[Audio/AudioData] Added a type dedicated to just audio data

- A Sound can be constructed with and load an AudioData object - The individual members of Sound are replaced with AudioData - Microphone's and WavFormat's functions now return & take AudioData objects instead of Sound ones - Removed Microphone::recoverSound(), now useless - Removed the Internal::SoundAccess type, now unneeded
Razakhel · Jan 10, 2025 · 5ada8a3 · 5ada8a3
1 parent 7a32866
commit 5ada8a3
Show file tree

Hide file tree

Showing 16 changed files with 223 additions and 232 deletions.
diff --git a/examples/audioDemo.cpp b/examples/audioDemo.cpp
@@ -50,14 +50,17 @@ int main() {
     // Audio //
     ///////////
 
+    static const Raz::AudioData knockAudio        = Raz::WavFormat::load(RAZ_ROOT "assets/sounds/knock.wav");
+    static const Raz::AudioData waveSeagullsAudio = Raz::WavFormat::load(RAZ_ROOT "assets/sounds/wave_seagulls.wav");
+
     auto& audio = world.addSystem<Raz::AudioSystem>();
 
     // The Listener entity requires a Transform component
     auto& listener = world.addEntityWithComponent<Raz::Transform>().addComponent<Raz::Listener>();
 
     Raz::Entity& sound = world.addEntity();
     auto& soundTrans   = sound.addComponent<Raz::Transform>();
-    auto& soundComp    = sound.addComponent<Raz::Sound>(Raz::WavFormat::load(RAZ_ROOT "assets/sounds/knock.wav"));
+    auto& soundComp    = sound.addComponent<Raz::Sound>(knockAudio);
 
     Raz::Microphone microphone(Raz::AudioFormat::MONO_U8, 16000, 1.f);
 
@@ -109,7 +112,6 @@ int main() {
       listener.setGain(listenerGain);
 
       soundComp.init();
-      soundComp.load();
       soundComp.setRepeat(isRepeating);
       soundComp.setGain(soundGain);
       soundComp.setPitch(soundPitch);
@@ -162,11 +164,11 @@ int main() {
       switch (i) {
         case 0:
         default:
-          soundComp = Raz::WavFormat::load(RAZ_ROOT "assets/sounds/knock.wav");
+          soundComp.load(knockAudio);
           break;
 
         case 1:
-          soundComp = Raz::WavFormat::load(RAZ_ROOT "assets/sounds/wave_seagulls.wav");
+          soundComp.load(waveSeagullsAudio);
           break;
       }
 
@@ -605,7 +607,7 @@ int main() {
     // Starting application //
     //////////////////////////
 
-    std::vector<uint8_t> captureData;
+    Raz::AudioData captureData;
 
     app.run([&] (const Raz::FrameTimeInfo& timeInfo) {
       soundTrans.setPosition((moveSource ? Raz::Vec3f(std::sin(timeInfo.globalTime) * 3.f, 0.f, 1.f) : Raz::Vec3f(0.f)));
@@ -624,23 +626,23 @@ int main() {
         constexpr float factorI16 = 1.f / 32767;
 
