better handling of chunk buffer

src/audio/CMakeLists.txt

@@ -1,6 +1,7 @@
 idf_component_register(
   SRCS "audio_decoder.cpp" "audio_task.cpp" "chunk.cpp" "fatfs_audio_input.cpp"
   "stream_info.cpp" "stream_message.cpp" "i2s_audio_output.cpp"
+  "stream_buffer.cpp" "audio_playback.cpp"
   INCLUDE_DIRS "include"
   REQUIRES "codecs" "drivers" "cbor" "result" "tasks" "span")
 

src/audio/audio_decoder.cpp

@@ -19,26 +19,9 @@ static const char* kTag = "DEC";
 
 namespace audio {
 
-AudioDecoder::AudioDecoder()
+AudioDecoder::AudioDecoder() : IAudioElement(), stream_info_({}) {}
-    : IAudioElement(),
-      stream_info_({}),
-      raw_chunk_buffer_(static_cast<std::byte*>(
-          heap_caps_malloc(kMaxChunkSize, MALLOC_CAP_SPIRAM))),
-      chunk_buffer_(raw_chunk_buffer_, kMaxChunkSize)
 
-{}
+AudioDecoder::~AudioDecoder() {}
-
-AudioDecoder::~AudioDecoder() {
-  free(raw_chunk_buffer_);
-}
-
-auto AudioDecoder::SetInputBuffer(MessageBufferHandle_t* buffer) -> void {
-  input_buffer_ = buffer;
-}
-
-auto AudioDecoder::SetOutputBuffer(MessageBufferHandle_t* buffer) -> void {
-  output_buffer_ = buffer;
-}
 
 auto AudioDecoder::ProcessStreamInfo(const StreamInfo& info)
     -> cpp::result<void, AudioProcessingError> {
@@ -75,7 +58,7 @@ auto AudioDecoder::ProcessChunk(const cpp::span<std::byte>& chunk)
   bool needs_more_input = false;
   std::optional<codecs::ICodec::ProcessingError> error = std::nullopt;
   WriteChunksToStream(
-      output_buffer_, chunk_buffer_,
+      output_buffer_,
       [&](cpp::span<std::byte> buffer) -> std::size_t {
         std::size_t bytes_written = 0;
         // Continue filling up the output buffer so long as we have samples

src/audio/audio_playback.cpp

@@ -4,322 +4,88 @@
 #include <cstdint>
 #include <memory>
 #include <string_view>
-
+#include "audio_decoder.hpp"
-#include "aac_decoder.h"
+#include "audio_task.hpp"
-#include "amr_decoder.h"
+#include "chunk.hpp"
-#include "audio_element.h"
+#include "fatfs_audio_input.hpp"
-#include "audio_event_iface.h"
+#include "freertos/portmacro.h"
-#include "audio_pipeline.h"
+#include "gpio_expander.hpp"
-#include "esp_err.h"
+#include "i2s_audio_output.hpp"
-#include "flac_decoder.h"
+#include "storage.hpp"
-#include "mp3_decoder.h"
+#include "stream_buffer.hpp"
-#include "ogg_decoder.h"
+#include "stream_info.hpp"
-#include "opus_decoder.h"
+#include "stream_message.hpp"
-#include "wav_decoder.h"
+
-
+namespace audio {
-#include "audio_output.hpp"
+
-
+// TODO: idk
-static const char* kTag = "PLAYBACK";
+static const std::size_t kMinElementBufferSize = 1024;
-static const char* kSource = "src";
+
-static const char* kDecoder = "dec";
+auto AudioPlayback::create(drivers::GpioExpander* expander,
-static const char* kSink = "sink";
+                           std::shared_ptr<drivers::SdStorage> storage)
-
-static audio_element_status_t toStatus(void* status) {
-  uintptr_t as_pointer_int = reinterpret_cast<uintptr_t>(status);
-  return static_cast<audio_element_status_t>(as_pointer_int);
-}
-
-static bool endsWith(std::string_view str, std::string_view suffix) {
-  return str.size() >= suffix.size() &&
-         0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
-}
-
-static void toLower(std::string& str) {
-  std::transform(str.begin(), str.end(), str.begin(),
-                 [](unsigned char c) { return std::tolower(c); });
-}
-
-namespace drivers {
-
-auto AudioPlayback::create(std::unique_ptr<IAudioOutput> output)
     -> cpp::result<std::unique_ptr<AudioPlayback>, Error> {
-  audio_pipeline_handle_t pipeline;
+  // Create everything
-  audio_element_handle_t fatfs_stream_reader;
+  auto source = std::make_shared<FatfsAudioInput>(storage);
-  audio_event_iface_handle_t event_interface;
+  auto codec = std::make_shared<AudioDecoder>();
-
-  audio_pipeline_cfg_t pipeline_config =
-      audio_pipeline_cfg_t(DEFAULT_AUDIO_PIPELINE_CONFIG());
-  pipeline = audio_pipeline_init(&pipeline_config);
-  if (pipeline == NULL) {
-    return cpp::fail(Error::PIPELINE_INIT);
-  }
 
-  fatfs_stream_cfg_t fatfs_stream_config =
+  auto sink_res = I2SAudioOutput::create(expander);
-      fatfs_stream_cfg_t(FATFS_STREAM_CFG_DEFAULT());
+  if (sink_res.has_error()) {
-  fatfs_stream_config.type = AUDIO_STREAM_READER;
+    return cpp::fail(ERR_INIT_ELEMENT);
-  fatfs_stream_reader = fatfs_stream_init(&fatfs_stream_config);
-  if (fatfs_stream_reader == NULL) {
-    return cpp::fail(Error::FATFS_INIT);
   }
+  auto sink = sink_res.value();
 
-  audio_event_iface_cfg_t event_config = AUDIO_EVENT_IFACE_DEFAULT_CFG();
+  auto playback = std::make_unique<AudioPlayback>();
-  event_interface = audio_event_iface_init(&event_config);
 
-  audio_pipeline_set_listener(pipeline, event_interface);
+  // Configure the pipeline
-  audio_element_msg_set_listener(fatfs_stream_reader, event_interface);
+  source->InputBuffer(&playback->stream_start_);
-  audio_element_msg_set_listener(output->GetAudioElement(), event_interface);
+  sink->OutputBuffer(&playback->stream_end_);
+  playback->ConnectElements(source.get(), codec.get());
+  playback->ConnectElements(codec.get(), sink.get());
 
