Flesh out opus decoder. it doesn't work! i hate opus.

src/audio/audio_task.cpp

@@ -125,7 +125,7 @@ AudioTask::AudioTask(IAudioSource* source, IAudioSink* sink)
       has_begun_decoding_(false),
       current_input_format_(),
       current_output_format_(),
-      codec_buffer_(new RawStream(kSampleBufferSize)) {}
+      codec_buffer_(new RawStream(kSampleBufferSize, MALLOC_CAP_8BIT)) {}
 
 void AudioTask::Main() {
   for (;;) {

src/audio/fatfs_audio_input.cpp

@@ -313,8 +313,8 @@ auto FatfsAudioInput::ContainerToStreamType(database::Encoding enc)
       return codecs::StreamType::kPcm;
     case database::Encoding::kFlac:
       return codecs::StreamType::kFlac;
-    case database::Encoding::kOgg:  // Misnamed; this is Ogg Vorbis.
-      return codecs::StreamType::kVorbis;
+    case database::Encoding::kOpus:
+      return codecs::StreamType::kOpus;
     case database::Encoding::kUnsupported:
     default:
       return {};

src/audio/i2s_audio_output.cpp

@@ -120,8 +120,7 @@ auto I2SAudioOutput::PrepareFormat(const StreamInfo::Pcm& orig)
   return StreamInfo::Pcm{
       .channels = std::min<uint8_t>(orig.channels, 2),
       .bits_per_sample = std::clamp<uint8_t>(orig.bits_per_sample, 16, 32),
-      .sample_rate = 44100,
-      //.sample_rate = std::clamp<uint32_t>(orig.sample_rate, 8000, 96000),
+      .sample_rate = std::clamp<uint32_t>(orig.sample_rate, 8000, 96000),
   };
 }
 

src/codecs/CMakeLists.txt

@@ -3,16 +3,18 @@
 # SPDX-License-Identifier: GPL-3.0-only
 
 idf_component_register(
-  SRCS "codec.cpp" "mad.cpp" "foxenflac.cpp"
+  SRCS "codec.cpp" "mad.cpp" "foxenflac.cpp" "opus.cpp"
   INCLUDE_DIRS "include"
   REQUIRES "result" "span" "libmad" "libfoxenflac")
 
 target_compile_options("${COMPONENT_LIB}" PRIVATE ${EXTRA_WARNINGS})
 
 set(OPUS_FIXED_POINT ON)
-set(OPUS_ENABLE_FLOAT_API ON)
+set(OPUS_ENABLE_FLOAT_API OFF)
 set(OPUS_INSTALL_PKG_CONFIG_MODULE OFF)
 set(OPUS_INSTALL_CMAKE_CONFIG_MODULE OFF)
+set(OPUS_BUILD_TESTING OFF)
+set(OPUS_BUILD_SHARED_LIBS OFF)
 
 set(CMAKE_POLICY_DEFAULT_CMP0077 NEW)
 

src/codecs/codec.cpp

@@ -10,6 +10,7 @@
 #include <optional>
 
 #include "foxenflac.hpp"
+#include "opus.hpp"
 #include "mad.hpp"
 #include "types.hpp"
 
@@ -21,6 +22,8 @@ auto CreateCodecForType(StreamType type) -> std::optional<ICodec*> {
       return new MadMp3Decoder();
     case StreamType::kFlac:
       return new FoxenFlacDecoder();
+    case StreamType::kOpus:
+      return new XiphOpusDecoder();
     default:
       return {};
   }

src/codecs/include/opus.hpp

@@ -14,6 +14,7 @@
 #include <utility>
 
 #include "opus.h"
+#include "sample.hpp"
 #include "span.hpp"
 
 #include "codec.hpp"
@@ -36,7 +37,7 @@ class XiphOpusDecoder : public ICodec {
    * Writes samples for the current frame.
    */
   auto ContinueStream(cpp::span<const std::byte> input,
-                      cpp::span<std::byte> output)
+                      cpp::span<sample::Sample> output)
       -> Result<OutputInfo> override;
 
   auto SeekStream(cpp::span<const std::byte> input, std::size_t target_sample)
@@ -44,6 +45,10 @@ class XiphOpusDecoder : public ICodec {
 
  private:
   OpusDecoder* opus_;
-  float *sample_buffer_;
-  std::size_t sample_buffer_len_;
+  cpp::span<int16_t> sample_buffer_;
+  int32_t pos_in_buffer_;
+  int32_t samples_in_buffer_;
+
+};
+
 }  // namespace codecs

src/codecs/include/types.hpp

@@ -13,8 +13,8 @@ namespace codecs {
 enum class StreamType {
   kMp3,
   kPcm,
-  kVorbis,
   kFlac,
+  kOpus,
 };
 
 }  // namespace codecs

src/codecs/opus.cpp

@@ -5,6 +5,7 @@
  */
 
 #include "opus.hpp"
+
 #include <stdint.h>
 #include <sys/_stdint.h>
 
@@ -12,36 +13,101 @@
 #include <cstring>
 #include <optional>
 
+#include "esp_heap_caps.h"
 #include "mad.h"
 