         if (captureBitDepth == 8) {
-          for (std::size_t i = 0; i < captureData.size(); ++i) {
+          for (std::size_t i = 0; i < captureData.buffer.size(); ++i) {
             if (isCaptureStereo) { // Stereo 8
-              leftCapturePlot.push(static_cast<float>(captureData[i]) * factorU8 - 1.f);
-              rightCapturePlot.push(static_cast<float>(captureData[i + 1]) * factorU8 - 1.f);
+              leftCapturePlot.push(static_cast<float>(captureData.buffer[i]) * factorU8 - 1.f);
+              rightCapturePlot.push(static_cast<float>(captureData.buffer[i + 1]) * factorU8 - 1.f);
               ++i;
             } else { // Mono 8
-              monoCapturePlot.push(static_cast<float>(captureData[i]) * factorU8 - 1.f);
+              monoCapturePlot.push(static_cast<float>(captureData.buffer[i]) * factorU8 - 1.f);
             }
           }
         } else {
-          for (std::size_t i = 0; i < captureData.size(); i += 2) {
+          for (std::size_t i = 0; i < captureData.buffer.size(); i += 2) {
             if (isCaptureStereo) { // Stereo 16
-              leftCapturePlot.push(static_cast<float>(static_cast<int16_t>((captureData[i] << 0u) | (captureData[i + 1] << 8u))) * factorI16);
-              rightCapturePlot.push(static_cast<float>(static_cast<int16_t>((captureData[i + 2] << 0u) | (captureData[i + 3] << 8u))) * factorI16);
+              leftCapturePlot.push(static_cast<float>(static_cast<int16_t>(captureData.buffer[i] | (captureData.buffer[i + 1] << 8u))) * factorI16);
+              rightCapturePlot.push(static_cast<float>(static_cast<int16_t>(captureData.buffer[i + 2] | (captureData.buffer[i + 3] << 8u))) * factorI16);
               i += 2;
             } else { // Mono 16
-              monoCapturePlot.push(static_cast<float>(static_cast<int16_t>((captureData[i] << 0u) | (captureData[i + 1] << 8u))) * factorI16);
+              monoCapturePlot.push(static_cast<float>(static_cast<int16_t>(captureData.buffer[i] | (captureData.buffer[i + 1] << 8u))) * factorI16);
             }
           }
         }

diff --git a/include/RaZ/Audio/AudioData.hpp b/include/RaZ/Audio/AudioData.hpp
@@ -0,0 +1,30 @@
+#pragma once
+
+#ifndef RAZ_AUDIODATA_HPP
+#define RAZ_AUDIODATA_HPP
+
+#include <cstdint>
+#include <vector>
+
+namespace Raz {
+
+enum class AudioFormat : int {
+  MONO_U8    = 4352  /* AL_FORMAT_MONO8             */, ///< Mono format on 8 unsigned bits (0 to 255).
+  STEREO_U8  = 4354  /* AL_FORMAT_STEREO8           */, ///< Stereo format on 8 unsigned bits (0 to 255).
+  MONO_I16   = 4353  /* AL_FORMAT_MONO16            */, ///< Mono format on 16 signed bits (-32768 to 32767).
+  STEREO_I16 = 4355  /* AL_FORMAT_STEREO16          */, ///< Stereo format on 16 signed bits (-32768 to 32767).
+  MONO_F32   = 65552 /* AL_FORMAT_MONO_FLOAT32      */, ///< Mono format on 32 floating-point bits (float).
+  STEREO_F32 = 65553 /* AL_FORMAT_STEREO_FLOAT32    */, ///< Stereo format on 32 floating-point bits (float).
+  MONO_F64   = 65554 /* AL_FORMAT_MONO_DOUBLE_EXT   */, ///< Mono format on 64 floating-point bits (double).
+  STEREO_F64 = 65555 /* AL_FORMAT_STEREO_DOUBLE_EXT */  ///< Stereo format on 64 floating-point bits (double).
+};
+
+struct AudioData {
+  AudioFormat format {};
+  unsigned int frequency {};
+  std::vector<uint8_t> buffer {};
+};
+
+} // namespace Raz
+
+#endif // RAZ_AUDIODATA_HPP
diff --git a/include/RaZ/Audio/AudioSystem.hpp b/include/RaZ/Audio/AudioSystem.hpp
@@ -9,17 +9,6 @@
 