-  audio_pipeline_register(pipeline, fatfs_stream_reader, kSource);
+  // Launch!
-  audio_pipeline_register(pipeline, output->GetAudioElement(), kSink);
+  StartAudioTask("src", source);
+  StartAudioTask("dec", codec);
+  StartAudioTask("sink", sink);
 
-  return std::make_unique<AudioPlayback>(output, pipeline, fatfs_stream_reader,
+  return playback;
-                                         event_interface);
 }
 
-AudioPlayback::AudioPlayback(std::unique_ptr<IAudioOutput>& output,
+// TODO(jacqueline): think about sizes
-                             audio_pipeline_handle_t pipeline,
+AudioPlayback::AudioPlayback()
-                             audio_element_handle_t source_element,
+    : stream_start_(128, 128), stream_end_(128, 128) {}
-                             audio_event_iface_handle_t event_interface)
-    : output_(std::move(output)),
-      pipeline_(pipeline),
-      source_element_(source_element),
-      event_interface_(event_interface) {}
 
 AudioPlayback::~AudioPlayback() {
-  audio_pipeline_remove_listener(pipeline_);
+  // TODO(jacqueline): signal the end of all things, and maybe wait for it?
-  audio_element_msg_remove_listener(source_element_, event_interface_);
-  audio_element_msg_remove_listener(output_->GetAudioElement(),
-                                    event_interface_);
-
-  audio_pipeline_stop(pipeline_);
-  audio_pipeline_wait_for_stop(pipeline_);
-  audio_pipeline_terminate(pipeline_);
-
-  ReconfigurePipeline(NONE);
-
-  audio_pipeline_unregister(pipeline_, source_element_);
-  audio_pipeline_unregister(pipeline_, output_->GetAudioElement());
-
-  audio_event_iface_destroy(event_interface_);
-
-  audio_pipeline_deinit(pipeline_);
-  audio_element_deinit(source_element_);
-}
-
-void AudioPlayback::Play(const std::string& filename) {
-  output_->SetSoftMute(true);
-
-  if (playback_state_ != STOPPED) {
-    audio_pipeline_stop(pipeline_);
-    audio_pipeline_wait_for_stop(pipeline_);
-    audio_pipeline_terminate(pipeline_);
-  }
-
-  playback_state_ = PLAYING;
-  Decoder decoder = GetDecoderForFilename(filename);
-  ReconfigurePipeline(decoder);
-  audio_element_set_uri(source_element_, filename.c_str());
-  audio_pipeline_reset_ringbuffer(pipeline_);
-  audio_pipeline_reset_elements(pipeline_);
-  audio_pipeline_run(pipeline_);
-
-  output_->SetSoftMute(false);
 }
 
-void AudioPlayback::Toggle() {
+auto AudioPlayback::Play(const std::string& filename) -> void {
-  if (playback_state_ == PLAYING) {
+  StreamInfo info;
-    Pause();
+  info.Path(filename);
-  } else if (playback_state_ == PAUSED) {
-    Resume();
-  }
-}
-
-void AudioPlayback::Resume() {
-  if (playback_state_ == PAUSED) {
-    ESP_LOGI(kTag, "resuming");
-    playback_state_ = PLAYING;
-    audio_pipeline_resume(pipeline_);
-    output_->SetSoftMute(false);
-  }
-}
-void AudioPlayback::Pause() {
-  if (GetPlaybackState() == PLAYING) {
-    ESP_LOGI(kTag, "pausing");
-    output_->SetSoftMute(true);
-    playback_state_ = PAUSED;
-    audio_pipeline_pause(pipeline_);
-  }
-}
-
-auto AudioPlayback::GetPlaybackState() const -> PlaybackState {
-  return playback_state_;
-}
-
-void AudioPlayback::ProcessEvents(uint16_t max_time_ms) {
-  if (playback_state_ == STOPPED) {
-    return;
-  }
-
-  while (1) {
-    audio_event_iface_msg_t event;
-    esp_err_t err = audio_event_iface_listen(event_interface_, &event,
-                                             pdMS_TO_TICKS(max_time_ms));
-    if (err != ESP_OK) {
-      // Error should only be timeouts, so use a 'failure' as an indication that
-      // we're out of events to process.
-      break;
-    }
-
-    if (event.source_type == AUDIO_ELEMENT_TYPE_ELEMENT &&
-        event.source == (void*)decoder_ &&
-        event.cmd == AEL_MSG_CMD_REPORT_MUSIC_INFO) {
-      audio_element_info_t music_info;
-      audio_element_getinfo(decoder_, &music_info);
-      ESP_LOGI(kTag, "sample_rate=%d, bits=%d, ch=%d", music_info.sample_rates,
-               music_info.bits, music_info.channels);
-    }
 
-    if (event.source_type == AUDIO_ELEMENT_TYPE_ELEMENT &&
+  std::array<std::byte, 128> dest;
-        event.source == (void*)source_element_ &&
+  auto len = WriteMessage(
-        event.cmd == AEL_MSG_CMD_REPORT_STATUS) {
+      TYPE_STREAM_INFO, [&](auto enc) { return info.Encode(enc); }, dest);
-      audio_element_status_t status = toStatus(event.data);
-      if (status == AEL_STATUS_STATE_FINISHED) {
-        // TODO: Could we change the uri here? hmm.
-        ESP_LOGI(kTag, "finished reading input.");
-      }
-    }
 
