Added RTP Read/Write and Opus Encode/Decode streams

src/Discord.Net/API/DiscordAPIClient.cs

@@ -91,7 +91,7 @@ namespace Discord.API
 
             _serializer = serializer ?? new JsonSerializer { ContractResolver = new DiscordContractResolver() };
         }
-        void Dispose(bool disposing)
+        private void Dispose(bool disposing)
         {
             if (!_isDisposed)
             {

src/Discord.Net/API/DiscordVoiceAPIClient.cs

@@ -83,7 +83,7 @@ namespace Discord.Audio
 
             _serializer = serializer ?? new JsonSerializer { ContractResolver = new DiscordContractResolver() };
         }
-        void Dispose(bool disposing)
+        private void Dispose(bool disposing)
         {
             if (!_isDisposed)
             {

src/Discord.Net/Audio/AudioClient.cs

@@ -14,6 +14,8 @@ namespace Discord.Audio
 {
     internal class AudioClient : IAudioClient, IDisposable
     {
+        public const int SampleRate = 48000;
+
         public event Func<Task> Connected
         {
             add { _connectedEvent.Add(value); }
@@ -57,7 +59,7 @@ namespace Discord.Audio
         private DiscordSocketClient Discord => Guild.Discord;
 
         /// <summary> Creates a new REST/WebSocket discord client. </summary>
-        internal AudioClient(CachedGuild guild, int id)
+        public AudioClient(CachedGuild guild, int id)
         {
             Guild = guild;
 
@@ -171,6 +173,22 @@ namespace Discord.Audio
             await _disconnectedEvent.InvokeAsync(ex).ConfigureAwait(false);
         }
 
+        public void Send(byte[] data, int count)
+        {
+            //TODO: Queue these?
+            ApiClient.SendAsync(data, count).ConfigureAwait(false);
+        }
+
+        public RTPWriteStream CreateOpusStream(int samplesPerFrame, int bufferSize = 4000)
+        {
+            return new RTPWriteStream(this, _secretKey, samplesPerFrame, _ssrc, bufferSize = 4000);
+        }
+        public OpusEncodeStream CreatePCMStream(int samplesPerFrame, int? bitrate = null, int channels = 2,
+            OpusApplication application = OpusApplication.MusicOrMixed, int bufferSize = 4000)
+        {
+            return new OpusEncodeStream(this, _secretKey, samplesPerFrame, _ssrc, SampleRate, bitrate, channels, application, bufferSize);
+        }
+
         private async Task ProcessMessageAsync(VoiceOpCode opCode, object payload)
         {
 #if BENCHMARK
@@ -253,7 +271,6 @@ namespace Discord.Audio
             }
 #endif
         }
-
         private async Task ProcessPacketAsync(byte[] packet)
         {
             if (!_connectTask.Task.IsCompleted)

src/Discord.Net/Audio/IAudioClient.cs

@@ -16,5 +16,9 @@ namespace Discord.Audio
         int Latency { get; }
 
         Task DisconnectAsync();
+
+        RTPWriteStream CreateOpusStream(int samplesPerFrame, int bufferSize = 4000);
+        OpusEncodeStream CreatePCMStream(int samplesPerFrame, int? bitrate = null, int channels = 2,
+            OpusApplication application = OpusApplication.MusicOrMixed, int bufferSize = 4000);
     }
 }

src/Discord.Net/Audio/Opus/OpusApplication.cs

@@ -1,6 +1,6 @@
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
-    internal enum OpusApplication : int
+    public enum OpusApplication : int
     {
         Voice = 2048,
         MusicOrMixed = 2049,

src/Discord.Net/Audio/Opus/OpusConverter.cs

@@ -1,6 +1,6 @@
 using System;
 
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
     internal abstract class OpusConverter : IDisposable
     {
@@ -8,17 +8,27 @@ namespace Discord.Audio.Opus
 
         /// <summary> Gets the bit rate of this converter. </summary>
         public const int BitsPerSample = 16;
+        /// <summary> Gets the bytes per sample. </summary>
+        public const int SampleSize = (BitsPerSample / 8) * MaxChannels;
+        /// <summary> Gets the maximum amount of channels this encoder supports. </summary>
+        public const int MaxChannels = 2;
+
         /// <summary> Gets the input sampling rate of this converter. </summary>
         public int SamplingRate { get; }
+        /// <summary> Gets the number of samples per second for this stream. </summary>
+        public int Channels { get; }
 
