl5%CIeApig(39g?25%LNK
zwGIFxeMs5>5%dZMwGIFxhe+B05%mfbr~&{Ykx1GA5%vmyr~&{Yn@HLK5%&sG-8BFr
zrAXQU5%>ylApig(aep$e01^2L2DJ_VM{7H603z#qKL8N}3kJ0g06rqlpke?K0}BSV
z4gdfEM}0eO03yzyVgL~r3kJ0g002jwJ8b|W>w7-{5hx1=wGIG2BF>;<01+q)2DJ_V
z002HB>w7-{5h)7>wGIG2BF>;<01+t*2DJ_V0002P0m^I600I%^mvc9TAwA^^2DJ_V
zc3d|ATrSvQ01@R12DJ_VuUt0(LJ6*)01@U22DJ_V=v+4d7P~ra05nVo0000005JRj
z01@ez(Km&T0SLEXIE9e}0SLG9JcW`10Th>oKZV-?5tm^=g=+y5m(4(hZvhmSJwb)<
z0Th=lLWTbU7MCqUg@*wYx3xos0|5c?m$^iRyaD)^c14B60ri(VMul7f__vrwg$n@z
z^|z!)g_Z&V@wYolg{uJp__yavg$Drv@we|xg|q>GmkI{84gdk_Q*8hr{FDv>5tj-E
zwGIG_0f)o)01?v)ivb$?%>WV33L++!DgY6f3UDC+03r#lp8ye=3I??f03v-z+5i!o
z3I??f03wG-+5i!p3KXaU03wk{+5i!q3V)~q03w@6+5i!r3R2xQ03xMG+5i!s3UDC+
z03vaJGOqv;pb7@H4gg1+Q*8hu>w7-{5vmFXwGIG2BF>;<01>JR2DJ_V002j=Q*8hu
z&Y)rd5xxorwGIFPN6k}h03z#qKL8QW3I??f06rqlpke?K&k6>$4gdfEJ|gRTKL8QX
z3I??f06rqlpke?K&<X~%4gdfE0K)+|0z?1;5sa6kPlX{pj0y&|4ggnnHvn8N*kJ$>
zj0y&|4gipLHvmEjuAcxAjS2>}4gkq^HvksvQf&Z~Mh5@@0000G^#A}7ke2~ag^mHQ
zw~$bUkpuy+w<1-Ak^%v{m$g@g+X1+jkywRm0lJs{ScPu^yO(WQh3^5omt9(g{{g+1
zU0a2R0lT;5TZIDw0hpKRT!p*=pO>axg~S1zmup^yTmhfA$X<mD0Rf!1)L(^`0s)w}
zYhs0~0Rf-47i5J80RfnYA7zDyA7zGzA7zJ!A7zM#A7zP$A7zTSA7zW+{Q;M@V~qI*
z0Zq61X^e#f0rt0mZH$rv0iCxKaE#m(0fx8mgp8^J0mzq|hm1i1%9q`Tj79;*w<L&+
za{~dkx0s8J69oa$m#mPCJOQ(p>5z<!0hgC{k&Gw-)0dx-j8Xx|m-3N}I03VlE0T<i
z0hgD!l8oyC^0%aujD!IJ`L`*Rj86dpX}5Wnj2R073zw6jjM4!Gmu8}jkO7yM_o9rZ
z0lc@Kqm20u0f3j7vW(XOgST+Aj9LK!(6`I9j4S~i*i`$wbN~%{FK7U<Km`B*00008
z0RR910000Hl(!VNjIIFz1ee3Nj4}b*xAwP;0s#ROm-)Dii~$z6hPjLl1OXeDl)a21
z0S>pfy^IY60UWo*!Hlv50ZX@d#*8Kb0brM($Bc6UpSKgpjL`xCsF&r+j7kARhcL^G
zw=m0%*=zw6mkKzJzyTGP%Q%j_0U(!%IgVEW7PsIzjtT<-a<}U|jspP!bGPq3jt2n&
zbeH!&j&uQZmux?dR{?~##6OM$0Re@#$Uu$<0Re`W&_RxL0fx6lLXIH;0T!2MLylbm
z6}QDhjtc<+6t~etj+X%e7ME#8j$Hv2x5q|~3jqNXhto%nx6?<D$SMKBmz;8sc>%DO
z7ITk30n)ctbB{;?55Cua2%Z#6Xh};;hMfa~x2<%KX9NM2mlb)BtpSO*!+DQ_0SNZ6
zeo6!p1B<s!dyhx~0g{)Ne2<g?lD8{;kC_4vs_lN(1Tq8X1WE)E1BRE<e~&35gYAB-
z1Q7rN1Ofx)k^zDQG5`Vu0t1`vemn#b00IO81Lc<;fRDHVwYS!Qk1+u^fsz680Rt1B
z18pt?6Cl_GDFc!0emn#b1M!mq)&wa7pzVIT1Q7$`k^!Cr;qZRQOQ1<hO9P&lK7x-g
z2hH$)@&lTWmn?&iJD0D5j}SM>?0yIm1Q7$dlL1NuDFessemn#b1G|#}bOXmMx&ymY
zkORdmJOjH^JOh{UelAO(NlQxu=#H1JgO4u;gYkay1LT)0gO5x$k@0>A5(E(g@s0sX
z1StcL@qRo65d-s%0dxc8JGuj&M34jGJ3IrPL_7oE^?oi(ph-(h1E80!gO4x=%k_Tp
z1DuMNEQ60bmmh?W5kUOyemn#b1BQ|T2&@Dt1B>r|<OC4|_mBZR1I+S%$V;F}OG^W@
zjRArH0t5jDW&i>N0tT4`DFch|e)0qn1NWD3gpVT#xbl7oW&^K{mn?&iGndYUkEb@a
z?tU%=5d+GR0ZIfb1C{Q6x&s$H1VIDw0D=S&1GmfpN&~w@tOId^1Jpb62q8;5O9U$e
zyqBwmk6;nEk^xu)qB{uW1Hwc~1S<o~mmG$VD*>~YNQRFoB%|+s$OJ(HOIQRE1DBBj
z<OC}NpYDEG0|{mXECZo`SOgIR-?x>9kKQf;+qZPJk7WY^rMKF+kN5!r+PBlXk7fe_
z%eOkdk9q_Fwzsvyk3Rzq-4^l$UIYATAOsNu<Cg)(k0Akzw<E@n{SzLK6bLQ@AW8#Z
zE(9wBiEPLN7oG$W1K<;o1S<oHY?p7LkR}P26j%rq1YZM~mu=dQXBD~>$OJ0`-fLI{
z5d)$W2ucJi1C@1`9iosOO1h;1f&?oApW=R=1Q7$_odJSdlmsgS_R)TH1Q7#@t^tAs
zD+B-0emn#$1EGId1Q7$i6j%eebdXD+NlQxu#|wf3c?diND+8RDP}`440hpJW+m9Os
zsC1qK!<V_+k0$|*m*LxwASs^|SOhBrz-(9q5d**zN(3tdw`^Dg30MRy1EGId1Q7$X
z6bMQLUIW^fD%_8>9_SPZE(0J+17I!$D+91=$O9Li1Q7$d6p#cf1F&nCZ=sMT3h)$I
z2o(fh1HWsRZrYD#6_XXn1S<oDYghyk1N#&RN(3tdymXfxqL3X+x}^bv1S<od;(ne4
z5d-0!0fJkU1S<pf(SCFU5d(^@0fGc81OL%}JOnHQp?_Ee5d)VMSOed4kV~LROG^Ww
z41xoB2s{KU1H)^VQQMD4BJmVh1MHar0x<yt&gOpd18i6W6apdv0wWZ=1J84w1Dp((
zy4#N?0<>$FA)}BVDfkpv1S<pQYghyk1DO;`1S<pHYghvbSOhEsp?_Ee5d(}B2ucKA
z1BI99-jB2%!4(KD10YHRU@inJ1J`QE0~ekI5d)GHkOV6O*J_t<p^zpD$rV@#6$D=c
z=4zL2+K*-x>=noaD+8=*SOgIR)D;Ly1S<pMbC(^WkR409r2&EjD+8b6ex3vo1L2(k
zf?JdXD+Bh?esly81B$Kzf&?oA|IvOt1S|uge^>+&1Mn4C1BP>uOQ1<hO9S}~f&+O7
zJOnEP>S~u!+mA;g$`x1xlbHbmF#!Xo=YH}7Y*+&n0wMtdBNV#>_Hv#B^$eG~+m9v!
z+iI5~qmUml(G^$(D+7*eSOgIR@)Sx0D+7gVSOW=I1S|uge^>+&1LYM6N(5d5!)uTe
zO9T-EpqH-Uk1i|LX$XP@Yy*%20R&zH@_lp!5d*>)kX%Ls4)OyH2ohZcNCWR@JWB)-
z1CW;o;*T8xtd}L?k3eP26$rWnD+7>a$O9Kx1VIDw0FVR`1Lzr81S<oOW%2|O1I5Py
zkO&q7AqesV4Q2zB(|)X=1S<ooZU{;Phj*+430MRy1EGJA1Q7$YkpWl)D+8kMeh72}
z30MRy1EGJ`1Q7#}m&)RgYB|Maf&&Ry1S|uge~<(b1Ew3ugINgv1S<ouWeB<h30MRy
z1EGJq1Q7$K8^{D+1EyvOSRp$EDFg49FXE3XJ>LLW1StdV16TwR1Cka9x&s@K17$n}
zD+9`8N&^=>1VIDw0FVR`1Fjla1S<o|Wby<N1JN6T2qFY41LT*~<Bu^)x}^bF1S<od
z;(ne45d-0!0a#xID+Bh?eslu~SOgIRfvy2c1S<po(SAGxECZo`SOgIR@EceJD+A_a
zkOK)y1S|uge~<(b1J{>N<c}U2|73y#2}%Sk1EGJA1Q7%58_0uM2>t{s1L~KD<c}T|
z>>Kg~UIU?K2!a#?6!H@YAp{Wv!k4<_j~^GQRRT2u69g#(^#D8sUIUDO2;>A21D==g
z<c}#B%&C6@H31U@DFf{QJOo|?lz#~11Q7$cmoep!DFL>ZTjh^)0o0ez<&Pl&n78re
zj~pWb-IoptkV!nbYY1it6Fc+s1Q7$*E&?n82v`FN@;eA20~%(bO9UwZ0t5mBsZL4+
z5dZ=N0t5Up2%ZQb2uuWD1Mg0kp$d>L7vM2i1Std6PzWvqfqqy7ECWiO1Q7$om)HrA
zcu?6;2;>8Sepm!714>u~5d+*Y@&sN3+fecd6Fc+s1Q7$vF9Iw92v`FN@;eA20~%(b
zO9UwZ0t5mBu}w+@5dZ=N0t56i2%ZQb2uuWD1NWDp3Xm@s*fLlIDFfY42rdJGepm!7
z14^C*5d*xJ<_eH_BH>R6<O6|zSOhEsN>~ID1Jp9|1YQH@Px2u<J*^)E5d*-NfD4eo
z59Uq+Gyx$yJ*^)E5d*<7mw^kA9hYnjkfIIcO#(ClOM^)S5d*?9mk=qCB9}c4kZlX4
zO#(Ur69f?h{y3KqDUc$U*A0+u0gkst4v>ukf6tZy<O3535(Fs&`sjX21Q7%CmH~7F
z6Y>LXbOb2_jp=?^1Q7#<mjMDR0p(dc1StdB?|wW45d*1^0ayen1K00<E(8$(0t5mB
z){p_Z00IOB2G#^A1_P?^eh63u5d+td0ayen1F7$RW&{xf*N_1`1Stcl?|x<k5dZ=N
z0|Eo!msAmuRT2E^ex3t?epm!714>u~5d*xJx)G3O0r{695|ADNh?hVTkX;@BR`LW9
z00IO81A!@?1Std8TGj*+1FI!G1StcPm%0*=J2ToR)&wa7fLLY(5d(lISOh5p=vY_-
zfqqy7ECWhd1Q7$=DOdz41KL)s1S|tekOUC}yO$^vkTEi?R#*fP1LY}LgINgv1StdO
zSO~fUfqsGnECWhd1StbH$OJ3{N>~ID1I{T3E(B-;=eLU!kgEbE#2N@ZTmu(M1Stdk
zWLN|d1LztEAOtA`m}T+=5d(u8kOW@?>z7*=kS;^y8dwA=1JP&X1A%^61S|teo&*sC
z%o<o(0~ZLM0}%o&0lNcNN(3nb=x2fhfqqy7ECWhd1Q7$^8pwlL2>t|K1B92%7La5C
zikC4LkRSnsmr@sydjkG7m(LfF9vAvZJOmK{0t5mBhBpE>0SF-oOaxy8<(C8)kS|Z_
zHdq4*2p|Yc2uuSH<O6Xo0}o~hT(AU41NBNg1Q7#^HA)B+2ut&p0xT2=paa`c@=Kse
zSxW<^E`kGT2v`FptO%5_1V{tTN<0J+1GBe+7?4x}1inZJN|(`Pl^FvO1Gks<8IUms
z?okMW1ERMr8jzy`0lb&d9FQLY=9lptkR}1zw-_CeiUKW|H;@RP0}1jw2q6O+W}!<2
zDF6Zl0t2f>N(2!A0t5mB`Z@@n2q6ee1YZO3m(L%NYGM1B0ayqy1Stch>wcaD5d+<n
z0SI&iDFdVHemn#b1KpGXSOZ1~bOb2_?(BXp1Q7$0lL1%=6a*;)+v|Qx1Q7$LlmV;*
zf-JfNfK-qJfGj)%fK)sKtMGm<OQ1<hO9RxGnjnxr0sfcTAdn>;!#Y?4vrYmp0ZX7s
zSxW<okO6}G1OD%R0yF^w(3cz`kQoTL@O}a~0Rx7YJt2^o0rR)kA&_VU0fLu5CXg5w
zvqVY+5dZ=N0t50p2#^RN2uuWD1NoP2CXi}4v6%r_2rvXG1Mlj7kOUC}la&DobOb2_
z{_B381Q7#-lmR>gzwdqs5=)>-OG^X5kO6}21TF))mk%e9C;_{dI46)H1(i*n1H6}B
zCy<c~l173cJ%dXG5d+Vc`X`Wb0nV3gD3BNd<(GyikVyfQm(VDX76IFr=O~aM2!v|n
z69f?h@RtuMkdO<*QUW>w69f?h<SmyFDUc$U%qfs<0mrvGDv(bB0f3i^E0A{xtjB&z
z1Q7$(mkBJ8#4Og3e~<(b1Gn)2@(9;32ulb{2ulJh69Ov}1StcXkADIy0R$`qN`eFt
z1Av$LERba**pPob1Q7$#?*Ir`16v55kB$Q<E(2L~1Std7kbjT_5d+_sfh~{|2-c8)
zkOUC}vX`7KkTDm}j(=DL5d*UE0Fc!LDFe@re^>+&1BsX9Es#72){uXY1Q7%8mnbff
zvk1Q6eyju$1HzZ%E|51C?dE=11Q7$1nE}WIDFf~1ex3vo1Cy69FOW|Gz?Xn8kQXPB
z@Bj!(1Stc@lYiC(5d*vI0D=T51A~u$SOgIR{qO(+Bmo2|1I3rVFOWDgi17dj$OI_^
z@Q{B9N(2!DknaG31Std9k$+eO5d*010D=T51KE#%W&|t)N>~ID1M8P4Fp%FIf{%X!
zECB=&1OD&;kO*Z2DFcI#e*!E41Q7%Mmm@Kd(>0Uhe!2sJeyjv614>u~5d+@v0D^;A
z2>t{q1K5v$0ww_jECWiO1Q7$?@Bj!d1Std8k$-fTzNwQI6SIH11YQG{1CRu41IGdZ
z1Q7$QmsB#4E&;-~gEElE0e`kV0xAImsNDbv5(C(xf2;$j-2e!>1Stdl<bFH^5d(so
z0Rn*m0p(eb1Stdk<bKEm5d({x0WJh71J>()W&@vjf&<|OkOUC}s+0j(^jFuC0~!dB
z1L}SPB>@B}1MTI0SOgIRlbQhnJOKnL1MTI0tOOAQlbQj#tywz+DLw<u?0%jE5d*f9
z0fGc61Ip}vE(8$+q`?7@10e`50}W;aky5M#DFdDCex3vo1LTtdkOV0Moa}yr1Q7$|
zmj^YFKmqTUNHvf)2#=cqSOh5plb41ykXsMungLh@UjyUieyju$1D%)fHIR}5f#sKk
zHjttS(LJ8n1StdfmmfWl6&3S60we(hDFe?&N(2!Du{mZ0DFe@!Nj;E(2ZlU?1YZO9
zm(e|t8khPvkbDQ&Jy--O1KpQ>IFMNZzn9lIkSz-M<bDD)0R#~P@s|!ckZTB!<$jO^
z5d(&ofjN*F0qU2VIglL_);(AQ6oRdTSpp{z1YrY(LI}E-5kQb32lze61YrY(m-9K0
zP7ATl0Rk)m*92Yz=C@2bkRJg?{5=9E0SG(;(cJ(DE(5Qkf2;&41K>ee1Q7%An*joN
z0p(eb1StcK<bJFK5d)%}0SJNvSqO9pOOgW`2oeJd2*?B}1Gty5JCG;{{+j_<1Stcg
zm*G2*VgZboL_Cl$0g{(=JdmRZmp)k71StcummfWl6&1NY0we(hDFcZ{N(2!D_c>++
zDFca@Nj;E(2h2Qz1YZNOm(e|tmRb6o0ayqy1Stci=YF095d++q0SI&iDFdYEemn#b
z1KgMaSOh5pwdj5bW&{xfk(U9G2rvXG1Kj6+N(2!DrkDY`yaTN4eh4m0ph-(h1JjcM
zg6#+t0}}|Bu^^Bdm)bs%4gtQG={}HC9fdwv1Mx}%Fab-TNm)w+iIf3?`vd>$egZ53
z1J9RPKae3Ux$J%dH~|BNlmV;*5rPA6N&^uBA_2Pty?zLg2tG)51StcF<bDD$0R#~P
z;FqUAkSIEf<bDVe1Q7#~oB;wO0fPiA1N-BC0xtmsUjyL)SOh5pYzRsOECWiO1Q7$f
zn*s6z(4v2Wodhfc-<JtMkSR~TKIDU02>t|N1HVB6C;<d11I++f0}z4)ECWhd1Q7$8
zKF9<q1Diyi1A%_51S|teSOi}K$^cjd5d*tD<OD1O<Ut6K1Q7$UK7s@?1I4$2K#*<_
zD~gi=0wn<iG6VMPeh54RA#?;01L=|hf&?-HnC*TDE(8$+=#l|g1Tq8mml{!!HUYht
zQBjaB2<Gs9x&#pe)t81*kS+<>?0&ifjsysFmkv^pDVJ_+lLwdWQIH4;{*wV%*90pA
z*O&fLkV^>vlL1%+D+9cjQBsi10lAkRQ;<Rdzn5WCkSYl4jRE8YD+9!rlT(mXMAwr6
z2y_4f1OWzg1TzK$<m-N(1Q7rN1Oo$@lmQ5400IO91JWZ1o&W*_12_Nz1Oo%ITs#Cp
z1LN&}<OC4_0t5pCo|h|BkW>M@mw;4|cRt|a00@Eu9SA%GG6UD0e{=(0JOd$u1Q7#`
z;sB5YG6VFTfAU(6Sp*RSsNw)v0wom#2?!DhOp*v81Tq7Nmlsx$Ef&_?0G<e02r$0{
zG6SNce+Uu;5d*Q@ms(bkR!OF#f5-$81KQjG2y_GTeu4uX2s{Kb1GAogbOT;I10jM0
z5d*Q|0FVSS1JIs-@>-5r1Q7$w;Q&|yB^3h+2oeZPk_aILG6Uq77gmrh75du%o(Ne8
zFuw#c1OB6b2oeMl1J##VR*+W-f~0@Q1Q7%N+n2dkkR2GwpMOdO5dZ=N0t4RQ0G<Rg
z00IO81Aw2G?pBaN72V(fSOy3Z1_(a`2?!F{lJf*I1Jjp8SCB0hi`xL62w4a(0xT5+
z2?!GAlJf*I1HhMuSCB3hp4$MP2w4a)=>#$Z+N6I75(E(g<(JA=kh@TurGJ715d)&z
z00JTbgINgv1NM9f)&m^~JOnZW=b?Xe1717>A%X-E1O43qkOVRVgra})T8>!+5d)as
z09XPg6$1$f5(rF^2q6SA1DL6o7*>!h7RJ~Bo(Ne8Fuw#c1G1-o2oeMl1EJWLT2_!&
zO17te$OI7s%Gdx1bOZ8!f&(20JOnZW!=it51717>A%X-E1EJjjkOVRV;G%!>T8>!+
z5d*y409XPg6$1$f5(rF^2q6SA1N5nv7*>!h7V6gko(Ne8Fuw#c1CFSF2oeMl1I5>u
zT2_!&36iLP$OI7s?$?*OR*)SS*`t3-1Q7rN1Ofxj+yI^gG5`Vu0t1kvm+n@OK^D#2
z09XbH5(WrA0|^Kc*pl-EG6UnPmqk~QECKtM5nGTH0??|Lh*yv;6^_>co(Ne8FzEy`
z1L~-M2oeMl1J##lTaddItEqp21Q7$0*8l<{0fSix{sb}ul(#utkQM|x-qrvJ<O3cE
zJOnZWrlx;%1717>A%X-E1HIV*kOVRV#HN4pOpaLu5d+!T09XPg6bMWNG6UqTe+WF5
ztB@F%P*so$mw;iA2pa#>0G<dx0|^Kc*a(A?1Tq8gtA7X}1Q7$_m(F02I#z|Oe+WDT
z5d-_v00?vhUIT@xe~^Pa^H~HD1Nhef2;>7E2s{Kb1B9i2bOT;I10jM05d+`a0FVSS
z1EQsW@=T6d1Q7%K+5lJrB@_ru1Tq7{tbYhRmaC8$mrzxZ3YUOkkO&*0)c~FdKLZH}
z64(fXk_0jX&Z>V1AOsNugVmSMV30X~<EnoMJOmK~oz(ycbOX@A0FVSJ1GnOS2s{Kb
z1NZ>s1Q7$#oB>z_DFe6Sepmz%1Ie8MkOVRVuj76QE(9q9h<~00G6VPk2#|vW5d-6#
z0fGcF1N!2AtOFr}y#x^h;+z4x<x7u~GXp4Q1Tq8b;(mez5d)Q-0fGZqSZ4?sOV|S%
z2v`6D1OWzE00IO8W(ETa<ODJUt(SCSkS+<amjMV81Tq7)m#bosY!ADa0ayew1I6fm
zJOmK~iI*m0kR<`Yw@hP@7Xbl+mt<s+Bmt?HiDZy050ah%SOd9#0yF{G1akx6m&9a{
zH35y6`ecwrHKyo(f&>u*?v?=vtOPOxlIec(1Q7#--2eh60SG(<G6R^Me+VuF5d-Mr
z09XSfSOh5p{hOC<XOJEOiQ|`zXOJBX+@ya95?QWSOG^Wym%U|>D*~3Bm*8iR9~ZLO
z09Xiv^aL^k&8L3|SOc?q@&pkBg_j^^kSYSxp_f5ukSqbkmw9H8BnR`Pe+YB~o|mL%
zkee-nqJN$Q5d)ju00<xh<^Tvv1M_;22s{Kb1E`^Y2rdK>1K8dGSOX(i1StcJmuzQ{
z9|8H7jAxJ@4D6_X2ohPYS4&F+l9#q;kSGDDm)~cQAQqt209Xiw1Tq8QtA7Ys1G9Sa
z1Q7%BmlbG`Dgow~KxmLH0kxNIXpkfagQ<TAbOVl;nP`xX6_TcZ<OC4|i`f7OAOqt7
z2ucKA1NpZ+X^=Al0k@ZMYmibGxmZdB5dZ=N0t50VSOhBo0t5mBrC4;Ak-3l?5b`IU
z1_%-c1Bqe?JP1nz`IiK2kUB!6SO}g230MRy1EGJA1Q7$XC|K77Ujvp{@&pkB>L=vb
z1StdTTS^1IdsqY!1J)=Ao&+fa>|06%5d)MYJOn8N?3aLSkW2~3Cs+h41E5&sm)~iV
z84nQy?kHFT39JMu1HV@Y5|@G7j}w;=ZIA>8^CwsX8P=D9w~!SJ1StdSRtUNT5d)T&
zI&F|13Hw$8B>@Bx1BI7iZIBpZ(pLgY0R#~Pt|%@8!~h8L1StcKR{~1`1Q7%BC@up^
z2ucJg1H@MXB>@Bx1H33M1EGci2oeMB`F`>P%m4x+0Ruq@$OI_^saFCo0R#~P*eH+#
zJpwoZ1StdSR-ObA1C^KCZIDY5^;cK~5d+RBSc6#z{sb!npO-gokT(jqC<q_~G6Sb+
ztd~7pkVUt>Zjed>NxJ0#0xAK91S<pdmVZhF5d)*?06YXM1N4@ESOgIR@#p}M1S<ph
zmVfdD5d)*?07?Ww1Lc>0JOnEP_<mRf5d)*?07?Wa1Dlsnagb<Kx}^bF1S<od;(ne4
z5d-0!0a#m<1S<pf(SCFT30MRX1A(pqN(3td|IvOt1S|uge^>+&1HkD3SOhBrmY08!
z0|`n5ECZo`kOUC}sp$Y%1S<oAmke@{Dm9?!00=Gv5dtg$y90~G0P+MY1NhQ@@&rKx
zOIQRE1BvJW@&glCgINgv1S<oYmwyQ40|`o(UIdU93K0Xm=>W(CUIUG{i*k@s0RiKe
z-gA&20jrn(bC6vDzn5HekT(LbVwO^n6ql)VkTn*j9#{xj2rmLG6$1$f67-Vu1Tq8N
zX_x=1kQtXyRge-7ff$|$SqLx#4+w$;G6U_GE_IL~4}cz610e!10R$@p#%2iQmtotH
z6$OF>5d*20cXf~_4zglch65ae17Ly#5d)r=taXq=4y|JXU;zX{151Jg5d(;q{B@9M
z1Gs2{moRmZ9u&PE$b(r3{sdkF>SF>U0TTof1A~{Dc94w$=C>zzkQyNYikCu|kU0U)
zmwcFz`T>TQcbSkm0p6FfnUMMc$d{~|kU0UQx8|CV4*?LOBLXx510=`;TrQpjG6S=h
z1)Pv34(}wo2owlR1YZMzTL==DZ`zL;mr|UNOc&85o&(#80ayd2^?sfNGXs@cW&{xf
z@0YclkS7$5WAaO&Nm)w+@En2zX9!9JG6UzA@SKn^0rHn2oseM%mtz7j0R!xpiJg#L
z8iHY-1Q7%J9`Xb-1BPLq1Q7$09|*bxUIV+fo{$XzFzO@%Gy%v1?4SYi1Stdmw|+_l
z5d(q90SJ)H10ez=0Rs(Y1O489te^y61H4%X5|?l$kY@*zCs+h21EQC!pO7p8`j^<B
zkZJ+yw?m+i4FMj^A_6o410l!*UoHeI1JPTa0~ekI5d)YckOV6O(U)$akS7b!Bv=R(
z2uuWD1KyXQp^#`4^d!gxD+8okSOgIR+$0D}1S<pEmmH#y9!t8V0fGc81E1o4o&*sC
z;hh13Ta*MV1NPB=bOaFtimm~I1S<po(SAGxECZo`SOgIR_#{{Z{$r3!ph-(h1A!cZ
z19=EM1S<pLTbEJWk4FK}m(8M(8wKuTo&)`t?V^w}0n?WtqmUpe*d$m4D+7mISOgIR
z`6EgMD+B&pSOW=I1S|uge^>+&1MMUTN(5d5l(%W4kPiVhkR%8^10D!w1Tq7oT66<m
zSOXz~1Q7$kC6EL%1HoGIOpaLu5d+*MSOO&!2uuVr1L0u^N|vjT7?)60kP<ibjsczo
zGXuQreh3l-5d+(k0aybP)(Bb%S_oMP69h5?mF#}<1Q7$%lL3MRG6Svbepmz%1Jjod
zrI0xX*<UUM5d-9xPNk4g4}xF_JOmK~{vQZ*1Xly2m#w9c_d2T|x&$i&i(d#l1Q7%E
zAc6!d1B+h>x&#pe^&m<FD+8imSOh@>@c_C6D+7;zf&>u*g{}d*1S<pNm)@q39~8|X
zkOV6O^<H!YG6Ot<2t5Q51KyVrr;tp4>Rt$f1Q7$4AxZ=*1L0m+0||lzECZo`SOgIR
z^&wc-1S<pSUI>r`5d)VYN(3td;9gh*34#PH1EGId1Q7%FAz0T0G6Rlk2rdJ&ddLJ3
z1HT*c*aHm$A^`|z1S<pBUI>r`5d)|pN(3td&t6yq34#PH1EGId1Q7#=B3Rc%1S<pA
zUI>B&5d*0qN(3td&R$pp34#PH1EGId1Q7#>B3Rc1G6Vc+@&mJa$OI7s(HqFv0}lcs
z0RsuF2t5c{1Tq7YmmsK+CjqIKNT`sY0lt^%sE{TBjF%9pklg|Bm+q;MZvlgsUaF8c
zY4u<Tf&>u*iyuk^D+BCcSOW=y1S|uge^>+&1L_}G2w4a(2)zg)S&>;;O9(#*Ay+-S
zlJf*B1L9x^kOUC}o*zmCD+AeJSOW=y1S|uge^>+&1N|RZ2w4a(0xT5+2?!GOlJf*I
z1BRFWs*o%J*q1A-kQxJ(W~`T>tdJlD$OI7s!MAOzkT3xO_LrWlkR}V2AjpGR2>t|D
z1D&_etdIf&0hE{0uaIZ~*_S)8kSi6t9FPPn00IO81FmCq1Q7rN1Ofw_mwT{~C<5DK
zmu|_BNCEqo?y!(K0mzpqv5-k1jvZJCSqLx#=&=C^JOiE1eu4wzu>m{-oz8wd1S<o;
zmyxlM9TU_XSOX$v0}}}31G9e!<OC}N@0Y=`kWc}FmlCp&ISJ%s0yqH#5d*51RI-pd
z9js*tW&{xf%^C=F2rmEv1OWzE1_LIn1S<o9WP+EmA(0sk5d)GP$b(r3{sd<O$+yw6
zkPHI>@|WJVkZTwFRZ0XA00IO81F<St1S<dn1Ofx*Rdkn;xsV$Wu_~Sh2oeSZ%3laP
z2ulOFmy5QLIsxRD#<q}2fAT6=1NEW-)}IKS1B~5%JlF&)1IAZa1GIiv1Iz(D1Q7$e
zC|Cq51EW<+1ET^;1KoW*1J?jp1Q7#<Dp&+71L0N30|{6JECZo`kOUC}tt(gq39JMw
z1E5t15(5cX1S|ugf5-$81F|Yu12KXFXb9v3<p6X8n}0k5<p2mgmyWlP8UezWske|Q
z0pOR?w~#mqidO<S0R#~Ptd|qGkZLibR#*cGSOhEsp?|sr5d-!rf`eHI{sb!n;8g-8
z0RstG1S|uge~<(b1NJKN1YQHBm#(;wUIE9K4Y`m~7pYWA1Q7rN1OfwsES>}_00IO8
z1C>;Cmyx-U6qky*kRlL*EaV0V5(WeLT?jk~69b8tyt$Acf1FeZW&{xf<SR-9D+7~M
zSOW=I1S|uge^>+&1H&v>*9f5q6HAdxOIZjp2q8&5yOIbK1S<o9R0y5}30MRy1EGJA
z1Q7$(ELaFx2rvXI1LIX#1GIiv1Iz(D1Q7$BDp&+71He;C1ET^;1KoW*1J?jp1Q7%2
zD_8_81O8Li76S=b1S|uge~<(b1CT9P0|~kWD+9fkaJrBn0fU#3x{%NV(o}+%fJBiT
zA`t`GEXadd2>t{s1N~E;0|{6JECZo`x&#pe*(}HeYy-HrqPvhl4J45lx(IXw3kWU)
zbvy$KSOPE;11e^sNlQzS1S<dn1Ofw+XONdYu#hPgfEkbkD*yrn0t4x0bOaFq0t5mB
zo0mk%kSGC`mu<<ANddr@x5<z>9gJrPf&>u*^cYG6D+7aPSOW=y1S|uge^>+&1Jjoc
z%8)4n?`D^=9g!Lp5d)eTSOXyfGyww>2;>AS1DR_G<d-0Zj}--21Q7$tmx0QVEDo4w
zSX=`fSOZ{K1Q7$Sm%GZ4P!62aeh72~K?6%z1Q7$Uml?~DZv(t%f|s!&kr@pU1KAnK
zgINgv1Z)GMx3SBR)E5ENw}#`82>}7%myqO;dI9T)3+0fv3+0jc92xe{0RaS=0Rz*p
zeo6!p1Fg>ibjSpm0Rz*w90HQ?0SWld0SKN1D+An@6$FyX0nmrw1d@l~1e1s01eAy1
z1eJ&21eS;31eb^41ek~51eu561e%B71e=H81e}N91f7TA1fGZB1fPfC1fYlD1fhrE
z1fqxF1fz%G1f+-H1f_@I1g3}J1gD4K1gMAL1gVGM1geMN1gnSO1gwYP1g(eQ1g?kR
z1h0qS1h9wT1hI$U1hR+V1ha?W1hj|X1ht3Y1h$9Z1h<Fa1h|Lb1i6Rc1iFXd1iOde
z1iXjf1igpg1ipvh1iy#i1i**j1i^>k1j2{l1jC2m1jL8n1jUEo1jdKp1jmQq1jvWr
z1j&cs1j>it1j~ou1k8uv1kH!w1kQ)x1kZ=y1ki`z1ks1!1k#7#1k;D$1k{J%1l5P&
z1lEV(1lNb)1lWh*1lfn+1lot-1lxz;1l)(<1l@<=1m1_>1mB0?1mK6@1mTC^1mcI_
z1mlO`1muU{1m%a|1m=g}1m}m~1n7t01nGz11nP(21nY<31nh_41nr051n!661n-C7
z1n`I81o4O91oDUA1oMaB1oVgC1oemD1onsE1owyF1o(&G1o?;H1p0^I1p9~J1pJ5K
z1pSBL1pbHM1pkZR1OboW1OWk$-~<8zkKhCY0gvDW1OboW1O)+)-~<K%kKhCc0gvDW
z2mz1a1PK9;-~<W*kKhCg0gvDW3;~be1PuX?-~<i<kKhCk0gvDW5CM<i1Q7v`-~<u@
zkKhCo0gvDW6akOm1Qh{~-~<){kKhCs0gvDW7y*yq1Q`L3-~<{0kKhCw0gvDW908Bu
z1RVj7-~=84kKhC!0gvDWAOVly1R(*B-~=K8kKhC&0gvDWBms}$1SJ8F-~=WCkKhC+
z0gvDWC;^Y)1StWJ-~=iGkKhC=0gvDWECG+;1T6uN-~=uKkKhC^0gvDWFaeL?1Tg`R
z-~=)OkKhC|0gvDWGy#v`1T_JV-~=`SkKhD10gvDWI028~1UUhZ-~>7WkKhD50gvDW
zJOPj31U&(d-~>JakKhD90gvDWKmm{71VI6h-~>VekKhDD0gvDWL;;WB1VsUl-~>hi
zkKhDH0gvDWNCA)F1W5sp-~>tmkKhDL0gvDWOaYJJ1Wf^t-~>(qkKhDP0gvDWPyvtN
z1W^Hx-~>_ukKhDT0gvDWQ~{6R1XTf#-~?6ykKhDX0gvDWSOJgV1X%%(-~?I$kKhDb
z0gvDWTmg^Z1YH4--~?U)kKhDf0gvDWU;&Td1YrS>-~?g;kKhDj0gvDWWC4%h1Z4q_
z-~?s?kKhDn0gvDWXaSGl1Ze?}-~?&`kKhDr0gvDWYypqp1Z@G2-~?^~kKhDv0gvDW
zZ~>3t1aSe6-~@63kKhDz0gvDWbODdx1a$$A-~@I7kKhD%0gvDWcma>#1bG3E-~@UB
zkKhD*0gvDWd;yQ(1bqRI-~@gFkKhD<0gvDWfB}!-1c3pM-~@sJkKhD@0gvDWgaMD>
z1cd>Q-~@&NkKhD{0gvDWhyjn_1c?EU-~@^RkKhE00gvDWi~*0}1dRcY-~^5VkKhE4
z0gvDWkO7b21d#!c-~^HZkKhE80gvDWlmU<61eF1g-~^TdkKhEC0gvDWm;sOA1epPk
z-~^fhkKhEG0gvDWoB@yE1f2no-~^rlkKhEK0gvDWpaGBI1fc<s-~^%pkKhEO0gvDW
zqydlM1f>Cw-~^@tkKhES0gvDWr~!}Q1gQa!-~_4xkKhEW0gvDWtO1YU1g!y&-~_G#
zkKhEa0gvDWumO+Y1hD~+-~_S(kKhEe0gvDWv;mLc1hoN=-~_e-kKhEi0gvDWxB-vg
z1i1l^-~_q>kKhEm0gvDWyaA8k1ib-|-~_$_kKhEq0gvDWzyXio1i=B1-~_?}kKhEu
z0gvDW!~u`s1jPZ5-~`42kKhEy0gvDW$N`Vw1jzx9-~`G6kKhE$0gvDW%mI(!1kC}D
z-~`SAkKhE)0gvDW&;gI&1knMH-~`eEkKhE;0gvDW)B%s+1l0kL-~`qIkKhE?0gvDW
z*a45=1la+P-~`$MkKhE`0gvDW+yRf^1l<9T-~`?QkKhE~0gvDW-~o@|1mOXX-~{3U
zkKhF30gvDW<N=T11myvb-~{FYkKhF70gvDW=mC%51nB{f-~{RckKhFB0gvDW>;aG9
z1nmKj-~{dgkKhFF0gvDW@BxqD1n~in-~{pkkKhFJ0gvDW^Z}3H1oZ)r-~{#okKhFN
k0gvDW_yLdL1o;7v-~{>skKhFR0gvDW`~kn<1pNUpdb8Y+r~m)}

