more voice work!!!

5 files changed, 241 insertions, 21 deletions

diff --git a/source/application/StarMainApplication_sdl.cpp b/source/application/StarMainApplication_sdl.cpp
index 8f47eeb..86b9a10 100644
--- a/source/application/StarMainApplication_sdl.cpp
+++ b/source/application/StarMainApplication_sdl.cpp
@@ -248,9 +248,15 @@ public:
     if (SDL_InitSubSystem(SDL_INIT_GAMECONTROLLER))
       throw ApplicationException(strf("Couldn't initialize SDL Controller: {}", SDL_GetError()));
 
-    Logger::info("Application: Initializing SDL Sound");
+#ifdef STAR_SYSTEM_WINDOWS // Newer SDL is defaulting to xaudio2, which does not support audio capture
+    SDL_setenv("SDL_AUDIODRIVER", "directsound", 1);
+#endif
+
+    Logger::info("Application: Initializing SDL Audio");
     if (SDL_InitSubSystem(SDL_INIT_AUDIO))
-      throw ApplicationException(strf("Couldn't initialize SDL Sound: {}", SDL_GetError()));
+      throw ApplicationException(strf("Couldn't initialize SDL Audio: {}", SDL_GetError()));
+
+    Logger::info("Application: using Audio Driver '{}'", SDL_GetCurrentAudioDriver());
 
     SDL_JoystickEventState(SDL_ENABLE);
 
@@ -336,17 +342,23 @@ public:
     closeAudioInputDevice();
 
     SDL_AudioSpec obtained = {};
-    return (m_sdlAudioInputDevice = SDL_OpenAudioDevice(name, 1, &desired, &obtained, 0)) != 0;
+    m_sdlAudioInputDevice = SDL_OpenAudioDevice(name, 1, &desired, &obtained, 0);
+
+    if (m_sdlAudioInputDevice)
+      Logger::info("Opened audio input device '{}'", SDL_GetAudioDeviceName(m_sdlAudioInputDevice, 1));
+    else
+      Logger::info("Failed to open audio input device: {}", SDL_GetError());
+
+    return m_sdlAudioInputDevice != 0;
   }
 
   bool closeAudioInputDevice() {
     if (m_sdlAudioInputDevice) {
+      Logger::info("Closing audio input device '{}'", SDL_GetAudioDeviceName(m_sdlAudioInputDevice, 1));
       SDL_CloseAudioDevice(m_sdlAudioInputDevice);
       m_sdlAudioInputDevice = 0;
-
       return true;
     }
-
     return false;
   }
 
diff --git a/source/client/StarClientApplication.cpp b/source/client/StarClientApplication.cpp
index c58b6b5..f1c7595 100644
--- a/source/client/StarClientApplication.cpp
+++ b/source/client/StarClientApplication.cpp
@@ -208,6 +208,11 @@ void ClientApplication::applicationInit(ApplicationControllerPtr appController)
 
   appController->setMaxFrameSkip(assets->json("/client.config:maxFrameSkip").toUInt());
   appController->setUpdateTrackWindow(assets->json("/client.config:updateTrackWindow").toFloat());
+
+  if (auto jVoice = configuration->get("voice"))
+    m_voice->loadJson(jVoice.toObject());
+
+  m_voice->init();
 }
 
 void ClientApplication::renderInit(RendererPtr renderer) {
diff --git a/source/core/StarCurve25519.cpp b/source/core/StarCurve25519.cpp
index 3fdddea..f00c193 100644
--- a/source/core/StarCurve25519.cpp
+++ b/source/core/StarCurve25519.cpp
@@ -20,8 +20,6 @@ struct KeySet {
     secret[31] |= 64;
 
     ed25519_CreateKeyPair(publicKey.data(), privateKey.data(), nullptr, secret.data());
-
-    Logger::info("Generated Curve25519 key-pair");
   }
 };
 
diff --git a/source/frontend/StarVoice.cpp b/source/frontend/StarVoice.cpp
index 436461b..e5cb299 100644
--- a/source/frontend/StarVoice.cpp
+++ b/source/frontend/StarVoice.cpp
@@ -1,8 +1,11 @@
 #include "StarVoice.hpp"
 #include "StarFormat.hpp"
 #include "StarApplicationController.hpp"
+#include "StarTime.hpp"
+#include "StarRoot.hpp"
 #include "opus/include/opus.h"
 