-        protected OpusConverter(int samplingRate)
+        protected OpusConverter(int samplingRate, int channels)
         {
             if (samplingRate != 8000 && samplingRate != 12000 &&
                 samplingRate != 16000 && samplingRate != 24000 &&
                 samplingRate != 48000)
                 throw new ArgumentOutOfRangeException(nameof(samplingRate));
+            if (channels != 1 && channels != 2)
+                throw new ArgumentOutOfRangeException(nameof(channels));
 
             SamplingRate = samplingRate;
+            Channels = channels;
         }
         
         private bool disposedValue = false; // To detect redundant calls

src/Discord.Net/Audio/Opus/Ctl.cs → src/Discord.Net/Audio/Opus/OpusCtl.cs

@@ -1,6 +1,6 @@
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
-    internal enum Ctl : int
+    internal enum OpusCtl : int
     {
         SetBitrateRequest = 4002,
         GetBitrateRequest = 4003,

src/Discord.Net/Audio/Opus/OpusDecoder.cs

@@ -1,7 +1,7 @@
 using System;
 using System.Runtime.InteropServices;
 
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
     internal unsafe class OpusDecoder : OpusConverter
     {
@@ -10,13 +10,13 @@ namespace Discord.Audio.Opus
         [DllImport("opus", EntryPoint = "opus_decoder_destroy", CallingConvention = CallingConvention.Cdecl)]
         private static extern void DestroyDecoder(IntPtr decoder);
         [DllImport("opus", EntryPoint = "opus_decode", CallingConvention = CallingConvention.Cdecl)]
-        private static extern int Decode(IntPtr st, byte* data, int len, byte[] pcm, int frame_size, int decode_fec);
+        private static extern int Decode(IntPtr st, byte* data, int len, byte* pcm, int max_frame_size, int decode_fec);
 
-        public OpusDecoder(int samplingRate)
-            : base(samplingRate)
+        public OpusDecoder(int samplingRate, int channels)
+            : base(samplingRate, channels)
         {
             OpusError error;
-            _ptr = CreateDecoder(samplingRate, 2, out error);
+            _ptr = CreateDecoder(samplingRate, channels, out error);
             if (error != OpusError.OK)
                 throw new InvalidOperationException($"Error occured while creating decoder: {error}");
         }
@@ -25,11 +25,12 @@ namespace Discord.Audio.Opus
         /// <param name="input">PCM samples to decode.</param>
         /// <param name="inputOffset">Offset of the frame in input.</param>
         /// <param name="output">Buffer to store the decoded frame.</param>
-        public unsafe int DecodeFrame(byte[] input, int inputOffset, int inputCount, byte[] output)
+        public unsafe int DecodeFrame(byte[] input, int inputOffset, int inputCount, byte[] output, int outputOffset)
         {
             int result = 0;
             fixed (byte* inPtr = input)
-                result = Decode(_ptr, inPtr + inputOffset, inputCount, output, inputCount, 0);
+            fixed (byte* outPtr = output)
+                result = Decode(_ptr, inPtr + inputOffset, inputCount, outPtr + outputOffset, (output.Length - outputOffset) / SampleSize / MaxChannels, 0);
 
             if (result < 0)
                 throw new Exception(((OpusError)result).ToString());

src/Discord.Net/Audio/Opus/OpusEncoder.cs

@@ -1,7 +1,7 @@
 using System;
 using System.Runtime.InteropServices;
 
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
     internal unsafe class OpusEncoder : OpusConverter
     {
@@ -10,49 +10,22 @@ namespace Discord.Audio.Opus
         [DllImport("opus", EntryPoint = "opus_encoder_destroy", CallingConvention = CallingConvention.Cdecl)]
         private static extern void DestroyEncoder(IntPtr encoder);
         [DllImport("opus", EntryPoint = "opus_encode", CallingConvention = CallingConvention.Cdecl)]
-        private static extern int Encode(IntPtr st, byte* pcm, int frame_size, byte[] data, int max_data_bytes);
+        private static extern int Encode(IntPtr st, byte* pcm, int frame_size, byte* data, int max_data_bytes);
         [DllImport("opus", EntryPoint = "opus_encoder_ctl", CallingConvention = CallingConvention.Cdecl)]
-        private static extern int EncoderCtl(IntPtr st, Ctl request, int value);
-
-        /// <summary> Gets the bit rate in kbit/s. </summary>
-        public int? BitRate { get; }
+        private static extern int EncoderCtl(IntPtr st, OpusCtl request, int value);
+        
         /// <summary> Gets the coding mode of the encoder. </summary>
         public OpusApplication Application { get; }
-        /// <summary> Gets the number of channels of this converter. </summary>
-        public int InputChannels { get; }
-        /// <summary> Gets the milliseconds per frame. </summary>
-        public int FrameMilliseconds { get; }
-
-        /// <summary> Gets the bytes per sample. </summary>
-        public int SampleSize => (BitsPerSample / 8) * InputChannels;
-        /// <summary> Gets the number of samples per frame. </summary>
-        public int SamplesPerFrame => SamplingRate / 1000 * FrameMilliseconds;
-        /// <summary> Gets the bytes per frame. </summary>
-        public int FrameSize => SamplesPerFrame * SampleSize;
 