diff --git a/gklearn/gedlib/gedlibpy.pyx b/gklearn/gedlib/gedlibpy.pyx
index 54f24b0..7fb7e84 100644
--- a/gklearn/gedlib/gedlibpy.pyx
+++ b/gklearn/gedlib/gedlibpy.pyx
@@ -112,6 +112,7 @@ cdef extern from "src/GedLibBind.hpp" namespace "pyged":
 ##CYTHON WRAPPER INTERFACES##
 #############################
 
+# import cython
 import numpy as np
 import networkx as nx
 from gklearn.ged.env import NodeMap
@@ -177,14 +178,16 @@ def get_dummy_node() :
 	return getDummyNode()
 	
 
+# @cython.auto_pickle(True)
 cdef class GEDEnv:
 	"""Cython wrapper class for C++ class PyGEDEnv
 	"""
-	# cdef PyGEDEnv c_env  # Hold a C++ instance which we're wrapping
+# 	cdef PyGEDEnv c_env  # Hold a C++ instance which we're wrapping
 	cdef PyGEDEnv* c_env  # hold a pointer to the C++ instance which we're wrapping
 
 
 	def __cinit__(self):
+# 		self.c_env = PyGEDEnv()
 		self.c_env = new PyGEDEnv()
 		
 	
@@ -192,6 +195,11 @@ cdef class GEDEnv:
 		del self.c_env
 
 
+# 	def __reduce__(self):
+# # 		return GEDEnv, (self.c_env,)
+# 		return GEDEnv, tuple()
+
+
 	def is_initialized(self) :
 		"""
 			Checks and returns if the computation environment is initialized or not.
diff --git a/gklearn/preimage/median_preimage_generator.py b/gklearn/preimage/median_preimage_generator.py
index fa1f4db..8753292 100644
--- a/gklearn/preimage/median_preimage_generator.py
+++ b/gklearn/preimage/median_preimage_generator.py
@@ -669,6 +669,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 		options = self.__mge_options.copy()
 		if not 'seed' in options:
 			options['seed'] = int(round(time.time() * 1000)) # @todo: may not work correctly for possible parallel usage.
+		options['parallel'] = self.__parallel
 		
 		# Select the GED algorithm.
 		self.__mge.set_options(mge_options_to_string(options))
@@ -676,8 +677,11 @@ class MedianPreimageGenerator(PreimageGenerator):
 					  edge_labels=self._dataset.edge_labels, 
 					  node_attrs=self._dataset.node_attrs, 
 					  edge_attrs=self._dataset.edge_attrs)
-		self.__mge.set_init_method(self.__ged_options['method'], ged_options_to_string(self.__ged_options))
-		self.__mge.set_descent_method(self.__ged_options['method'], ged_options_to_string(self.__ged_options))
+		ged_options = self.__ged_options.copy()
+		if self.__parallel:
+			ged_options['threads'] = 1
+		self.__mge.set_init_method(self.__ged_options['method'], ged_options_to_string(ged_options))
+		self.__mge.set_descent_method(self.__ged_options['method'], ged_options_to_string(ged_options))
 		
 		# Run the estimator.
 		self.__mge.run(graph_ids, set_median_id, gen_median_id)

From 85ad732a887efb1c41053a5882df66cc28bf48dc Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Fri, 8 May 2020 16:43:28 +0200
Subject: [PATCH 08/17] Allow zero costs and no triangle rule when updating
 edit costs in MPG.

---
 gklearn/preimage/median_preimage_generator.py | 261 +++++++++++++++++++++++---
 1 file changed, 236 insertions(+), 25 deletions(-)

diff --git a/gklearn/preimage/median_preimage_generator.py b/gklearn/preimage/median_preimage_generator.py
index 8753292..20fbec6 100644
--- a/gklearn/preimage/median_preimage_generator.py
+++ b/gklearn/preimage/median_preimage_generator.py
@@ -39,6 +39,8 @@ class MedianPreimageGenerator(PreimageGenerator):
 		self.__max_itrs_without_update = 3
 		self.__epsilon_residual = 0.01
 		self.__epsilon_ec = 0.1
+		self.__allow_zeros = False
+		self.__triangle_rule = True
 		# values to compute.
 		self.__runtime_optimize_ec = None
 		self.__runtime_generate_preimage = None
@@ -79,6 +81,8 @@ class MedianPreimageGenerator(PreimageGenerator):
 		self.__epsilon_ec = kwargs.get('epsilon_ec', 0.1)
 		self.__gram_matrix_unnorm = kwargs.get('gram_matrix_unnorm', None)
 		self.__runtime_precompute_gm = kwargs.get('runtime_precompute_gm', None)
+		self.__allow_zeros = kwargs.get('allow_zeros', False)
+		self.__triangle_rule = kwargs.get('triangle_rule', True)
 		
 		
 	def run(self):
@@ -382,7 +386,8 @@ class MedianPreimageGenerator(PreimageGenerator):
 
 	def __update_ecc(self, nb_cost_mat, dis_k_vec, rw_constraints='inequality'):
 	#	if self.__ds_name == 'Letter-high':
-		if self.__ged_options['edit_cost'] == 'LETTER':			
+		if self.__ged_options['edit_cost'] == 'LETTER':
+			raise Exception('Cannot compute for cost "LETTER".')
 			pass
 	#		# method 1: set alpha automatically, just tune c_vir and c_eir by 
 	#		# LMS using cvxpy.
@@ -461,32 +466,34 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#			edit_costs_new = [x.value[0], x.value[0], x.value[1], x.value[2], x.value[2]]
 	#			edit_costs_new = np.array(edit_costs_new)
 	#			residual = np.sqrt(prob.value)
-			if rw_constraints == 'inequality':
-				# c_vs <= c_vi + c_vr.
+			if not self.__triangle_rule and self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
 				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
-				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
-							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
+				constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+							   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 0.0, 1.0]).T@x >= 0.01]
 				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 				self.__execute_cvx(prob)
 				edit_costs_new = x.value
 				residual = np.sqrt(prob.value)
-			elif rw_constraints == '2constraints':
-				# c_vs <= c_vi + c_vr and c_vi == c_vr, c_ei == c_er.
+			elif self.__triangle_rule and self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
 				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
-				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
-							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0,
-							   np.array([1.0, -1.0, 0.0, 0.0, 0.0]).T@x == 0.0,
-							   np.array([0.0, 0.0, 0.0, 1.0, -1.0]).T@x == 0.0]
+				constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+							   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 0.0, 1.0]).T@x >= 0.01,
+							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
 				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
-				prob.solve()
+				self.__execute_cvx(prob)
 				edit_costs_new = x.value
 				residual = np.sqrt(prob.value)
-			elif rw_constraints == 'no-constraint':
-				# no constraint.
+			elif not self.__triangle_rule and not self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
 				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
@@ -508,11 +515,36 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#				edit_costs_new = [x.value[0], x.value[0], x.value[1], x.value[2], x.value[2]]
 	#				edit_costs_new = np.array(edit_costs_new)
 	#				residual = np.sqrt(prob.value)
+			elif self.__triangle_rule and not self.__allow_zeros:
+				# c_vs <= c_vi + c_vr.
+				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
+				x = cp.Variable(nb_cost_mat_new.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
+							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				self.__execute_cvx(prob)
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)				
+			elif rw_constraints == '2constraints': # @todo: rearrange it later.
+				# c_vs <= c_vi + c_vr and c_vi == c_vr, c_ei == c_er.
+				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
+				x = cp.Variable(nb_cost_mat_new.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
+							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0,
+							   np.array([1.0, -1.0, 0.0, 0.0, 0.0]).T@x == 0.0,
+							   np.array([0.0, 0.0, 0.0, 1.0, -1.0]).T@x == 0.0]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				prob.solve()
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)
+
 		elif self.__ged_options['edit_cost'] == 'NON_SYMBOLIC':
 			is_n_attr = np.count_nonzero(nb_cost_mat[:,2])
 			is_e_attr = np.count_nonzero(nb_cost_mat[:,5])
 			
-			if self.__ds_name == 'SYNTHETICnew':
+			if self.__ds_name == 'SYNTHETICnew': # @todo: rearrenge this later.
 	#			nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 				nb_cost_mat_new = nb_cost_mat[:,[2,3,4]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
@@ -529,7 +561,149 @@ class MedianPreimageGenerator(PreimageGenerator):
 												 np.array([0.0])))
 				residual = np.sqrt(prob.value)
 				
-			elif rw_constraints == 'inequality':
+			elif not self.__triangle_rule and self.__allow_zeros:
+				if is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = x.value
+					residual = np.sqrt(prob.value)
+				elif is_n_attr and not is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 0.0, 1.0]).T@x >= 0.01]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value, np.array([0.0])))
+					residual = np.sqrt(prob.value)
+				elif not is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), x.value[2:]))
+					residual = np.sqrt(prob.value)
+				else:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), 
+													 x.value[2:], np.array([0.0])))
+					residual = np.sqrt(prob.value)
+			elif self.__triangle_rule and self.__allow_zeros:
+				if is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+								   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
+								   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = x.value
+					residual = np.sqrt(prob.value)
+				elif is_n_attr and not is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 0.0, 1.0]).T@x >= 0.01,
+								   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value, np.array([0.0])))
+					residual = np.sqrt(prob.value)
+				elif not is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
+								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+								   np.array([0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), x.value[2:]))
+					residual = np.sqrt(prob.value)
+				else:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), 
+													 x.value[2:], np.array([0.0])))
+					residual = np.sqrt(prob.value)
+			elif not self.__triangle_rule and not self.__allow_zeros:
+				if is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = x.value
+					residual = np.sqrt(prob.value)
+				elif is_n_attr and not is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value, np.array([0.0])))
+					residual = np.sqrt(prob.value)
+				elif not is_n_attr and is_e_attr:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), x.value[2:]))
+					residual = np.sqrt(prob.value)
+				else:
+					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
+					x = cp.Variable(nb_cost_mat_new.shape[1])
+					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
+					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+					self.__execute_cvx(prob)
+					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), 
+													 x.value[2:], np.array([0.0])))
+					residual = np.sqrt(prob.value)
+			elif self.__triangle_rule and not self.__allow_zeros:
 				# c_vs <= c_vi + c_vr.
 				if is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
@@ -572,17 +746,54 @@ class MedianPreimageGenerator(PreimageGenerator):
 					edit_costs_new = np.concatenate((x.value[0:2], np.array([0.0]), 
 													 x.value[2:], np.array([0.0])))
 					residual = np.sqrt(prob.value)
+
 		elif self.__ged_options['edit_cost'] == 'CONSTANT': # @todo: node/edge may not labeled.
-			x = cp.Variable(nb_cost_mat.shape[1])
-			cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
-			constraints = [x >= [0.01 for i in range(nb_cost_mat.shape[1])],
-						   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
-						   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
-			prob = cp.Problem(cp.Minimize(cost_fun), constraints)
-			self.__execute_cvx(prob)
-			edit_costs_new = x.value
-			residual = np.sqrt(prob.value)
+			if not self.__triangle_rule and self.__allow_zeros:
+				x = cp.Variable(nb_cost_mat.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				constraints = [x >= [0.0 for i in range(nb_cost_mat.shape[1])],
+							   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				self.__execute_cvx(prob)
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)
+			elif self.__triangle_rule and self.__allow_zeros:
+				x = cp.Variable(nb_cost_mat.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				constraints = [x >= [0.0 for i in range(nb_cost_mat.shape[1])],
+							   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0]).T@x >= 0.01,
+							   np.array([0.0, 0.0, 0.0, 0.0, 1.0, 0.0]).T@x >= 0.01,
+							   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
+							   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				self.__execute_cvx(prob)
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)
+			elif not self.__triangle_rule and not self.__allow_zeros:
+				x = cp.Variable(nb_cost_mat.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				constraints = [x >= [0.01 for i in range(nb_cost_mat.shape[1])]]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				self.__execute_cvx(prob)
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)
+			elif self.__triangle_rule and not self.__allow_zeros:
+				x = cp.Variable(nb_cost_mat.shape[1])
+				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				constraints = [x >= [0.01 for i in range(nb_cost_mat.shape[1])],
+							   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
+							   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
+				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
+				self.__execute_cvx(prob)
+				edit_costs_new = x.value
+				residual = np.sqrt(prob.value)
 		else:
+			raise Exception('The edit cost "', self.__ged_options['edit_cost'], '" is not supported for update progress.')
 	#	# method 1: simple least square method.
 	#	edit_costs_new, residual, _, _ = np.linalg.lstsq(nb_cost_mat, dis_k_vec,
 	#													 rcond=None)

From 7e94ecaf31f35a07fb006c794c147c2a6f044669 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Wed, 13 May 2020 16:31:37 +0200
Subject: [PATCH 09/17] Add k-nn test.

---
 gklearn/__init__.py                           |   2 +-
 gklearn/ged/util/util.py                      |  13 +-
 gklearn/kernels/graph_kernel.py               |   3 +
 gklearn/preimage/__init__.py                  |   1 +
 gklearn/preimage/experiments/xp_1nn.py        | 103 +++++++
 gklearn/preimage/kernel_knn_cv.py             | 418 ++++++++++++++++++++++++++
 gklearn/preimage/median_preimage_generator.py |   4 +-
 gklearn/preimage/utils.py                     |  52 ++--
 gklearn/utils/__init__.py                     |   2 +
 gklearn/utils/dataset.py                      |  20 +-
 gklearn/utils/knn.py                          | 141 +++++++++
 gklearn/utils/utils.py                        |  30 +-
 12 files changed, 756 insertions(+), 33 deletions(-)
 create mode 100644 gklearn/preimage/experiments/xp_1nn.py
 create mode 100644 gklearn/preimage/kernel_knn_cv.py
 create mode 100644 gklearn/utils/knn.py

diff --git a/gklearn/__init__.py b/gklearn/__init__.py
index c607b26..08ca4ed 100644
--- a/gklearn/__init__.py
+++ b/gklearn/__init__.py
@@ -18,4 +18,4 @@ __date__    = "November 2017"
 # import sub modules
 # from gklearn import c_ext
 # from gklearn import ged
-from gklearn import utils
+# from gklearn import utils
diff --git a/gklearn/ged/util/util.py b/gklearn/ged/util/util.py
index d72c2e6..c41ca86 100644
--- a/gklearn/ged/util/util.py
+++ b/gklearn/ged/util/util.py
@@ -46,7 +46,7 @@ def compute_ged(g1, g2, options):
 	return dis, pi_forward, pi_backward
 