+#include <queue>
 #include "SDL.h"
 
 constexpr int VOICE_SAMPLE_RATE = 48000;
@@ -25,10 +28,129 @@ EnumMap<VoiceChannelMode> const VoiceChannelModeNames{
   {VoiceChannelMode::Stereo, "Stereo"}
 };
 
+float getAudioChunkLoudness(int16_t* data, size_t samples) {
+	if (!samples)
+		return 0.f;
+
+	double rms = 0.;
+	for (size_t i = 0; i != samples; ++i) {
+		float sample = (float)data[i] / 32767.f;
+		rms += (double)(sample * sample);
+	}
+
+  float fRms = sqrtf((float)(rms / samples));
+
+	if (fRms > 0)
+		return std::clamp<float>(20.f * log10f(fRms), -127.f, 0.f);
+	else
+		return -127.f;
+}
+
+float getAudioLoudness(int16_t* data, size_t samples) {
+	constexpr size_t CHUNK_SIZE = 50;
+
+	float highest = -127.f;
+	for (size_t i = 0; i < samples; i += CHUNK_SIZE) {
+		float level = getAudioChunkLoudness(data + i, std::min<size_t>(i + CHUNK_SIZE, samples) - i);
+		if (level > highest)
+      highest = level;
+	}
+
+	return highest;
+}
+
+struct VoiceAudioChunk {
+	std::unique_ptr<int16_t[]> data;
+	size_t remaining;
+	size_t offset = 0;
+
+  VoiceAudioChunk(int16_t* ptr, size_t size) {
+		data.reset(ptr);
+		remaining = size;
+		offset = 0;
+	}
+
+	inline size_t takeSamples(std::vector<int16_t>& out, size_t count) {
+		size_t toRead = std::min<size_t>(count, remaining);
+		int16_t* start = data.get() + offset;
+		out.insert(out.end(), start, start + toRead);
+		offset += toRead;
+		remaining -= toRead;
+		return toRead;
+	}
+
+	//this one's unsafe
+	inline int16_t takeSample() {
+		--remaining;
+		return *(data.get() + offset++);
+	}
+
+	inline bool exhausted() {
+		return remaining == 0;
+	}
+};
+
+struct VoiceAudioStream {
+  // TODO: This should really be a ring buffer instead.
+  std::queue<VoiceAudioChunk> chunks{};
+  size_t samples = 0;
+  atomic<bool> muted = false;
+  atomic<bool> playing = false;
+  atomic<float> decibelLevel = 0.0f;
+  atomic<Array<float, 2>> channelVolumes = Array<float, 2>::filled(1.0f);
+
+  Mutex mutex;
+
+  inline int16_t getSample() {
+		int16_t sample = 0;
+		while (!chunks.empty()) {
+			auto& front = chunks.front();
+			if (front.exhausted()) {
+        chunks.pop();
+				continue;
+			}
+			--samples;
+			return front.takeSample();
+		}
+		return 0;
+	}
+
+	void nukeSamples(size_t count) {
+		while (!chunks.empty() && count > 0) {
+			auto& front = chunks.front();
+			if (count >= front.remaining) {
+				count -= front.remaining;
+				samples -= front.remaining;
+        chunks.pop();
+			}
+			else {
+				for (size_t i = 0; i != count; ++i) {
+					--samples;
+					front.takeSample();
+				}
+				break;
+			}
+		}
+	}
+
+	inline bool empty() { return chunks.empty(); }
+
+	void take(int16_t* ptr, size_t size) {
+    MutexLocker lock(mutex);
+	  while (samples > 22050 && !chunks.empty()) {
+		  samples -= chunks.front().remaining;
+		  chunks.pop();
+    }
+    chunks.emplace(ptr, size);
+		samples += size;
+	}
+};
+
 Voice::Speaker::Speaker(SpeakerId id)
   : decoderMono  (createDecoder(1), opus_decoder_destroy)
   , decoderStereo(createDecoder(2), opus_decoder_destroy) {
   speakerId = id;
+  audioStream = make_shared<VoiceAudioStream>();
 }
 