-    if (event.source_type == AUDIO_ELEMENT_TYPE_ELEMENT &&
+  if (len.has_error()) {
-        event.source == (void*)output_->GetAudioElement() &&
+    // TODO.
-        event.cmd == AEL_MSG_CMD_REPORT_STATUS) {
-      audio_element_status_t status = toStatus(event.data);
-      if (status == AEL_STATUS_STATE_FINISHED) {
-        if (next_filename_ != "") {
-          ESP_LOGI(kTag, "finished writing output. enqueing next.");
-          Decoder decoder = GetDecoderForFilename(next_filename_);
-          if (decoder == decoder_type_) {
-            output_->SetSoftMute(true);
-            audio_element_set_uri(source_element_, next_filename_.c_str());
-            audio_pipeline_reset_ringbuffer(pipeline_);
-            audio_pipeline_reset_elements(pipeline_);
-            audio_pipeline_change_state(pipeline_, AEL_STATE_INIT);
-            audio_pipeline_run(pipeline_);
-            output_->SetSoftMute(true);
-          } else {
-            Play(next_filename_);
-          }
-          next_filename_ = "";
-        } else {
-          ESP_LOGI(kTag, "finished writing output. stopping.");
-          audio_pipeline_wait_for_stop(pipeline_);
-          audio_pipeline_terminate(pipeline_);
-          playback_state_ = STOPPED;
-        }
     return;
   }
-    }
-
-    if (event.need_free_data) {
-      // AFAICT this never happens in practice, but it doesn't hurt to follow
-      // the api here anyway.
-      free(event.data);
-    }
-  }
-}
-
-void AudioPlayback::SetNextFile(const std::string& filename) {
-  next_filename_ = filename;
-}
-
-void AudioPlayback::SetVolume(uint8_t volume) {
-  output_->SetVolume(volume);
-}
-
-auto AudioPlayback::GetVolume() const -> uint8_t {
-  return output_->GetVolume();
-}
-
-auto AudioPlayback::GetDecoderForFilename(std::string filename) const
-    -> Decoder {
-  toLower(filename);
-  if (endsWith(filename, "mp3")) {
-    return MP3;
-  }
-  if (endsWith(filename, "amr") || endsWith(filename, "wamr")) {
-    return AMR;
-  }
-  if (endsWith(filename, "opus")) {
-    return OPUS;
-  }
-  if (endsWith(filename, "ogg")) {
-    return OGG;
-  }
-  if (endsWith(filename, "flac")) {
-    return FLAC;
-  }
-  if (endsWith(filename, "wav")) {
-    return WAV;
-  }
-  if (endsWith(filename, "aac") || endsWith(filename, "m4a") ||
-      endsWith(filename, "ts") || endsWith(filename, "mp4")) {
-    return AAC;
-  }
-  return NONE;
-}
 
-auto AudioPlayback::CreateDecoder(Decoder decoder) const
+  // TODO: short delay, return error on fail
-    -> audio_element_handle_t {
+  xMessageBufferSend(*stream_start_.Handle(), dest.data(), len.value(),
-  if (decoder == MP3) {
+                     portMAX_DELAY);
-    mp3_decoder_cfg_t config = DEFAULT_MP3_DECODER_CONFIG();
-    return mp3_decoder_init(&config);
-  }
-  if (decoder == AMR) {
-    amr_decoder_cfg_t config = DEFAULT_AMR_DECODER_CONFIG();
-    return amr_decoder_init(&config);
-  }
-  if (decoder == OPUS) {
-    opus_decoder_cfg_t config = DEFAULT_OPUS_DECODER_CONFIG();
-    return decoder_opus_init(&config);
-  }
-  if (decoder == OGG) {
-    ogg_decoder_cfg_t config = DEFAULT_OGG_DECODER_CONFIG();
-    return ogg_decoder_init(&config);
-  }
-  if (decoder == FLAC) {
-    flac_decoder_cfg_t config = DEFAULT_FLAC_DECODER_CONFIG();
-    return flac_decoder_init(&config);
-  }
-  if (decoder == WAV) {
-    wav_decoder_cfg_t config = DEFAULT_WAV_DECODER_CONFIG();
-    return wav_decoder_init(&config);
-  }
-  if (decoder == AAC) {
-    aac_decoder_cfg_t config = DEFAULT_AAC_DECODER_CONFIG();
-    return aac_decoder_init(&config);
-  }
-  return nullptr;
 }
 
-void AudioPlayback::ReconfigurePipeline(Decoder decoder) {
+auto AudioPlayback::ConnectElements(IAudioElement* src, IAudioElement* sink)
-  if (decoder_type_ == decoder) {
+    -> void {
-    return;
+  std::size_t chunk_size =
-  }
+      std::max(src->InputMinChunkSize(), sink->InputMinChunkSize());
-
+  std::size_t buffer_size = std::max(kMinElementBufferSize, chunk_size * 2);
-  if (decoder_type_ != NONE) {
-    audio_pipeline_unlink(pipeline_);
-    audio_element_msg_remove_listener(decoder_, event_interface_);
-    audio_pipeline_unregister(pipeline_, decoder_);
-    audio_element_deinit(decoder_);
-  }
 
-  if (decoder != NONE) {
+  auto buffer = std::make_unique<StreamBuffer>(chunk_size, buffer_size);
-    decoder_ = CreateDecoder(decoder);
+  src->OutputBuffer(buffer.get());
-    decoder_type_ = decoder;
+  sink->OutputBuffer(buffer.get());
-    audio_pipeline_register(pipeline_, decoder_, kDecoder);
+  element_buffers_.push_back(std::move(buffer));
-    audio_element_msg_set_listener(decoder_, event_interface_);
-    static const char* link_tag[3] = {kSource, kDecoder, kSink};
-    audio_pipeline_link(pipeline_, &link_tag[0], 3);
-  }
 }
 
-}  // namespace drivers
+}  // namespace audio

src/audio/audio_task.cpp

@@ -19,12 +19,8 @@
 
 namespace audio {
 
-static const TickType_t kCommandWaitTicks = 1;
-static const TickType_t kIdleTaskDelay = 1;
-static const size_t kChunkBufferSize = kMaxChunkSize * 1.5;
-
 auto StartAudioTask(const std::string& name,
-                    std::shared_ptr<IAudioElement>& element) -> void {
+                    std::shared_ptr<IAudioElement> element) -> void {
   AudioTaskArgs* args = new AudioTaskArgs{.element = element};
   xTaskCreate(&AudioTaskMain, name.c_str(), element->StackSizeBytes(), args,
               kTaskPriorityAudio, NULL);
@@ -45,7 +41,6 @@ void AudioTaskMain(void* args) {
       // processing any chunks from it. Try doing this first, then fall back to
       // the other cases.
       bool has_received_message = false;
-      if (element->InputBuffer() != nullptr) {
       ChunkReadResult chunk_res = chunk_reader.ReadChunkFromStream(
           [&](cpp::span<std::byte> data) -> std::optional<size_t> {
             process_res = element->ProcessChunk(data);
@@ -63,7 +58,6 @@ void AudioTaskMain(void* args) {
       } else if (chunk_res == CHUNK_STREAM_ENDED) {
         has_received_message = true;
       }
-      }
 
       if (has_received_message) {
         auto message = chunk_reader.GetLastMessage();

src/audio/chunk.cpp

@@ -8,31 +8,19 @@
 
 #include "cbor.h"
 