 
-def compute_geds(graphs, options={}, parallel=False):
+def compute_geds(graphs, options={}, parallel=False, verbose=True):
 	# initialize ged env.
 	ged_env = gedlibpy.GEDEnv()
 	ged_env.set_edit_cost(options['edit_cost'], edit_cost_constant=options['edit_cost_constants'])
@@ -81,8 +81,11 @@ def compute_geds(graphs, options={}, parallel=False):
 			G_listID = listID_toshare
 		do_partial = partial(_wrapper_compute_ged_parallel, neo_options)
 		pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(graphs, ged_env, listID))
-		iterator = tqdm(pool.imap_unordered(do_partial, itr, chunksize),
+		if verbose:
+			iterator = tqdm(pool.imap_unordered(do_partial, itr, chunksize),
 						desc='computing GEDs', file=sys.stdout)
+		else:
+			iterator = pool.imap_unordered(do_partial, itr, chunksize)
 #		iterator = pool.imap_unordered(do_partial, itr, chunksize)
 		for i, j, dis, n_eo_tmp in iterator:
 			idx_itr = int(len(graphs) * i + j - (i + 1) * (i + 2) / 2)
@@ -98,7 +101,11 @@ def compute_geds(graphs, options={}, parallel=False):
 	else:
 		ged_vec = []
 		n_edit_operations = []
-		for i in tqdm(range(len(graphs)), desc='computing GEDs', file=sys.stdout):
+		if verbose:
+			iterator = tqdm(range(len(graphs)), desc='computing GEDs', file=sys.stdout)
+		else:
+			iterator = range(len(graphs))
+		for i in iterator:
 #		for i in range(len(graphs)):
 			for j in range(i + 1, len(graphs)):
 				dis, pi_forward, pi_backward = _compute_ged(ged_env, listID[i], listID[j], graphs[i], graphs[j])
diff --git a/gklearn/kernels/graph_kernel.py b/gklearn/kernels/graph_kernel.py
index 6f667e1..db4abf8 100644
--- a/gklearn/kernels/graph_kernel.py
+++ b/gklearn/kernels/graph_kernel.py
@@ -67,6 +67,9 @@ class GraphKernel(object):
 			
 			
 	def normalize_gm(self, gram_matrix):
+		import warnings
+		warnings.warn('gklearn.kernels.graph_kernel.normalize_gm will be deprecated, use gklearn.utils.normalize_gram_matrix instead', DeprecationWarning)
+
 		diag = gram_matrix.diagonal().copy()
 		for i in range(len(gram_matrix)):
 			for j in range(i, len(gram_matrix)):
diff --git a/gklearn/preimage/__init__.py b/gklearn/preimage/__init__.py
index f04b5cc..21e688e 100644
--- a/gklearn/preimage/__init__.py
+++ b/gklearn/preimage/__init__.py
@@ -12,3 +12,4 @@ __date__ = "March 2020"
 
 from gklearn.preimage.preimage_generator import PreimageGenerator
 from gklearn.preimage.median_preimage_generator import MedianPreimageGenerator
+from gklearn.preimage.kernel_knn_cv import kernel_knn_cv
diff --git a/gklearn/preimage/experiments/xp_1nn.py b/gklearn/preimage/experiments/xp_1nn.py
new file mode 100644
index 0000000..872be03
--- /dev/null
+++ b/gklearn/preimage/experiments/xp_1nn.py
@@ -0,0 +1,103 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Mon May 11 14:15:11 2020
+
+@author: ljia
+"""
+import functools
+import multiprocessing
+import os
+import sys
+import logging
+from gklearn.utils.kernels import deltakernel, gaussiankernel, kernelproduct
+from gklearn.preimage import kernel_knn_cv
+
+dir_root = '../results/xp_1nn.init1.no_triangle_rule.allow_zeros/'
+num_random = 10
+initial_solutions = 1
+triangle_rule = False
+allow_zeros = True
+update_order = False
+test_sizes = [0.9, 0.7] # , 0.5, 0.3, 0.1]
+
+def xp_knn_1_1():
+	for test_size in test_sizes:
+		ds_name = 'Letter-high'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+# 		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+# 		for train_examples in ['expert']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+# 			try:
+			kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=None, edge_required=False, cut_range=None)
+# 			except Exception as exp:
+# 				print('An exception occured when running this experiment:')
+# 				LOG_FILENAME = dir_save + 'error.txt'
+# 				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+# 				logging.exception('')
+# 				print(repr(exp))
+
+
+if __name__ == '__main__':
+	xp_knn_1_1()
\ No newline at end of file
diff --git a/gklearn/preimage/kernel_knn_cv.py b/gklearn/preimage/kernel_knn_cv.py
new file mode 100644
index 0000000..ce822aa
--- /dev/null
+++ b/gklearn/preimage/kernel_knn_cv.py
@@ -0,0 +1,418 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue May 12 12:52:15 2020
+
+@author: ljia
+"""
+import numpy as np
+import csv
+import os
+import os.path
+from gklearn.utils import Dataset
+from sklearn.model_selection import ShuffleSplit
+from gklearn.preimage import MedianPreimageGenerator
+from gklearn.utils import normalize_gram_matrix, compute_distance_matrix
+from gklearn.preimage.utils import get_same_item_indices
+from gklearn.utils.knn import knn_classification
+from gklearn.preimage.utils import compute_k_dis
+	
+
+def kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=True, load_gm='auto', dir_save='', irrelevant_labels=None, edge_required=False, cut_range=None):
+	
+	# 1. get dataset.
+	print('1. getting dataset...')
+	dataset_all = Dataset()
+	dataset_all.load_predefined_dataset(ds_name)
+	dataset_all.trim_dataset(edge_required=edge_required)
+	if irrelevant_labels is not None:
+		dataset_all.remove_labels(**irrelevant_labels)
+	if cut_range is not None:
+		dataset_all.cut_graphs(cut_range)
+#	datasets = split_dataset_by_target(dataset_all)
+
+	if save_results:
+		# create result files.
+		print('creating output files...')
+		fn_output_detail, fn_output_summary = __init_output_file_knn(ds_name, kernel_options['name'], mpg_options['fit_method'], dir_save)
+	else:
+		fn_output_detail, fn_output_summary = None, None
+		
+	# 2. compute/load Gram matrix a priori.
+	print('2. computing/loading Gram matrix...')
+	gram_matrix_unnorm, time_precompute_gm = __get_gram_matrix(load_gm, dir_save, ds_name, kernel_options, dataset_all)
+	
+	# 3. perform k-nn CV.
+	print('3. performing k-nn CV...')
+	if train_examples == 'k-graphs' or train_examples == 'expert' or train_examples == 'random':
+		__kernel_knn_cv_median(dataset_all, ds_name, knn_options, mpg_options, kernel_options, mge_options, ged_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary)
+	
+	elif train_examples == 'best-dataset':
+		__kernel_knn_cv_best_ds(dataset_all, ds_name, knn_options, kernel_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary)
+		
+	elif train_examples == 'trainset':
+		__kernel_knn_cv_trainset(dataset_all, ds_name, knn_options, kernel_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary)
+
+	print('\ncomplete.\n')	
+	
+	
+def __kernel_knn_cv_median(dataset_all, ds_name, knn_options, mpg_options, kernel_options, mge_options, ged_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary):
+	Gn = dataset_all.graphs
+	y_all = dataset_all.targets
+	n_neighbors, n_splits, test_size = knn_options['n_neighbors'], knn_options['n_splits'], knn_options['test_size']
+
+	# get shuffles. 
+	train_indices, test_indices, train_nums, y_app = __get_shuffles(y_all, n_splits, test_size)
+	
+	accuracies = [[], [], []]
+	for trial in range(len(train_indices)):
+		print('\ntrial =', trial)
+		
+		train_index = train_indices[trial]
+		test_index = test_indices[trial]
+		G_app = [Gn[i] for i in train_index]
+		G_test = [Gn[i] for i in test_index]
+		y_test = [y_all[i] for i in test_index]
+		gm_unnorm_trial = gram_matrix_unnorm[train_index,:][:,train_index].copy()
+		
+		# compute pre-images for each class.
+		medians = [[], [], []]
+		train_nums_tmp = [0] + train_nums
+		print('\ncomputing pre-image for each class...\n')
+		for i_class in range(len(train_nums_tmp) - 1):
+			print(i_class + 1, 'of', len(train_nums_tmp) - 1, 'classes:')
+			i_start = int(np.sum(train_nums_tmp[0:i_class + 1]))
+			i_end = i_start + train_nums_tmp[i_class + 1]
+			median_set = G_app[i_start:i_end]
+			
+			dataset = dataset_all.copy()
+			dataset.load_graphs(median_set.copy(), targets=None)
+			mge_options['update_order'] = True
+			mpg_options['gram_matrix_unnorm'] = gm_unnorm_trial[i_start:i_end,i_start:i_end].copy()
+			mpg_options['runtime_precompute_gm'] = 0
+			set_median, gen_median_uo = __generate_median_preimages(dataset, mpg_options, kernel_options, ged_options, mge_options)
+			mge_options['update_order'] = False
+			mpg_options['gram_matrix_unnorm'] = gm_unnorm_trial[i_start:i_end,i_start:i_end].copy()
+			mpg_options['runtime_precompute_gm'] = 0
+			_, gen_median = __generate_median_preimages(dataset, mpg_options, kernel_options, ged_options, mge_options)
+			medians[0].append(set_median)
+			medians[1].append(gen_median)
+			medians[2].append(gen_median_uo)
+			
+		# for each set of medians.
+		print('\nperforming k-nn...')
+		for i_app, G_app in enumerate(medians):
+			# compute dis_mat between medians.
+			dataset = dataset_all.copy()
+			dataset.load_graphs(G_app.copy(), targets=None)
+			gm_app_unnorm, _ = __compute_gram_matrix_unnorm(dataset, kernel_options.copy())
+			
+			# compute the entire Gram matrix.
+			graph_kernel = __get_graph_kernel(dataset.copy(), kernel_options.copy())
+			kernels_to_medians = []
+			for g in G_app:
+				kernels_to_median, _ = graph_kernel.compute(g, G_test, **kernel_options.copy()) 
+				kernels_to_medians.append(kernels_to_median)
+			kernels_to_medians = np.array(kernels_to_medians)
+			gm_all = np.concatenate((gm_app_unnorm, kernels_to_medians), axis=1)
+			gm_all = np.concatenate((gm_all, np.concatenate((kernels_to_medians.T, gram_matrix_unnorm[test_index,:][:,test_index].copy()), axis=1)), axis=0)
+			
+			gm_all = normalize_gram_matrix(gm_all.copy())
+			dis_mat, _, _, _ = compute_distance_matrix(gm_all)
+			
+			N = len(G_app)
+			
+			d_app = dis_mat[range(N),:][:,range(N)].copy()
+			
+			d_test = np.zeros((N, len(test_index)))
+			for i in range(N):
+				for j in range(len(test_index)):
+					d_test[i, j] = dis_mat[i, j]
+					
+			accuracies[i_app].append(knn_classification(d_app, d_test, y_app, y_test, n_neighbors, verbose=True, text=train_examples))
+			
+		# write result detail.
+		if save_results:
+			f_detail = open(dir_save + fn_output_detail, 'a')
+			print('writing results to files...')
+			for i, median_type in enumerate(['set-median', 'gen median', 'gen median uo']):
+				csv.writer(f_detail).writerow([ds_name, kernel_options['name'], 
+					  train_examples + ': ' + median_type, trial, 
+					  knn_options['n_neighbors'],
+					  len(gm_all), knn_options['test_size'], 
+					  accuracies[i][-1][0], accuracies[i][-1][1]])
+			f_detail.close()
+		
+	results = {}
+	results['ave_perf_train'] = [np.mean([i[0] for i in j], axis=0) for j in accuracies]
+	results['std_perf_train'] = [np.std([i[0] for i in j], axis=0, ddof=1) for j in accuracies]
+	results['ave_perf_test'] = [np.mean([i[1] for i in j], axis=0) for j in accuracies]
+	results['std_perf_test'] = [np.std([i[1] for i in j], axis=0, ddof=1) for j in accuracies]
+
+	# write result summary for each letter. 
+	if save_results:
+		f_summary = open(dir_save + fn_output_summary, 'a')
+		for i, median_type in enumerate('set-median', 'gen median', 'gen median uo'):
+			csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
+				  train_examples + ': ' + median_type, 
+				  knn_options['n_neighbors'],
+				  knn_options['test_size'], results['ave_perf_train'][i],
+				  results['ave_perf_test'][i], results['std_perf_train'][i],
+				  results['std_perf_test'][i], time_precompute_gm])
+		f_summary.close()
+		
+		
+def __kernel_knn_cv_best_ds(dataset_all, ds_name, knn_options, kernel_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary):
+	Gn = dataset_all.graphs
+	y_all = dataset_all.targets
+	n_neighbors, n_splits, test_size = knn_options['n_neighbors'], knn_options['n_splits'], knn_options['test_size']
+
+	# get shuffles. 
+	train_indices, test_indices, train_nums, y_app = __get_shuffles(y_all, n_splits, test_size)
+	
+	accuracies = []
+	for trial in range(len(train_indices)):
+		print('\ntrial =', trial)
+		
+		train_index = train_indices[trial]
+		test_index = test_indices[trial]
+		G_app = [Gn[i] for i in train_index]
+		G_test = [Gn[i] for i in test_index]
+		y_test = [y_all[i] for i in test_index]
+		gm_unnorm_trial = gram_matrix_unnorm[train_index,:][:,train_index].copy()
+		
+		# get best graph from trainset according to distance in kernel space for each class.
+		best_graphs = []
+		train_nums_tmp = [0] + train_nums
+		print('\ngetting best graph from trainset for each class...')
+		for i_class in range(len(train_nums_tmp) - 1):
+			print(i_class + 1, 'of', len(train_nums_tmp) - 1, 'classes.')
+			i_start = int(np.sum(train_nums_tmp[0:i_class + 1]))
+			i_end = i_start + train_nums_tmp[i_class + 1]
+			G_class = G_app[i_start:i_end]
+			gm_unnorm_class = gm_unnorm_trial[i_start:i_end,i_start:i_end]
+			gm_class = normalize_gram_matrix(gm_unnorm_class.copy())
+			
+			k_dis_list = []
+			for idx in range(len(G_class)):
+				k_dis_list.append(compute_k_dis(idx, range(0, len(G_class)), [1 / len(G_class)] * len(G_class), gm_class, withterm3=False))
+			idx_k_dis_min = np.argmin(k_dis_list)
+			best_graphs.append(G_class[idx_k_dis_min].copy())
+			
+			
+		# perform k-nn.
+		print('\nperforming k-nn...')
+		# compute dis_mat between medians.
+		dataset = dataset_all.copy()
+		dataset.load_graphs(best_graphs.copy(), targets=None)
+		gm_app_unnorm, _ = __compute_gram_matrix_unnorm(dataset, kernel_options.copy())
+		
+		# compute the entire Gram matrix.
+		graph_kernel = __get_graph_kernel(dataset.copy(), kernel_options.copy())
+		kernels_to_best_graphs = []
+		for g in best_graphs:
+			kernels_to_best_graph, _ = graph_kernel.compute(g, G_test, **kernel_options.copy()) 
+			kernels_to_best_graphs.append(kernels_to_best_graph)
+		kernels_to_best_graphs = np.array(kernels_to_best_graphs)
+		gm_all = np.concatenate((gm_app_unnorm, kernels_to_best_graphs), axis=1)
+		gm_all = np.concatenate((gm_all, np.concatenate((kernels_to_best_graphs.T, gram_matrix_unnorm[test_index,:][:,test_index].copy()), axis=1)), axis=0)
+		
+		gm_all = normalize_gram_matrix(gm_all.copy())
+		dis_mat, _, _, _ = compute_distance_matrix(gm_all)
+		
+		N = len(best_graphs)
+		
+		d_app = dis_mat[range(N),:][:,range(N)].copy()
+		
+		d_test = np.zeros((N, len(test_index)))
+		for i in range(N):
+			for j in range(len(test_index)):
+				d_test[i, j] = dis_mat[i, j]
+				
+		accuracies.append(knn_classification(d_app, d_test, y_app, y_test, n_neighbors, verbose=True, text=train_examples))
+			
+		# write result detail.
+		if save_results:
+			f_detail = open(dir_save + fn_output_detail, 'a')
+			print('writing results to files...')
+			csv.writer(f_detail).writerow([ds_name, kernel_options['name'], 
+				  train_examples, trial, 
+				  knn_options['n_neighbors'],
+				  len(gm_all), knn_options['test_size'], 
+				  accuracies[-1][0], accuracies[-1][1]])
+			f_detail.close()
+		
+	results = {}
+	results['ave_perf_train'] = np.mean([i[0] for i in accuracies], axis=0)
+	results['std_perf_train'] = np.std([i[0] for i in accuracies], axis=0, ddof=1)
+	results['ave_perf_test'] = np.mean([i[1] for i in accuracies], axis=0)
+	results['std_perf_test'] = np.std([i[1] for i in accuracies], axis=0, ddof=1)
+	
+	# write result summary for each letter. 
+	if save_results:
+		f_summary = open(dir_save + fn_output_summary, 'a')
+		csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
+				  train_examples, 
+				  knn_options['n_neighbors'],
+				  knn_options['test_size'], results['ave_perf_train'],
+				  results['ave_perf_test'], results['std_perf_train'],
+				  results['std_perf_test'], time_precompute_gm])
+		f_summary.close()
+	
+	
+def __kernel_knn_cv_trainset(dataset_all, ds_name, knn_options, kernel_options, gram_matrix_unnorm, time_precompute_gm, train_examples, save_results, dir_save, fn_output_detail, fn_output_summary):
+	y_all = dataset_all.targets
+	n_neighbors, n_splits, test_size = knn_options['n_neighbors'], knn_options['n_splits'], knn_options['test_size']
+	
+	# compute distance matrix.
+	gram_matrix = normalize_gram_matrix(gram_matrix_unnorm.copy())
+	dis_mat, _, _, _ = compute_distance_matrix(gram_matrix)
+
+	# get shuffles. 
+	train_indices, test_indices, _, _ = __get_shuffles(y_all, n_splits, test_size)
+			 
+	accuracies = []
+	for trial in range(len(train_indices)):
+		print('\ntrial =', trial)
+		
+		train_index = train_indices[trial]
+		test_index = test_indices[trial]
+		y_app = [y_all[i] for i in train_index]
+		y_test = [y_all[i] for i in test_index]
+		
+		N = len(train_index)
+		
+		d_app = dis_mat[train_index,:][:,train_index].copy()
+		
+		d_test = np.zeros((N, len(test_index)))
+		for i in range(N):
+			for j in range(len(test_index)):
+				d_test[i, j] = dis_mat[train_index[i], test_index[j]]
+				
+		accuracies.append(knn_classification(d_app, d_test, y_app, y_test, n_neighbors, verbose=True, text=train_examples))
+		
+		# write result detail.
+		if save_results:
+			print('writing results to files...')
+			f_detail = open(dir_save + fn_output_detail, 'a')
+			csv.writer(f_detail).writerow([ds_name, kernel_options['name'], 
+					  train_examples, trial, knn_options['n_neighbors'],
+					  len(gram_matrix), knn_options['test_size'], 
+					  accuracies[-1][0], accuracies[-1][1]])
+			f_detail.close()
+		
+	results = {}
+	results['ave_perf_train'] = np.mean([i[0] for i in accuracies], axis=0)
+	results['std_perf_train'] = np.std([i[0] for i in accuracies], axis=0, ddof=1)
+	results['ave_perf_test'] = np.mean([i[1] for i in accuracies], axis=0)
+	results['std_perf_test'] = np.std([i[1] for i in accuracies], axis=0, ddof=1)
+
+	# write result summary for each letter. 
+	if save_results:
+		f_summary = open(dir_save + fn_output_summary, 'a')
+		csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
+				  train_examples, knn_options['n_neighbors'],
+				  knn_options['test_size'], results['ave_perf_train'],
+				  results['ave_perf_test'], results['std_perf_train'],
+				  results['std_perf_test'], time_precompute_gm])
+		f_summary.close()
+	
+	
+def __get_shuffles(y_all, n_splits, test_size):
+	rs = ShuffleSplit(n_splits=n_splits, test_size=test_size, random_state=0)
+	train_indices = [[] for _ in range(n_splits)] 
+	test_indices = [[] for _ in range(n_splits)]
+	idx_targets = get_same_item_indices(y_all)
+	train_nums = []
+	keys = []
+	for key, item in idx_targets.items():
+		i = 0
+		for train_i, test_i in rs.split(item): # @todo: careful when parallel.
+			train_indices[i] += [item[idx] for idx in train_i]
+			test_indices[i] += [item[idx] for idx in test_i]
+			i += 1
+		train_nums.append(len(train_i))
+		keys.append(key)
+	return train_indices, test_indices, train_nums, keys
+	
+	
+def __generate_median_preimages(dataset, mpg_options, kernel_options, ged_options, mge_options):
+	mpg = MedianPreimageGenerator()
+	mpg.dataset = dataset.copy()
+	mpg.set_options(**mpg_options.copy())
+	mpg.kernel_options = kernel_options.copy()
+	mpg.ged_options = ged_options.copy()
+	mpg.mge_options = mge_options.copy()
+	mpg.run()
+	return mpg.set_median, mpg.gen_median
+
+
+def __get_gram_matrix(load_gm, dir_save, ds_name, kernel_options, dataset_all):
+	if load_gm == 'auto':
+		gm_fname = dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm.npz'
+		gmfile_exist = os.path.isfile(os.path.abspath(gm_fname))
+		if gmfile_exist:
+			gmfile = np.load(gm_fname, allow_pickle=True) # @todo: may not be safe.
+			gram_matrix_unnorm = gmfile['gram_matrix_unnorm']
+			time_precompute_gm = float(gmfile['run_time'])
+		else:
+			gram_matrix_unnorm, time_precompute_gm = __compute_gram_matrix_unnorm(dataset_all, kernel_options)
+			np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm=gram_matrix_unnorm, run_time=time_precompute_gm)
+	elif not load_gm:
+		gram_matrix_unnorm, time_precompute_gm = __compute_gram_matrix_unnorm(dataset_all, kernel_options)
+		np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm=gram_matrix_unnorm, run_time=time_precompute_gm)
+	else:
+		gmfile = np.load()
+		gram_matrix_unnorm = gmfile['gram_matrix_unnorm']
+		time_precompute_gm = float(gmfile['run_time'])
+		
+	return gram_matrix_unnorm, time_precompute_gm
+
+
+def __get_graph_kernel(dataset, kernel_options):
+	from gklearn.utils.utils import get_graph_kernel_by_name
+	graph_kernel = get_graph_kernel_by_name(kernel_options['name'], 
+				  node_labels=dataset.node_labels,
+				  edge_labels=dataset.edge_labels, 
+				  node_attrs=dataset.node_attrs,
+				  edge_attrs=dataset.edge_attrs,
+				  ds_infos=dataset.get_dataset_infos(keys=['directed']),
+				  kernel_options=kernel_options)
+	return graph_kernel
+		
+		
+def __compute_gram_matrix_unnorm(dataset, kernel_options):
+	from gklearn.utils.utils import get_graph_kernel_by_name
+	graph_kernel = get_graph_kernel_by_name(kernel_options['name'], 
+				  node_labels=dataset.node_labels,
+				  edge_labels=dataset.edge_labels, 
+				  node_attrs=dataset.node_attrs,
+				  edge_attrs=dataset.edge_attrs,
+				  ds_infos=dataset.get_dataset_infos(keys=['directed']),
+				  kernel_options=kernel_options)
+	
+	gram_matrix, run_time = graph_kernel.compute(dataset.graphs, **kernel_options)
+	gram_matrix_unnorm = graph_kernel.gram_matrix_unnorm
+		
+	return gram_matrix_unnorm, run_time
+		
+		
+def __init_output_file_knn(ds_name, gkernel, fit_method, dir_output):
+	if not os.path.exists(dir_output):
+		os.makedirs(dir_output)
+	fn_output_detail = 'results_detail_knn.' + ds_name + '.' + gkernel + '.csv'
+	f_detail = open(dir_output + fn_output_detail, 'a')
+	csv.writer(f_detail).writerow(['dataset', 'graph kernel',
+			  'train examples', 'trial', 'num neighbors', 'num graphs', 'test size',
+			  'perf train', 'perf test'])
+	f_detail.close()
+	
+	fn_output_summary = 'results_summary_knn.' + ds_name + '.' + gkernel + '.csv'
+	f_summary = open(dir_output + fn_output_summary, 'a')
+	csv.writer(f_summary).writerow(['dataset', 'graph kernel',
+		      'train examples', 'num neighbors', 'test size',
+			  'ave perf train', 'ave perf test',
+			  'std perf train', 'std perf test', 'time precompute gm'])
+	f_summary.close()
+	
+	return fn_output_detail, fn_output_summary
\ No newline at end of file
diff --git a/gklearn/preimage/median_preimage_generator.py b/gklearn/preimage/median_preimage_generator.py
index 20fbec6..6c66de0 100644
--- a/gklearn/preimage/median_preimage_generator.py
+++ b/gklearn/preimage/median_preimage_generator.py
@@ -281,7 +281,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 		options['edge_labels'] = self._dataset.edge_labels
 		options['node_attrs'] = self._dataset.node_attrs
 		options['edge_attrs'] = self._dataset.edge_attrs
-		ged_vec_init, ged_mat, n_edit_operations = compute_geds(graphs, options=options, parallel=self.__parallel)
+		ged_vec_init, ged_mat, n_edit_operations = compute_geds(graphs, options=options, parallel=self.__parallel, verbose=(self._verbose > 1))
 		residual_list = [np.sqrt(np.sum(np.square(np.array(ged_vec_init) - dis_k_vec)))]	
 		time_list = [time.time() - time0]
 		edit_cost_list = [self.__init_ecc]  
@@ -323,7 +323,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 			options['edge_labels'] = self._dataset.edge_labels
 			options['node_attrs'] = self._dataset.node_attrs
 			options['edge_attrs'] = self._dataset.edge_attrs
-			ged_vec, ged_mat, n_edit_operations = compute_geds(graphs, options=options, parallel=self.__parallel)
+			ged_vec, ged_mat, n_edit_operations = compute_geds(graphs, options=options, parallel=self.__parallel, verbose=(self._verbose > 1))
 			residual_list.append(np.sqrt(np.sum(np.square(np.array(ged_vec) - dis_k_vec))))
 			time_list.append(time.time() - time0)
 			edit_cost_list.append(self.__edit_cost_constants)
diff --git a/gklearn/preimage/utils.py b/gklearn/preimage/utils.py
index f99ab8a..55e1e9e 100644
--- a/gklearn/preimage/utils.py
+++ b/gklearn/preimage/utils.py
@@ -45,7 +45,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 	if save_results:
 		# create result files.
 		print('creating output files...')
-		fn_output_detail, fn_output_summary = __init_output_file(ds_name, kernel_options['name'], mpg_options['fit_method'], dir_save)
+		fn_output_detail, fn_output_summary = __init_output_file_preimage(ds_name, kernel_options['name'], mpg_options['fit_method'], dir_save)
 		
 	sod_sm_list = []
 	sod_gm_list = []
@@ -82,22 +82,22 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 		gram_matrix_unnorm_list = []
 		time_precompute_gm_list = []
 	else:
-		gmfile = np.load()
-		gram_matrix_unnorm_list = gmfile['gram_matrix_unnorm_list']
-		time_precompute_gm_list = gmfile['run_time_list']
-# 	repeats_better_sod_sm2gm = []
-# 	repeats_better_dis_k_sm2gm = []
-# 	repeats_better_dis_k_gi2sm = []
-# 	repeats_better_dis_k_gi2gm = []
+		gmfile = np.load(gm_fname, allow_pickle=True) # @todo: may not be safe.
+		gram_matrix_unnorm_list = [item for item in gmfile['gram_matrix_unnorm_list']]
+		time_precompute_gm_list = gmfile['run_time_list'].tolist()
+#	repeats_better_sod_sm2gm = []
+#	repeats_better_dis_k_sm2gm = []
+#	repeats_better_dis_k_gi2sm = []
+#	repeats_better_dis_k_gi2gm = []
 		
 		
-	print('start generating preimage for each class of target...')
+	print('starting generating preimage for each class of target...')
 	idx_offset = 0
 	for idx, dataset in enumerate(datasets):
 		target = dataset.targets[0]
 		print('\ntarget =', target, '\n')
-# 		if target != 1:
-#  			continue
+#		if target != 1:
+# 			continue
 		
 		num_graphs = len(dataset.graphs)
 		if num_graphs < 2:
@@ -148,7 +148,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 					  results['sod_set_median'], results['sod_gen_median'],
 					  results['k_dis_set_median'], results['k_dis_gen_median'], 
 					  results['k_dis_dataset'], sod_sm2gm, dis_k_sm2gm, 
-					  dis_k_gi2sm, dis_k_gi2gm, 	results['edit_cost_constants'],
+					  dis_k_gi2sm, dis_k_gi2gm,	results['edit_cost_constants'],
 					  results['runtime_precompute_gm'], results['runtime_optimize_ec'], 
 					  results['runtime_generate_preimage'], results['runtime_total'],
 					  results['itrs'], results['converged'],
@@ -177,7 +177,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 			# # SOD SM -> GM
 			if results['sod_set_median'] > results['sod_gen_median']:
 				nb_sod_sm2gm[0] += 1
-	# 			repeats_better_sod_sm2gm.append(1)
+	#			repeats_better_sod_sm2gm.append(1)
 			elif results['sod_set_median'] == results['sod_gen_median']:
 				nb_sod_sm2gm[1] += 1
 			elif results['sod_set_median'] < results['sod_gen_median']:
@@ -185,7 +185,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 			# # dis_k SM -> GM
 			if results['k_dis_set_median'] > results['k_dis_gen_median']:
 				nb_dis_k_sm2gm[0] += 1
-	# 			repeats_better_dis_k_sm2gm.append(1)
+	#			repeats_better_dis_k_sm2gm.append(1)
 			elif results['k_dis_set_median'] == results['k_dis_gen_median']:
 				nb_dis_k_sm2gm[1] += 1
 			elif results['k_dis_set_median'] < results['k_dis_gen_median']:
@@ -193,7 +193,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 			# # dis_k gi -> SM
 			if results['k_dis_dataset'] > results['k_dis_set_median']:
 				nb_dis_k_gi2sm[0] += 1
-	# 			repeats_better_dis_k_gi2sm.append(1)
+	#			repeats_better_dis_k_gi2sm.append(1)
 			elif results['k_dis_dataset'] == results['k_dis_set_median']:
 				nb_dis_k_gi2sm[1] += 1
 			elif results['k_dis_dataset'] < results['k_dis_set_median']:
@@ -201,7 +201,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 			# # dis_k gi -> GM
 			if results['k_dis_dataset'] > results['k_dis_gen_median']:
 				nb_dis_k_gi2gm[0] += 1
-	# 			repeats_better_dis_k_gi2gm.append(1)
+	#			repeats_better_dis_k_gi2gm.append(1)
 			elif results['k_dis_dataset'] == results['k_dis_gen_median']:
 				nb_dis_k_gi2gm[1] += 1
 			elif results['k_dis_dataset'] < results['k_dis_gen_median']:
@@ -225,7 +225,7 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 					  results['mge']['num_increase_order'] > 0,
 					  results['mge']['num_converged_descents'] > 0, 
 					  nb_sod_sm2gm, 
- 					  nb_dis_k_sm2gm, nb_dis_k_gi2sm, nb_dis_k_gi2gm])
+					  nb_dis_k_sm2gm, nb_dis_k_gi2sm, nb_dis_k_gi2gm])
 			f_summary.close()
 			 