 namespace Raz {
 
-enum class AudioFormat : int {
-  MONO_U8    = 4352  /* AL_FORMAT_MONO8             */, ///< Mono format on 8 unsigned bits (0 to 255).
-  STEREO_U8  = 4354  /* AL_FORMAT_STEREO8           */, ///< Stereo format on 8 unsigned bits (0 to 255).
-  MONO_I16   = 4353  /* AL_FORMAT_MONO16            */, ///< Mono format on 16 signed bits (-32768 to 32767).
-  STEREO_I16 = 4355  /* AL_FORMAT_STEREO16          */, ///< Stereo format on 16 signed bits (-32768 to 32767).
-  MONO_F32   = 65552 /* AL_FORMAT_MONO_FLOAT32      */, ///< Mono format on 32 floating-point bits (float).
-  STEREO_F32 = 65553 /* AL_FORMAT_STEREO_FLOAT32    */, ///< Stereo format on 32 floating-point bits (float).
-  MONO_F64   = 65554 /* AL_FORMAT_MONO_DOUBLE_EXT   */, ///< Mono format on 64 floating-point bits (double).
-  STEREO_F64 = 65555 /* AL_FORMAT_STEREO_DOUBLE_EXT */  ///< Stereo format on 64 floating-point bits (double).
-};
-
 class AudioSystem final : public System {
 public:
   /// Creates a system handling audio.

diff --git a/include/RaZ/Audio/Microphone.hpp b/include/RaZ/Audio/Microphone.hpp
@@ -3,11 +3,11 @@
 #ifndef RAZ_MICROPHONE_HPP
 #define RAZ_MICROPHONE_HPP
 
-#include "RaZ/Audio/AudioSystem.hpp"
+#include "RaZ/Audio/AudioData.hpp"
 
-namespace Raz {
+#include <string>
 
-class Sound;
+namespace Raz {
 
 class Microphone {
 public:
@@ -45,21 +45,16 @@ class Microphone {
   /// Recovers the amount of currently captured time.
   /// \return Available captured duration, in seconds.
   float recoverAvailableDuration() const noexcept;
-  /// Recovers captured samples.
-  /// \note This flushes the recovered captured data; if recovering something, the available sample count right after this call will be less than it was before.
-  /// \param maxDuration Maximum amount of time to recover, in seconds. Giving a negative value will result in recovering all available samples.
-  /// \return Captured samples.
-  std::vector<uint8_t> recoverData(float maxDuration = -1.f) const;
-  /// Recovers captured samples. This overload can be used to avoid reallocating the whole memory range on each call.
-  /// \note This flushes the recovered captured data; if recovering something, the available sample count right after this call will be less than it was before.
-  /// \param data Data to be filled with the captured samples.
+  /// Recovers captured audio data.
+  /// \note This flushes the captured data; if recovering something, the available sample count right after this call will be less than it was before.
   /// \param maxDuration Maximum amount of time to recover, in seconds. Giving a negative value will result in recovering all available samples.
-  void recoverData(std::vector<uint8_t>& data, float maxDuration = -1.f) const;
-  /// Recovers captured samples as a Sound object.
-  /// \note This flushes the recovered captured data; if recovering something, the available sample count right after this call will be less than it was before.
+  /// \return Captured audio data. The format & frequency are those of the current audio input device.
+  AudioData recoverData(float maxDuration = -1.f) const;
+  /// Recovers captured audio data. This overload can be used to avoid reallocating the whole memory range on each call.
+  /// \note This flushes the captured data; if recovering something, the available sample count right after this call will be less than it was before.
+  /// \param data Data to be filled with the captured audio.
   /// \param maxDuration Maximum amount of time to recover, in seconds. Giving a negative value will result in recovering all available samples.
-  /// \return Captured sound.
-  Sound recoverSound(float maxDuration = -1.f) const;
+  void recoverData(AudioData& data, float maxDuration = -1.f) const;
 
   Microphone& operator=(const Microphone&) = delete;
   Microphone& operator=(Microphone&&) = delete;

diff --git a/include/RaZ/Audio/Sound.hpp b/include/RaZ/Audio/Sound.hpp
@@ -4,17 +4,13 @@
 #define RAZ_SOUND_HPP
 