+#include "stream_buffer.hpp"
 #include "stream_message.hpp"
 
 namespace audio {
 
-/*
+auto WriteChunksToStream(StreamBuffer* stream,
- * The maximum size of a single chunk of stream data. This should be comfortably
- * larger than the largest size of a frame of audio we should expect to handle.
- *
- * 128 kbps MPEG-1 @ 44.1 kHz is approx. 418 bytes according to the internet.
- *
- * TODO(jacqueline): tune as more codecs are added.
- */
-const std::size_t kMaxChunkSize = 2048;
-
-// TODO: tune
-static const std::size_t kWorkingBufferSize = kMaxChunkSize * 1.5;
-
-auto WriteChunksToStream(MessageBufferHandle_t* stream,
-                         cpp::span<std::byte> working_buffer,
                          std::function<size_t(cpp::span<std::byte>)> callback,
                          TickType_t max_wait) -> ChunkWriteResult {
+  cpp::span<std::byte> write_buffer = stream->WriteBuffer();
   while (1) {
     // First, write out our chunk header so we know how much space to give to
     // the callback.
-    auto header_size = WriteTypeOnlyMessage(TYPE_CHUNK_HEADER, working_buffer);
+    auto header_size = WriteTypeOnlyMessage(TYPE_CHUNK_HEADER, write_buffer);
     if (header_size.has_error()) {
       return CHUNK_ENCODING_ERROR;
     }
@@ -40,8 +28,8 @@ auto WriteChunksToStream(MessageBufferHandle_t* stream,
     // Now we can ask the callback to fill the remaining space.
     size_t chunk_size = std::invoke(
         callback,
-        working_buffer.subspan(header_size.value(),
+        write_buffer.subspan(header_size.value(),
-                               working_buffer.size() - header_size.value()));
+                             write_buffer.size() - header_size.value()));
 
     if (chunk_size == 0) {
       // They had nothing for us, so bail out.
@@ -51,7 +39,7 @@ auto WriteChunksToStream(MessageBufferHandle_t* stream,
     // Try to write to the buffer. Note the return type here will be either 0 or
     // header_size + chunk_size, as MessageBuffer doesn't allow partial writes.
     size_t actual_write_size =
-        xMessageBufferSend(*stream, working_buffer.data(),
+        xMessageBufferSend(stream->Handle(), write_buffer.data(),
                            header_size.value() + chunk_size, max_wait);
 
     if (actual_write_size == 0) {
@@ -63,15 +51,9 @@ auto WriteChunksToStream(MessageBufferHandle_t* stream,
   }
 }
 
-ChunkReader::ChunkReader(MessageBufferHandle_t* stream)
+ChunkReader::ChunkReader(StreamBuffer* stream) : stream_(stream) {}
-    : stream_(stream),
-      raw_working_buffer_(static_cast<std::byte*>(
-          heap_caps_malloc(kWorkingBufferSize, MALLOC_CAP_SPIRAM))),
-      working_buffer_(raw_working_buffer_, kWorkingBufferSize) {}
 
-ChunkReader::~ChunkReader() {
+ChunkReader::~ChunkReader() {}
-  free(raw_working_buffer_);
-}
 
 auto ChunkReader::Reset() -> void {
   leftover_bytes_ = 0;
@@ -79,16 +61,17 @@ auto ChunkReader::Reset() -> void {
 }
 
 auto ChunkReader::GetLastMessage() -> cpp::span<std::byte> {
-  return working_buffer_.subspan(leftover_bytes_, last_message_size_);
+  return stream_->ReadBuffer().subspan(leftover_bytes_, last_message_size_);
 }
 
 auto ChunkReader::ReadChunkFromStream(
     std::function<std::optional<size_t>(cpp::span<std::byte>)> callback,
     TickType_t max_wait) -> ChunkReadResult {
   // First, wait for a message to arrive over the buffer.
-  last_message_size_ =
+  cpp::span<std::byte> new_data_dest = stream_->ReadBuffer().last(
-      xMessageBufferReceive(*stream_, raw_working_buffer_ + leftover_bytes_,
+      stream_->ReadBuffer().size() - leftover_bytes_);
-                            working_buffer_.size() - leftover_bytes_, max_wait);
+  last_message_size_ = xMessageBufferReceive(
+      stream_->Handle(), new_data_dest.data(), new_data_dest.size(), max_wait);
 
   if (last_message_size_ == 0) {
     return CHUNK_READ_TIMEOUT;
@@ -109,7 +92,8 @@ auto ChunkReader::ReadChunkFromStream(
   // Now we need to stick the end of the last chunk (if it exists) onto the
   // front of the new chunk. Do it this way around bc we assume the old chunk
   // is shorter, and therefore faster to move.
-  cpp::span<std::byte> leftover_data = working_buffer_.first(leftover_bytes_);
+  cpp::span<std::byte> leftover_data =
+      stream_->ReadBuffer().first(leftover_bytes_);
   cpp::span<std::byte> combined_data(chunk_data.data() - leftover_data.size(),
                                      leftover_data.size() + chunk_data.size());
   if (leftover_bytes_ > 0) {
@@ -127,7 +111,7 @@ auto ChunkReader::ReadChunkFromStream(
   leftover_bytes_ = combined_data.size() - amount_processed.value();
   if (leftover_bytes_ > 0) {
     std::copy(combined_data.begin() + amount_processed.value(),
-              combined_data.end(), working_buffer_.begin());
+              combined_data.end(), stream_->ReadBuffer().begin());
     return CHUNK_LEFTOVER_DATA;
   }
 

src/audio/fatfs_audio_input.cpp

@@ -19,7 +19,6 @@ static const TickType_t kServiceInterval = pdMS_TO_TICKS(50);
 
 static const std::size_t kFileBufferSize = 1024 * 128;
 static const std::size_t kMinFileReadSize = 1024 * 4;
-static const std::size_t kOutputBufferSize = 1024 * 4;
 