 		# save median graphs.
@@ -235,15 +235,15 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 			print('Saving median graphs to files...')
 			fn_pre_sm = dir_save + 'medians/set_median.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
 			saveGXL(mpg.set_median, fn_pre_sm + '.gxl', method='default', 
-					node_labels=dataset.node_labels, edge_labels=dataset.edge_labels, 	
+					node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
 					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
 			fn_pre_gm = dir_save + 'medians/gen_median.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
 			saveGXL(mpg.gen_median, fn_pre_gm + '.gxl', method='default',
-		   			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels, 	
+		  			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
 					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
 			fn_best_dataset = dir_save + 'medians/g_best_dataset.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
 			saveGXL(mpg.best_from_dataset, fn_best_dataset + '.gxl', method='default',
-		   			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels, 	
+		  			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
 					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
 		
 		# plot median graphs.
@@ -304,10 +304,10 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 	if (load_gm == 'auto' and not gmfile_exist) or not load_gm:
 		np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm_list=gram_matrix_unnorm_list, run_time_list=time_precompute_gm_list)	
 
-	print('\ncomplete.')	
+	print('\ncomplete.\n')	
 
 	
-def __init_output_file(ds_name, gkernel, fit_method, dir_output):
+def __init_output_file_preimage(ds_name, gkernel, fit_method, dir_output):
 	if not os.path.exists(dir_output):
 		os.makedirs(dir_output)
 #	fn_output_detail = 'results_detail.' + ds_name + '.' + gkernel + '.' + fit_method + '.csv'
@@ -335,9 +335,9 @@ def __init_output_file(ds_name, gkernel, fit_method, dir_output):
 			  'num updates ecc', 'mge num decrease order', 'mge num increase order', 
 			  'mge num converged', '# SOD SM -> GM', '# dis_k SM -> GM', 
 			  '# dis_k gi -> SM', '# dis_k gi -> GM'])
-# 			   'repeats better SOD SM -> GM', 
-# 			  'repeats better dis_k SM -> GM', 'repeats better dis_k gi -> SM', 
-# 			  'repeats better dis_k gi -> GM'])
+#			   'repeats better SOD SM -> GM', 
+#			  'repeats better dis_k SM -> GM', 'repeats better dis_k gi -> SM', 
+#			  'repeats better dis_k gi -> GM'])
 	f_summary.close()
 	
 	return fn_output_detail, fn_output_summary
@@ -462,6 +462,8 @@ def gram2distances(Kmatrix):
 
 def kernel_distance_matrix(Gn, node_label, edge_label, Kmatrix=None, 
 						   gkernel=None, verbose=True):
+	import warnings
+	warnings.warn('gklearn.preimage.utils.kernel_distance_matrix is deprecated, use gklearn.kernels.graph_kernel.compute_distance_matrix or gklearn.utils.compute_distance_matrix instead', DeprecationWarning)
 	dis_mat = np.empty((len(Gn), len(Gn)))
 	if Kmatrix is None:
 		Kmatrix = compute_kernel(Gn, gkernel, node_label, edge_label, verbose)
diff --git a/gklearn/utils/__init__.py b/gklearn/utils/__init__.py
index 198bf75..af9c751 100644
--- a/gklearn/utils/__init__.py
+++ b/gklearn/utils/__init__.py
@@ -21,4 +21,6 @@ from gklearn.utils.timer import Timer
 from gklearn.utils.utils import get_graph_kernel_by_name
 from gklearn.utils.utils import compute_gram_matrices_by_class
 from gklearn.utils.utils import SpecialLabel
+from gklearn.utils.utils import normalize_gram_matrix, compute_distance_matrix
 from gklearn.utils.trie import Trie
+from gklearn.utils.knn import knn_cv, knn_classification
diff --git a/gklearn/utils/dataset.py b/gklearn/utils/dataset.py
index c499ce2..78d8841 100644
--- a/gklearn/utils/dataset.py
+++ b/gklearn/utils/dataset.py
@@ -522,6 +522,20 @@ class Dataset(object):
 		self.__targets = [self.__targets[i] for i in idx]
 		self.clean_labels()
 		
+		
+	def copy(self):
+		dataset = Dataset()
+		graphs = self.__graphs.copy() if self.__graphs is not None else None
+		target = self.__targets.copy() if self.__targets is not None else None
+		node_labels = self.__node_labels.copy() if self.__node_labels is not None else None
+		node_attrs = self.__node_attrs.copy() if self.__node_attrs is not None else None
+		edge_labels = self.__edge_labels.copy() if self.__edge_labels is not None else None
+		edge_attrs = self.__edge_attrs.copy() if self.__edge_attrs is not None else None
+		dataset.load_graphs(graphs, target)
+		dataset.set_labels(node_labels=node_labels, node_attrs=node_attrs, edge_labels=edge_labels, edge_attrs=edge_attrs)
+		# @todo: clean_labels and add other class members?
+		return dataset
+		
 	
 	def __get_dataset_size(self):
 		return len(self.__graphs)
@@ -721,7 +735,11 @@ def split_dataset_by_target(dataset):
 		sub_graphs = [graphs[i] for i in val]
 		sub_dataset = Dataset()
 		sub_dataset.load_graphs(sub_graphs, [key] * len(val))
-		sub_dataset.set_labels(node_labels=dataset.node_labels, node_attrs=dataset.node_attrs, edge_labels=dataset.edge_labels, edge_attrs=dataset.edge_attrs)
+		node_labels = dataset.node_labels.copy() if dataset.node_labels is not None else None
+		node_attrs = dataset.node_attrs.copy() if dataset.node_attrs is not None else None
+		edge_labels = dataset.edge_labels.copy() if dataset.edge_labels is not None else None
+		edge_attrs = dataset.edge_attrs.copy() if dataset.edge_attrs is not None else None
+		sub_dataset.set_labels(node_labels=node_labels, node_attrs=node_attrs, edge_labels=edge_labels, edge_attrs=edge_attrs)
 		datasets.append(sub_dataset)
 		# @todo: clean_labels?
 	return datasets
\ No newline at end of file
diff --git a/gklearn/utils/knn.py b/gklearn/utils/knn.py
new file mode 100644
index 0000000..81419be
--- /dev/null
+++ b/gklearn/utils/knn.py
@@ -0,0 +1,141 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Mon May 11 11:03:01 2020
+
+@author: ljia
+"""
+import numpy as np
+from sklearn.model_selection import ShuffleSplit
+from sklearn.neighbors import KNeighborsClassifier
+from sklearn.metrics import accuracy_score
+from gklearn.utils.utils import get_graph_kernel_by_name
+# from gklearn.preimage.utils import get_same_item_indices
+
+def sum_squares(a, b):
+	"""
+	Return the sum of squares of the difference between a and b, aka MSE
+	"""
+	return np.sum([(a[i] - b[i])**2 for i in range(len(a))])
+
+
+def euclid_d(x, y):
+	"""
+	1D euclidean distance
+	"""
+	return np.sqrt((x-y)**2)
+
+
+def man_d(x, y):
+	"""
+	1D manhattan distance
+	"""
+	return np.abs((x-y))
+
+
+def knn_regression(D_app, D_test, y_app, y_test, n_neighbors, verbose=True, text=None):
+
+	from sklearn.neighbors import KNeighborsRegressor
+	knn = KNeighborsRegressor(n_neighbors=n_neighbors, metric='precomputed')
+	knn.fit(D_app, y_app)
+	y_pred = knn.predict(D_app)
+	y_pred_test = knn.predict(D_test.T)
+	perf_app = np.sqrt(sum_squares(y_pred, y_app)/len(y_app))
+	perf_test = np.sqrt(sum_squares(y_pred_test, y_test)/len(y_test))
+
+	if (verbose):
+		print("Learning error with {} train examples : {}".format(text, perf_app))
+		print("Test error with {} train examples : {}".format(text, perf_test))
+
+	return perf_app, perf_test
+
+
+def knn_classification(d_app, d_test, y_app, y_test, n_neighbors, verbose=True, text=None):
+	knn = KNeighborsClassifier(n_neighbors=n_neighbors, metric='precomputed')
+	knn.fit(d_app, y_app)
+	y_pred = knn.predict(d_app)
+	y_pred_test = knn.predict(d_test.T)
+	perf_app = accuracy_score(y_app, y_pred)
+	perf_test = accuracy_score(y_test, y_pred_test)
+
+	if (verbose):
+		print("Learning accuracy with {} costs : {}".format(text, perf_app))
+		print("Test accuracy with {} costs : {}".format(text, perf_test))
+		
+	return perf_app, perf_test
+	
+
+def knn_cv(dataset, kernel_options, trainset=None, n_neighbors=1, n_splits=50, test_size=0.9, verbose=True):
+	'''
+	Perform a knn classification cross-validation on given dataset.
+	'''
+# 	Gn = dataset.graphs
+	y_all = dataset.targets
+	
+	# compute kernel distances.
+	dis_mat = __compute_kernel_distances(dataset, kernel_options, trainset=trainset)
+		
+	
+	rs = ShuffleSplit(n_splits=n_splits, test_size=test_size, random_state=0)
+# 	train_indices = [[] for _ in range(n_splits)] 
+# 	test_indices = [[] for _ in range(n_splits)]
+# 	idx_targets = get_same_item_indices(y_all)
+# 	for key, item in idx_targets.items():
+# 		i = 0
+# 		for train_i, test_i in rs.split(item): # @todo: careful when parallel.
+# 			train_indices[i] += [item[idx] for idx in train_i]
+# 			test_indices[i] += [item[idx] for idx in test_i]
+# 			i += 1
+	
+	accuracies = []
+# 	for trial in range(len(train_indices)):
+# 		train_index = train_indices[trial]
+# 		test_index = test_indices[trial] 
+	for train_index, test_index in rs.split(y_all):
+# 		print(train_index, test_index)
+# 		G_app = [Gn[i] for i in train_index]
+# 		G_test = [Gn[i] for i in test_index]
+		y_app = [y_all[i] for i in train_index]
+		y_test = [y_all[i] for i in test_index]
+		
+		N = len(train_index)
+		
+		d_app = dis_mat.copy()
+		d_app = d_app[train_index,:]
+		d_app = d_app[:,train_index]
+		
+		d_test = np.zeros((N, len(test_index)))
+		
+		for i in range(N):
+			for j in range(len(test_index)):
+				d_test[i, j] = dis_mat[train_index[i], test_index[j]]
+				
+		accuracies.append(knn_classification(d_app, d_test, y_app, y_test, n_neighbors, verbose=verbose, text=''))
+		
+	results = {}
+	results['ave_perf_train'] = np.mean([i[0] for i in accuracies], axis=0)
+	results['std_perf_train'] = np.std([i[0] for i in accuracies], axis=0, ddof=1)
+	results['ave_perf_test'] = np.mean([i[1] for i in accuracies], axis=0)
+	results['std_perf_test'] = np.std([i[1] for i in accuracies], axis=0, ddof=1)
+
+	return results
+		
+		
+def __compute_kernel_distances(dataset, kernel_options, trainset=None):
+	graph_kernel = get_graph_kernel_by_name(kernel_options['name'], 
+				  node_labels=dataset.node_labels,
+				  edge_labels=dataset.edge_labels, 
+				  node_attrs=dataset.node_attrs,
+				  edge_attrs=dataset.edge_attrs,
+				  ds_infos=dataset.get_dataset_infos(keys=['directed']),
+				  kernel_options=kernel_options)
+	
+	gram_matrix, run_time = graph_kernel.compute(dataset.graphs, **kernel_options)
+
+	dis_mat, _, _, _ = graph_kernel.compute_distance_matrix()
+	
+	if trainset is not None:
+		gram_matrix_unnorm = graph_kernel.gram_matrix_unnorm
+		
+
+	return dis_mat
\ No newline at end of file
diff --git a/gklearn/utils/utils.py b/gklearn/utils/utils.py
index 14dc7c1..868f0f6 100644
--- a/gklearn/utils/utils.py
+++ b/gklearn/utils/utils.py
@@ -467,9 +467,37 @@ def get_mlti_dim_edge_attrs(G, attr_names):
 		attributes.append(tuple(attrs[aname] for aname in attr_names))
 	return attributes
 
+
 @unique
 class SpecialLabel(Enum):
 	"""can be used to define special labels.
 	"""
 	DUMMY = 1 # The dummy label.
-	# DUMMY = auto # enum.auto does not exist in Python 3.5.
\ No newline at end of file
+	# DUMMY = auto # enum.auto does not exist in Python 3.5.
+	
+	
+def normalize_gram_matrix(gram_matrix):
+	diag = gram_matrix.diagonal().copy()
+	for i in range(len(gram_matrix)):
+		for j in range(i, len(gram_matrix)):
+			gram_matrix[i][j] /= np.sqrt(diag[i] * diag[j])
+			gram_matrix[j][i] = gram_matrix[i][j]
+	return gram_matrix
+	
+	
+def compute_distance_matrix(gram_matrix):
+	dis_mat = np.empty((len(gram_matrix), len(gram_matrix)))
+	for i in range(len(gram_matrix)):
+		for j in range(i, len(gram_matrix)):
+			dis = gram_matrix[i, i] + gram_matrix[j, j] - 2 * gram_matrix[i, j]
+			if dis < 0:
+				if dis > -1e-10:
+					dis = 0
+				else:
+					raise ValueError('The distance is negative.')
+			dis_mat[i, j] = np.sqrt(dis)
+			dis_mat[j, i] = dis_mat[i, j]
+	dis_max = np.max(np.max(dis_mat))
+	dis_min = np.min(np.min(dis_mat[dis_mat != 0]))
+	dis_mean = np.mean(np.mean(dis_mat))
+	return dis_mat, dis_max, dis_min, dis_mean
\ No newline at end of file

From 7293234d6c48d24ab0be0ce8d31a496867ab2edb Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Thu, 14 May 2020 18:39:49 +0200
Subject: [PATCH 10/17] Update and fix bugs in MGE: 1. avoid decreasing all
 median nodes. 2. in __compute_initial_node_labels(), stop appending elements
 to median_labels when all node_labels are already selected. 3. return False
 directly in __update_node_label() when node_labels is empty. 4. solve the bug
 that median is not modified in place in __delete_node_from_median().

---
 gklearn/ged/median/median_graph_estimator.py  | 104 +++++++++++---------------
 gklearn/preimage/experiments/xp_1nn.py        |  10 ++-
 gklearn/preimage/kernel_knn_cv.py             |   7 +-
 gklearn/preimage/median_preimage_generator.py |   4 +-
 gklearn/utils/dataset.py                      |   2 +-
 5 files changed, 59 insertions(+), 68 deletions(-)

diff --git a/gklearn/ged/median/median_graph_estimator.py b/gklearn/ged/median/median_graph_estimator.py
index 9e0db71..c4291ce 100644
--- a/gklearn/ged/median/median_graph_estimator.py
+++ b/gklearn/ged/median/median_graph_estimator.py
@@ -370,7 +370,9 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			self.__ged_env.init(self.__ged_env.get_init_type())
 			
 			# Compute node maps and sum of distances for initial median.
+# 			xxx = self.__node_maps_from_median
 			self.__compute_init_node_maps(graph_ids, gen_median_id)
+# 			yyy = self.__node_maps_from_median
 			
 			self.__best_init_sum_of_distances = min(self.__best_init_sum_of_distances, self.__sum_of_distances)
 			self.__ged_env.load_nx_graph(median, set_median_id)
@@ -557,7 +559,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			
 	
 	def __median_available(self):
-		return self.__gen_median_id != np.inf
+		return self.__median_id != np.inf
 	
 	
 	def get_state(self):
@@ -827,6 +829,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		
 		
 	def __update_node_labels(self, graphs, median):
+# 		print('----------------------------')
 		
 		# Print information about current iteration.
 		if self.__print_to_stdout == 2:
@@ -834,14 +837,15 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			
 		# Iterate through all nodes of the median.
 		for i in range(0, nx.number_of_nodes(median)):
-#			print('i: ', i)
+# 			print('i: ', i)
 			# Collect the labels of the substituted nodes.
 			node_labels = []
 			for graph_id, graph in graphs.items():
-#				print('graph_id: ', graph_id)
-#				print(self.__node_maps_from_median[graph_id])
+# 				print('graph_id: ', graph_id)
+# 				print(self.__node_maps_from_median[graph_id])
+# 				print(self.__node_maps_from_median[graph_id].get_forward_map(), self.__node_maps_from_median[graph_id].get_backward_map())
 				k = self.__node_maps_from_median[graph_id].image(i)
-#				print('k: ', k)
+# 				print('k: ', k)
 				if k != np.inf:
 					node_labels.append(graph.nodes[k])
 					
@@ -961,6 +965,11 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		if self.__print_to_stdout == 2:
 			print('Trying to decrease order: ... ', end='')
 			
+		if nx.number_of_nodes(median) <= 1:
+			if self.__print_to_stdout == 2:
+				print('median graph has only 1 node, skip decrease.')
+			return False
+			
 		# Initialize ID of the node that is to be deleted.
 		id_deleted_node = [None] # @todo: or np.inf
 		decreased_order = False
@@ -968,7 +977,11 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		# Decrease the order as long as the best deletion delta is negative.
 		while self.__compute_best_deletion_delta(graphs, median, id_deleted_node) < -self.__epsilon:
 			decreased_order = True
-			median = self.__delete_node_from_median(id_deleted_node[0], median)
+			self.__delete_node_from_median(id_deleted_node[0], median)
+			if nx.number_of_nodes(median) <= 1:
+				if self.__print_to_stdout == 2:
+					print('decrease stopped because median graph remains only 1 node. ', end='')
+				break
 			
 		# Print information about current iteration.
 		if self.__print_to_stdout == 2:
@@ -1011,16 +1024,22 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 	
 	def __delete_node_from_median(self, id_deleted_node, median):
 		# Update the median.
+		mapping = {}
+		for i in range(0, nx.number_of_nodes(median)):
+			if i != id_deleted_node:
+				new_i = (i if i < id_deleted_node else (i - 1))
+				mapping[i] = new_i
 		median.remove_node(id_deleted_node)
-		median = nx.convert_node_labels_to_integers(median, first_label=0, ordering='default', label_attribute=None) # @todo:  This doesn't guarantee that the order is the same as in G.
+		nx.relabel_nodes(median, mapping, copy=False)
 		
 		# Update the node maps.
+# 		xxx = self.__node_maps_from_median
 		for key, node_map in self.__node_maps_from_median.items():
 			new_node_map = NodeMap(nx.number_of_nodes(median), node_map.num_target_nodes())
 			is_unassigned_target_node = [True] * node_map.num_target_nodes()
 			for i in range(0, nx.number_of_nodes(median) + 1):
 				if i != id_deleted_node:
-					new_i = (i if i < id_deleted_node else i - 1)
+					new_i = (i if i < id_deleted_node else (i - 1))
 					k = node_map.image(i)
 					new_node_map.add_assignment(new_i, k)
 					if k != np.inf:
@@ -1028,13 +1047,12 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			for k in range(0, node_map.num_target_nodes()):
 				if is_unassigned_target_node[k]:
 					new_node_map.add_assignment(np.inf, k)
+# 			print(self.__node_maps_from_median[key].get_forward_map(), self.__node_maps_from_median[key].get_backward_map())
 # 			print(new_node_map.get_forward_map(), new_node_map.get_backward_map())
 			self.__node_maps_from_median[key] = new_node_map
 			
 		# Increase overall number of decreases.
 		self.__num_decrease_order += 1
-		
-		return median
 	
 	
 	def __increase_order(self, graphs, median):
@@ -1230,15 +1248,22 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 						continue
 					for label in median_labels:
 						weights[label_id] = min(weights[label_id], self.__ged_env.get_node_rel_cost(dict(label), dict(node_labels[label_id])))
-				sum_weight = np.sum(weights)
-				if sum_weight == 0:
-					p = np.array([1 / len(weights)] * len(weights))
-				else:
-					p = np.array(weights) / np.sum(weights)
-				selected_label_id = urng.choice(range(0, len(weights)), size=1, p=p)[0] # for c++ test: xxx[iii] 
+				
+				# get non-zero weights.
+				weights_p, idx_p = [], []
+				for i, w in enumerate(weights):
+					if w != 0:
+						weights_p.append(w)
+						idx_p.append(i)
+				if len(weights_p) > 0:
+					p = np.array(weights_p) / np.sum(weights_p)
+					selected_label_id = urng.choice(range(0, len(weights_p)), size=1, p=p)[0] # for c++ test: xxx[iii] 
+					selected_label_id = idx_p[selected_label_id]
 # 				iii += 1 for c++ test
-				median_labels.append(node_labels[selected_label_id])
-				already_selected[selected_label_id] = True
+					median_labels.append(node_labels[selected_label_id])
+					already_selected[selected_label_id] = True
+				else: # skip the loop when all node_labels are selected. This happens when len(node_labels) <= self.__num_inits_increase_order.
+					break
 		else:
 			# Compute the initial node medians as the medians of randomly generated clusters of (roughly) equal size.
 			# @todo: go through and test.
@@ -1315,6 +1340,8 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 				 
 	
 	def __update_node_label(self, node_labels, node_label):
+		if len(node_labels) == 0: # @todo: check if this is the correct solution. Especially after calling __update_config().
+			return False
 		new_node_label = self.__get_median_node_label(node_labels)
 		if self.__ged_env.get_node_rel_cost(new_node_label, node_label) > self.__epsilon:
 			node_label.clear()
@@ -1360,47 +1387,6 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			
 		# Increase overall number of increases.
 		self.__num_increase_order += 1
-			
-	
-	def __improve_sum_of_distances(self, timer):
-		pass
-	
-	
-	def __median_available(self):
-		return self.__median_id != np.inf
-		
-				
-# 	def __get_node_image_from_map(self, node_map, node):
-# 		"""
-# 		Return ID of the node mapping of `node` in `node_map`.
-
-# 		Parameters
-# 		----------
-# 		node_map : list[tuple(int, int)]
-# 			List of node maps where the mapping node is found.
-# 		
-# 		node : int
-# 			The mapping node of this node is returned
-
-# 		Raises
-# 		------
-# 		Exception
-# 			If the node with ID `node` is not contained in the source nodes of the node map.
-
-# 		Returns
-# 		-------
-# 		int
-# 			ID of the mapping of `node`.
-# 			
-# 		Notes
-# 		-----
-# 		This function is not implemented in the `ged::MedianGraphEstimator` class of the `GEDLIB` library. Instead it is a Python implementation of the `ged::NodeMap::image` function.
-# 		"""
-# 		if node < len(node_map):
-# 			return node_map[node][1] if node_map[node][1] < len(node_map) else np.inf
-# 		else:
-# 			raise Exception('The node with ID ', str(node), ' is not contained in the source nodes of the node map.')
-# 		return np.inf
 				
 	
 	def __are_graphs_equal(self, g1, g2):
diff --git a/gklearn/preimage/experiments/xp_1nn.py b/gklearn/preimage/experiments/xp_1nn.py
index 872be03..a45a9c3 100644
--- a/gklearn/preimage/experiments/xp_1nn.py
+++ b/gklearn/preimage/experiments/xp_1nn.py
@@ -31,7 +31,7 @@ def xp_knn_1_1():
 		mpg_options = {'fit_method': 'k-graphs',
 					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
 					   'ds_name': ds_name,
-					   'parallel': True, # False
+					   'parallel': False, # False
 					   'time_limit_in_sec': 0,
 					   'max_itrs': 100,
 					   'max_itrs_without_update': 3,
@@ -100,4 +100,10 @@ def xp_knn_1_1():
 
 
 if __name__ == '__main__':
-	xp_knn_1_1()
\ No newline at end of file
+	import pdb, traceback
+	try:
+		xp_knn_1_1()
+	except:
+		extype, value, tb = sys.exc_info()
+		traceback.print_exc()
+		pdb.post_mortem(tb)
\ No newline at end of file
diff --git a/gklearn/preimage/kernel_knn_cv.py b/gklearn/preimage/kernel_knn_cv.py
index ce822aa..2faf4ba 100644
--- a/gklearn/preimage/kernel_knn_cv.py
+++ b/gklearn/preimage/kernel_knn_cv.py
@@ -29,7 +29,6 @@ def kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_opti
 		dataset_all.remove_labels(**irrelevant_labels)
 	if cut_range is not None:
 		dataset_all.cut_graphs(cut_range)
-#	datasets = split_dataset_by_target(dataset_all)
 
 	if save_results:
 		# create result files.
@@ -86,7 +85,7 @@ def __kernel_knn_cv_median(dataset_all, ds_name, knn_options, mpg_options, kerne
 			median_set = G_app[i_start:i_end]
 			
 			dataset = dataset_all.copy()
-			dataset.load_graphs(median_set.copy(), targets=None)
+			dataset.load_graphs([g.copy() for g in median_set], targets=None)
 			mge_options['update_order'] = True
 			mpg_options['gram_matrix_unnorm'] = gm_unnorm_trial[i_start:i_end,i_start:i_end].copy()
 			mpg_options['runtime_precompute_gm'] = 0
@@ -104,7 +103,7 @@ def __kernel_knn_cv_median(dataset_all, ds_name, knn_options, mpg_options, kerne
 		for i_app, G_app in enumerate(medians):
 			# compute dis_mat between medians.
 			dataset = dataset_all.copy()
-			dataset.load_graphs(G_app.copy(), targets=None)
+			dataset.load_graphs([g.copy() for g in G_app], targets=None)
 			gm_app_unnorm, _ = __compute_gram_matrix_unnorm(dataset, kernel_options.copy())
 			
 			# compute the entire Gram matrix.
@@ -204,7 +203,7 @@ def __kernel_knn_cv_best_ds(dataset_all, ds_name, knn_options, kernel_options, g
 		print('\nperforming k-nn...')
 		# compute dis_mat between medians.
 		dataset = dataset_all.copy()
-		dataset.load_graphs(best_graphs.copy(), targets=None)
+		dataset.load_graphs([g.copy() for g in best_graphs], targets=None)
 		gm_app_unnorm, _ = __compute_gram_matrix_unnorm(dataset, kernel_options.copy())
 		
 		# compute the entire Gram matrix.
diff --git a/gklearn/preimage/median_preimage_generator.py b/gklearn/preimage/median_preimage_generator.py
index 6c66de0..9aaa88a 100644
--- a/gklearn/preimage/median_preimage_generator.py
+++ b/gklearn/preimage/median_preimage_generator.py
@@ -891,8 +891,8 @@ class MedianPreimageGenerator(PreimageGenerator):
 		ged_options = self.__ged_options.copy()
 		if self.__parallel:
 			ged_options['threads'] = 1
-		self.__mge.set_init_method(self.__ged_options['method'], ged_options_to_string(ged_options))
-		self.__mge.set_descent_method(self.__ged_options['method'], ged_options_to_string(ged_options))
+		self.__mge.set_init_method(ged_options['method'], ged_options_to_string(ged_options))
+		self.__mge.set_descent_method(ged_options['method'], ged_options_to_string(ged_options))
 		
 		# Run the estimator.
 		self.__mge.run(graph_ids, set_median_id, gen_median_id)
diff --git a/gklearn/utils/dataset.py b/gklearn/utils/dataset.py
index 78d8841..90bb886 100644
--- a/gklearn/utils/dataset.py
+++ b/gklearn/utils/dataset.py
@@ -525,7 +525,7 @@ class Dataset(object):
 		
 	def copy(self):
 		dataset = Dataset()
-		graphs = self.__graphs.copy() if self.__graphs is not None else None
+		graphs = [g.copy() for g in self.__graphs] if self.__graphs is not None else None
 		target = self.__targets.copy() if self.__targets is not None else None
 		node_labels = self.__node_labels.copy() if self.__node_labels is not None else None
 		node_attrs = self.__node_attrs.copy() if self.__node_attrs is not None else None

From 34506302ce7365292b9254272ffd87307ab8f66a Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Fri, 15 May 2020 12:18:38 +0200
Subject: [PATCH 11/17] 1. Solve uncolsed file warning in graph_file.py. 2.
 Solve CXVPY * and @ warning in MPG.