 #include "RaZ/Component.hpp"
-#include "RaZ/Audio/AudioSystem.hpp"
+#include "RaZ/Audio/AudioData.hpp"
 #include "RaZ/Data/OwnerValue.hpp"
 
-#include <cstddef>
 #include <limits>
-#include <vector>
 
 namespace Raz {
 
-namespace Internal { class SoundAccess; }
-
 class SoundEffectSlot;
 
 template <typename T, std::size_t Size>
@@ -29,24 +25,25 @@ enum class SoundState : int {
 };
 
 class Sound final : public Component {
-  friend Internal::SoundAccess;
-  friend class Microphone;
-
 public:
   Sound() { init(); }
+  explicit Sound(AudioData data) : Sound() { load(std::move(data)); }
   Sound(const Sound&) = delete;
   Sound(Sound&&) noexcept = default;
 
-  constexpr unsigned int getBufferIndex() const noexcept { return m_buffer; }
-  constexpr AudioFormat getFormat() const noexcept { return m_format; }
-  constexpr int getFrequency() const noexcept { return m_frequency; }
+  constexpr unsigned int getBufferIndex() const noexcept { return m_bufferIndex; }
+  constexpr const AudioData& getData() const noexcept { return m_data; }
 
-  /// Initializes the sound.
+  /// Initializes the sound. If there is audio data, also loads it into memory.
   /// \note A Sound must be initialized again after opening an audio device.
   /// \see AudioSystem::open()
   void init();
-  /// Loads the sound's data into memory.
-  void load();
+  /// Loads the given audio data into memory.
+  /// @param data Data to be loaded.
+  void load(AudioData data) {
+    m_data = std::move(data);
+    load();
+  }
   /// Sets the sound's pitch multiplier.
   /// \param pitch Sound's pitch multiplier; must be positive. 1 is the default.
   void setPitch(float pitch) const noexcept;
@@ -128,30 +125,15 @@ class Sound final : public Component {
   ~Sound() override { destroy(); }
 
 private:
-  OwnerValue<unsigned int, std::numeric_limits<unsigned int>::max()> m_buffer {};
-  OwnerValue<unsigned int, std::numeric_limits<unsigned int>::max()> m_source {};
-
-  AudioFormat m_format {};
-  int m_frequency {};
-  std::vector<uint8_t> m_data {};
-};
-
-namespace Internal {
+  /// Loads the audio data into memory.
+  void load();
 
-/// Class giving direct access to a Sound's private members; useful for file importers.
-/// \note This class is not meant to be used in user code.
-class SoundAccess {
-public:
-  SoundAccess() = delete;
+  OwnerValue<unsigned int, std::numeric_limits<unsigned int>::max()> m_bufferIndex {};
+  OwnerValue<unsigned int, std::numeric_limits<unsigned int>::max()> m_sourceIndex {};
 
-  static AudioFormat& getFormat(Sound& sound) { return sound.m_format; }
-  static int& getFrequency(Sound& sound) { return sound.m_frequency; }
-  static const std::vector<uint8_t>& getData(const Sound& sound) { return sound.m_data; }
-  static std::vector<uint8_t>& getData(Sound& sound) { return sound.m_data; }
+  AudioData m_data {};
 };
 
-} // namespace Internal
-
 } // namespace Raz
 
 #endif // RAZ_SOUND_HPP
diff --git a/include/RaZ/Data/WavFormat.hpp b/include/RaZ/Data/WavFormat.hpp
@@ -5,20 +5,20 @@
 
 namespace Raz {
 
+struct AudioData;
 class FilePath;
-class Sound;
 
 namespace WavFormat {
 
-/// Loads a [WAV](https://en.wikipedia.org/wiki/WAV) audio from a file.
+/// Loads audio data from a [WAV](https://en.wikipedia.org/wiki/WAV) file.
 /// \param filePath File from which to load the audio.
-/// \return Imported sound.
-Sound load(const FilePath& filePath);
+/// \return Imported audio data.
+AudioData load(const FilePath& filePath);
 