-        public OpusEncoder(int samplingRate, int channels, int frameMillis, 
-            int? bitrate = null, OpusApplication application = OpusApplication.MusicOrMixed)
-            : base(samplingRate)
+        public OpusEncoder(int samplingRate, int channels, OpusApplication application = OpusApplication.MusicOrMixed)
+            : base(samplingRate, channels)
         {
-            if (channels != 1 && channels != 2)
-                throw new ArgumentOutOfRangeException(nameof(channels));
-            if (bitrate != null && (bitrate < 1 || bitrate > DiscordVoiceAPIClient.MaxBitrate))
-                throw new ArgumentOutOfRangeException(nameof(bitrate));
+            Application = application;
 
             OpusError error;
             _ptr = CreateEncoder(samplingRate, channels, (int)application, out error);
             if (error != OpusError.OK)
                 throw new InvalidOperationException($"Error occured while creating encoder: {error}");
-
-
-            BitRate = bitrate;
-            Application = application;
-            InputChannels = channels;
-            FrameMilliseconds = frameMillis;
-
-            SetForwardErrorCorrection(true);
-            if (bitrate != null)
-                SetBitrate(bitrate.Value);
         }
 
 
@@ -61,11 +34,12 @@ namespace Discord.Audio.Opus
         /// <param name="inputOffset">Offset of the frame in pcmSamples.</param>
         /// <param name="output">Buffer to store the encoded frame.</param>
         /// <returns>Length of the frame contained in outputBuffer.</returns>
-        public unsafe int EncodeFrame(byte[] input, int inputOffset, byte[] output)
+        public unsafe int EncodeFrame(byte[] input, int inputOffset, int inputCount, byte[] output, int outputOffset)
         {
             int result = 0;
             fixed (byte* inPtr = input)
-                result = Encode(_ptr, inPtr + inputOffset, SamplesPerFrame, output, output.Length);
+            fixed (byte* outPtr = output)
+                result = Encode(_ptr, inPtr + inputOffset, inputCount / SampleSize, outPtr + outputOffset, output.Length - outputOffset);
 
             if (result < 0)
                 throw new Exception(((OpusError)result).ToString());
@@ -75,7 +49,7 @@ namespace Discord.Audio.Opus
         /// <summary> Gets or sets whether Forward Error Correction is enabled. </summary>
         public void SetForwardErrorCorrection(bool value)
         {
-            var result = EncoderCtl(_ptr, Ctl.SetInbandFECRequest, value ? 1 : 0);
+            var result = EncoderCtl(_ptr, OpusCtl.SetInbandFECRequest, value ? 1 : 0);
             if (result < 0)
                 throw new Exception(((OpusError)result).ToString());
         }
@@ -83,7 +57,10 @@ namespace Discord.Audio.Opus
         /// <summary> Gets or sets whether Forward Error Correction is enabled. </summary>
         public void SetBitrate(int value)
         {
-            var result = EncoderCtl(_ptr, Ctl.SetBitrateRequest, value * 1000);
+            if (value < 1 || value > DiscordVoiceAPIClient.MaxBitrate)
+                throw new ArgumentOutOfRangeException(nameof(value));
+
+            var result = EncoderCtl(_ptr, OpusCtl.SetBitrateRequest, value * 1000);
             if (result < 0)
                 throw new Exception(((OpusError)result).ToString());
         }

src/Discord.Net/Audio/Opus/OpusError.cs

@@ -1,4 +1,4 @@
-namespace Discord.Audio.Opus
+namespace Discord.Audio
 {
     internal enum OpusError : int
     {

src/Discord.Net/Audio/Sodium/SecretBox.cs

@@ -1,23 +1,25 @@
 using System.Runtime.InteropServices;
 