---
 gklearn/preimage/experiments/xp_1nn.py             |  109 -
 .../experiments/xp_1nn_init10_trianglerule.py      | 3022 ++++++++++++++++++++
 gklearn/preimage/median_preimage_generator.py      |   60 +-
 gklearn/utils/graph_files.py                       |   37 +-
 4 files changed, 3076 insertions(+), 152 deletions(-)
 delete mode 100644 gklearn/preimage/experiments/xp_1nn.py
 create mode 100644 gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py

diff --git a/gklearn/preimage/experiments/xp_1nn.py b/gklearn/preimage/experiments/xp_1nn.py
deleted file mode 100644
index a45a9c3..0000000
--- a/gklearn/preimage/experiments/xp_1nn.py
+++ /dev/null
@@ -1,109 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-"""
-Created on Mon May 11 14:15:11 2020
-
-@author: ljia
-"""
-import functools
-import multiprocessing
-import os
-import sys
-import logging
-from gklearn.utils.kernels import deltakernel, gaussiankernel, kernelproduct
-from gklearn.preimage import kernel_knn_cv
-
-dir_root = '../results/xp_1nn.init1.no_triangle_rule.allow_zeros/'
-num_random = 10
-initial_solutions = 1
-triangle_rule = False
-allow_zeros = True
-update_order = False
-test_sizes = [0.9, 0.7] # , 0.5, 0.3, 0.1]
-
-def xp_knn_1_1():
-	for test_size in test_sizes:
-		ds_name = 'Letter-high'
-		knn_options = {'n_neighbors': 1,
-					   'n_splits': 30,
-					   'test_size': test_size,
-					   'verbose': True}
-		mpg_options = {'fit_method': 'k-graphs',
-					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
-					   'ds_name': ds_name,
-					   'parallel': False, # False
-					   'time_limit_in_sec': 0,
-					   'max_itrs': 100,
-					   'max_itrs_without_update': 3,
-					   'epsilon_residual': 0.01,
-					   'epsilon_ec': 0.1,
-					   'allow_zeros': allow_zeros,
-					   'triangle_rule': triangle_rule,
-					   'verbose': 1}
-		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
-		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
-		kernel_options = {'name': 'StructuralSP',
-						  'edge_weight': None,
-						  'node_kernels': sub_kernels,
-						  'edge_kernels': sub_kernels, 
-						  'compute_method': 'naive',
-						  'parallel': 'imap_unordered', 
-# 						  'parallel': None, 
-						  'n_jobs': multiprocessing.cpu_count(),
-						  'normalize': True,
-						  'verbose': 0}
-		ged_options = {'method': 'IPFP',
-					   'initialization_method': 'RANDOM', # 'NODE'
-					   'initial_solutions': initial_solutions, # 1
-					   'edit_cost': 'LETTER2',
-					   'attr_distance': 'euclidean',
-					   'ratio_runs_from_initial_solutions': 1,
-					   'threads': multiprocessing.cpu_count(),
-					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
-		mge_options = {'init_type': 'MEDOID',
-					   'random_inits': 10,
-					   'time_limit': 0,
-					   'verbose': 1,
-					   'update_order': update_order,
-					   'randomness': 'REAL',
-					   'refine': False}
-		save_results = True
-		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
-		
-		if not os.path.exists(dir_save):
-			os.makedirs(dir_save)
-		file_output = open(dir_save + 'output.txt', 'a')
-# 		sys.stdout = file_output
-
-		# print settings.
-		print('parameters:')
-		print('dataset name:', ds_name)
-		print('mpg_options:', mpg_options)
-		print('kernel_options:', kernel_options)
-		print('ged_options:', ged_options)
-		print('mge_options:', mge_options)
-		print('save_results:', save_results)
-		
-		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
-# 		for train_examples in ['expert']:
-			print('\n-------------------------------------')
-			print('train examples used:', train_examples, '\n')
-			mpg_options['fit_method'] = train_examples
-# 			try:
-			kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=None, edge_required=False, cut_range=None)
-# 			except Exception as exp:
-# 				print('An exception occured when running this experiment:')
-# 				LOG_FILENAME = dir_save + 'error.txt'
-# 				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
-# 				logging.exception('')
-# 				print(repr(exp))
-
-
-if __name__ == '__main__':
-	import pdb, traceback
-	try:
-		xp_knn_1_1()
-	except:
-		extype, value, tb = sys.exc_info()
-		traceback.print_exc()
-		pdb.post_mortem(tb)
\ No newline at end of file
diff --git a/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
new file mode 100644
index 0000000..9af5828
--- /dev/null
+++ b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
@@ -0,0 +1,3022 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Fri May 15 10:50:46 2020
+
+@author: ljia
+"""
+import multiprocessing
+import functools
+import sys
+import os
+import logging
+from gklearn.utils.kernels import deltakernel, gaussiankernel, kernelproduct
+from gklearn.preimage import kernel_knn_cv
+from gklearn.utils import compute_gram_matrices_by_class
+
+
+dir_root = '../results/xp_1nn.init10.triangle_rule/'
+initial_solutions = 10
+triangle_rule = True
+allow_zeros = False
+update_order = False
+test_sizes = [0.9, 0.7] # , 0.5, 0.3, 0.1]
+
+
+def xp_median_preimage_14_1():
+	"""xp 14_1: DD, PathUpToH, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'DD' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 2, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+	
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+	# 	# compute gram matrices for each class a priori.
+	# 	print('Compute gram matrices for each class a priori.')
+	# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=save_results, dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_13_1():
+	"""xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_13_2():
+	"""xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '') # 
+		irrelevant_labels = None # 
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']: # 
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_12_1():
+	"""xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 0, 1, 1, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_2():
+	"""xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 1, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+
+def xp_median_preimage_12_3():
+	"""xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+	"""
+	from gklearn.utils.kernels import gaussiankernel
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		pkernel = functools.partial(gaussiankernel, gamma=None) # @todo
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_4():
+	"""xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 14,
+						  'base_kernel': 'subtree',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+	# 	# compute gram matrices for each class a priori.
+	# 	print('Compute gram matrices for each class a priori.')
+	# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_5():
+	"""xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'PAH' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 0, 1, 1, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' # 
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']: # 
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_9_1():
+	"""xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MAO' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_2():
+	"""xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MAO' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 9, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_3():
+	"""xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'MAO' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		pkernel = functools.partial(polynomialkernel, d=4, c=1e+7)
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_4():
+	"""xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MAO' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 6,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_8_1():
+	"""xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_2():
+	"""xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 7, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_3():
+	"""xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		pkernel = functools.partial(polynomialkernel, d=2, c=1e+5)
+		kernel_options = {'name': 'Treelet',
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_4():
+	"""xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 4,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_7_1():
+	"""xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_2():
+	"""xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 2, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_7_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_3():
+	"""xp 7_3: MUTAG, Treelet, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'MUTAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		pkernel = functools.partial(polynomialkernel, d=3, c=1e+8)
+		kernel_options = {'name': 'Treelet',
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_4():
+	"""xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 1,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_6_1():
+	"""xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'COIL-RAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_6_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_6_2():
+	"""xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'COIL-RAG' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_6_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+
+def xp_median_preimage_5_1():
+	"""xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'FRANKENSTEIN' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_4_1():
+	"""xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'COLORS-3' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_3_2():
+	"""xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Fingerprint' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.01, 0.125, 0.125], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'edge_attrs': ['orient', 'angle']} #
+		edge_required = True #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_3_1():
+	"""xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Fingerprint' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.01, 0.125, 0.125], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'edge_attrs': ['orient', 'angle']} #
+		edge_required = False #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_2_1():
+	"""xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'COIL-DEL' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.node_attrs/'
+		irrelevant_labels = {'edge_labels': ['valence']}
+		edge_required = False
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_1_1():
+	"""xp 1_1: Letter-high, StructuralSP.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-high'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None
+		edge_required = False
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_1_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_1_2():
+	"""xp 1_2: Letter-high, ShortestPath.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-high'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_1_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_10_1():
+	"""xp 10_1: Letter-med, StructuralSP.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-med'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.75, 0.475, 0.475],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None
+		edge_required = False
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_10_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_10_2():
+	"""xp 10_2: Letter-med, ShortestPath.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-med'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.75, 0.475, 0.475],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_10_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+
+def xp_median_preimage_11_1():
+	"""xp 11_1: Letter-low, StructuralSP.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-low'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.075, 0.075, 0.25, 0.075, 0.075],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None
+		edge_required = False
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_11_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_11_2():
+	"""xp 11_2: Letter-low, ShortestPath.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'Letter-low'
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.075, 0.075, 0.25, 0.075, 0.075],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_11_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+if __name__ == "__main__":
+	
+# 	#### xp 1_1: Letter-high, StructuralSP.
+ 	# xp_median_preimage_1_1()
+
+# 	#### xp 1_2: Letter-high, ShortestPath.
+ 	# xp_median_preimage_1_2()
+
+# 	#### xp 10_1: Letter-med, StructuralSP.
+ 	# xp_median_preimage_10_1()
+
+# 	#### xp 10_2: Letter-med, ShortestPath.
+ 	# xp_median_preimage_10_2()
+
+# 	#### xp 11_1: Letter-low, StructuralSP.
+ 	# xp_median_preimage_11_1()
+
+# 	#### xp 11_2: Letter-low, ShortestPath.
+ 	# xp_median_preimage_11_2()
+# 	
+# 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+# # 	xp_median_preimage_2_1()
+# 	
+# 	#### xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+#  	# xp_median_preimage_3_1()
+
+# 	#### xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+ 	# xp_median_preimage_3_2()
+
+# 	#### xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_4_1()
+# 	
+# 	#### xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_5_1()
+# 	
+# 	#### xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_1()
+
+# 	#### xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_2()
+
+# 	#### xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+ 	# xp_median_preimage_7_1()
+
+# 	#### xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+ 	# xp_median_preimage_7_2()
+
+# 	#### xp 7_3: MUTAG, Treelet, using CONSTANT.
+ 	# xp_median_preimage_7_3()
+
+# 	#### xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+ 	# xp_median_preimage_7_4()
+# 	 
+#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+ 	# xp_median_preimage_8_1()
+
+# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+ 	# xp_median_preimage_8_2()
+
+# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+ 	# xp_median_preimage_8_3()
+
+# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+ 	# xp_median_preimage_8_4()
+
+# 	#### xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_1()
+
+# 	#### xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_2()
+
+# 	#### xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_3()
+
+# 	#### xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_4()
+
+	#### xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+ 	# xp_median_preimage_12_1()
+
+	#### xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_2()
+
+	#### xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_3()
+
+	#### xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_4()
+
+	#### xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+ 	# xp_median_preimage_12_5()
+
+	#### xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+ 	# xp_median_preimage_13_1()
+
+	#### xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+# 	xp_median_preimage_13_2()
+
+	#### xp 14_1: DD, PathUpToH, using CONSTANT.
+# 	xp_median_preimage_14_1()
+
+
+
+
+
+
+
+
+# 	#### xp 1_1: Letter-high, StructuralSP.
+ 	xp_median_preimage_1_1()
+
+# 	#### xp 1_2: Letter-high, ShortestPath.
+ 	xp_median_preimage_1_2()
+
+# 	#### xp 10_1: Letter-med, StructuralSP.
+ 	xp_median_preimage_10_1()
+
+# 	#### xp 10_2: Letter-med, ShortestPath.
+ 	xp_median_preimage_10_2()
+
+# 	#### xp 11_1: Letter-low, StructuralSP.
+ 	xp_median_preimage_11_1()
+
+# 	#### xp 11_2: Letter-low, ShortestPath.
+ 	xp_median_preimage_11_2()
+	 
+	#### xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+ 	xp_median_preimage_13_1()
+
+	#### xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+ 	xp_median_preimage_13_2()
+	 
+# 	#### xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+ 	xp_median_preimage_7_2()
+
+# 	#### xp 7_3: MUTAG, Treelet, using CONSTANT.
+ 	xp_median_preimage_7_3()
+
+# 	#### xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+ 	xp_median_preimage_7_4()
+# 	 
+# 	#### xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+ 	xp_median_preimage_7_1()
+
+# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+ 	xp_median_preimage_8_2()
+
+# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+ 	xp_median_preimage_8_3()
+
+# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+ 	xp_median_preimage_8_4()
+	 
+#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+ 	xp_median_preimage_8_1()
+
+# 	#### xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_2()
+
+# 	#### xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_3()
+
+# 	#### xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_4()
+	 
+# 	#### xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_1()
+
+	#### xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+ 	xp_median_preimage_12_1()
+
+	#### xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_2()
+
+	#### xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_3()
+
+	#### xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_4()
+
+	#### xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+ 	xp_median_preimage_12_5()
+	 
+# 	#### xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+ 	xp_median_preimage_6_1()
+# 	
+# 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+ 	xp_median_preimage_2_1()
+# 	
+# 	#### xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+#  	# xp_median_preimage_3_1()
+
+# 	#### xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+ 	# xp_median_preimage_3_2()
+
+# 	#### xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_4_1()
+# 	
+# 	#### xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_5_1()
+# 	
+# 	#### xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_2()
+
+	#### xp 14_1: DD, PathUpToH, using CONSTANT.
+# 	xp_median_preimage_14_1()
\ No newline at end of file
diff --git a/gklearn/preimage/median_preimage_generator.py b/gklearn/preimage/median_preimage_generator.py
index 9aaa88a..9deabe0 100644
--- a/gklearn/preimage/median_preimage_generator.py
+++ b/gklearn/preimage/median_preimage_generator.py
@@ -443,7 +443,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#			# 1. if c_vi != c_vr, c_ei != c_er.
 	#			nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 	#			x = cp.Variable(nb_cost_mat_new.shape[1])
-	#			cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+	#			cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 	##			# 1.1 no constraints.
 	##			constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])]]
 	#			# 1.2 c_vs <= c_vi + c_vr.
@@ -454,7 +454,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 	##			nb_cost_mat_new[:,0] += nb_cost_mat[:,1]
 	##			nb_cost_mat_new[:,2] += nb_cost_mat[:,5]
 	##			x = cp.Variable(nb_cost_mat_new.shape[1])
-	##			cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+	##			cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 	##			# 2.1 no constraints.
 	##			constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])]]
 	###			# 2.2 c_vs <= c_vi + c_vr.
@@ -469,7 +469,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 			if not self.__triangle_rule and self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 				constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 							   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 							   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -482,7 +482,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 			elif self.__triangle_rule and self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 				constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 							   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 							   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -496,7 +496,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 			elif not self.__triangle_rule and not self.__allow_zeros:
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 				prob.solve()
@@ -506,7 +506,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#				# c_vs <= c_vi + c_vr.
 	#				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 	#				x = cp.Variable(nb_cost_mat_new.shape[1])
-	#				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+	#				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 	#				constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 	#							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
 	#				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
@@ -519,7 +519,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				# c_vs <= c_vi + c_vr.
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
 							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
 				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
@@ -530,7 +530,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				# c_vs <= c_vi + c_vr and c_vi == c_vr, c_ei == c_er.
 				nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 				constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
 							   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0,
 							   np.array([1.0, -1.0, 0.0, 0.0, 0.0]).T@x == 0.0,
@@ -548,7 +548,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#			nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 				nb_cost_mat_new = nb_cost_mat[:,[2,3,4]]
 				x = cp.Variable(nb_cost_mat_new.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 	#			constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 	#						   np.array([0.0, 0.0, 0.0, 1.0, -1.0]).T@x == 0.0]
 	#			constraints = [x >= [0.0001 for i in range(nb_cost_mat_new.shape[1])]]
@@ -565,7 +565,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				if is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -578,7 +578,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif is_n_attr and not is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -591,7 +591,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif not is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -604,7 +604,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				else:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -615,7 +615,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				if is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -630,7 +630,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif is_n_attr and not is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -644,7 +644,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif not is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.0 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 								   np.array([0.0, 1.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -658,7 +658,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				else:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -669,7 +669,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				if is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -678,7 +678,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif is_n_attr and not is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -687,7 +687,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif not is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -696,7 +696,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				else:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -708,7 +708,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				if is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
 								   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
@@ -719,7 +719,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif is_n_attr and not is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,2,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
@@ -729,7 +729,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				elif not is_n_attr and is_e_attr:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4,5]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])],
 								   np.array([0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
@@ -739,7 +739,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				else:
 					nb_cost_mat_new = nb_cost_mat[:,[0,1,3,4]]
 					x = cp.Variable(nb_cost_mat_new.shape[1])
-					cost_fun = cp.sum_squares(nb_cost_mat_new * x - dis_k_vec)
+					cost_fun = cp.sum_squares(nb_cost_mat_new @ x - dis_k_vec)
 					constraints = [x >= [0.01 for i in range(nb_cost_mat_new.shape[1])]]
 					prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 					self.__execute_cvx(prob)
@@ -750,7 +750,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 		elif self.__ged_options['edit_cost'] == 'CONSTANT': # @todo: node/edge may not labeled.
 			if not self.__triangle_rule and self.__allow_zeros:
 				x = cp.Variable(nb_cost_mat.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat @ x - dis_k_vec)
 				constraints = [x >= [0.0 for i in range(nb_cost_mat.shape[1])],
 							   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 							   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -762,7 +762,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				residual = np.sqrt(prob.value)
 			elif self.__triangle_rule and self.__allow_zeros:
 				x = cp.Variable(nb_cost_mat.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat @ x - dis_k_vec)
 				constraints = [x >= [0.0 for i in range(nb_cost_mat.shape[1])],
 							   np.array([1.0, 0.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
 							   np.array([0.0, 1.0, 0.0, 0.0, 0.0, 0.0]).T@x >= 0.01,
@@ -776,7 +776,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				residual = np.sqrt(prob.value)
 			elif not self.__triangle_rule and not self.__allow_zeros:
 				x = cp.Variable(nb_cost_mat.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat @ x - dis_k_vec)
 				constraints = [x >= [0.01 for i in range(nb_cost_mat.shape[1])]]
 				prob = cp.Problem(cp.Minimize(cost_fun), constraints)
 				self.__execute_cvx(prob)
@@ -784,7 +784,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 				residual = np.sqrt(prob.value)
 			elif self.__triangle_rule and not self.__allow_zeros:
 				x = cp.Variable(nb_cost_mat.shape[1])
-				cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+				cost_fun = cp.sum_squares(nb_cost_mat @ x - dis_k_vec)
 				constraints = [x >= [0.01 for i in range(nb_cost_mat.shape[1])],
 							   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
 							   np.array([0.0, 0.0, 0.0, 1.0, 1.0, -1.0]).T@x >= 0.0]
@@ -818,7 +818,7 @@ class MedianPreimageGenerator(PreimageGenerator):
 	#	G = -1 * np.identity(nb_cost_mat.shape[1])
 	#	h = np.array([0 for i in range(nb_cost_mat.shape[1])])
 			x = cp.Variable(nb_cost_mat.shape[1])
-			cost_fun = cp.sum_squares(nb_cost_mat * x - dis_k_vec)
+			cost_fun = cp.sum_squares(nb_cost_mat @ x - dis_k_vec)
 			constraints = [x >= [0.0 for i in range(nb_cost_mat.shape[1])],
 		#				   np.array([1.0, 1.0, -1.0, 0.0, 0.0]).T@x >= 0.0]
 						   np.array([1.0, 1.0, -1.0, 0.0, 0.0, 0.0]).T@x >= 0.0,
diff --git a/gklearn/utils/graph_files.py b/gklearn/utils/graph_files.py
index d977b73..7de4ba0 100644
--- a/gklearn/utils/graph_files.py
+++ b/gklearn/utils/graph_files.py
@@ -494,7 +494,8 @@ def load_tud(filename):
 					   'edge_labels': [], 'edge_attrs': []}
 		class_label_map = None
 		class_label_map_strings = []
-		content_rm = open(frm).read().splitlines()
+		with open(frm) as rm:
+			content_rm = rm.read().splitlines()
 		i = 0
 		while i < len(content_rm):
 			line = content_rm[i].strip()
@@ -558,16 +559,20 @@ def load_tud(filename):
 		label_names = {'node_labels': [], 'node_attrs': [], 
 					   'edge_labels': [], 'edge_attrs': []}
 		class_label_map = None
-
-	content_gi = open(fgi).read().splitlines()  # graph indicator
-	content_am = open(fam).read().splitlines()  # adjacency matrix
+	
+	with open(fgi) as gi:
+		content_gi = gi.read().splitlines()  # graph indicator
+	with open(fam) as am:
+		content_am = am.read().splitlines()  # adjacency matrix
 	
 	# load targets.
 	if 'fgl' in locals():
-		content_targets = open(fgl).read().splitlines()  # targets (classification)
+		with open(fgl) as gl:
+			content_targets = gl.read().splitlines()  # targets (classification)
 		targets = [float(i) for i in content_targets]
 	elif 'fga' in locals():
-		content_targets = open(fga).read().splitlines()  # targets (regression)
+		with open(fga) as ga:
+			content_targets = ga.read().splitlines()  # targets (regression)
 		targets = [int(i) for i in content_targets]
 	else:
 		raise Exception('Can not find targets file. Please make sure there is a "', ds_name, '_graph_labels.txt" or "', ds_name, '_graph_attributes.txt"', 'file in your dataset folder.')
@@ -577,7 +582,8 @@ def load_tud(filename):
 	# create graphs and add nodes
 	data = [nx.Graph(name=str(i)) for i in range(0, len(content_targets))]
 	if 'fnl' in locals():
-		content_nl = open(fnl).read().splitlines()  # node labels
+		with open(fnl) as nl:
+			content_nl = nl.read().splitlines()  # node labels
 		for idx, line in enumerate(content_gi):
 			# transfer to int first in case of unexpected blanks
 			data[int(line) - 1].add_node(idx)
@@ -605,7 +611,8 @@ def load_tud(filename):
 
 	# add edge labels
 	if 'fel' in locals():
-		content_el = open(fel).read().splitlines()
+		with open(fel) as el:
+			content_el = el.read().splitlines()
 		for idx, line in enumerate(content_el):
 			labels = [l.strip() for l in line.split(',')]
 			n = [int(i) - 1 for i in content_am[idx].split(',')]
@@ -621,7 +628,8 @@ def load_tud(filename):
 
 	# add node attributes
 	if 'fna' in locals():
-		content_na = open(fna).read().splitlines()
+		with open(fna) as na:
+			content_na = na.read().splitlines()
 		for idx, line in enumerate(content_na):
 			attrs = [a.strip() for a in line.split(',')]
 			g = int(content_gi[idx]) - 1
@@ -636,7 +644,8 @@ def load_tud(filename):
 
 	# add edge attributes
 	if 'fea' in locals():
-		content_ea = open(fea).read().splitlines()
+		with open(fea) as ea:
+			content_ea = ea.read().splitlines()
 		for idx, line in enumerate(content_ea):
 			attrs = [a.strip() for a in line.split(',')]
 			n = [int(i) - 1 for i in content_am[idx].split(',')]
@@ -669,7 +678,8 @@ def load_from_ds(filename, filename_targets):
 	data = []
 	y = []
 	label_names = {'node_labels': [], 'edge_labels': [], 'node_attrs': [], 'edge_attrs': []}
-	content = open(filename).read().splitlines()
+	with open(filename) as fn:
+		content = fn.read().splitlines()
 	extension = splitext(content[0].split(' ')[0])[1][1:]
 	if extension == 'ct':
 		load_file_fun = load_ct
@@ -691,8 +701,9 @@ def load_from_ds(filename, filename_targets):
 			g, l_names = load_file_fun(dirname_dataset + '/' + tmp.replace('#', '', 1))
 			data.append(g)
 			__append_label_names(label_names, l_names)
-															   
-		content_y = open(filename_targets).read().splitlines()
+		
+		with open(filename_targets) as fnt:
+			content_y = fnt.read().splitlines()
 		# assume entries in filename and filename_targets have the same order.
 		for item in content_y:
 			tmp = item.split(' ')

From cac0cfa114000ba3becb53bcae7715a299dc7acf Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Tue, 19 May 2020 10:00:42 +0200
Subject: [PATCH 12/17] Add tools to analyze preimage results.

