audio_pipeline.h

Go to the documentation of this file.

#pragma once
 
#ifdef USE_ESP32
 
#include "esphome/components/audio/audio.h"
#include "esphome/components/audio/audio_reader.h"
#include "esphome/components/audio/audio_decoder.h"
#include "esphome/components/ring_buffer/ring_buffer.h"
#include "esphome/components/speaker/speaker.h"
 
#include "esphome/core/static_task.h"
 
#include "esp_err.h"
 
#include <freertos/event_groups.h>
#include <freertos/queue.h>
 
namespace esphome::speaker {
 
// Internal sink/source buffers for reader and decoder
static const size_t DEFAULT_TRANSFER_BUFFER_SIZE = 24 * 1024;
 
enum class AudioPipelineType : uint8_t {
  MEDIA,
  ANNOUNCEMENT,
};
 
enum class AudioPipelineState : uint8_t {
  STARTING_FILE,
  STARTING_URL,
  PLAYING,
  STOPPING,
  STOPPED,
  PAUSED,
  ERROR_READING,
  ERROR_DECODING,
};
 
enum class InfoErrorSource : uint8_t {
  READER = 0,
  DECODER,
};
 
enum class DecodingError : uint8_t {
  FAILED_HEADER = 0,
  INCOMPATIBLE_BITS_PER_SAMPLE,
  INCOMPATIBLE_CHANNELS,
};
 
// Used to pass information from each task.
struct InfoErrorEvent {
  InfoErrorSource source;
  optional<esp_err_t> err;
  optional<audio::AudioFileType> file_type;
  optional<audio::AudioStreamInfo> audio_stream_info;
  optional<DecodingError> decoding_err;
};
 
class AudioPipeline {
 public:
  AudioPipeline(speaker::Speaker *speaker, size_t buffer_size, bool task_stack_in_psram, std::string base_name,
                UBaseType_t priority);
 
  void start_url(const std::string &uri);
 
  void start_file(audio::AudioFile *audio_file);
 
  esp_err_t stop();
 
  AudioPipelineState process_state();
 
  void suspend_tasks();
  void resume_tasks();
 
  uint32_t get_playback_ms() { return this->playback_ms_; }
 
  void set_pause_state(bool pause_state);
 
 protected:
  esp_err_t allocate_communications_();
 
  esp_err_t start_tasks_();
 
  std::string base_name_;
  UBaseType_t priority_;
 
  uint32_t playback_ms_{0};
 
  bool hard_stop_{false};
  bool is_playing_{false};
  bool is_finishing_{false};
  bool pause_state_{false};
  bool task_stack_in_psram_;
 
  // Pending file start state used to ensure the pipeline fully stops before attempting to start the next file
  bool pending_url_{false};
  bool pending_file_{false};
 
  speaker::Speaker *speaker_{nullptr};
 
  std::string current_uri_{};
  audio::AudioFile *current_audio_file_{nullptr};
 
  audio::AudioFileType current_audio_file_type_;
  audio::AudioStreamInfo current_audio_stream_info_;
 
  size_t buffer_size_;           // Ring buffer between reader and decoder
  size_t transfer_buffer_size_;  // Internal source/sink buffers for the audio reader and decoder
 
  std::weak_ptr<ring_buffer::RingBuffer> raw_file_ring_buffer_;
 
  // Handles basic control/state of the three tasks
  EventGroupHandle_t event_group_{nullptr};
 
  // Receives detailed info (file type, stream info, resampling info) or specific errors from the three tasks
  QueueHandle_t info_error_queue_{nullptr};
 
  // Handles reading the media file from flash or a url
  static void read_task(void *params);
  StaticTask read_task_;
 
  // Decodes the media file into PCM audio
  static void decode_task(void *params);
  StaticTask decode_task_;
};
 
}  // namespace esphome::speaker
 
#endif