-namespace Discord.Net.Audio.Sodium
+namespace Discord.Audio
 {
     public unsafe static class SecretBox
     {
         [DllImport("libsodium", EntryPoint = "crypto_secretbox_easy", CallingConvention = CallingConvention.Cdecl)]
-        private static extern int SecretBoxEasy(byte* output, byte[] input, long inputLength, byte[] nonce, byte[] secret);
+        private static extern int SecretBoxEasy(byte* output, byte* input, long inputLength, byte[] nonce, byte[] secret);
         [DllImport("libsodium", EntryPoint = "crypto_secretbox_open_easy", CallingConvention = CallingConvention.Cdecl)]
-        private static extern int SecretBoxOpenEasy(byte[] output, byte* input, long inputLength, byte[] nonce, byte[] secret);
+        private static extern int SecretBoxOpenEasy(byte* output, byte* input, long inputLength, byte[] nonce, byte[] secret);
 
-        public static int Encrypt(byte[] input, long inputLength, byte[] output, int outputOffset, byte[] nonce, byte[] secret)
+        public static int Encrypt(byte[] input, int inputOffset, long inputLength, byte[] output, int outputOffset, byte[] nonce, byte[] secret)
         {
+            fixed (byte* inPtr = input)
             fixed (byte* outPtr = output)
-                return SecretBoxEasy(outPtr + outputOffset, input, inputLength, nonce, secret);
+                return SecretBoxEasy(outPtr + outputOffset, inPtr + inputOffset, inputLength, nonce, secret);
         }
-        public static int Decrypt(byte[] input, int inputOffset, long inputLength, byte[] output, byte[] nonce, byte[] secret)
+        public static int Decrypt(byte[] input, int inputOffset, long inputLength, byte[] output, int outputOffset, byte[] nonce, byte[] secret)
         {
             fixed (byte* inPtr = input)
-                return SecretBoxOpenEasy(output, inPtr + inputLength, inputLength, nonce, secret);
+            fixed (byte* outPtr = output)
+                return SecretBoxOpenEasy(outPtr + outputOffset, inPtr + inputOffset, inputLength, nonce, secret);
         }
     }
 }

src/Discord.Net/Audio/Streams/OpusDecodeStream.cs

@@ -0,0 +1,30 @@
+namespace Discord.Audio
+{
+    public class OpusDecodeStream : RTPReadStream
+    {
+        private readonly byte[] _buffer;
+        private readonly OpusDecoder _decoder;
+
+        internal OpusDecodeStream(AudioClient audioClient, byte[] secretKey, int samplingRate, 
+            int channels = OpusConverter.MaxChannels, int bufferSize = 4000)
+            : base(audioClient, secretKey)
+        {
+            _buffer = new byte[bufferSize];
+            _decoder = new OpusDecoder(samplingRate, channels);
+        }
+
+        public override int Read(byte[] buffer, int offset, int count)
+        {
+            count = _decoder.DecodeFrame(buffer, offset, count, _buffer, 0);
+            return base.Read(_buffer, 0, count);
+        }
+
+        protected override void Dispose(bool disposing)
+        {
+            base.Dispose(disposing);
+
+            if (disposing)
+                _decoder.Dispose();
+        }
+    }
+}

src/Discord.Net/Audio/Streams/OpusEncodeStream.cs

@@ -0,0 +1,34 @@
+namespace Discord.Audio
+{
+    public class OpusEncodeStream : RTPWriteStream
+    {
+        private readonly byte[] _buffer;
+        private readonly OpusEncoder _encoder;
+
+        internal OpusEncodeStream(AudioClient audioClient, byte[] secretKey, int samplesPerFrame, uint ssrc, int samplingRate, int? bitrate = null, 
+            int channels = OpusConverter.MaxChannels,  OpusApplication application = OpusApplication.MusicOrMixed, int bufferSize = 4000)
+            : base(audioClient, secretKey, samplesPerFrame, ssrc)
+        {
+            _buffer = new byte[bufferSize];
+            _encoder = new OpusEncoder(samplingRate, channels);
+
+            _encoder.SetForwardErrorCorrection(true);
+            if (bitrate != null)
+                _encoder.SetBitrate(bitrate.Value);
+        }
+
+        public override void Write(byte[] buffer, int offset, int count)
+        {
+            count = _encoder.EncodeFrame(buffer, offset, count, _buffer, 0);
+            base.Write(_buffer, 0, count);
+        }
+
+        protected override void Dispose(bool disposing)
+        {
+            base.Dispose(disposing);
+
+            if (disposing)
+                _encoder.Dispose();
+        }
+    }
+}