 Voice* Voice::s_singleton;
@@ -53,19 +175,58 @@ Voice::Voice(ApplicationControllerPtr appController) : m_encoder(nullptr, opus_e
   m_channelMode = VoiceChannelMode::Mono;
   m_applicationController = appController;
 
-  resetEncoder();
   s_singleton = this;
 }
 
 Voice::~Voice() {
+  save();
+
   s_singleton = nullptr;
 }
 
-void Voice::load(Json const& config) {
-  // do stuff
+void Voice::init() {
+  resetEncoder();
+  if (m_inputEnabled)
+    openDevice();
+}
+
+void Voice::loadJson(Json const& config) {
+  m_enabled      = config.getBool("enabled",         m_enabled);
+  m_inputEnabled = config.getBool("inputEnabled",    m_inputEnabled);
+  m_deviceName   = config.optQueryString("inputDevice");
+  m_threshold    = config.getFloat("threshold", m_threshold);
+  m_inputVolume  = config.getFloat("inputVolume", m_inputVolume);
+  m_outputVolume = config.getFloat("outputVolume",   m_outputVolume);
+  m_inputMode    = VoiceInputModeNames.getLeft(config.getString("inputMode", "pushToTalk"));
+  m_channelMode  = VoiceChannelModeNames.getLeft(config.getString("channelMode", "mono"));
+}
+
+
+
+Json Voice::saveJson() const {
+  return JsonObject{
+    {"enabled",      m_enabled},
+    {"inputEnabled", m_inputEnabled},
+    {"inputDevice",  m_deviceName ? *m_deviceName : Json()},
+    {"threshold",    m_threshold},
+    {"inputVolume",  m_inputVolume},
+    {"outputVolume", m_outputVolume},
+    {"inputMode",    VoiceInputModeNames.getRight(m_inputMode)},
+    {"channelMode",  VoiceChannelModeNames.getRight(m_channelMode)},
+    {"version",      1}
+  };
+}
+
+void Voice::save() const {
+  if (Root* root = Root::singletonPtr()) {
+    if (auto config = root->configuration())
+      config->set("voice", saveJson());
+  }
 }
-Json Voice::save() const {
-  return JsonObject{};
+
+void Voice::scheduleSave() {
+  if (nextSaveTime == 0.0)
+    nextSaveTime = Time::monotonicTime() + 2.0;
 }
 
 Voice::SpeakerPtr Voice::setLocalSpeaker(SpeakerId speakerId) {
@@ -87,6 +248,10 @@ Voice::SpeakerPtr Voice::speaker(SpeakerId speakerId) {
   }
 }
 
+void Voice::getAudioData(uint8_t* stream, int len) {
+
+}
+
 void Voice::mix(int16_t* buffer, size_t frames, unsigned channels) {
 
 }
@@ -95,13 +260,29 @@ void Voice::update(PositionalAttenuationFunction positionalAttenuationFunction)
   if (positionalAttenuationFunction) {
     for (auto& entry : m_speakers) {
       if (SpeakerPtr& speaker = entry.second) {
-        speaker->channelVolumes = {
+        speaker->audioStream->channelVolumes = {
           positionalAttenuationFunction(0, speaker->position, 1.0f),
           positionalAttenuationFunction(1, speaker->position, 1.0f)
         };
       }
     }
   }
+
+  auto now = Time::monotonicTime();
+  if (now > nextSaveTime) {
+    nextSaveTime = 0.0;
+    save();
+  }
+}
+
+
+void Voice::setDeviceName(Maybe<String> deviceName) {
+  if (m_deviceName == deviceName)
+    return;
+
+  m_deviceName = deviceName;
+  if (m_deviceOpen)
+    openDevice();
 }
 