---
 .gitignore                                         |   1 +
 .../tools/analyze_results_of_random_edit_costs.py  |  86 ++++++++
 .../tools/preimage_results_to_latex_tables.py      | 228 +++++++++++++++++++++
 3 files changed, 315 insertions(+)
 create mode 100644 gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
 create mode 100644 gklearn/preimage/experiments/tools/preimage_results_to_latex_tables.py

diff --git a/.gitignore b/.gitignore
index 0146aaf..13141a3 100644
--- a/.gitignore
+++ b/.gitignore
@@ -32,6 +32,7 @@ gklearn/kernels/*_sym.py
 gklearn/preimage/*
 !gklearn/preimage/*.py
 !gklearn/preimage/experiments/*.py
+!gklearn/preimage/experiments/tools/*.py
 
 __pycache__
 ##*#
diff --git a/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py b/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
new file mode 100644
index 0000000..4dc20d3
--- /dev/null
+++ b/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
@@ -0,0 +1,86 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Apr 14 16:57:18 2020
+
+@author: ljia
+"""
+import pandas as pd
+import numpy as np
+import os
+import math
+
+def summarize_results_of_random_edit_costs(data_dir, ds_name, gkernel):
+	sod_sm_list = []
+	sod_gm_list = []
+	dis_k_sm_list = []
+	dis_k_gm_list = []
+	dis_k_min_gi = []
+	time_total_list = []
+	mge_dec_order_list = []
+	mge_inc_order_list = []
+	
+	# get results from .csv.
+	file_name = data_dir + 'results_summary.' + ds_name + '.' + gkernel + '.csv'
+	try:
+		df = pd.read_csv(file_name)
+	except FileNotFoundError:
+		return
+	for index, row in df.iterrows():
+		if row['target'] == 'all' and row['fit method'] == 'random':
+			if not math.isnan(float(row['SOD SM'])):
+				sod_sm_list.append(float(row['SOD SM']))
+			if not math.isnan(float(row['SOD GM'])):
+				sod_gm_list.append(float(row['SOD GM']))
+			if not math.isnan(float(row['dis_k SM'])):
+				dis_k_sm_list.append(float(row['dis_k SM']))
+			if not math.isnan(float(row['dis_k GM'])):
+				dis_k_gm_list.append(float(row['dis_k GM']))
+			if not math.isnan(float(row['min dis_k gi'])):
+				dis_k_min_gi.append(float(row['min dis_k gi']))
+			if not math.isnan(float(row['time total'])):
+				time_total_list.append(float(row['time total']))
+			if 'mge num decrease order' in row:
+				mge_dec_order_list.append(int(row['mge num decrease order']))
+			if 'mge num increase order' in row:
+				mge_inc_order_list.append(int(row['mge num increase order']))
+	# return if no results.
+	if len(sod_sm_list) == 0:
+		return
+	
+	# construct output results.
+	op = {}
+	op['measure'] = ['max', 'min', 'mean']
+	op['SOD SM'] = [np.max(sod_sm_list), np.min(sod_sm_list), np.mean(sod_sm_list)]
+	op['SOD GM'] = [np.max(sod_gm_list), np.min(sod_gm_list), np.mean(sod_gm_list)]
+	op['dis_k SM'] = [np.max(dis_k_sm_list), np.min(dis_k_sm_list), np.mean(dis_k_sm_list)]
+	op['dis_k GM'] = [np.max(dis_k_gm_list), np.min(dis_k_gm_list), np.mean(dis_k_gm_list)]
+	op['min dis_k gi'] = [np.max(dis_k_min_gi), np.min(dis_k_min_gi), np.mean(dis_k_min_gi)]
+	op['time total'] = [np.max(time_total_list), np.min(time_total_list), np.mean(time_total_list)]
+	if len(mge_dec_order_list) > 0:
+		op['mge num decrease order'] = [np.max(mge_dec_order_list), np.min(mge_dec_order_list), np.mean(mge_dec_order_list)]
+	if len(mge_inc_order_list) > 0:
+		op['mge num increase order'] = [np.max(mge_inc_order_list), np.min(mge_inc_order_list), np.mean(mge_inc_order_list)]
+	df = pd.DataFrame(data=op)
+	
+	# write results to .csv
+	df.to_csv(data_dir + 'summary_for_random_edit_costs.csv', index=False, header=True)
+	
+	
+def compute_for_all_experiments(data_dir):
+	dir_list = [i for i in os.listdir(data_dir) if os.path.isdir(data_dir + i)]
+	for dir_name in dir_list:
+		sp_tmp = dir_name.split('.')
+		ds_name = sp_tmp[0].strip('[error]')
+		gkernel = sp_tmp[1]
+		summarize_results_of_random_edit_costs(data_dir + dir_name + '/',
+										 ds_name, gkernel)
+		if os.path.exists(data_dir + dir_name + '/update_order/'):
+			summarize_results_of_random_edit_costs(data_dir + dir_name + '/update_order/',
+										 ds_name, gkernel)
+
+
+if __name__ == '__main__':
+# 	data_dir = '../results/xp_median_preimage.update_order/'
+	data_dir = '../../results/CRIANN/xp_median_preimage.init10/'
+	compute_for_all_experiments(data_dir)
\ No newline at end of file
diff --git a/gklearn/preimage/experiments/tools/preimage_results_to_latex_tables.py b/gklearn/preimage/experiments/tools/preimage_results_to_latex_tables.py
new file mode 100644
index 0000000..301f87f
--- /dev/null
+++ b/gklearn/preimage/experiments/tools/preimage_results_to_latex_tables.py
@@ -0,0 +1,228 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Thu Apr 30 10:16:33 2020
+
+@author: ljia
+"""
+import pandas as pd
+import numpy as np
+import os
+
+
+DS_SYMB = ['MUTAG', 'Monoterpenoides', 'MAO_symb']
+DS_NON_SYMB = ['Letter-high', 'Letter-med', 'Letter-low', 'COIL-RAG', 'PAH']
+DS_UNLABELED = ['PAH_unlabeled']
+
+
+def rounder(x, decimals):
+	x_strs = str(x).split('.')
+	if len(x_strs) == 2:
+		before = x_strs[0]
+		after = x_strs[1]
+		if len(after) > decimals:
+			if int(after[decimals]) >= 5:
+				after0s = ''
+				for c in after:
+					if c == '0':
+						after0s += '0'
+					elif c != '0':
+						break
+				after = after0s + str(int(after[0:decimals]) + 1)[-decimals:]
+			else:
+				after = after[0:decimals]
+		elif len(after) < decimals:
+			after += '0' * (decimals - len(after))
+		return before + '.' + after
+
+	elif len(x_strs) == 1:
+		return x_strs[0]
+	
+
+def replace_nth(string, sub, wanted, n):
+	import re
+	where = [m.start() for m in re.finditer(sub, string)][n-1]
+	before = string[:where]
+	after = string[where:]
+	after = after.replace(sub, wanted, 1)
+	newString = before + after
+	return newString
+
+
+def df_to_latex_table(df):
+	ltx = df.to_latex(index=True, escape=False, multirow=True)
+	
+	# modify middle lines.
+	ltx = ltx.replace('\\cline{1-9}\n\\cline{2-9}', 	'\\toprule')
+	ltx = ltx.replace('\\cline{2-9}', '\\cmidrule(l){2-9}')
+	
+	# modify first row.
+	i_start = ltx.find('\n\\toprule\n')
+	i_end = ltx.find('\\\\\n\\midrule\n')
+	ltx = ltx.replace(ltx[i_start:i_end+12], '\n\\toprule\nDatasets & Graph Kernels & Algorithms & $d_\\mathcal{F}$ SM & $d_\\mathcal{F}$ SM (UO) & $d_\\mathcal{F}$ GM & $d_\\mathcal{F}$ GM (UO) & Runtime & Runtime (UO) \\\\\n\\midrule\n', 1)
+	
+	# add row numbers.
+	ltx = ltx.replace('lllllllll', 'lllllllll|@{\\makebox[2em][r]{\\textit{\\rownumber\\space}}}', 1)
+	ltx = replace_nth(ltx, '\\\\\n', '\\gdef\\rownumber{\\stepcounter{magicrownumbers}\\arabic{magicrownumbers}} \\\\\n', 1)
+		
+	return ltx
+
+
+def beautify_df(df):
+	df = df.sort_values(by=['Datasets', 'Graph Kernels'])
+	df = df.set_index(['Datasets', 'Graph Kernels', 'Algorithms'])
+# 	index = pd.MultiIndex.from_frame(df[['Datasets', 'Graph Kernels', 'Algorithms']])
+
+	# bold the best results.
+	for ds in df.index.get_level_values('Datasets').unique():
+		for gk in df.loc[ds].index.get_level_values('Graph Kernels').unique():
+			min_val = np.inf
+			min_indices = []
+			min_labels = []
+			for index, row in df.loc[(ds, gk)].iterrows():
+				for label in ['$d_\mathcal{F}$ SM', '$d_\mathcal{F}$ GM', '$d_\mathcal{F}$ GM (UO)']:
+					value = row[label]
+					if value != '-':
+						value = float(value.strip('/same'))
+						if value < min_val:
+							min_val = value
+							min_indices = [index]
+							min_labels = [label]
+						elif value == min_val:
+							min_indices.append(index)
+							min_labels.append(label)
+			for idx, index in enumerate(min_indices):
+				df.loc[(ds, gk, index), min_labels[idx]] = '\\textbf{' + df.loc[(ds, gk, index), min_labels[idx]] + '}'
+	
+	return df
+
+
+def get_results(data_dir, ds_name, gkernel):
+	# get results from .csv.
+	file_name = data_dir + 'results_summary.' + ds_name + '.' + gkernel + '.csv'
+	try:
+		df_summary = pd.read_csv(file_name)
+	except FileNotFoundError:
+		return None
+
+	df_results = pd.DataFrame(index=None, columns=['d_F SM', 'd_F GM', 'runtime'])
+	for index, row in df_summary.iterrows():
+		if row['target'] == 'all' and row['fit method'] == 'k-graphs':
+			df_results.loc['From median set'] = ['-', rounder(row['min dis_k gi'], 3), '-']
+			if_uo = (int(row['mge num decrease order']) > 0 or int(row['mge num increase order']) > 0)
+			df_results.loc['Optimized'] = [rounder(row['dis_k SM'], 3), 
+								  rounder(row['dis_k GM'], 3) if if_uo else (rounder(row['dis_k GM'], 3) + '/same'),
+								  rounder(row['time total'], 2)]
+		if row['target'] == 'all' and row['fit method'] == 'expert':
+			if_uo = (int(row['mge num decrease order']) > 0 or int(row['mge num increase order']) > 0)
+			df_results.loc['IAM: expert costs'] = [rounder(row['dis_k SM'], 3), 
+								  rounder(row['dis_k GM'], 3) if if_uo else (rounder(row['dis_k GM'], 3) + '/same'),
+								  rounder(row['time total'], 2)]
+								  
+	# get results from random summary .csv.
+	random_fini = True
+	file_name = data_dir + 'summary_for_random_edit_costs.csv'
+	try:
+		df_random = pd.read_csv(file_name)
+	except FileNotFoundError:
+		random_fini = False
+
+	if random_fini:								  
+		for index, row in df_random.iterrows():
+			if row['measure'] == 'mean':
+				if_uo = (float(row['mge num decrease order']) > 0 or float(row['mge num increase order']) > 0)
+				df_results.loc['IAM: random costs'] = [rounder(row['dis_k SM'], 3), 
+								  rounder(row['dis_k GM'], 3) if if_uo else (rounder(row['dis_k GM'], 3) + '/same'),
+								  rounder(row['time total'], 2)]
+				
+	# sort index.
+	df_results = df_results.reindex([item for item in ['From median set', 'IAM: random costs', 'IAM: expert costs', 'Optimized'] if item in df_results.index])
+				
+	return df_results
+	
+
+def get_results_of_one_xp(data_dir, ds_name, gkernel):
+	df_results = pd.DataFrame()
+	
+	df_tmp_uo = None
+	if not os.path.isfile(data_dir + 'update_order/error.txt'):
+		df_tmp_uo = get_results(data_dir + 'update_order/', ds_name, gkernel)
+
+	df_tmp = None
+	if not os.path.isfile(data_dir + 'error.txt'):
+		df_tmp = get_results(data_dir, ds_name, gkernel)
+
+	if (df_tmp_uo is not None and not df_tmp_uo.empty) or (df_tmp is not None and not df_tmp.empty):
+		df_results = pd.DataFrame(index=['From median set', 'IAM: random costs', 'IAM: expert costs', 'Optimized'], columns=['$d_\mathcal{F}$ SM', '$d_\mathcal{F}$ SM (UO)', '$d_\mathcal{F}$ GM', '$d_\mathcal{F}$ GM (UO)', 'Runtime', 'Runtime (UO)'])
+		if df_tmp_uo is not None and not df_tmp_uo.empty:
+			for index, row in df_tmp_uo.iterrows():
+				for algo in df_results.index:
+					if index == algo:
+						df_results.at[algo, '$d_\mathcal{F}$ SM (UO)'] = row['d_F SM']
+						df_results.at[algo, '$d_\mathcal{F}$ GM (UO)'] = row['d_F GM']
+						df_results.at[algo, 'Runtime (UO)'] = row['runtime']
+		if df_tmp is not None and not df_tmp.empty:
+			for index, row in df_tmp.iterrows():
+				for algo in df_results.index:
+					if index == algo:
+						df_results.at[algo, '$d_\mathcal{F}$ SM'] = row['d_F SM']
+						df_results.at[algo, '$d_\mathcal{F}$ GM'] = row['d_F GM'].strip('/same')
+						df_results.at[algo, 'Runtime'] = row['runtime']	
+	
+	df_results = df_results.dropna(axis=0, how='all')
+	df_results = df_results.fillna(value='-')
+	df_results = df_results.reset_index().rename(columns={'index': 'Algorithms'})
+		
+	return df_results
+
+
+def get_results_for_all_experiments(root_dir):
+	columns=['Datasets', 'Graph Kernels', 'Algorithms', '$d_\mathcal{F}$ SM', '$d_\mathcal{F}$ SM (UO)', '$d_\mathcal{F}$ GM', '$d_\mathcal{F}$ GM (UO)', 'Runtime', 'Runtime (UO)']
+	df_symb = pd.DataFrame(columns=columns)
+	df_nonsymb = pd.DataFrame(columns=columns)
+	df_unlabeled = pd.DataFrame(columns=columns)
+	
+	dir_list = [i for i in os.listdir(root_dir) if os.path.isdir(root_dir + i)]
+	for dir_name in dir_list:
+		sp_tmp = dir_name.split('.')
+		gkernel = sp_tmp[1]
+		ds_name = sp_tmp[0].strip('[error]')
+		suffix = ''
+		if sp_tmp[-1] == 'unlabeled':
+ 			suffix = '_unlabeled'
+		elif sp_tmp[-1] == 'symb':
+ 			suffix = '_symb'
+			 
+		df_results = get_results_of_one_xp(root_dir + dir_name + '/', ds_name, gkernel)
+		
+		if not df_results.empty:
+			ds_name += suffix
+			if ds_name in DS_SYMB:
+				for index, row in df_results.iterrows():
+					df_symb.loc[len(df_symb)] = [ds_name.replace('_', '\_'), gkernel] + row.tolist()
+			elif ds_name in DS_NON_SYMB:
+				for index, row in df_results.iterrows():
+					df_nonsymb.loc[len(df_nonsymb)] = [ds_name.replace('_', '\_'), gkernel] + row.tolist()
+			elif ds_name in DS_UNLABELED:
+				for index, row in df_results.iterrows():
+					df_unlabeled.loc[len(df_unlabeled)] = [ds_name.replace('_', '\_'), gkernel] + row.tolist()
+			else:
+				raise Exception('dataset' + ds_name + 'is not pre-defined.')
+				
+	# sort.
+	df_symb = beautify_df(df_symb)
+	df_nonsymb = beautify_df(df_nonsymb)
+	df_unlabeled = beautify_df(df_unlabeled)
+	
+	# convert dfs to latex strings.
+	ltx_symb = df_to_latex_table(df_symb)
+	ltx_nonsymb = df_to_latex_table(df_nonsymb)
+	ltx_unlabeled = df_to_latex_table(df_unlabeled)
+	
+	return ltx_symb, ltx_nonsymb, ltx_unlabeled
+
+
+if __name__ == '__main__':
+# 	root_dir = '../results/xp_median_preimage.init20/'
+	root_dir = '../../results/CRIANN/xp_median_preimage.init10/'
+	ltx_symb, ltx_nonsymb, ltx_unlabeled = get_results_for_all_experiments(root_dir)

From 06c35201b264218baac2cb05dc058e2314a45858 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Tue, 26 May 2020 15:55:58 +0200
Subject: [PATCH 13/17] Add option to sort graphs by size when computing GEDs
 for the sake of the stability.

---
 gklearn/ged/env/node_map.py                        |  28 ++--
 gklearn/ged/median/median_graph_estimator.py       | 146 ++++++++++++++++-----
 gklearn/ged/median/utils.py                        |   2 +
 gklearn/ged/util/util.py                           |  24 ++--
 .../tools/analyze_results_of_random_edit_costs.py  |   9 +-
 .../experiments/xp_1nn_init10_trianglerule.py      |  24 ++--
 gklearn/preimage/kernel_knn_cv.py                  |   2 +-
 7 files changed, 166 insertions(+), 69 deletions(-)

diff --git a/gklearn/ged/env/node_map.py b/gklearn/ged/env/node_map.py
index 4812486..dc3e3bf 100644
--- a/gklearn/ged/env/node_map.py
+++ b/gklearn/ged/env/node_map.py
@@ -39,14 +39,6 @@ class NodeMap(object):
 		return np.inf
 	
 	
-	def get_forward_map(self):
-		return self.__forward_map
-	
-	
-	def get_backward_map(self):
-		return self.__backward_map
-	
-	
 	def as_relation(self, relation):
 		relation.clear()
 		for i in range(0, len(self.__forward_map)):
@@ -77,4 +69,22 @@ class NodeMap(object):
 		
 		
 	def induced_cost(self):
-		return self.__induced_cost
\ No newline at end of file
+		return self.__induced_cost
+	
+	
+	@property
+	def forward_map(self):
+		return self.__forward_map
+
+	@forward_map.setter
+	def forward_map(self, value):
+		self.__forward_map = value	
+		
+		
+	@property
+	def backward_map(self):
+		return self.__backward_map
+
+	@backward_map.setter
+	def backward_map(self, value):
+		self.__backward_map = value	
\ No newline at end of file
diff --git a/gklearn/ged/median/median_graph_estimator.py b/gklearn/ged/median/median_graph_estimator.py
index c4291ce..03c7892 100644
--- a/gklearn/ged/median/median_graph_estimator.py
+++ b/gklearn/ged/median/median_graph_estimator.py
@@ -52,6 +52,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		self.__seed = 0
 		self.__parallel = True
 		self.__update_order = True
+		self.__sort_graphs = True # sort graphs by size when computing GEDs.
 		self.__refine = True
 		self.__time_limit_in_sec = 0
 		self.__epsilon = 0.0001
@@ -150,6 +151,16 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 				else:
 					raise Exception('Invalid argument "' + opt_val  + '" for option update-order. Usage: options = "[--update-order TRUE|FALSE] [...]"')
 					
+			elif opt_name == 'sort-graphs':
+				if opt_val == 'TRUE':
+					self.__sort_graphs = True
+	
+				elif opt_val == 'FALSE':
+					self.__sort_graphs = False
+	
+				else:
+					raise Exception('Invalid argument "' + opt_val  + '" for option sort-graphs. Usage: options = "[--sort-graphs TRUE|FALSE] [...]"')
+					
 			elif opt_name == 'refine':
 				if opt_val == 'TRUE':
 					self.__refine = True
@@ -537,18 +548,31 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			progress.update(1)
 			
 		# Improving the node maps.
+		nb_nodes_median = self.__ged_env.get_graph_num_nodes(self.__gen_median_id)
 		for graph_id, node_map in self.__node_maps_from_median.items():
 			if time.expired():
 				if self.__state == AlgorithmState.TERMINATED:
 					self.__state = AlgorithmState.CONVERGED
 				break
-			self.__ged_env.run_method(self.__gen_median_id, graph_id)
-			if self.__ged_env.get_upper_bound(self.__gen_median_id, graph_id) < node_map.induced_cost():
-				self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__gen_median_id, graph_id)
-			self.__sum_of_distances += self.__node_maps_from_median[graph_id].induced_cost()
+
+			nb_nodes_g = self.__ged_env.get_graph_num_nodes(graph_id)
+			if nb_nodes_median <= nb_nodes_g or not self.__sort_graphs:			
+				self.__ged_env.run_method(self.__gen_median_id, graph_id)
+				if self.__ged_env.get_upper_bound(self.__gen_median_id, graph_id) < node_map.induced_cost():
+					self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__gen_median_id, graph_id)
+			else:
+				self.__ged_env.run_method(graph_id, self.__gen_median_id)
+				if self.__ged_env.get_upper_bound(graph_id, self.__gen_median_id) < node_map.induced_cost():
+						node_map_tmp = self.__ged_env.get_node_map(graph_id, self.__gen_median_id)
+						node_map_tmp.forward_map, node_map_tmp.backward_map = node_map_tmp.backward_map, node_map_tmp.forward_map
+						self.__node_maps_from_median[graph_id] = node_map_tmp	
+			
+			self.__sum_of_distances += self.__node_maps_from_median[graph_id].induced_cost()				
+
 			# Print information.
 			if self.__print_to_stdout == 2:
 				progress.update(1)
+
 		self.__sum_of_distances = 0.0
 		for key, val in self.__node_maps_from_median.items():
 			self.__sum_of_distances += val.induced_cost()
@@ -636,6 +660,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 		self.__seed = 0
 		self.__parallel = True
 		self.__update_order = True
+		self.__sort_graphs = True
 		self.__refine = True
 		self.__time_limit_in_sec = 0
 		self.__epsilon = 0.0001
@@ -695,7 +720,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			def init_worker(ged_env_toshare):
 				global G_ged_env
 				G_ged_env = ged_env_toshare
-			do_fun = partial(_compute_medoid_parallel, graph_ids)
+			do_fun = partial(_compute_medoid_parallel, graph_ids, self.__sort_graphs)
 			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
 			if self.__print_to_stdout == 2:
 				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
@@ -723,10 +748,16 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 				if timer.expired():
 					self.__state = AlgorithmState.CALLED
 					break
+				nb_nodes_g = self.__ged_env.get_graph_num_nodes(g_id)
 				sum_of_distances = 0
-				for h_id in graph_ids:
-					self.__ged_env.run_method(g_id, h_id)
-					sum_of_distances += self.__ged_env.get_upper_bound(g_id, h_id)
+				for h_id in graph_ids:					
+					nb_nodes_h = self.__ged_env.get_graph_num_nodes(h_id)
+					if nb_nodes_g <= nb_nodes_h or not self.__sort_graphs:
+						self.__ged_env.run_method(g_id, h_id)
+						sum_of_distances += self.__ged_env.get_upper_bound(g_id, h_id)
+					else:
+						self.__ged_env.run_method(h_id, g_id)
+						sum_of_distances += self.__ged_env.get_upper_bound(h_id, g_id)	
 				if sum_of_distances < best_sum_of_distances:
 					best_sum_of_distances = sum_of_distances
 					medoid_id = g_id
@@ -760,7 +791,8 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			def init_worker(ged_env_toshare):
 				global G_ged_env
 				G_ged_env = ged_env_toshare
-			do_fun = partial(_compute_init_node_maps_parallel, gen_median_id)
+			nb_nodes_median = self.__ged_env.get_graph_num_nodes(gen_median_id)
+			do_fun = partial(_compute_init_node_maps_parallel, gen_median_id, self.__sort_graphs, nb_nodes_median)
 			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
 			if self.__print_to_stdout == 2:
 				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
@@ -783,9 +815,17 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 				
 			self.__sum_of_distances = 0
 			self.__node_maps_from_median.clear()
+			nb_nodes_median = self.__ged_env.get_graph_num_nodes(gen_median_id)
 			for graph_id in graph_ids:
-				self.__ged_env.run_method(gen_median_id, graph_id)
-				self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(gen_median_id, graph_id)
+				nb_nodes_g = self.__ged_env.get_graph_num_nodes(graph_id)
+				if nb_nodes_median <= nb_nodes_g or not self.__sort_graphs:
+					self.__ged_env.run_method(gen_median_id, graph_id)
+					self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(gen_median_id, graph_id)
+				else:
+					self.__ged_env.run_method(graph_id, gen_median_id)
+					node_map_tmp = self.__ged_env.get_node_map(graph_id, gen_median_id)
+					node_map_tmp.forward_map, node_map_tmp.backward_map = node_map_tmp.backward_map, node_map_tmp.forward_map
+					self.__node_maps_from_median[graph_id] = node_map_tmp
 	# 				print(self.__node_maps_from_median[graph_id])
 				self.__sum_of_distances += self.__node_maps_from_median[graph_id].induced_cost()
 	# 				print(self.__sum_of_distances)
@@ -843,7 +883,7 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			for graph_id, graph in graphs.items():
 # 				print('graph_id: ', graph_id)
 # 				print(self.__node_maps_from_median[graph_id])
-# 				print(self.__node_maps_from_median[graph_id].get_forward_map(), self.__node_maps_from_median[graph_id].get_backward_map())
+# 				print(self.__node_maps_from_median[graph_id].forward_map, self.__node_maps_from_median[graph_id].backward_map)
 				k = self.__node_maps_from_median[graph_id].image(i)
 # 				print('k: ', k)
 				if k != np.inf:
@@ -920,7 +960,8 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			def init_worker(ged_env_toshare):
 				global G_ged_env
 				G_ged_env = ged_env_toshare
-			do_fun = partial(_update_node_maps_parallel, self.__median_id, self.__epsilon)
+			nb_nodes_median = self.__ged_env.get_graph_num_nodes(self.__median_id)
+			do_fun = partial(_update_node_maps_parallel, self.__median_id, self.__epsilon, self.__sort_graphs, nb_nodes_median)
 			pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(self.__ged_env,))
 			if self.__print_to_stdout == 2:
 				iterator = tqdm(pool.imap_unordered(do_fun, itr, chunksize),
@@ -941,13 +982,25 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 				progress = tqdm(desc='Updating node maps', total=len(self.__node_maps_from_median), file=sys.stdout)
 				
 			node_maps_were_modified = False
+			nb_nodes_median = self.__ged_env.get_graph_num_nodes(self.__median_id)
 			for graph_id, node_map in self.__node_maps_from_median.items():
-				self.__ged_env.run_method(self.__median_id, graph_id)
-				if self.__ged_env.get_upper_bound(self.__median_id, graph_id) < node_map.induced_cost() - self.__epsilon:
-	# 				xxx = self.__node_maps_from_median[graph_id]
-					self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__median_id, graph_id)
-	# 				yyy = self.__node_maps_from_median[graph_id]
-					node_maps_were_modified = True
+				nb_nodes_g = self.__ged_env.get_graph_num_nodes(graph_id)
+				
+				if nb_nodes_median <= nb_nodes_g or not self.__sort_graphs:
+					self.__ged_env.run_method(self.__median_id, graph_id)
+					if self.__ged_env.get_upper_bound(self.__median_id, graph_id) < node_map.induced_cost() - self.__epsilon:
+		# 				xxx = self.__node_maps_from_median[graph_id]
+						self.__node_maps_from_median[graph_id] = self.__ged_env.get_node_map(self.__median_id, graph_id)
+						node_maps_were_modified = True
+						
+				else:
+					self.__ged_env.run_method(graph_id, self.__median_id)
+					if self.__ged_env.get_upper_bound(graph_id, self.__median_id) < node_map.induced_cost() - self.__epsilon:
+						node_map_tmp = self.__ged_env.get_node_map(graph_id, self.__median_id)
+						node_map_tmp.forward_map, node_map_tmp.backward_map = node_map_tmp.backward_map, node_map_tmp.forward_map
+						self.__node_maps_from_median[graph_id] = node_map_tmp	
+						node_maps_were_modified = True
+					
 				# Print information about current iteration.
 				if self.__print_to_stdout == 2:
 					progress.update(1)
@@ -1047,8 +1100,8 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 			for k in range(0, node_map.num_target_nodes()):
 				if is_unassigned_target_node[k]:
 					new_node_map.add_assignment(np.inf, k)
-# 			print(self.__node_maps_from_median[key].get_forward_map(), self.__node_maps_from_median[key].get_backward_map())
-# 			print(new_node_map.get_forward_map(), new_node_map.get_backward_map())
+# 			print(self.__node_maps_from_median[key].forward_map, self.__node_maps_from_median[key].backward_map)
+# 			print(new_node_map.forward_map, new_node_map.backward_map
 			self.__node_maps_from_median[key] = new_node_map
 			
 		# Increase overall number of decreases.
@@ -1599,37 +1652,58 @@ class MedianGraphEstimator(object): # @todo: differ dummy_node from undifined no
 #			return median_label
 
 
-def _compute_medoid_parallel(graph_ids, itr):
+def _compute_medoid_parallel(graph_ids, sort, itr):
 	g_id = itr[0]
 	i = itr[1]
 	# @todo: timer not considered here.
 # 			if timer.expired():
 # 				self.__state = AlgorithmState.CALLED
 # 				break
+	nb_nodes_g = G_ged_env.get_graph_num_nodes(g_id)
 	sum_of_distances = 0
 	for h_id in graph_ids:
-		G_ged_env.run_method(g_id, h_id)
-		sum_of_distances += G_ged_env.get_upper_bound(g_id, h_id)
+		nb_nodes_h = G_ged_env.get_graph_num_nodes(h_id)
+		if nb_nodes_g <= nb_nodes_h or not sort:
+			G_ged_env.run_method(g_id, h_id)
+			sum_of_distances += G_ged_env.get_upper_bound(g_id, h_id)
+		else:
+			G_ged_env.run_method(h_id, g_id)
+			sum_of_distances += G_ged_env.get_upper_bound(h_id, g_id)
 	return i, sum_of_distances
+				
 
-
-def _compute_init_node_maps_parallel(gen_median_id, itr):
+def _compute_init_node_maps_parallel(gen_median_id, sort, nb_nodes_median, itr):
 	graph_id = itr
-	G_ged_env.run_method(gen_median_id, graph_id)
-	node_maps_from_median = G_ged_env.get_node_map(gen_median_id, graph_id)
+	nb_nodes_g = G_ged_env.get_graph_num_nodes(graph_id)
+	if nb_nodes_median <= nb_nodes_g or not sort:
+		G_ged_env.run_method(gen_median_id, graph_id)
+		node_map = G_ged_env.get_node_map(gen_median_id, graph_id)
 # 				print(self.__node_maps_from_median[graph_id])
-	sum_of_distance = node_maps_from_median.induced_cost()
+	else:
+		G_ged_env.run_method(graph_id, gen_median_id)
+		node_map = G_ged_env.get_node_map(graph_id, gen_median_id)
+		node_map.forward_map, node_map.backward_map = node_map.backward_map, node_map.forward_map
+	sum_of_distance = node_map.induced_cost()
 # 				print(self.__sum_of_distances)
-	return graph_id, sum_of_distance, node_maps_from_median
-
+	return graph_id, sum_of_distance, node_map
+					
 