 #include "codec.hpp"
 #include "esp_log.h"
 #include "opus.h"
+#include "opus_types.h"
 #include "result.hpp"
+#include "sample.hpp"
 #include "types.hpp"
 
 namespace codecs {
 
-  static constexpr std::size_t kSampleBufferSize = 5760 * sizeof(float);
+static constexpr char kTag[] = "opus";
+
+// "If this is less than the maximum packet duration (120ms; 5760 for 48kHz),
+// this function will not be capable of decoding some packets"
+static constexpr size_t kSampleBufferSize = 5760;
 
 XiphOpusDecoder::XiphOpusDecoder() {
   int err;
   opus_ = opus_decoder_create(48000, 2, &err);
   assert(err == OPUS_OK);
+
+  pos_in_buffer_ = 0;
+  sample_buffer_ = {reinterpret_cast<opus_int16*>(
+                        heap_caps_calloc(kSampleBufferSize, sizeof(opus_int16),
+                                         MALLOC_CAP_8BIT | MALLOC_CAP_SPIRAM)),
+                    kSampleBufferSize};
 }
 XiphOpusDecoder::~XiphOpusDecoder() {
   opus_decoder_destroy(opus_);
+  heap_caps_free(sample_buffer_.data());
 }
 
 auto XiphOpusDecoder::BeginStream(const cpp::span<const std::byte> input)
-    -> Result<OutputFormat> {}
+    -> Result<OutputFormat> {
+  return {0, OutputFormat{
+                 .num_channels = 2,
+                 .sample_rate_hz = 48000,
+             }};
+}
+
+auto read_uint32(cpp::span<const std::byte> src) -> uint32_t {
+  return static_cast<uint32_t>(src[0] << 24) |
+         static_cast<uint32_t>(src[1] << 16) |
+         static_cast<uint32_t>(src[2] << 8) |
+         static_cast<uint32_t>(src[3] << 0);
+}
 
 auto XiphOpusDecoder::ContinueStream(cpp::span<const std::byte> input,
-                                     cpp::span<std::byte> output)
+                                     cpp::span<sample::Sample> output)
     -> Result<OutputInfo> {
-  int samples_decoded = opus_decode_float(
-      opus_, reinterpret_cast<const unsigned char*>(input.data()),
-      input.size_bytes(), sample_buffer_, sample_buffer_len_, 0);
+  size_t bytes_used = 0;
+  if (pos_in_buffer_ >= samples_in_buffer_) {
+    ESP_LOGI(kTag, "sample buffer is empty. parsing more.");
+    if (input.size() < 4) {
+      return {0, cpp::fail(Error::kOutOfInput)};
+    }
+    uint32_t payload_length = read_uint32(input);
+    ESP_LOGI(kTag, "payload length is %lu", payload_length);
+
+    if (input.size() - 4 < payload_length) {
+      ESP_LOGI(kTag, "input too small for payload");
+      return {0, cpp::fail(Error::kOutOfInput)};
+    }
+
+    // Next 4 bytes are the 'final range'.
+    // uint32_t enc_final_range = read_uint32(input.subspan(4));
+
+    bytes_used = payload_length + 8;
+
+    pos_in_buffer_ = 0;
+    samples_in_buffer_ = opus_decode(
+        opus_, reinterpret_cast<const unsigned char*>(input.data() + 8),
+        payload_length, sample_buffer_.data(), sample_buffer_.size(), 0);
+
+    if (samples_in_buffer_ < 0) {
+      ESP_LOGE(kTag, "error decoding stream");
+      return {bytes_used, cpp::fail(Error::kMalformedData)};
+    }
+  }
+
+  size_t samples_written = 0;
+  while (pos_in_buffer_ < samples_in_buffer_ &&
+         samples_written < output.size()) {
+    output[samples_written++] =
+        sample::FromSigned(sample_buffer_[pos_in_buffer_++], 16);
+  }
+
+  return {bytes_used,
+          OutputInfo{
+              .samples_written = samples_written,
+              .is_finished_writing = pos_in_buffer_ >= samples_in_buffer_,
+          }};
 }
 
 auto XiphOpusDecoder::SeekStream(cpp::span<const std::byte> input,

src/database/include/track.hpp

@@ -43,6 +43,7 @@ enum class Encoding {
   kWav = 2,
   kOgg = 3,
   kFlac = 4,
+  kOpus = 5,
 };
 
 enum class Tag {

src/database/tag_parser.cpp

@@ -153,6 +153,9 @@ auto TagParserImpl::ReadAndParseTags(const std::string& path, TrackTags* out)
     case Fwav:
       out->encoding(Encoding::kWav);
       break;
+    case Fopus:
+      out->encoding(Encoding::kOpus);
+      break;
     default:
       out->encoding(Encoding::kUnsupported);
   }