 FatfsAudioInput::FatfsAudioInput(std::shared_ptr<drivers::SdStorage> storage)
     : IAudioElement(),
@@ -29,24 +28,11 @@ FatfsAudioInput::FatfsAudioInput(std::shared_ptr<drivers::SdStorage> storage)
       file_buffer_(raw_file_buffer_, kFileBufferSize),
       file_buffer_read_pos_(file_buffer_.begin()),
       file_buffer_write_pos_(file_buffer_.begin()),
-      raw_chunk_buffer_(static_cast<std::byte*>(
-          heap_caps_malloc(kMaxChunkSize, MALLOC_CAP_SPIRAM))),
-      chunk_buffer_(raw_chunk_buffer_, kMaxChunkSize),
       current_file_(),
-      is_file_open_(false),
+      is_file_open_(false) {}
-      output_buffer_memory_(static_cast<uint8_t*>(
-          heap_caps_malloc(kOutputBufferSize, MALLOC_CAP_SPIRAM))) {
-  output_buffer_ = new MessageBufferHandle_t;
-  *output_buffer_ = xMessageBufferCreateStatic(
-      kOutputBufferSize, output_buffer_memory_, &output_buffer_metadata_);
-}
 
 FatfsAudioInput::~FatfsAudioInput() {
   free(raw_file_buffer_);
-  free(raw_chunk_buffer_);
-  vMessageBufferDelete(output_buffer_);
-  free(output_buffer_memory_);
-  free(output_buffer_);
 }
 
 auto FatfsAudioInput::ProcessStreamInfo(const StreamInfo& info)
@@ -70,13 +56,13 @@ auto FatfsAudioInput::ProcessStreamInfo(const StreamInfo& info)
   auto write_size =
       WriteMessage(TYPE_STREAM_INFO,
                    std::bind(&StreamInfo::Encode, info, std::placeholders::_1),
-                   chunk_buffer_);
+                   output_buffer_->WriteBuffer());
 
   if (write_size.has_error()) {
     return cpp::fail(IO_ERROR);
   } else {
-    xMessageBufferSend(output_buffer_, chunk_buffer_.data(), write_size.value(),
+    xMessageBufferSend(output_buffer_, output_buffer_->WriteBuffer().data(),
-                       portMAX_DELAY);
+                       write_size.value(), portMAX_DELAY);
   }
 
   return {};
@@ -142,8 +128,8 @@ auto FatfsAudioInput::ProcessIdle() -> cpp::result<void, AudioProcessingError> {
   // Now stream data into the output buffer until it's full.
   pending_read_pos_ = file_buffer_read_pos_;
   ChunkWriteResult result = WriteChunksToStream(
-      output_buffer_, chunk_buffer_,
+      output_buffer_, [&](cpp::span<std::byte> d) { return SendChunk(d); },
-      [&](cpp::span<std::byte> d) { return SendChunk(d); }, kServiceInterval);
+      kServiceInterval);
 
   switch (result) {
     case CHUNK_WRITE_TIMEOUT:

src/audio/i2s_audio_output.cpp

@@ -16,7 +16,7 @@ static const char* kTag = "I2SOUT";
 namespace audio {
 
 auto I2SAudioOutput::create(drivers::GpioExpander* expander)
-    -> cpp::result<std::unique_ptr<I2SAudioOutput>, Error> {
+    -> cpp::result<std::shared_ptr<I2SAudioOutput>, Error> {
   // First, we need to perform initial configuration of the DAC chip.
   auto dac_result = drivers::AudioDac::create(expander);
   if (dac_result.has_error()) {
@@ -27,9 +27,10 @@ auto I2SAudioOutput::create(drivers::GpioExpander* expander)
 
   // Soft mute immediately, in order to minimise any clicks and pops caused by
   // the initial output element and pipeline configuration.
-  dac->WriteVolume(255);
+  // dac->WriteVolume(255);
+  dac->WriteVolume(120);  // for testing
 
-  return std::make_unique<I2SAudioOutput>(expander, std::move(dac));
+  return std::make_shared<I2SAudioOutput>(expander, std::move(dac));
 }
 
 I2SAudioOutput::I2SAudioOutput(drivers::GpioExpander* expander,

src/audio/include/audio_decoder.hpp

@@ -21,11 +21,15 @@ class AudioDecoder : public IAudioElement {
   AudioDecoder();
   ~AudioDecoder();
 
-  auto SetInputBuffer(MessageBufferHandle_t*) -> void;
-  auto SetOutputBuffer(MessageBufferHandle_t*) -> void;
-
   auto StackSizeBytes() const -> std::size_t override { return 8196; };
 
+  auto InputMinChunkSize() const -> std::size_t override {
+    // 128 kbps MPEG-1 @ 44.1 kHz is approx. 418 bytes according to the
+    // internet.
+    // TODO(jacqueline): tune as more codecs are added.
+    return 1024;
+  }
+
   auto ProcessStreamInfo(const StreamInfo& info)
       -> cpp::result<void, AudioProcessingError> override;
   auto ProcessChunk(const cpp::span<std::byte>& chunk)
@@ -38,9 +42,6 @@ class AudioDecoder : public IAudioElement {
  private:
   std::unique_ptr<codecs::ICodec> current_codec_;
   std::optional<StreamInfo> stream_info_;
-
-  std::byte* raw_chunk_buffer_;
-  cpp::span<std::byte> chunk_buffer_;
 };
 
 }  // namespace audio

src/audio/include/audio_element.hpp

@@ -2,6 +2,7 @@
 
 #include <cstdint>
 
+#include "chunk.hpp"
 #include "freertos/FreeRTOS.h"
 
 #include "freertos/message_buffer.h"
@@ -9,6 +10,7 @@
 #include "result.hpp"
 #include "span.hpp"
 
+#include "stream_buffer.hpp"
 #include "stream_info.hpp"
 #include "types.hpp"
 
@@ -41,7 +43,7 @@ enum AudioProcessingError {
 class IAudioElement {
  public:
   IAudioElement() : input_buffer_(nullptr), output_buffer_(nullptr) {}
-  virtual ~IAudioElement();
+  virtual ~IAudioElement() {}
 
   /*
    * Returns the stack size in bytes that this element requires. This should
@@ -57,11 +59,17 @@ class IAudioElement {
    */
   virtual auto IdleTimeout() const -> TickType_t { return portMAX_DELAY; }
 
