Do some prep cleanup for multiple filetypes

src/audio/audio_decoder.cpp

@@ -50,18 +50,20 @@ auto AudioDecoder::ProcessStreamInfo(const StreamInfo& info) -> bool {
   // Reuse the existing codec if we can. This will help with gapless playback,
   // since we can potentially just continue to decode as we were before,
   // without any setup overhead.
-  if (current_codec_ != nullptr &&
-      current_codec_->CanHandleType(encoded.type)) {
-    current_codec_->ResetForNewStream();
-    ESP_LOGI(kTag, "reusing existing decoder");
-    return true;
+  if (current_codec_ != nullptr && current_input_format_) {
+    auto cur_encoding = std::get<StreamInfo::Encoded>(*current_input_format_);
+    if (cur_encoding.type == encoded.type) {
+      ESP_LOGI(kTag, "reusing existing decoder");
+      current_input_format_ = info.format;
+      return true;
+    }
   }
+  current_input_format_ = info.format;
 
-  // TODO: use audio type from stream
+  ESP_LOGI(kTag, "creating new decoder");
   auto result = codecs::CreateCodecForType(encoded.type);
   if (result.has_value()) {
-    ESP_LOGI(kTag, "creating new decoder");
-    current_codec_ = std::move(result.value());
+    current_codec_.reset(result.value());
   } else {
     ESP_LOGE(kTag, "no codec for this file");
     return false;
@@ -88,9 +90,7 @@ auto AudioDecoder::Process(const std::vector<InputStream>& inputs,
   if (!current_input_format_ || *current_input_format_ != info.format) {
     // The input stream has changed! Immediately throw everything away and
     // start from scratch.
-    current_input_format_ = info.format;
     has_samples_to_send_ = false;
-
     ProcessStreamInfo(info);
   }
 

src/audio/fatfs_audio_input.cpp

@@ -5,6 +5,7 @@
  */
 
 #include "fatfs_audio_input.hpp"
+#include <stdint.h>
 
 #include <algorithm>
 #include <cstdint>
@@ -23,10 +24,12 @@
 
 #include "audio_element.hpp"
 #include "chunk.hpp"
+#include "song.hpp"
 #include "stream_buffer.hpp"
 #include "stream_event.hpp"
 #include "stream_info.hpp"
 #include "stream_message.hpp"
+#include "tag_parser.hpp"
 #include "types.hpp"
 
 static const char* kTag = "SRC";
@@ -34,7 +37,11 @@ static const char* kTag = "SRC";
 namespace audio {
 
 FatfsAudioInput::FatfsAudioInput()
-    : IAudioElement(), current_file_(), is_file_open_(false) {}
+    : IAudioElement(),
+      current_file_(),
+      is_file_open_(false),
+      current_container_(),
+      current_format_() {}
 
 FatfsAudioInput::~FatfsAudioInput() {}
 
@@ -44,6 +51,36 @@ auto FatfsAudioInput::OpenFile(const std::string& path) -> bool {
     is_file_open_ = false;
   }
   ESP_LOGI(kTag, "opening file %s", path.c_str());
+
+  database::TagParserImpl tag_parser;
+  database::SongTags tags;
+  if (!tag_parser.ReadAndParseTags(path, &tags)) {
+    ESP_LOGE(kTag, "failed to read tags");
+    return false;
+  }
+
+  auto stream_type = ContainerToStreamType(tags.encoding);
+  if (!stream_type.has_value()) {
+    return false;
+  }
+
+  current_container_ = tags.encoding;
+
+  if (*stream_type == codecs::StreamType::kPcm && tags.channels &&
+      tags.bits_per_sample && tags.channels) {
+    // WAV files are a special case bc they contain raw PCM streams. These don't
+    // need decoding, but we *do* need to parse the PCM format from the header.
+    // TODO(jacqueline): Maybe we should have a decoder for this just to deal
+    // with endianness differences?
+    current_format_ = StreamInfo::Pcm{
+        .channels = static_cast<uint8_t>(*tags.channels),
+        .bits_per_sample = static_cast<uint8_t>(*tags.bits_per_sample),
+        .sample_rate = static_cast<uint32_t>(*tags.sample_rate),
+    };
+  } else {
+    current_format_ = StreamInfo::Encoded{*stream_type};
+  }
+
   FRESULT res = f_open(&current_file_, path.c_str(), FA_READ);
   if (res != FR_OK) {
     ESP_LOGE(kTag, "failed to open file! res: %i", res);
@@ -61,13 +98,10 @@ auto FatfsAudioInput::NeedsToProcess() const -> bool {
 auto FatfsAudioInput::Process(const std::vector<InputStream>& inputs,
                               OutputStream* output) -> void {
   if (!is_file_open_) {
-    // TODO(jacqueline): should we clear the stream format?
-    // output->prepare({});
     return;
   }
 