-def _update_node_maps_parallel(median_id, epsilon, itr):
+def _update_node_maps_parallel(median_id, epsilon, sort, nb_nodes_median, itr):
 	graph_id = itr[0]
 	node_map = itr[1]
 
 	node_maps_were_modified = False
-	G_ged_env.run_method(median_id, graph_id)
-	if G_ged_env.get_upper_bound(median_id, graph_id) < node_map.induced_cost() - epsilon:
-		node_map = G_ged_env.get_node_map(median_id, graph_id)
-		node_maps_were_modified = True
+	nb_nodes_g = G_ged_env.get_graph_num_nodes(graph_id)
+	if nb_nodes_median <= nb_nodes_g or not sort:
+		G_ged_env.run_method(median_id, graph_id)
+		if G_ged_env.get_upper_bound(median_id, graph_id) < node_map.induced_cost() - epsilon:
+			node_map = G_ged_env.get_node_map(median_id, graph_id)
+			node_maps_were_modified = True			
+	else:
+		G_ged_env.run_method(graph_id, median_id)
+		if G_ged_env.get_upper_bound(graph_id, median_id) < node_map.induced_cost() - epsilon:
+			node_map = G_ged_env.get_node_map(graph_id, median_id)
+			node_map.forward_map, node_map.backward_map = node_map.backward_map, node_map.forward_map
+			node_maps_were_modified = True	
+			
 	return graph_id, node_map, node_maps_were_modified
\ No newline at end of file
diff --git a/gklearn/ged/median/utils.py b/gklearn/ged/median/utils.py
index 5c4c52f..d27c86d 100644
--- a/gklearn/ged/median/utils.py
+++ b/gklearn/ged/median/utils.py
@@ -34,6 +34,8 @@ def mge_options_to_string(options):
 			opt_str += '--parallel ' + ('TRUE' if val else 'FALSE') + ' '
 		elif key == 'update_order':
 			opt_str += '--update-order ' + ('TRUE' if val else 'FALSE') + ' '
+		elif key == 'sort_graphs':
+			opt_str += '--sort-graphs ' + ('TRUE' if val else 'FALSE') + ' '
 		elif key == 'refine':
  			opt_str += '--refine ' + ('TRUE' if val else 'FALSE') + ' '
 		elif key == 'time_limit':
diff --git a/gklearn/ged/util/util.py b/gklearn/ged/util/util.py
index c41ca86..7032345 100644
--- a/gklearn/ged/util/util.py
+++ b/gklearn/ged/util/util.py
@@ -46,7 +46,7 @@ def compute_ged(g1, g2, options):
 	return dis, pi_forward, pi_backward
 
 
-def compute_geds(graphs, options={}, parallel=False, verbose=True):
+def compute_geds(graphs, options={}, sort=True, parallel=False, verbose=True):
 	# initialize ged env.
 	ged_env = gedlibpy.GEDEnv()
 	ged_env.set_edit_cost(options['edit_cost'], edit_cost_constant=options['edit_cost_constants'])
@@ -79,7 +79,7 @@ def compute_geds(graphs, options={}, parallel=False, verbose=True):
 			G_graphs = graphs_toshare
 			G_ged_env = ged_env_toshare
 			G_listID = listID_toshare
-		do_partial = partial(_wrapper_compute_ged_parallel, neo_options)
+		do_partial = partial(_wrapper_compute_ged_parallel, neo_options, sort)
 		pool = Pool(processes=n_jobs, initializer=init_worker, initargs=(graphs, ged_env, listID))
 		if verbose:
 			iterator = tqdm(pool.imap_unordered(do_partial, itr, chunksize),
@@ -108,25 +108,31 @@ def compute_geds(graphs, options={}, parallel=False, verbose=True):
 		for i in iterator:
 #		for i in range(len(graphs)):
 			for j in range(i + 1, len(graphs)):
-				dis, pi_forward, pi_backward = _compute_ged(ged_env, listID[i], listID[j], graphs[i], graphs[j])
+				if nx.number_of_nodes(graphs[i]) <= nx.number_of_nodes(graphs[j]) or not sort:
+					dis, pi_forward, pi_backward = _compute_ged(ged_env, listID[i], listID[j], graphs[i], graphs[j])
+				else:
+					dis, pi_backward, pi_forward = _compute_ged(ged_env, listID[j], listID[i], graphs[j], graphs[i])
 				ged_vec.append(dis)
 				ged_mat[i][j] = dis
 				ged_mat[j][i] = dis
-				n_eo_tmp = get_nb_edit_operations(graphs[i], graphs[j], pi_forward, pi_backward, **neo_options)
+				n_eo_tmp = get_nb_edit_operations(graphs[i], graphs[j], pi_forward, pi_backward, 	**neo_options)
 				n_edit_operations.append(n_eo_tmp)
-					
+
 	return ged_vec, ged_mat, n_edit_operations
 
 
-def _wrapper_compute_ged_parallel(options, itr):
+def _wrapper_compute_ged_parallel(options, sort, itr):
 	i = itr[0]
 	j = itr[1]
-	dis, n_eo_tmp = _compute_ged_parallel(G_ged_env, G_listID[i], G_listID[j], G_graphs[i], G_graphs[j], options)
+	dis, n_eo_tmp = _compute_ged_parallel(G_ged_env, G_listID[i], G_listID[j], G_graphs[i], G_graphs[j], options, sort)
 	return i, j, dis, n_eo_tmp
 
 
-def _compute_ged_parallel(env, gid1, gid2, g1, g2, options):
-	dis, pi_forward, pi_backward = _compute_ged(env, gid1, gid2, g1, g2)
+def _compute_ged_parallel(env, gid1, gid2, g1, g2, options, sort):
+	if nx.number_of_nodes(g1) <= nx.number_of_nodes(g2) or not sort:
+		dis, pi_forward, pi_backward = _compute_ged(env, gid1, gid2, g1, g2)
+	else:
+		dis, pi_backward, pi_forward = _compute_ged(env, gid2, gid1, g2, g1)
 	n_eo_tmp = get_nb_edit_operations(g1, g2, pi_forward, pi_backward, **options) # [0,0,0,0,0,0]
 	return dis, n_eo_tmp
 
diff --git a/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py b/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
index 4dc20d3..bbde39e 100644
--- a/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
+++ b/gklearn/preimage/experiments/tools/analyze_results_of_random_edit_costs.py
@@ -82,5 +82,10 @@ def compute_for_all_experiments(data_dir):
 
 if __name__ == '__main__':
 # 	data_dir = '../results/xp_median_preimage.update_order/'
-	data_dir = '../../results/CRIANN/xp_median_preimage.init10/'
-	compute_for_all_experiments(data_dir)
\ No newline at end of file
+	root_dir_tnz = '../../results/CRIANN/xp_median_preimage.init10/'
+	root_dir_ntnz = '../../results/CRIANN/xp_median_preimage.init10.no_triangle_rule/'
+	root_dir_tz = '../../results/CRIANN/xp_median_preimage.init10.triangle_rule.allow_zeros/'
+	root_dir_ntz = '../../results/CRIANN/xp_median_preimage.init10.no_triangle_rule.allow_zeros/'
+	data_dirs = [root_dir_tnz, root_dir_ntnz, root_dir_tz, root_dir_ntz]
+	for data_dir in data_dirs:
+		compute_for_all_experiments(data_dir)
\ No newline at end of file
diff --git a/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
index 9af5828..9d7809d 100644
--- a/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
+++ b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
@@ -2958,18 +2958,6 @@ if __name__ == "__main__":
 # 	#### xp 7_1: MUTAG, StructuralSP, using CONSTANT.
  	xp_median_preimage_7_1()
 
-# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
- 	xp_median_preimage_8_2()
-
-# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
- 	xp_median_preimage_8_3()
-
-# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
- 	xp_median_preimage_8_4()
-	 
-#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
- 	xp_median_preimage_8_1()
-
 # 	#### xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
  	xp_median_preimage_9_2()
 
@@ -2999,6 +2987,18 @@ if __name__ == "__main__":
 	 
 # 	#### xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
  	xp_median_preimage_6_1()
+	 
+# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+ 	xp_median_preimage_8_2()
+
+# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+ 	xp_median_preimage_8_3()
+
+# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+ 	xp_median_preimage_8_4()
+	 
+#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+ 	xp_median_preimage_8_1()
 # 	
 # 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
  	xp_median_preimage_2_1()
diff --git a/gklearn/preimage/kernel_knn_cv.py b/gklearn/preimage/kernel_knn_cv.py
index 2faf4ba..3e5e88b 100644
--- a/gklearn/preimage/kernel_knn_cv.py
+++ b/gklearn/preimage/kernel_knn_cv.py
@@ -151,7 +151,7 @@ def __kernel_knn_cv_median(dataset_all, ds_name, knn_options, mpg_options, kerne
 	# write result summary for each letter. 
 	if save_results:
 		f_summary = open(dir_save + fn_output_summary, 'a')
-		for i, median_type in enumerate('set-median', 'gen median', 'gen median uo'):
+		for i, median_type in enumerate(['set-median', 'gen median', 'gen median uo']):
 			csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
 				  train_examples + ': ' + median_type, 
 				  knn_options['n_neighbors'],

From dff2c8449e805117c4038f0721edb0a31dcbcfd2 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Tue, 26 May 2020 15:57:28 +0200
Subject: [PATCH 14/17] Update setup.py.

---
 setup.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index 0d7fbba..932002c 100644
--- a/setup.py
+++ b/setup.py
@@ -11,7 +11,7 @@ setuptools.setup(
     version="0.2b1",
     author="Linlin Jia",
     author_email="linlin.jia@insa-rouen.fr",
-    description="A Python library for graph kernels based on linear patterns",
+    description="A Python library for graph kernels, graph edit distances, and graph pre-images",
     long_description=long_description,
     long_description_content_type="text/markdown",
     url="https://github.com/jajupmochi/graphkit-learn",

From 3bbab5506986cafb16770a2357d199a44306e93d Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Wed, 27 May 2020 17:13:03 +0200
Subject: [PATCH 15/17] Package cvxpy cannot be installed from wheel, remove it
 from PyPI setup temporally.

---
 requirements.txt      |  3 ++-
 requirements_pypi.txt | 11 +++++++++++
 setup.py              |  4 ++--
 3 files changed, 15 insertions(+), 3 deletions(-)
 create mode 100644 requirements_pypi.txt

diff --git a/requirements.txt b/requirements.txt
index 47a91a1..24d6efe 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -5,6 +5,7 @@ networkx>=2.2
 scikit-learn>=0.20.0
 tabulate>=0.8.2
 tqdm>=4.26.0
-cvxpy>=1.0.31 # for preimage.
+cvxpy>=1.0.31 # for preimage. Does not work for "pip install graphkit-learn".
+# -e https://files.pythonhosted.org/packages/11/d0/d900870dc2d02ea74961b90c353666c6528a33ea61a10aa59a0d5574ae59/cvxpy-1.0.31.tar.gz # for preimage.
 cvxopt>=1.2.5 # for preimage.
 mosek>=9.2.5; python_version >= '3.6' # for preimage.
diff --git a/requirements_pypi.txt b/requirements_pypi.txt
new file mode 100644
index 0000000..f4854fc
--- /dev/null
+++ b/requirements_pypi.txt
@@ -0,0 +1,11 @@
+numpy>=1.16.2
+scipy>=1.1.0
+matplotlib>=3.0.0
+networkx>=2.2
+scikit-learn>=0.20.0
+tabulate>=0.8.2
+tqdm>=4.26.0
+# cvxpy>=1.0.31 # for preimage. Does not work for "pip install graphkit-learn".
+# -e https://files.pythonhosted.org/packages/11/d0/d900870dc2d02ea74961b90c353666c6528a33ea61a10aa59a0d5574ae59/cvxpy-1.0.31.tar.gz # for preimage.
+cvxopt>=1.2.5 # for preimage.
+mosek>=9.2.5; python_version >= '3.6' # for preimage.
diff --git a/setup.py b/setup.py
index 932002c..a84679e 100644
--- a/setup.py
+++ b/setup.py
@@ -3,12 +3,12 @@ import setuptools
 with open("README.md", "r") as fh:
     long_description = fh.read()
 
-with open('requirements.txt') as fp:
+with open('requirements_pypi.txt') as fp:
     install_requires = fp.read()
 
 setuptools.setup(
     name="graphkit-learn",
-    version="0.2b1",
+    version="0.2b2",
     author="Linlin Jia",
     author_email="linlin.jia@insa-rouen.fr",
     description="A Python library for graph kernels, graph edit distances, and graph pre-images",

From 9adaebb680837c2e2f4cd746886db93a47a650ae Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Wed, 27 May 2020 17:14:33 +0200
Subject: [PATCH 16/17] Add xp: remove the best graph from median set.

---
 .../experiments/xp_1nn_init10_trianglerule.py      | 360 ++++++++++++++++++
 gklearn/preimage/kernel_knn_cv.py                  |   3 +-
 gklearn/preimage/remove_best_graph.py              | 423 +++++++++++++++++++++
 gklearn/preimage/utils.py                          |   4 +-
 4 files changed, 787 insertions(+), 3 deletions(-)
 create mode 100644 gklearn/preimage/remove_best_graph.py

diff --git a/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
index 9d7809d..d23e119 100644
--- a/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
+++ b/gklearn/preimage/experiments/xp_1nn_init10_trianglerule.py
@@ -23,6 +23,344 @@ update_order = False
 test_sizes = [0.9, 0.7] # , 0.5, 0.3, 0.1]
 
 
+def xp_median_preimage_15_1():
+	"""xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'AIDS' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_15_2():
+	"""xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'AIDS' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 1, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_15_3():
+	"""xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'AIDS' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		pkernel = functools.partial(polynomialkernel, d=1, c=1e+2)
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_15_4():
+	"""xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+	"""
+	for test_size in test_sizes:
+		# set parameters.
+		ds_name = 'AIDS' #
+		knn_options = {'n_neighbors': 1,
+					   'n_splits': 30,
+					   'test_size': test_size,
+					   'verbose': True}
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'allow_zeros': allow_zeros,
+					   'triangle_rule': triangle_rule,
+					   'verbose': 1}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 10,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 0}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 1,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/'
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('knn_options:', knn_options)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for train_examples in ['k-graphs', 'expert', 'random', 'best-dataset', 'trainset']:
+			print('\n-------------------------------------')
+			print('train examples used:', train_examples, '\n')
+			mpg_options['fit_method'] = train_examples
+			try:
+				kernel_knn_cv(ds_name, train_examples, knn_options, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
 def xp_median_preimage_14_1():
 	"""xp 14_1: DD, PathUpToH, using CONSTANT.
 	"""
@@ -2915,7 +3253,17 @@ if __name__ == "__main__":
 	#### xp 14_1: DD, PathUpToH, using CONSTANT.
 # 	xp_median_preimage_14_1()
 
+# 	#### xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_1()
+
+# 	#### xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_2()
 
+# 	#### xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_3()
+
+# 	#### xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_4()
 
 
 
@@ -2999,6 +3347,18 @@ if __name__ == "__main__":
 	 
 #     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
  	xp_median_preimage_8_1()
+	 
+	 # 	#### xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_1()
+
+# 	#### xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_2()
+
+# 	#### xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_3()
+
+# 	#### xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_4()
 # 	
 # 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
  	xp_median_preimage_2_1()
diff --git a/gklearn/preimage/kernel_knn_cv.py b/gklearn/preimage/kernel_knn_cv.py
index 3e5e88b..073fa31 100644
--- a/gklearn/preimage/kernel_knn_cv.py
+++ b/gklearn/preimage/kernel_knn_cv.py
@@ -361,7 +361,8 @@ def __get_gram_matrix(load_gm, dir_save, ds_name, kernel_options, dataset_all):
 		gram_matrix_unnorm, time_precompute_gm = __compute_gram_matrix_unnorm(dataset_all, kernel_options)
 		np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm=gram_matrix_unnorm, run_time=time_precompute_gm)
 	else:
-		gmfile = np.load()
+		gm_fname = dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm.npz'
+		gmfile = np.load(gm_fname, allow_pickle=True)
 		gram_matrix_unnorm = gmfile['gram_matrix_unnorm']
 		time_precompute_gm = float(gmfile['run_time'])
 		
diff --git a/gklearn/preimage/remove_best_graph.py b/gklearn/preimage/remove_best_graph.py
new file mode 100644
index 0000000..d6be2a6
--- /dev/null
+++ b/gklearn/preimage/remove_best_graph.py
@@ -0,0 +1,423 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Wen May 27 14:27:15 2020
+
+@author: ljia
+"""
+import numpy as np
+import csv
+import os
+import os.path
+from gklearn.utils import Dataset
+from gklearn.preimage import MedianPreimageGenerator
+from gklearn.utils import normalize_gram_matrix
+from gklearn.utils import split_dataset_by_target
+from gklearn.preimage.utils import compute_k_dis
+from gklearn.utils.graphfiles import saveGXL
+import networkx as nx
+	
+
+def remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=True, save_medians=True, plot_medians=True, load_gm='auto', dir_save='', irrelevant_labels=None, edge_required=False, cut_range=None):
+	"""Remove the best graph from the median set w.r.t. distance in kernel space, and to see if it is possible to generate the removed graph using the graphs left in the median set.
+	"""
+	# 1. get dataset.
+	print('1. getting dataset...')
+	dataset_all = Dataset()
+	dataset_all.load_predefined_dataset(ds_name)
+	dataset_all.trim_dataset(edge_required=edge_required)
+	if irrelevant_labels is not None:
+		dataset_all.remove_labels(**irrelevant_labels)
+	if cut_range is not None:
+		dataset_all.cut_graphs(cut_range)
+	datasets = split_dataset_by_target(dataset_all)
+
+	if save_results:
+		# create result files.
+		print('creating output files...')
+		fn_output_detail, fn_output_summary = __init_output_file(ds_name, kernel_options['name'], mpg_options['fit_method'], dir_save)
+	else:
+		fn_output_detail, fn_output_summary = None, None
+		
+	# 2. compute/load Gram matrix a priori.
+	print('2. computing/loading Gram matrix...')
+	gram_matrix_unnorm_list, time_precompute_gm_list = __get_gram_matrix(load_gm, dir_save, ds_name, kernel_options, datasets)
+		
+	sod_sm_list = []
+	sod_gm_list = []
+	dis_k_sm_list = []
+	dis_k_gm_list = []
+	dis_k_gi_min_list = []
+	time_optimize_ec_list = []
+	time_generate_list = []
+	time_total_list = []
+	itrs_list = []
+	converged_list = []
+	num_updates_ecc_list = []
+	mge_decrease_order_list = []
+	mge_increase_order_list = []
+	mge_converged_order_list = []
+	nb_sod_sm2gm = [0, 0, 0]
+	nb_dis_k_sm2gm = [0, 0, 0]
+	nb_dis_k_gi2sm = [0, 0, 0]
+	nb_dis_k_gi2gm = [0, 0, 0]
+	dis_k_max_list = []	
+	dis_k_min_list = []
+	dis_k_mean_list = []
+	best_dis_list = []
+	print('starting experiment for each class of target...')
+	idx_offset = 0
+	for idx, dataset in enumerate(datasets):
+		target = dataset.targets[0]
+		print('\ntarget =', target, '\n')
+#		if target != 1:
+# 			continue
+		
+		num_graphs = len(dataset.graphs)
+		if num_graphs < 2:
+			print('\nnumber of graphs = ', num_graphs, ', skip.\n')
+			idx_offset += 1
+			continue
+
+		# 3. get the best graph and remove it from median set.
+		print('3. getting and removing the best graph...')
+		gram_matrix_unnorm = gram_matrix_unnorm_list[idx - idx_offset]
+		best_index, best_dis, best_graph = __get_best_graph([g.copy() for g in dataset.graphs], normalize_gram_matrix(gram_matrix_unnorm.copy()))
+		median_set_new = [dataset.graphs[i] for i in range(len(dataset.graphs)) if i != best_index]
+		num_graphs -= 1
+		if num_graphs == 1:
+			continue		
+		best_dis_list.append(best_dis)
+		
+		dataset.load_graphs(median_set_new, targets=None)
+		gram_matrix_unnorm_new = np.delete(gram_matrix_unnorm, best_index, axis=0)
+		gram_matrix_unnorm_new = np.delete(gram_matrix_unnorm_new, best_index, axis=1)
+			
+		# 4. set parameters.
+		print('4. initializing mpg and setting parameters...')	
+		mpg_options['gram_matrix_unnorm'] = gram_matrix_unnorm_new
+		mpg_options['runtime_precompute_gm'] = time_precompute_gm_list[idx - idx_offset]
+		mpg = MedianPreimageGenerator()
+		mpg.dataset = dataset
+		mpg.set_options(**mpg_options.copy())
+		mpg.kernel_options = kernel_options.copy()
+		mpg.ged_options = ged_options.copy()
+		mpg.mge_options = mge_options.copy()
+
+		# 5. compute median preimage.
+		print('5. computing median preimage...')
+		mpg.run()
+		results = mpg.get_results()
+		
+		# 6. compute pairwise kernel distances.
+		print('6. computing pairwise kernel distances...')
+		_, dis_k_max, dis_k_min, dis_k_mean = mpg.graph_kernel.compute_distance_matrix()
+		dis_k_max_list.append(dis_k_max)
+		dis_k_min_list.append(dis_k_min)
+		dis_k_mean_list.append(dis_k_mean)
+		
+		# 7. save results (and median graphs).
+		print('7. saving results (and median graphs)...')
+		# write result detail.
+		if save_results:
+			print('writing results to files...')
+			sod_sm2gm = get_relations(np.sign(results['sod_gen_median'] - results['sod_set_median']))
+			dis_k_sm2gm = get_relations(np.sign(results['k_dis_gen_median'] - results['k_dis_set_median']))
+			dis_k_gi2sm = get_relations(np.sign(results['k_dis_set_median'] - results['k_dis_dataset']))
+			dis_k_gi2gm = get_relations(np.sign(results['k_dis_gen_median'] - results['k_dis_dataset']))
+
+			f_detail = open(dir_save + fn_output_detail, 'a')
+			csv.writer(f_detail).writerow([ds_name, kernel_options['name'], 
+					  ged_options['edit_cost'], ged_options['method'], 
+					  ged_options['attr_distance'], mpg_options['fit_method'], 
+					  num_graphs, target, 1, 
+					  results['sod_set_median'], results['sod_gen_median'],
+					  results['k_dis_set_median'], results['k_dis_gen_median'], 
+					  results['k_dis_dataset'], best_dis, best_index,
+					  sod_sm2gm, dis_k_sm2gm, 
+					  dis_k_gi2sm, dis_k_gi2gm,	results['edit_cost_constants'],
+					  results['runtime_precompute_gm'], results['runtime_optimize_ec'], 
+					  results['runtime_generate_preimage'], results['runtime_total'],
+					  results['itrs'], results['converged'],
+					  results['num_updates_ecc'],
+					  results['mge']['num_decrease_order'] > 0, # @todo: not suitable for multi-start mge
+					  results['mge']['num_increase_order'] > 0,
+					  results['mge']['num_converged_descents'] > 0])
+			f_detail.close()
+		
+			# compute result summary.
+			sod_sm_list.append(results['sod_set_median'])
+			sod_gm_list.append(results['sod_gen_median'])
+			dis_k_sm_list.append(results['k_dis_set_median'])
+			dis_k_gm_list.append(results['k_dis_gen_median'])
+			dis_k_gi_min_list.append(results['k_dis_dataset'])
+			time_precompute_gm_list.append(results['runtime_precompute_gm'])
+			time_optimize_ec_list.append(results['runtime_optimize_ec'])
+			time_generate_list.append(results['runtime_generate_preimage'])
+			time_total_list.append(results['runtime_total'])
+			itrs_list.append(results['itrs'])
+			converged_list.append(results['converged'])
+			num_updates_ecc_list.append(results['num_updates_ecc'])
+			mge_decrease_order_list.append(results['mge']['num_decrease_order'] > 0)
+			mge_increase_order_list.append(results['mge']['num_increase_order'] > 0)
+			mge_converged_order_list.append(results['mge']['num_converged_descents'] > 0)
+			# # SOD SM -> GM
+			if results['sod_set_median'] > results['sod_gen_median']:
+				nb_sod_sm2gm[0] += 1
+	#			repeats_better_sod_sm2gm.append(1)
+			elif results['sod_set_median'] == results['sod_gen_median']:
+				nb_sod_sm2gm[1] += 1
+			elif results['sod_set_median'] < results['sod_gen_median']:
+				nb_sod_sm2gm[2] += 1
+			# # dis_k SM -> GM
+			if results['k_dis_set_median'] > results['k_dis_gen_median']:
+				nb_dis_k_sm2gm[0] += 1
+	#			repeats_better_dis_k_sm2gm.append(1)
+			elif results['k_dis_set_median'] == results['k_dis_gen_median']:
+				nb_dis_k_sm2gm[1] += 1
+			elif results['k_dis_set_median'] < results['k_dis_gen_median']:
+				nb_dis_k_sm2gm[2] += 1
+			# # dis_k gi -> SM
+			if results['k_dis_dataset'] > results['k_dis_set_median']:
+				nb_dis_k_gi2sm[0] += 1
+	#			repeats_better_dis_k_gi2sm.append(1)
+			elif results['k_dis_dataset'] == results['k_dis_set_median']:
+				nb_dis_k_gi2sm[1] += 1
+			elif results['k_dis_dataset'] < results['k_dis_set_median']:
+				nb_dis_k_gi2sm[2] += 1
+			# # dis_k gi -> GM
+			if results['k_dis_dataset'] > results['k_dis_gen_median']:
+				nb_dis_k_gi2gm[0] += 1
+	#			repeats_better_dis_k_gi2gm.append(1)
+			elif results['k_dis_dataset'] == results['k_dis_gen_median']:
+				nb_dis_k_gi2gm[1] += 1
+			elif results['k_dis_dataset'] < results['k_dis_gen_median']:
+				nb_dis_k_gi2gm[2] += 1
+
+			# write result summary for each letter.
+			f_summary = open(dir_save + fn_output_summary, 'a')
+			csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
+					  ged_options['edit_cost'], ged_options['method'], 
+					  ged_options['attr_distance'], mpg_options['fit_method'], 
+					  num_graphs, target,
+					  results['sod_set_median'], results['sod_gen_median'],
+					  results['k_dis_set_median'], results['k_dis_gen_median'], 
+					  results['k_dis_dataset'], best_dis, best_index,
+					  sod_sm2gm, dis_k_sm2gm, 
+					  dis_k_gi2sm, dis_k_gi2gm, 
+					  results['runtime_precompute_gm'], results['runtime_optimize_ec'], 
+					  results['runtime_generate_preimage'], results['runtime_total'],
+					  results['itrs'], results['converged'],
+					  results['num_updates_ecc'], 
+					  results['mge']['num_decrease_order'] > 0, # @todo: not suitable for multi-start mge
+					  results['mge']['num_increase_order'] > 0,
+					  results['mge']['num_converged_descents'] > 0, 
+					  nb_sod_sm2gm, 
+					  nb_dis_k_sm2gm, nb_dis_k_gi2sm, nb_dis_k_gi2gm])
+			f_summary.close()
+			 
+		# save median graphs.
+		if save_medians:
+			if not os.path.exists(dir_save + 'medians/'):
+				os.makedirs(dir_save + 'medians/')
+			print('Saving median graphs to files...')
+			fn_pre_sm = dir_save + 'medians/set_median.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
+			saveGXL(mpg.set_median, fn_pre_sm + '.gxl', method='default', 
+					node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
+					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
+			fn_pre_gm = dir_save + 'medians/gen_median.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
+			saveGXL(mpg.gen_median, fn_pre_gm + '.gxl', method='default',
+		  			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
+					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
+			fn_best_dataset = dir_save + 'medians/g_best_dataset.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
+			saveGXL(best_graph, fn_best_dataset + '.gxl', method='default',
+		  			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
+					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
+			fn_best_median_set = dir_save + 'medians/g_best_median_set.' + mpg_options['fit_method'] + '.nbg' + str(num_graphs) + '.y' + str(target) + '.repeat' + str(1)
+			saveGXL(mpg.best_from_dataset, fn_best_median_set + '.gxl', method='default',
+		  			node_labels=dataset.node_labels, edge_labels=dataset.edge_labels,	
+					node_attrs=dataset.node_attrs, edge_attrs=dataset.edge_attrs)
+		
+		# plot median graphs.
+		if plot_medians and save_medians:
+			if ged_options['edit_cost'] == 'LETTER2' or ged_options['edit_cost'] == 'LETTER' or ds_name == 'Letter-high' or ds_name == 'Letter-med' or ds_name == 'Letter-low':			
+				draw_Letter_graph(mpg.set_median, fn_pre_sm)
+				draw_Letter_graph(mpg.gen_median, fn_pre_gm)
+				draw_Letter_graph(mpg.best_from_dataset, fn_best_dataset)
+
+	# write result summary for each letter. 
+	if save_results:
+		sod_sm_mean = np.mean(sod_sm_list)
+		sod_gm_mean = np.mean(sod_gm_list)
+		dis_k_sm_mean = np.mean(dis_k_sm_list)
+		dis_k_gm_mean = np.mean(dis_k_gm_list)
+		dis_k_gi_min_mean = np.mean(dis_k_gi_min_list)
+		best_dis_mean = np.mean(best_dis_list)
+		time_precompute_gm_mean = np.mean(time_precompute_gm_list)
+		time_optimize_ec_mean = np.mean(time_optimize_ec_list)
+		time_generate_mean = np.mean(time_generate_list)
+		time_total_mean = np.mean(time_total_list)
+		itrs_mean = np.mean(itrs_list)
+		num_converged = np.sum(converged_list)
+		num_updates_ecc_mean = np.mean(num_updates_ecc_list)
+		num_mge_decrease_order = np.sum(mge_decrease_order_list)
+		num_mge_increase_order = np.sum(mge_increase_order_list)
+		num_mge_converged = np.sum(mge_converged_order_list)
+		sod_sm2gm_mean = get_relations(np.sign(sod_gm_mean - sod_sm_mean))
+		dis_k_sm2gm_mean = get_relations(np.sign(dis_k_gm_mean - dis_k_sm_mean))
+		dis_k_gi2sm_mean = get_relations(np.sign(dis_k_sm_mean - dis_k_gi_min_mean))
+		dis_k_gi2gm_mean = get_relations(np.sign(dis_k_gm_mean - dis_k_gi_min_mean))
+		f_summary = open(dir_save + fn_output_summary, 'a')
+		csv.writer(f_summary).writerow([ds_name, kernel_options['name'], 
+				  ged_options['edit_cost'], ged_options['method'], 
+				  ged_options['attr_distance'], mpg_options['fit_method'], 
+				  num_graphs, 'all',
+				  sod_sm_mean, sod_gm_mean, dis_k_sm_mean, dis_k_gm_mean,
+				  dis_k_gi_min_mean, best_dis_mean, '-', 
+				  sod_sm2gm_mean, dis_k_sm2gm_mean, 
+				  dis_k_gi2sm_mean, dis_k_gi2gm_mean,
+				  time_precompute_gm_mean, time_optimize_ec_mean,
+				  time_generate_mean, time_total_mean, itrs_mean, 
+				  num_converged, num_updates_ecc_mean,
+				  num_mge_decrease_order, num_mge_increase_order,
+				  num_mge_converged])
+		f_summary.close()
+		
+	# save total pairwise kernel distances.
+	dis_k_max = np.max(dis_k_max_list)
+	dis_k_min = np.min(dis_k_min_list)
+	dis_k_mean = np.mean(dis_k_mean_list)
+	print('The maximum pairwise distance in kernel space:', dis_k_max)
+	print('The minimum pairwise distance in kernel space:', dis_k_min)
+	print('The average pairwise distance in kernel space:', dis_k_mean)
+
+	print('\ncomplete.\n')	
+
+
+def __get_best_graph(Gn, gram_matrix):
+	k_dis_list = []
+	for idx in range(len(Gn)):
+		k_dis_list.append(compute_k_dis(idx, range(0, len(Gn)), [1 / len(Gn)] * len(Gn), gram_matrix, withterm3=False))
+	best_index = np.argmin(k_dis_list)
+	best_dis = k_dis_list[best_index]
+	best_graph = Gn[best_index].copy()
+	return best_index, best_dis, best_graph 
+
+
+def get_relations(sign):
+	if sign == -1:
+		return 'better'
+	elif sign == 0:
+		return 'same'
+	elif sign == 1:
+		return 'worse'
+
+
+def __get_gram_matrix(load_gm, dir_save, ds_name, kernel_options, datasets):
+	if load_gm == 'auto':
+		gm_fname = dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm.npz'
+		gmfile_exist = os.path.isfile(os.path.abspath(gm_fname))
+		if gmfile_exist:
+			gmfile = np.load(gm_fname, allow_pickle=True) # @todo: may not be safe.
+			gram_matrix_unnorm_list = [item for item in gmfile['gram_matrix_unnorm_list']]
+			time_precompute_gm_list = gmfile['run_time_list'].tolist()
+		else:
+			gram_matrix_unnorm_list = []
+			time_precompute_gm_list = []
+			for dataset in datasets:
+				gram_matrix_unnorm, time_precompute_gm = __compute_gram_matrix_unnorm(dataset, kernel_options)
+				gram_matrix_unnorm_list.append(gram_matrix_unnorm)
+				time_precompute_gm_list.append(time_precompute_gm)
+			np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm_list=gram_matrix_unnorm_list, run_time_list=time_precompute_gm_list)
+	elif not load_gm:
+		gram_matrix_unnorm_list = []
+		time_precompute_gm_list = []
+		for dataset in datasets:
+			gram_matrix_unnorm, time_precompute_gm = __compute_gram_matrix_unnorm(dataset, kernel_options)
+			gram_matrix_unnorm_list.append(gram_matrix_unnorm)
+			time_precompute_gm_list.append(time_precompute_gm)
+		np.savez(dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm', gram_matrix_unnorm_list=gram_matrix_unnorm_list, run_time_list=time_precompute_gm_list)
+	else:
+		gm_fname = dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm.npz'
+		gmfile = np.load(gm_fname, allow_pickle=True) # @todo: may not be safe.
+		gram_matrix_unnorm_list = [item for item in gmfile['gram_matrix_unnorm_list']]
+		time_precompute_gm_list = gmfile['run_time_list'].tolist()
+		
+	return gram_matrix_unnorm_list, time_precompute_gm_list
+
+
+def __get_graph_kernel(dataset, kernel_options):
+	from gklearn.utils.utils import get_graph_kernel_by_name
+	graph_kernel = get_graph_kernel_by_name(kernel_options['name'], 
+				  node_labels=dataset.node_labels,
+				  edge_labels=dataset.edge_labels, 
+				  node_attrs=dataset.node_attrs,
+				  edge_attrs=dataset.edge_attrs,
+				  ds_infos=dataset.get_dataset_infos(keys=['directed']),
+				  kernel_options=kernel_options)
+	return graph_kernel
+		
+		
+def __compute_gram_matrix_unnorm(dataset, kernel_options):
+	from gklearn.utils.utils import get_graph_kernel_by_name
+	graph_kernel = get_graph_kernel_by_name(kernel_options['name'], 
+				  node_labels=dataset.node_labels,
+				  edge_labels=dataset.edge_labels, 
+				  node_attrs=dataset.node_attrs,
+				  edge_attrs=dataset.edge_attrs,
+				  ds_infos=dataset.get_dataset_infos(keys=['directed']),
+				  kernel_options=kernel_options)
+	
+	gram_matrix, run_time = graph_kernel.compute(dataset.graphs, **kernel_options)
+	gram_matrix_unnorm = graph_kernel.gram_matrix_unnorm
+		
+	return gram_matrix_unnorm, run_time
+		
+		
+def __init_output_file(ds_name, gkernel, fit_method, dir_output):
+	if not os.path.exists(dir_output):
+		os.makedirs(dir_output)
+	fn_output_detail = 'results_detail.' + ds_name + '.' + gkernel + '.csv'
+	f_detail = open(dir_output + fn_output_detail, 'a')
+	csv.writer(f_detail).writerow(['dataset', 'graph kernel', 'edit cost', 
+			  'GED method', 'attr distance', 'fit method', 'num graphs', 
+			  'target', 'repeat', 'SOD SM', 'SOD GM', 'dis_k SM', 'dis_k GM',
+			  'min dis_k gi', 'best kernel dis', 'best graph index', 
+			  'SOD SM -> GM', 'dis_k SM -> GM', 'dis_k gi -> SM', 
+			  'dis_k gi -> GM', 'edit cost constants', 'time precompute gm',
+			  'time optimize ec', 'time generate preimage', 'time total',
+			  'itrs', 'converged', 'num updates ecc', 'mge decrease order', 
+			  'mge increase order', 'mge converged'])
+	f_detail.close()
+	
+	fn_output_summary = 'results_summary.' + ds_name + '.' + gkernel + '.csv'
+	f_summary = open(dir_output + fn_output_summary, 'a')
+	csv.writer(f_summary).writerow(['dataset', 'graph kernel', 'edit cost', 
+			  'GED method', 'attr distance', 'fit method', 'num graphs', 
+			  'target', 'SOD SM', 'SOD GM', 'dis_k SM', 'dis_k GM',
+			  'min dis_k gi', 'best kernel dis', 'best graph index',
+			  'SOD SM -> GM', 'dis_k SM -> GM', 'dis_k gi -> SM', 
+			  'dis_k gi -> GM', 'time precompute gm', 'time optimize ec', 
+			  'time generate preimage', 'time total', 'itrs', 'num converged', 
+			  'num updates ecc', 'mge num decrease order', 'mge num increase order', 
+			  'mge num converged', '# SOD SM -> GM', '# dis_k SM -> GM', 
+			  '# dis_k gi -> SM', '# dis_k gi -> GM'])
+	f_summary.close()
+	
+	return fn_output_detail, fn_output_summary
+
+
+#Dessin median courrant
+def draw_Letter_graph(graph, file_prefix):
+	import matplotlib
+	matplotlib.use('agg')
+	import matplotlib.pyplot as plt
+	plt.figure()
+	pos = {}
+	for n in graph.nodes:
+		pos[n] = np.array([float(graph.nodes[n]['x']),float(graph.nodes[n]['y'])])
+	nx.draw_networkx(graph, pos)
+	plt.savefig(file_prefix + '.eps', format='eps', dpi=300)
+#	plt.show()
+	plt.clf()
+	plt.close()
\ No newline at end of file
diff --git a/gklearn/preimage/utils.py b/gklearn/preimage/utils.py
index 55e1e9e..5ca0c1e 100644
--- a/gklearn/preimage/utils.py
+++ b/gklearn/preimage/utils.py
@@ -82,14 +82,14 @@ def generate_median_preimages_by_class(ds_name, mpg_options, kernel_options, ged
 		gram_matrix_unnorm_list = []
 		time_precompute_gm_list = []
 	else:
+		gm_fname = dir_save + 'gram_matrix_unnorm.' + ds_name + '.' + kernel_options['name'] + '.gm.npz'
 		gmfile = np.load(gm_fname, allow_pickle=True) # @todo: may not be safe.
 		gram_matrix_unnorm_list = [item for item in gmfile['gram_matrix_unnorm_list']]
 		time_precompute_gm_list = gmfile['run_time_list'].tolist()
 #	repeats_better_sod_sm2gm = []
 #	repeats_better_dis_k_sm2gm = []
 #	repeats_better_dis_k_gi2sm = []
-#	repeats_better_dis_k_gi2gm = []
-		
+#	repeats_better_dis_k_gi2gm = []		
 		