+  virtual auto InputMinChunkSize() const -> std::size_t { return 0; }
+
   /* Returns this element's input buffer. */
-  auto InputBuffer() const -> MessageBufferHandle_t* { return input_buffer_; }
+  auto InputBuffer() const -> StreamBuffer* { return input_buffer_; }
 
   /* Returns this element's output buffer. */
-  auto OutputBuffer() const -> MessageBufferHandle_t* { return output_buffer_; }
+  auto OutputBuffer() const -> StreamBuffer* { return output_buffer_; }
+
+  auto InputBuffer(StreamBuffer* b) -> void { input_buffer_ = b; }
+
+  auto OutputBuffer(StreamBuffer* b) -> void { output_buffer_ = b; }
 
   /*
    * Called when a StreamInfo message is received. Used to configure this
@@ -87,8 +95,8 @@ class IAudioElement {
   virtual auto ProcessIdle() -> cpp::result<void, AudioProcessingError> = 0;
 
  protected:
-  MessageBufferHandle_t* input_buffer_;
+  StreamBuffer* input_buffer_;
-  MessageBufferHandle_t* output_buffer_;
+  StreamBuffer* output_buffer_;
 };
 
 }  // namespace audio

src/audio/include/audio_playback.hpp

@@ -3,95 +3,44 @@
 #include <cstdint>
 #include <memory>
 #include <string>
+#include <vector>
 
-#include "audio_common.h"
+#include "audio_element.hpp"
-#include "audio_element.h"
-#include "audio_event_iface.h"
-#include "audio_pipeline.h"
 #include "esp_err.h"
-#include "fatfs_stream.h"
+#include "gpio_expander.hpp"
-#include "i2s_stream.h"
-#include "mp3_decoder.h"
 #include "result.hpp"
-
+#include "span.hpp"
-#include "audio_output.hpp"
-#include "dac.hpp"
 #include "storage.hpp"
+#include "stream_buffer.hpp"
 
-namespace drivers {
+namespace audio {
 
 /*
- * Sends an I2S audio stream to the DAC. Includes basic controls for pausing
+ * TODO.
- * and resuming the stream, as well as support for gapless playback of the next
- * queued song, but does not implement any kind of sophisticated queing or
- * playback control; these should be handled at a higher level.
  */
 class AudioPlayback {
  public:
-  enum Error { FATFS_INIT, I2S_INIT, PIPELINE_INIT };
+  enum Error { ERR_INIT_ELEMENT, ERR_MEM };
-  static auto create(std::unique_ptr<IAudioOutput> output)
+  static auto create(drivers::GpioExpander* expander,
+                     std::shared_ptr<drivers::SdStorage> storage)
       -> cpp::result<std::unique_ptr<AudioPlayback>, Error>;
 
-  AudioPlayback(std::unique_ptr<IAudioOutput>& output,
+  // TODO(jacqueline): configure on the fly once we have things to configure.
-                audio_pipeline_handle_t pipeline,
+  AudioPlayback();
-                audio_element_handle_t source_element,
-                audio_event_iface_handle_t event_interface);
   ~AudioPlayback();
 
-  /*
-   * Replaces any currently playing file with the one given, and begins
-   * playback.
-   *
-   * Any value set in `set_next_file` is cleared by this method.
-   */
   auto Play(const std::string& filename) -> void;
-  /* Toogle between resumed and paused. */
-  auto Toggle() -> void;
-  auto Resume() -> void;
-  auto Pause() -> void;
-
-  enum PlaybackState { PLAYING, PAUSED, STOPPED };
-  auto GetPlaybackState() const -> PlaybackState;
-
-  /*
-   * Handles any pending events from the underlying audio pipeline. This must
-   * be called regularly in order to handle configuring the I2S stream for
-   * different audio types (e.g. sample rate, bit depth), and for gapless
-   * playback.
-   */
-  auto ProcessEvents(uint16_t max_time_ms) -> void;
-
-  /*
-   * Sets the file that should be played immediately after the current file
-   * finishes. This is used for gapless playback
-   */
-  auto SetNextFile(const std::string& filename) -> void;
-
-  auto SetVolume(uint8_t volume) -> void;
-  auto GetVolume() const -> uint8_t;
 
   // Not copyable or movable.
   AudioPlayback(const AudioPlayback&) = delete;
   AudioPlayback& operator=(const AudioPlayback&) = delete;
 
  private:
-  PlaybackState playback_state_;
+  auto ConnectElements(IAudioElement* src, IAudioElement* sink) -> void;
-
-  enum Decoder { NONE, MP3, AMR, OPUS, OGG, FLAC, WAV, AAC };
-  auto GetDecoderForFilename(std::string filename) const -> Decoder;
-  auto CreateDecoder(Decoder decoder) const -> audio_element_handle_t;
-  auto ReconfigurePipeline(Decoder decoder) -> void;
-
-  std::unique_ptr<IAudioOutput> output_;
-
-  std::string next_filename_ = "";
-
-  audio_pipeline_handle_t pipeline_;
-  audio_element_handle_t source_element_;
-  audio_event_iface_handle_t event_interface_;
 
-  audio_element_handle_t decoder_ = nullptr;
+  StreamBuffer stream_start_;
-  Decoder decoder_type_ = NONE;
+  StreamBuffer stream_end_;
+  std::vector<std::unique_ptr<StreamBuffer>> element_buffers_;
 };
 
-}  // namespace drivers
+}  // namespace audio

src/audio/include/audio_task.hpp

@@ -10,7 +10,8 @@ struct AudioTaskArgs {
   std::shared_ptr<IAudioElement>& element;
 };
 
-auto StartAudioTask(std::shared_ptr<IAudioElement>& element) -> void;
+auto StartAudioTask(const std::string& name,
+                    std::shared_ptr<IAudioElement> element) -> void;
 
 void AudioTaskMain(void* args);
 

src/audio/include/chunk.hpp

@@ -13,11 +13,10 @@
 #include "freertos/queue.h"
 #include "result.hpp"
 #include "span.hpp"
+#include "stream_buffer.hpp"
 
 namespace audio {
 
-extern const std::size_t kMaxChunkSize;
-
 enum ChunkWriteResult {
   // Returned when the callback does not write any data.
   CHUNK_OUT_OF_DATA,
@@ -37,8 +36,7 @@ enum ChunkWriteResult {
  * number of bytes it wrote. Return a value of 0 to indicate that there is no
  * more input to read.
  */
-auto WriteChunksToStream(MessageBufferHandle_t* stream,
+auto WriteChunksToStream(StreamBuffer* stream,
-                         cpp::span<std::byte> working_buffer,
                          std::function<size_t(cpp::span<std::byte>)> callback,
                          TickType_t max_wait) -> ChunkWriteResult;
 