-  StreamInfo::Format format = StreamInfo::Encoded{codecs::STREAM_MP3};
-  if (!output->prepare(format)) {
+  if (!output->prepare(*current_format_)) {
     return;
   }
 
@@ -91,12 +125,31 @@ auto FatfsAudioInput::Process(const std::vector<InputStream>& inputs,
     f_close(&current_file_);
     is_file_open_ = false;
 
-    // TODO(jacqueline): MP3 only
-    std::fill_n(output->data().begin(), 8, std::byte(0));
-    output->add(8);
+    // HACK: libmad requires an 8 byte padding at the end of each file.
+    if (current_container_ == database::Encoding::kMp3) {
+      std::fill_n(output->data().begin(), 8, std::byte(0));
+      output->add(8);
+    }
 
     events::Dispatch<InputFileFinished, AudioState>({});
   }
 }
 
+auto FatfsAudioInput::ContainerToStreamType(database::Encoding enc)
+    -> std::optional<codecs::StreamType> {
+  switch (enc) {
+    case database::Encoding::kMp3:
+      return codecs::StreamType::kMp3;
+    case database::Encoding::kWav:
+      return codecs::StreamType::kPcm;
+    case database::Encoding::kFlac:
+      return codecs::StreamType::kFlac;
+    case database::Encoding::kOgg:
+      return codecs::StreamType::kOgg;
+    case database::Encoding::kUnsupported:
+    default:
+      return {};
+  }
+}
+
 }  // namespace audio

src/audio/include/fatfs_audio_input.hpp

@@ -18,11 +18,13 @@
 #include "ff.h"
 #include "freertos/message_buffer.h"
 #include "freertos/queue.h"
+#include "song.hpp"
 #include "span.hpp"
 
 #include "audio_element.hpp"
 #include "stream_buffer.hpp"
 #include "stream_info.hpp"
+#include "types.hpp"
 
 namespace audio {
 
@@ -42,8 +44,14 @@ class FatfsAudioInput : public IAudioElement {
   FatfsAudioInput& operator=(const FatfsAudioInput&) = delete;
 
  private:
+  auto ContainerToStreamType(database::Encoding)
+      -> std::optional<codecs::StreamType>;
+
   FIL current_file_;
   bool is_file_open_;
+
+  std::optional<database::Encoding> current_container_;
+  std::optional<StreamInfo::Format> current_format_;
 };
 
 }  // namespace audio

src/codecs/codec.cpp

@@ -7,13 +7,19 @@
 #include "codec.hpp"
 
 #include <memory>
+#include <optional>
 #include "mad.hpp"
+#include "types.hpp"
 
 namespace codecs {
 
-auto CreateCodecForType(StreamType type)
-    -> cpp::result<std::unique_ptr<ICodec>, CreateCodecError> {
-  return std::make_unique<MadMp3Decoder>();  // TODO.
+auto CreateCodecForType(StreamType type) -> std::optional<ICodec*> {
+  switch (type) {
+    case StreamType::kMp3:
+      return new MadMp3Decoder();
+    default:
+      return {};
+  }
 }
 
 }  // namespace codecs

src/codecs/include/codec.hpp

@@ -25,8 +25,6 @@ class ICodec {
  public:
   virtual ~ICodec() {}
 
-  virtual auto CanHandleType(StreamType type) -> bool = 0;
-
   struct OutputFormat {
     uint8_t num_channels;
     uint8_t bits_per_sample;
@@ -37,8 +35,6 @@ class ICodec {
 
   enum ProcessingError { MALFORMED_DATA };
 