-/// Saves a sound to a WAV file.
+/// Saves audio data to a [WAV](https://en.wikipedia.org/wiki/WAV) file.
 /// \param filePath File to which to save the sound.
-/// \param sound Sound to export data from.
-void save(const FilePath& filePath, const Sound& sound);
+/// \param data Audio data to export.
+void save(const FilePath& filePath, const AudioData& data);
 
 } // namespace WavFormat
 

diff --git a/include/RaZ/RaZ.hpp b/include/RaZ/RaZ.hpp
@@ -9,6 +9,7 @@
 #include "System.hpp"
 #include "World.hpp"
 #include "Animation/Skeleton.hpp"
+#include "Audio/AudioData.hpp"
 #include "Audio/AudioSystem.hpp"
 #include "Audio/Listener.hpp"
 #include "Audio/Microphone.hpp"

diff --git a/src/RaZ/Audio/Microphone.cpp b/src/RaZ/Audio/Microphone.cpp
@@ -7,6 +7,7 @@
 #include <AL/al.h>
 #include <AL/alc.h>
 
+#include <stdexcept>
 #include <string>
 
 namespace Raz {
@@ -52,7 +53,7 @@ constexpr int recoverFrameSize(AudioFormat format) {
       break;
 
     default:
-      throw std::invalid_argument("Error: Unhandled audio format");
+      throw std::invalid_argument("[Microphone] Unhandled audio format");
   }
 
   switch (format) {
@@ -77,7 +78,7 @@ constexpr int recoverFrameSize(AudioFormat format) {
       break;
 
     default:
-      throw std::invalid_argument("Error: Unhandled audio format");
+      throw std::invalid_argument("[Microphone] Unhandled audio format");
   }
 
   return channelCount * bitCount / 8;
@@ -170,17 +171,19 @@ float Microphone::recoverAvailableDuration() const noexcept {
   return (static_cast<float>(recoverAvailableSampleCount()) / static_cast<float>(m_frequency));
 }
 
-std::vector<uint8_t> Microphone::recoverData(float maxDuration) const {
-  std::vector<uint8_t> data;
+AudioData Microphone::recoverData(float maxDuration) const {
+  AudioData data {};
   recoverData(data, maxDuration);
 
   return data;
 }
 
-void Microphone::recoverData(std::vector<uint8_t>& data, float maxDuration) const {
+void Microphone::recoverData(AudioData& data, float maxDuration) const {
   ZoneScopedN("Microphone::recoverData");
 
-  data.clear();
+  data.format    = m_format;
+  data.frequency = m_frequency;
+  data.buffer.clear();
 
   if (maxDuration == 0.f)
     return;
@@ -193,25 +196,12 @@ void Microphone::recoverData(std::vector<uint8_t>& data, float maxDuration) cons
   if (maxDuration > 0.f)
     sampleCount = std::min(sampleCount, static_cast<int>(maxDuration * static_cast<float>(m_frequency)));
 
-  data.resize(recoverFrameSize(m_format) * sampleCount);
+  data.buffer.resize(recoverFrameSize(m_format) * sampleCount);
 
-  alcCaptureSamples(static_cast<ALCdevice*>(m_device), data.data(), sampleCount);
+  alcCaptureSamples(static_cast<ALCdevice*>(m_device), data.buffer.data(), sampleCount);
   checkError(m_device, "Failed to recover captured data");
 }
 
-Sound Microphone::recoverSound(float maxDuration) const {
-  ZoneScopedN("Microphone::recoverSound");
-
-  Sound sound;
-
-  sound.m_format    = m_format;
-  sound.m_frequency = static_cast<int>(m_frequency);
-  sound.m_data      = recoverData(maxDuration);
-  sound.load();
-
-  return sound;
-}
-
 void Microphone::destroy() {
   ZoneScopedN("Microphone::destroy");