@@ -59,7 +57,7 @@ enum ChunkReadResult {
 
 class ChunkReader {
  public:
-  ChunkReader(MessageBufferHandle_t* stream);
+  explicit ChunkReader(StreamBuffer* buffer);
   ~ChunkReader();
 
   auto Reset() -> void;
@@ -83,10 +81,7 @@ class ChunkReader {
       TickType_t max_wait) -> ChunkReadResult;
 
  private:
-  MessageBufferHandle_t* stream_;
+  StreamBuffer* stream_;
-  std::byte* raw_working_buffer_;
-  cpp::span<std::byte> working_buffer_;
-
   std::size_t leftover_bytes_ = 0;
   std::size_t last_message_size_ = 0;
 };

src/audio/include/fatfs_audio_input.hpp

@@ -4,6 +4,7 @@
 #include <memory>
 #include <string>
 
+#include "chunk.hpp"
 #include "freertos/FreeRTOS.h"
 
 #include "freertos/message_buffer.h"
@@ -12,6 +13,7 @@
 
 #include "audio_element.hpp"
 #include "storage.hpp"
+#include "stream_buffer.hpp"
 
 namespace audio {
 
@@ -28,6 +30,9 @@ class FatfsAudioInput : public IAudioElement {
 
   auto SendChunk(cpp::span<std::byte> dest) -> size_t;
 
+  FatfsAudioInput(const FatfsAudioInput&) = delete;
+  FatfsAudioInput& operator=(const FatfsAudioInput&) = delete;
+
  private:
   auto GetRingBufferDistance() const -> size_t;
 
@@ -39,14 +44,8 @@ class FatfsAudioInput : public IAudioElement {
   cpp::span<std::byte>::iterator pending_read_pos_;
   cpp::span<std::byte>::iterator file_buffer_write_pos_;
 
-  std::byte* raw_chunk_buffer_;
-  cpp::span<std::byte> chunk_buffer_;
-
   FIL current_file_;
   bool is_file_open_;
-
-  uint8_t* output_buffer_memory_;
-  StaticMessageBuffer_t output_buffer_metadata_;
 };
 
 }  // namespace audio

src/audio/include/i2s_audio_output.hpp

@@ -16,13 +16,17 @@ class I2SAudioOutput : public IAudioElement {
  public:
   enum Error { DAC_CONFIG, I2S_CONFIG, STREAM_INIT };
   static auto create(drivers::GpioExpander* expander)
-      -> cpp::result<std::unique_ptr<I2SAudioOutput>, Error>;
+      -> cpp::result<std::shared_ptr<I2SAudioOutput>, Error>;
 
   I2SAudioOutput(drivers::GpioExpander* expander,
                  std::unique_ptr<drivers::AudioDac> dac);
   ~I2SAudioOutput();
 
-  auto SetInputBuffer(MessageBufferHandle_t* in) -> void { input_buffer_ = in; }
+  auto InputMinChunkSize() const -> std::size_t override {
+    // TODO(jacqueline): work out a good value here. Maybe similar to the total
+    // DMA buffer size?
+    return 128;
+  }
 
   auto IdleTimeout() const -> TickType_t override;
   auto ProcessStreamInfo(const StreamInfo& info)

src/audio/include/stream_buffer.hpp

@@ -0,0 +1,37 @@
+#pragma once
+
+#include <cstddef>
+#include <cstdint>
+
+#include "freertos/FreeRTOS.h"
+
+#include "freertos/message_buffer.h"
+#include "span.hpp"
+
+namespace audio {
+
+class StreamBuffer {
+ public:
+  explicit StreamBuffer(std::size_t chunk_size, std::size_t buffer_size);
+  ~StreamBuffer();
+
+  auto Handle() -> MessageBufferHandle_t* { return &handle_; }
+  auto ReadBuffer() -> cpp::span<std::byte> { return input_chunk_; }
+  auto WriteBuffer() -> cpp::span<std::byte> { return output_chunk_; }
+
+  StreamBuffer(const StreamBuffer&) = delete;
+  StreamBuffer& operator=(const StreamBuffer&) = delete;
+
+ private:
+  std::byte* raw_memory_;
+  StaticMessageBuffer_t metadata_;
+  MessageBufferHandle_t handle_;
+
+  std::byte* raw_input_chunk_;
+  cpp::span<std::byte> input_chunk_;
+
+  std::byte* raw_output_chunk_;
+  cpp::span<std::byte> output_chunk_;
+};
+
+}  // namespace audio

src/audio/include/stream_info.hpp

@@ -3,6 +3,7 @@
 #include <cstdint>
 #include <optional>
 #include <string>
+#include <string_view>
 
 #include "cbor.h"
 #include "result.hpp"
@@ -19,10 +20,14 @@ class StreamInfo {
   ~StreamInfo() = default;
 
   auto Path() const -> const std::optional<std::string>& { return path_; }
+  auto Path(const std::string_view& d) -> void { path_ = d; }
+
   auto Channels() const -> const std::optional<uint8_t>& { return channels_; }
+
   auto BitsPerSample() const -> const std::optional<uint8_t>& {
     return bits_per_sample_;
   }
+
   auto SampleRate() const -> const std::optional<uint16_t>& {
     return sample_rate_;
   }

src/audio/stream_buffer.cpp

@@ -0,0 +1,26 @@
+#include "stream_buffer.hpp"
+
+namespace audio {
+
+StreamBuffer::StreamBuffer(std::size_t chunk_size, std::size_t buffer_size)
+    : raw_memory_(static_cast<std::byte*>(
+          heap_caps_malloc(buffer_size, MALLOC_CAP_SPIRAM))),
+      handle_(
+          xMessageBufferCreateStatic(buffer_size,
+                                     reinterpret_cast<uint8_t*>(raw_memory_),
+                                     &metadata_)),
+      raw_input_chunk_(static_cast<std::byte*>(
+          heap_caps_malloc(chunk_size * 1.5, MALLOC_CAP_SPIRAM))),
+      input_chunk_(raw_input_chunk_, chunk_size * 1.5),
+      raw_output_chunk_(static_cast<std::byte*>(
+          heap_caps_malloc(chunk_size, MALLOC_CAP_SPIRAM))),
+      output_chunk_(raw_output_chunk_, chunk_size) {}
+
+StreamBuffer::~StreamBuffer() {
+  vMessageBufferDelete(handle_);
+  free(raw_memory_);
+  free(raw_input_chunk_);
+  free(raw_output_chunk_);
+}
+
+}  // namespace audio