-  virtual auto ResetForNewStream() -> void = 0;
-
   virtual auto SetInput(cpp::span<const std::byte> input) -> void = 0;
 
   /*
@@ -69,9 +65,6 @@ class ICodec {
       -> std::pair<std::size_t, bool> = 0;
 };
 
-enum CreateCodecError { UNKNOWN_EXTENSION };
-
-auto CreateCodecForType(StreamType type)
-    -> cpp::result<std::unique_ptr<ICodec>, CreateCodecError>;
+auto CreateCodecForType(StreamType type) -> std::optional<ICodec*>;
 
 }  // namespace codecs

src/codecs/include/mad.hpp

@@ -24,9 +24,7 @@ class MadMp3Decoder : public ICodec {
   MadMp3Decoder();
   ~MadMp3Decoder();
 
-  auto CanHandleType(StreamType type) -> bool override;
   auto GetOutputFormat() -> std::optional<OutputFormat> override;
-  auto ResetForNewStream() -> void override;
   auto SetInput(cpp::span<const std::byte> input) -> void override;
   auto GetInputPosition() -> std::size_t override;
   auto ProcessNextFrame() -> cpp::result<bool, ProcessingError> override;

src/codecs/include/types.hpp

@@ -10,9 +10,11 @@
 
 namespace codecs {
 
-enum StreamType {
-  STREAM_MP3,
+enum class StreamType {
+  kMp3,
+  kPcm,
+  kOgg,
+  kFlac,
 };
 
-auto GetStreamTypeFromFilename(std::string filename);
 }  // namespace codecs

src/codecs/mad.cpp

@@ -42,10 +42,6 @@ MadMp3Decoder::~MadMp3Decoder() {
   mad_synth_finish(&synth_);
 }
 
-auto MadMp3Decoder::CanHandleType(StreamType type) -> bool {
-  return type == STREAM_MP3;
-}
-
 auto MadMp3Decoder::GetOutputFormat() -> std::optional<OutputFormat> {
   if (synth_.pcm.channels == 0 || synth_.pcm.samplerate == 0) {
     return {};
@@ -57,8 +53,6 @@ auto MadMp3Decoder::GetOutputFormat() -> std::optional<OutputFormat> {
   });
 }
 
-auto MadMp3Decoder::ResetForNewStream() -> void {}
-
 auto MadMp3Decoder::SetInput(cpp::span<const std::byte> input) -> void {
   mad_stream_buffer(&stream_,
                     reinterpret_cast<const unsigned char*>(input.data()),
@@ -115,8 +109,6 @@ auto MadMp3Decoder::WriteOutputSamples(cpp::span<std::byte> output)
     }
 
     for (int channel = 0; channel < synth_.pcm.channels; channel++) {
-      // TODO(jacqueline): output 24 bit samples when (if?) we have a downmix
-      // step in the pipeline.
       uint32_t sample_24 =
           scaleToBits(synth_.pcm.samples[channel][current_sample_], 24);
       output[output_byte++] = static_cast<std::byte>((sample_24 >> 16) & 0xFF);

src/database/include/song.hpp

@@ -36,6 +36,9 @@ typedef uint32_t SongId;
 enum class Encoding {
   kUnsupported = 0,
   kMp3 = 1,
+  kWav = 2,
+  kOgg = 3,
+  kFlac = 4,
 };
 
 /*
@@ -53,6 +56,10 @@ struct SongTags {
   std::optional<std::string> artist;
   std::optional<std::string> album;
 
+  std::optional<int> channels;
+  std::optional<int> sample_rate;
+  std::optional<int> bits_per_sample;
+
   /*
    * Returns a hash of the 'identifying' tags of this song. That is, a hash that
    * can be used to determine if one song is likely the same as another, across

src/database/tag_parser.cpp

@@ -107,6 +107,16 @@ auto TagParserImpl::ReadAndParseTags(const std::string& path, SongTags* out)
       out->encoding = Encoding::kUnsupported;
   }
 
+  if (ctx.channels > 0) {
+    out->channels = ctx.channels;
+  }
+  if (ctx.samplerate > 0) {
+    out->sample_rate = ctx.samplerate;
+  }
+  if (ctx.bitrate > 0) {
+    out->bits_per_sample = ctx.bitrate;
+  }
+
   return true;
 }