 	print('starting generating preimage for each class of target...')
 	idx_offset = 0

From 7e3da6b0bc97a800f20c587748d689c5836e32b4 Mon Sep 17 00:00:00 2001
From: jajupmochi <jajupmochi@gmail.com>
Date: Wed, 27 May 2020 17:19:11 +0200
Subject: [PATCH 17/17] Add xp script: remove the best graph from median set.

---
 .../experiments/xp_remove_best_graph_init10.py     | 3085 ++++++++++++++++++++
 1 file changed, 3085 insertions(+)
 create mode 100644 gklearn/preimage/experiments/xp_remove_best_graph_init10.py

diff --git a/gklearn/preimage/experiments/xp_remove_best_graph_init10.py b/gklearn/preimage/experiments/xp_remove_best_graph_init10.py
new file mode 100644
index 0000000..8b722ae
--- /dev/null
+++ b/gklearn/preimage/experiments/xp_remove_best_graph_init10.py
@@ -0,0 +1,3085 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Jan 14 15:39:29 2020
+
+@author: ljia
+"""
+import multiprocessing
+import functools
+import sys
+import os
+import logging
+from gklearn.utils.kernels import deltakernel, gaussiankernel, kernelproduct
+from gklearn.preimage.remove_best_graph import remove_best_graph
+from gklearn.utils import compute_gram_matrices_by_class
+
+
+dir_root = '../results/xp_remove_best_graph.init10/'
+num_random = 10
+initial_solutions = 10
+
+
+def xp_median_preimage_15_1():
+	"""xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'AIDS' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+
+def xp_median_preimage_15_2():
+	"""xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'AIDS' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 1, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_15_3():
+	"""xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'AIDS' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		pkernel = functools.partial(polynomialkernel, d=1, c=1e+2)
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_15_4():
+	"""xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'AIDS' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 10,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['chem', 'charge', 'x', 'y']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_14_1():
+	"""xp 14_1: DD, PathUpToH, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'DD' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 2, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+	
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+	# 	# compute gram matrices for each class a priori.
+	# 	print('Compute gram matrices for each class a priori.')
+	# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=save_results, dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_13_1():
+	"""xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_13_2():
+	"""xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '') # 
+		irrelevant_labels = None # 
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random: # 
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_12_1():
+	"""xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 0, 1, 1, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_2():
+	"""xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 1, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+
+def xp_median_preimage_12_3():
+	"""xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+	"""
+	from gklearn.utils.kernels import gaussiankernel
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		pkernel = functools.partial(gaussiankernel, gamma=None) # @todo
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_4():
+	"""xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 14,
+						  'base_kernel': 'subtree',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+	# 	# compute gram matrices for each class a priori.
+	# 	print('Compute gram matrices for each class a priori.')
+	# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+		
+		
+def xp_median_preimage_12_5():
+	"""xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'PAH' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 0, 1, 1, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.unlabeled/' + ('update_order/' if update_order else '') # 
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random: # 
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_9_1():
+	"""xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MAO' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_2():
+	"""xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MAO' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 9, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_3():
+	"""xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'MAO' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		pkernel = functools.partial(polynomialkernel, d=4, c=1e+7)
+		kernel_options = {'name': 'Treelet', #
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_9_4():
+	"""xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MAO' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 6,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.symb/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'node_attrs': ['x', 'y', 'z'], 'edge_labels': ['bond_stereo']} # 
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_8_1():
+	"""xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_2():
+	"""xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 7, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_3():
+	"""xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		pkernel = functools.partial(polynomialkernel, d=2, c=1e+5)
+		kernel_options = {'name': 'Treelet',
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_8_4():
+	"""xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Monoterpenoides' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 4,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_7_1():
+	"""xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_2():
+	"""xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'PathUpToH',
+						  'depth': 2, #
+						  'k_func': 'MinMax', #
+						  'compute_method': 'trie',
+	 					  'parallel': 'imap_unordered', 
+	                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required, cut_range=None)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_7_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_3():
+	"""xp 7_3: MUTAG, Treelet, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		from gklearn.utils.kernels import polynomialkernel
+		# set parameters.
+		ds_name = 'MUTAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		pkernel = functools.partial(polynomialkernel, d=3, c=1e+8)
+		kernel_options = {'name': 'Treelet',
+					      'sub_kernel': pkernel,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_7_4():
+	"""xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'MUTAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [4, 4, 2, 1, 1, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100, # 
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		kernel_options = {'name': 'WeisfeilerLehman',
+					      'height': 1,
+						  'base_kernel': 'subtree',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'CONSTANT', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_6_1():
+	"""xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'COIL-RAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_6_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_6_2():
+	"""xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'COIL-RAG' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 1], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC', # 
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_6_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+
+def xp_median_preimage_5_1():
+	"""xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'FRANKENSTEIN' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_4_1():
+	"""xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'COLORS-3' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3, 0], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'NON_SYMBOLIC',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = False #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_3_2():
+	"""xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Fingerprint' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.01, 0.125, 0.125], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'edge_attrs': ['orient', 'angle']} #
+		edge_required = True #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_3_1():
+	"""xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Fingerprint' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.01, 0.125, 0.125], #
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'edge_attrs': ['orient', 'angle']} #
+		edge_required = False #
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+def xp_median_preimage_2_1():
+	"""xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'COIL-DEL' #
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [3, 3, 1, 3, 3],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+ 					  'parallel': 'imap_unordered', 
+                      # 'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '.node_attrs/' + ('update_order/' if update_order else '')
+		irrelevant_labels = {'edge_labels': ['valence']}
+
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+		
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+# 	# compute gram matrices for each class a priori.
+# 	print('Compute gram matrices for each class a priori.')
+# 	compute_gram_matrices_by_class(ds_name, kernel_options, save_results=True, dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels)
+			except Exception as exp:
+				print('An exception occured when running this experiment:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_1_1():
+	"""xp 1_1: Letter-high, StructuralSP.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-high'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_1_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+			
+def xp_median_preimage_1_2():
+	"""xp 1_2: Letter-high, ShortestPath.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-high'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.675, 0.675, 0.75, 0.425, 0.425],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_1_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+
+
+def xp_median_preimage_10_1():
+	"""xp 10_1: Letter-med, StructuralSP.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-med'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.75, 0.475, 0.475],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_10_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_10_2():
+	"""xp 10_2: Letter-med, ShortestPath.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-med'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.525, 0.525, 0.75, 0.475, 0.475],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_10_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+
+def xp_median_preimage_11_1():
+	"""xp 11_1: Letter-low, StructuralSP.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-low'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.075, 0.075, 0.25, 0.075, 0.075],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'StructuralSP',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'edge_kernels': sub_kernels, 
+						  'compute_method': 'naive',
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_11_1:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+				
+			
+def xp_median_preimage_11_2():
+	"""xp 11_2: Letter-low, ShortestPath.
+	"""
+	for update_order in [True, False]:
+		# set parameters.
+		ds_name = 'Letter-low'
+		mpg_options = {'fit_method': 'k-graphs',
+					   'init_ecc': [0.075, 0.075, 0.25, 0.075, 0.075],
+					   'ds_name': ds_name,
+					   'parallel': True, # False
+					   'time_limit_in_sec': 0,
+					   'max_itrs': 100,
+					   'max_itrs_without_update': 3,
+					   'epsilon_residual': 0.01,
+					   'epsilon_ec': 0.1,
+					   'verbose': 2}
+		mixkernel = functools.partial(kernelproduct, deltakernel, gaussiankernel)
+		sub_kernels = {'symb': deltakernel, 'nsymb': gaussiankernel, 'mix': mixkernel}
+		kernel_options = {'name': 'ShortestPath',
+						  'edge_weight': None,
+						  'node_kernels': sub_kernels,
+						  'parallel': 'imap_unordered', 
+# 						  'parallel': None, 
+						  'n_jobs': multiprocessing.cpu_count(),
+						  'normalize': True,
+						  'verbose': 2}
+		ged_options = {'method': 'IPFP',
+					   'initialization_method': 'RANDOM', # 'NODE'
+					   'initial_solutions': initial_solutions, # 1
+					   'edit_cost': 'LETTER2',
+					   'attr_distance': 'euclidean',
+					   'ratio_runs_from_initial_solutions': 1,
+					   'threads': multiprocessing.cpu_count(),
+					   'init_option': 'EAGER_WITHOUT_SHUFFLED_COPIES'}
+		mge_options = {'init_type': 'MEDOID',
+					   'random_inits': 10,
+					   'time_limit': 0,
+					   'verbose': 2,
+					   'update_order': update_order,
+					   'randomness': 'REAL',
+					   'refine': False}
+		save_results = True
+		dir_save = dir_root + ds_name + '.' + kernel_options['name'] + '/' + ('update_order/' if update_order else '')
+		irrelevant_labels = None #
+		edge_required = True #
+		
+		if not os.path.exists(dir_save):
+			os.makedirs(dir_save)
+		file_output = open(dir_save + 'output.txt', 'a')
+		sys.stdout = file_output
+
+		# print settings.
+		print('parameters:')
+		print('dataset name:', ds_name)
+		print('mpg_options:', mpg_options)
+		print('kernel_options:', kernel_options)
+		print('ged_options:', ged_options)
+		print('mge_options:', mge_options)
+		print('save_results:', save_results)
+		print('irrelevant_labels:', irrelevant_labels)
+		print()
+		
+		# generate preimages.
+		for fit_method in ['k-graphs', 'expert'] + ['random'] * num_random:
+			print('\n-------------------------------------')
+			print('fit method:', fit_method, '\n')
+			mpg_options['fit_method'] = fit_method
+			try:
+				remove_best_graph(ds_name, mpg_options, kernel_options, ged_options, mge_options, save_results=save_results, save_medians=True, plot_medians=True, load_gm='auto', dir_save=dir_save, irrelevant_labels=irrelevant_labels, edge_required=edge_required)
+			except Exception as exp:
+				print('An exception occured when running experiment on xp_median_preimage_11_2:')
+				LOG_FILENAME = dir_save + 'error.txt'
+				logging.basicConfig(filename=LOG_FILENAME, level=logging.DEBUG)
+				logging.exception('')
+				print(repr(exp))
+			
+
+if __name__ == "__main__":
+	
+# 	#### xp 1_1: Letter-high, StructuralSP.
+ 	# xp_median_preimage_1_1()
+
+# 	#### xp 1_2: Letter-high, ShortestPath.
+ 	# xp_median_preimage_1_2()
+
+# 	#### xp 10_1: Letter-med, StructuralSP.
+ 	# xp_median_preimage_10_1()
+
+# 	#### xp 10_2: Letter-med, ShortestPath.
+ 	# xp_median_preimage_10_2()
+
+# 	#### xp 11_1: Letter-low, StructuralSP.
+ 	# xp_median_preimage_11_1()
+
+# 	#### xp 11_2: Letter-low, ShortestPath.
+ 	# xp_median_preimage_11_2()
+# 	
+# 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+# # 	xp_median_preimage_2_1()
+# 	
+# 	#### xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+#  	# xp_median_preimage_3_1()
+
+# 	#### xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+ 	# xp_median_preimage_3_2()
+
+# 	#### xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_4_1()
+# 	
+# 	#### xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_5_1()
+# 	
+# 	#### xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_1()
+
+# 	#### xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_2()
+
+# 	#### xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+ 	# xp_median_preimage_7_1()
+
+# 	#### xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+ 	# xp_median_preimage_7_2()
+
+# 	#### xp 7_3: MUTAG, Treelet, using CONSTANT.
+ 	# xp_median_preimage_7_3()
+
+# 	#### xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+ 	# xp_median_preimage_7_4()
+# 	 
+#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+ 	# xp_median_preimage_8_1()
+
+# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+ 	# xp_median_preimage_8_2()
+
+# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+ 	# xp_median_preimage_8_3()
+
+# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+ 	# xp_median_preimage_8_4()
+
+# 	#### xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_1()
+
+# 	#### xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_2()
+
+# 	#### xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_3()
+
+# 	#### xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_9_4()
+
+	#### xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+ 	# xp_median_preimage_12_1()
+
+	#### xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_2()
+
+	#### xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_3()
+
+	#### xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+ 	# xp_median_preimage_12_4()
+
+	#### xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+ 	# xp_median_preimage_12_5()
+
+	#### xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+ 	# xp_median_preimage_13_1()
+
+	#### xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+# 	xp_median_preimage_13_2()
+
+	#### xp 14_1: DD, PathUpToH, using CONSTANT.
+# 	xp_median_preimage_14_1()
+
+# 	#### xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_1()
+
+# 	#### xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_2()
+
+# 	#### xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_3()
+
+# 	#### xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	# xp_median_preimage_15_4()
+
+
+
+
+
+
+
+# 	#### xp 7_2: MUTAG, PathUpToH, using CONSTANT.
+ 	xp_median_preimage_7_2()
+
+# 	#### xp 7_3: MUTAG, Treelet, using CONSTANT.
+ 	xp_median_preimage_7_3()
+
+# 	#### xp 7_4: MUTAG, WeisfeilerLehman, using CONSTANT.
+ 	xp_median_preimage_7_4()
+# 	 
+# 	#### xp 7_1: MUTAG, StructuralSP, using CONSTANT.
+ 	xp_median_preimage_7_1()
+
+# 	#### xp 8_2: Monoterpenoides, PathUpToH, using CONSTANT.
+ 	xp_median_preimage_8_2()
+
+# 	#### xp 8_3: Monoterpenoides, Treelet, using CONSTANT.
+ 	xp_median_preimage_8_3()
+
+# 	#### xp 8_4: Monoterpenoides, WeisfeilerLehman, using CONSTANT.
+ 	xp_median_preimage_8_4()
+	 
+#     #### xp 8_1: Monoterpenoides, StructuralSP, using CONSTANT.
+ 	xp_median_preimage_8_1()
+
+# 	#### xp 9_2: MAO, PathUpToH, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_2()
+
+# 	#### xp 9_3: MAO, Treelet, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_3()
+
+# 	#### xp 9_4: MAO, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_4()
+	 
+# 	#### xp 9_1: MAO, StructuralSP, using CONSTANT, symbolic only.
+ 	xp_median_preimage_9_1()
+
+ 	#### xp 12_1: PAH, StructuralSP, using NON_SYMBOLIC, unlabeled.
+ 	xp_median_preimage_12_1()
+
+ 	#### xp 12_2: PAH, PathUpToH, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_2()
+
+ 	#### xp 12_3: PAH, Treelet, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_3()
+
+ 	#### xp 12_4: PAH, WeisfeilerLehman, using CONSTANT, unlabeled.
+ 	xp_median_preimage_12_4()
+
+ 	#### xp 12_5: PAH, ShortestPath, using NON_SYMBOLIC, unlabeled.
+ 	xp_median_preimage_12_5()
+	 
+# 	#### xp 1_1: Letter-high, StructuralSP.
+ 	xp_median_preimage_1_1()
+
+# 	#### xp 1_2: Letter-high, ShortestPath.
+ 	xp_median_preimage_1_2()
+
+# 	#### xp 10_1: Letter-med, StructuralSP.
+ 	xp_median_preimage_10_1()
+
+# 	#### xp 10_2: Letter-med, ShortestPath.
+ 	xp_median_preimage_10_2()
+
+# 	#### xp 11_1: Letter-low, StructuralSP.
+ 	xp_median_preimage_11_1()
+
+# 	#### xp 11_2: Letter-low, ShortestPath.
+ 	xp_median_preimage_11_2()
+
+ 	#### xp 13_1: PAH, StructuralSP, using NON_SYMBOLIC.
+ 	xp_median_preimage_13_1()
+
+ 	#### xp 13_2: PAH, ShortestPath, using NON_SYMBOLIC.
+ 	xp_median_preimage_13_2()
+	 
+# 	#### xp 6_1: COIL-RAG, StructuralSP, using NON_SYMBOLIC.
+ 	xp_median_preimage_6_1()
+	 
+	 # 	#### xp 15_1: AIDS, StructuralSP, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_1()
+
+# 	#### xp 15_2: AIDS, PathUpToH, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_2()
+
+# 	#### xp 15_3: AIDS, Treelet, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_3()
+
+# 	#### xp 15_4: AIDS, WeisfeilerLehman, using CONSTANT, symbolic only.
+ 	xp_median_preimage_15_4()
+# 	
+# 	#### xp 2_1: COIL-DEL, StructuralSP, using LETTER2, only node attrs.
+ 	xp_median_preimage_2_1()
+# 	
+# 	#### xp 3_1: Fingerprint, StructuralSP, using LETTER2, only node attrs.
+#  	# xp_median_preimage_3_1()
+
+# 	#### xp 3_2: Fingerprint, ShortestPath, using LETTER2, only node attrs.
+ 	# xp_median_preimage_3_2()
+
+# 	#### xp 4_1: COLORS-3, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_4_1()
+# 	
+# 	#### xp 5_1: FRANKENSTEIN, StructuralSP, using NON_SYMBOLIC.
+# # 	xp_median_preimage_5_1()
+# 	
+# 	#### xp 6_2: COIL-RAG, ShortestPath, using NON_SYMBOLIC.
+ 	# xp_median_preimage_6_2()
+
+	#### xp 14_1: DD, PathUpToH, using CONSTANT.
+# 	xp_median_preimage_14_1()