src/codecs/codec.cpp

@@ -5,7 +5,7 @@
 
 namespace codecs {
 
-auto CreateCodecForExtension(std::string extension)
+auto CreateCodecForFile(const std::string& file)
     -> cpp::result<std::unique_ptr<ICodec>, CreateCodecError> {
   return std::make_unique<MadMp3Decoder>();  // TODO.
 }

src/codecs/include/codec.hpp

@@ -63,7 +63,7 @@ class ICodec {
 
 enum CreateCodecError { UNKNOWN_EXTENSION };
 
-auto CreateCodecForFile(const std::string& extension)
+auto CreateCodecForFile(const std::string& file)
     -> cpp::result<std::unique_ptr<ICodec>, CreateCodecError>;
 
 }  // namespace codecs

src/drivers/include/storage.hpp

@@ -26,7 +26,7 @@ class SdStorage {
   };
 
   static auto create(GpioExpander* gpio)
-      -> cpp::result<std::unique_ptr<SdStorage>, Error>;
+      -> cpp::result<std::shared_ptr<SdStorage>, Error>;
 
   SdStorage(GpioExpander* gpio,
             esp_err_t (*do_transaction)(sdspi_dev_handle_t, sdmmc_command_t*),

src/drivers/storage.cpp

@@ -50,7 +50,7 @@ static esp_err_t do_transaction(sdspi_dev_handle_t handle,
 }  // namespace callback
 
 auto SdStorage::create(GpioExpander* gpio)
-    -> cpp::result<std::unique_ptr<SdStorage>, Error> {
+    -> cpp::result<std::shared_ptr<SdStorage>, Error> {
   // Acquiring the bus will also flush the mux switch change.
   gpio->set_pin(GpioExpander::SD_MUX_SWITCH, GpioExpander::SD_MUX_ESP);
 

src/main/app_console.cpp

@@ -8,11 +8,14 @@
 #include <iostream>
 #include <string>
 
+#include "audio_playback.hpp"
 #include "esp_console.h"
 
 namespace console {
 
-std::string toSdPath(std::string filepath) {
+static AppConsole* sInstance = nullptr;
+
+std::string toSdPath(const std::string& filepath) {
   return std::string(drivers::kStoragePath) + "/" + filepath;
 }
 
@@ -57,12 +60,7 @@ int CmdPlayFile(int argc, char** argv) {
     return 1;
   }
 
-  /*
   sInstance->playback_->Play(toSdPath(argv[1]));
-  if (argc == 3) {
-    sInstance->playback_->SetNextFile(toSdPath(argv[2]));
-  }
-  */
 
   return 0;
 }
@@ -125,14 +123,12 @@ void RegisterVolume() {
   esp_console_cmd_register(&cmd);
 }
 
-/*
+AppConsole::AppConsole(audio::AudioPlayback* playback) : playback_(playback) {
-AppConsole::AppConsole() {
   sInstance = this;
 }
 AppConsole::~AppConsole() {
   sInstance = nullptr;
 }
-*/
 
 auto AppConsole::RegisterExtraComponents() -> void {
   RegisterListDir();

src/main/app_console.hpp

@@ -2,6 +2,7 @@
 
 #include <memory>
 
+#include "audio_playback.hpp"
 #include "console.hpp"
 #include "storage.hpp"
 
@@ -9,8 +10,10 @@ namespace console {
 
 class AppConsole : public Console {
  public:
-  AppConsole() {}
+  explicit AppConsole(audio::AudioPlayback* playback);
-  virtual ~AppConsole() {}
+  virtual ~AppConsole();
+
+  audio::AudioPlayback* playback_;
 
  protected:
   virtual auto RegisterExtraComponents() -> void;

src/main/main.cpp

@@ -24,6 +24,7 @@
 #include "widgets/lv_label.h"
 
 #include "app_console.hpp"
+#include "audio_playback.hpp"
 #include "battery.hpp"
 #include "dac.hpp"
 #include "display.hpp"
@@ -102,7 +103,7 @@ extern "C" void app_main(void) {
     ESP_LOGE(TAG, "Failed: %d", storage_res.error());
     return;
   }
-  std::unique_ptr<drivers::SdStorage> storage = std::move(storage_res.value());
+  std::shared_ptr<drivers::SdStorage> storage = std::move(storage_res.value());
 
   LvglArgs* lvglArgs = (LvglArgs*)calloc(1, sizeof(LvglArgs));
   lvglArgs->gpio_expander = expander;
@@ -110,32 +111,20 @@ extern "C" void app_main(void) {
                                 (void*)lvglArgs, 1, sLvglStack,
                                 &sLvglTaskBuffer, 1);
 
-  /*
-  ESP_LOGI(TAG, "Init audio output (I2S)");
-  auto sink_res = drivers::I2SAudioOutput::create(expander);
-  if (sink_res.has_error()) {
-    ESP_LOGE(TAG, "Failed: %d", sink_res.error());
-    return;
-  }
-  std::unique_ptr<drivers::IAudioOutput> sink = std::move(sink_res.value());
-
   ESP_LOGI(TAG, "Init audio pipeline");
-  auto playback_res = drivers::AudioPlayback::create(std::move(sink));
+  auto playback_res = audio::AudioPlayback::create(expander, storage);
   if (playback_res.has_error()) {
     ESP_LOGE(TAG, "Failed: %d", playback_res.error());
     return;
   }
-  std::unique_ptr<drivers::AudioPlayback> playback =
+  std::shared_ptr<audio::AudioPlayback> playback =
       std::move(playback_res.value());
-  playback->SetVolume(130);
-  */
 
   ESP_LOGI(TAG, "Launch console");
-  console::AppConsole console;
+  console::AppConsole console(playback.get());
   console.Launch();
 
   while (1) {
-    // playback->ProcessEvents(5);
     vTaskDelay(pdMS_TO_TICKS(100));
   }
 }