src/Discord.Net/Audio/Streams/RTPReadStream.cs

@@ -0,0 +1,53 @@
+using System;
+using System.Collections.Concurrent;
+using System.IO;
+
+namespace Discord.Audio
+{
+    public class RTPReadStream : Stream
+    {
+        private readonly BlockingCollection<byte[]> _queuedData; //TODO: Replace with max-length ring buffer
+        private readonly AudioClient _audioClient;
+        private readonly byte[] _buffer, _nonce, _secretKey;
+
+        public override bool CanRead => true;
+        public override bool CanSeek => false;
+        public override bool CanWrite => true;
+
+        internal RTPReadStream(AudioClient audioClient, byte[] secretKey, int bufferSize = 4000)
+        {
+            _audioClient = audioClient;
+            _secretKey = secretKey;
+            _buffer = new byte[bufferSize];
+            _queuedData = new BlockingCollection<byte[]>(100);
+            _nonce = new byte[24];
+        }
+
+        public override int Read(byte[] buffer, int offset, int count)
+        {
+            var queuedData = _queuedData.Take();
+            Buffer.BlockCopy(queuedData, 0, buffer, offset, Math.Min(queuedData.Length, count));
+            return queuedData.Length;
+        }
+        public override void Write(byte[] buffer, int offset, int count)
+        {
+            Buffer.BlockCopy(buffer, 0, _nonce, 0, 12);
+            count = SecretBox.Decrypt(buffer, offset, count, _buffer, 0, _nonce, _secretKey);
+            var newBuffer = new byte[count];
+            Buffer.BlockCopy(_buffer, 0, newBuffer, 0, count);
+            _queuedData.Add(newBuffer);
+        }
+
+        public override void Flush() { throw new NotSupportedException(); }
+
+        public override long Length { get { throw new NotSupportedException(); } }
+        public override long Position
+        {
+            get { throw new NotSupportedException(); }
+            set { throw new NotSupportedException(); }
+        }
+
+        public override void SetLength(long value) { throw new NotSupportedException(); }
+        public override long Seek(long offset, SeekOrigin origin) { throw new NotSupportedException(); }
+    }
+}

src/Discord.Net/Audio/Streams/RTPWriteStream.cs

@@ -0,0 +1,66 @@
+using System;
+using System.IO;
+
+namespace Discord.Audio
+{
+    public class RTPWriteStream : Stream
+    {
+        private readonly AudioClient _audioClient;
+        private readonly byte[] _buffer, _nonce, _secretKey;
+        private int _samplesPerFrame;
+        private uint _ssrc, _timestamp = 0;
+
+        public override bool CanRead => false;
+        public override bool CanSeek => false;
+        public override bool CanWrite => true;
+
+
+        internal RTPWriteStream(AudioClient audioClient, byte[] secretKey, int samplesPerFrame, uint ssrc, int bufferSize = 4000)
+        {
+            _audioClient = audioClient;
+            _secretKey = secretKey;
+            _samplesPerFrame = samplesPerFrame;
+            _ssrc = ssrc;
+            _nonce = new byte[24];
+            _buffer = new byte[bufferSize];
+            _buffer[0] = 0x80;
+            _buffer[1] = 0x78;
+            _buffer[8] = (byte)(_ssrc >> 24);
+            _buffer[9] = (byte)(_ssrc >> 16);
+            _buffer[10] = (byte)(_ssrc >> 8);
+            _buffer[11] = (byte)(_ssrc >> 0);
+        }
+
+        public override void Write(byte[] buffer, int offset, int count)
+        {
+            unchecked
+            {
+                if (_buffer[3]++ == byte.MaxValue)
+                    _buffer[4]++;
+
+                _timestamp += (uint)_samplesPerFrame;
+                _buffer[4] = (byte)(_timestamp >> 24);
+                _buffer[5] = (byte)(_timestamp >> 16);
+                _buffer[6] = (byte)(_timestamp >> 8);
+                _buffer[7] = (byte)(_timestamp >> 0);
+            }
+
+            Buffer.BlockCopy(buffer, 0, _nonce, 0, 12);
+            count = SecretBox.Encrypt(buffer, offset, count, _buffer, 12, _nonce, _secretKey);
+            _audioClient.Send(_buffer, count);
+        }
+
+        public override void Flush() { }
+
+        public override long Length { get { throw new NotSupportedException(); } }
+        public override long Position
+        {
+            get { throw new NotSupportedException(); }
+            set { throw new NotSupportedException(); }
+        }
+
+        public override int Read(byte[] buffer, int offset, int count) { throw new NotSupportedException(); }
+        public override void SetLength(long value) { throw new NotSupportedException(); }
+        public override long Seek(long offset, SeekOrigin origin) { throw new NotSupportedException(); }
+    }
+}