 OpusDecoder* Voice::createDecoder(int channels) {
@@ -131,6 +312,10 @@ void Voice::resetEncoder() {
 void Voice::openDevice() {
   closeDevice();
 
+  m_applicationController->openAudioInputDevice(m_deviceName ? m_deviceName->utf8Ptr() : nullptr, VOICE_SAMPLE_RATE, encoderChannels(), this, [](void* userdata, uint8_t* stream, int len) {
+    ((Voice*)(userdata))->getAudioData(stream, len);
+  });
+
   m_deviceOpen = true;
 }
 
diff --git a/source/frontend/StarVoice.hpp b/source/frontend/StarVoice.hpp
index 0d485db..269adb4 100644
--- a/source/frontend/StarVoice.hpp
+++ b/source/frontend/StarVoice.hpp
@@ -5,6 +5,7 @@
 #include "StarException.hpp"
 #include "StarGameTypes.hpp"
 #include "StarMaybe.hpp"
+#include "StarThread.hpp"
 #include "StarApplicationController.hpp"
 
 struct OpusDecoder;
@@ -23,6 +24,7 @@ enum class VoiceChannelMode: uint8_t { Mono = 1, Stereo = 2 };
 extern EnumMap<VoiceChannelMode> const VoiceChannelModeNames;
 
 STAR_CLASS(Voice);
+STAR_CLASS(VoiceAudioStream);
 STAR_CLASS(ApplicationController);
 
 class Voice {
@@ -30,7 +32,8 @@ public:
   // Individual speakers are represented by their connection ID.
   typedef ConnectionId SpeakerId;
 
-  struct Speaker {
+  class Speaker {
+  public:
     SpeakerId speakerId = 0;
     EntityId entityId = 0;
 
@@ -39,10 +42,8 @@ public:
 
     OpusDecoderPtr decoderMono;
     OpusDecoderPtr decoderStereo;
-
-    atomic<bool> active = false;
-    atomic<float> currentLoudness = 0.0f;
-    atomic<Array<float, 2>> channelVolumes = Array<float, 2>::filled(1.0f);
+    VoiceAudioStreamPtr audioStream;
+    Mutex mutex;
 
     Speaker(SpeakerId speakerId);
   };
@@ -63,19 +64,29 @@ public:
   Voice(Voice const&) = delete;
   Voice& operator=(Voice const&) = delete;
 
-  void load(Json const& config);
-  Json save() const;
-  
+  void init();
+
+  void loadJson(Json const& config);
+  Json saveJson() const;
+
+  void save() const;
+  void scheduleSave();
+
   // Sets the local speaker ID and returns the local speaker. Must be called upon loading into a world.
   SpeakerPtr setLocalSpeaker(SpeakerId speakerId);
   SpeakerPtr speaker(SpeakerId speakerId);
 
+  // Called when receiving input audio data from SDL, on its own thread.
+  void getAudioData(uint8_t* stream, int len);
+
   // Called to mix voice audio with the game.
   void mix(int16_t* buffer, size_t frames, unsigned channels);
 
   typedef function<float(unsigned, Vec2F, float)> PositionalAttenuationFunction;
   void update(PositionalAttenuationFunction positionalAttenuationFunction = {});
 
+  void setDeviceName(Maybe<String> device);
+
   inline int encoderChannels() const {
     return m_channelMode == VoiceChannelMode::Mono ? 1 : 2;
   }
@@ -96,12 +107,21 @@ private:
 
   OpusEncoderPtr m_encoder;
 
+  float m_outputVolume = 1.0f;
+  float m_inputVolume = 1.0f;
+  float m_threshold = -50.0f;
+
+  bool m_enabled = true;
+  bool m_inputEnabled = true;
+
   bool m_deviceOpen = false;
   Maybe<String> m_deviceName;
   VoiceInputMode m_inputMode;
   VoiceChannelMode m_channelMode;
 
   ApplicationControllerPtr m_applicationController;
+
+  double nextSaveTime = 0.0f;
 };
   
 }