[audio_file] Keep audio files in flash instead of building them in setup() (#20072)

esphome/components/audio/audio_reader.cpp

@@ -72,7 +72,7 @@ esp_err_t AudioReader::add_sink(const std::weak_ptr<ring_buffer::RingBuffer> &ou
   return ESP_ERR_INVALID_STATE;
 }
 
-esp_err_t AudioReader::start(AudioFile *audio_file, AudioFileType &file_type) {
+esp_err_t AudioReader::start(const AudioFile *audio_file, AudioFileType &file_type) {
   file_type = AudioFileType::NONE;
 
   this->current_audio_file_ = audio_file;

esphome/components/audio/audio_reader.h

@@ -47,7 +47,7 @@ class AudioReader {
   /// @param audio_file AudioFile struct containing the file.
   /// @param file_type AudioFileType variable passed-by-reference indicating the type of file being read.
   /// @return ESP_OK
-  esp_err_t start(AudioFile *audio_file, AudioFileType &file_type);
+  esp_err_t start(const AudioFile *audio_file, AudioFileType &file_type);
 
   /// @brief Reads new file data from the source and sends to the ring buffer sink.
   /// @return AudioReaderState
@@ -69,7 +69,7 @@ class AudioReader {
 
   esp_http_client_handle_t client_{nullptr};
 
-  AudioFile *current_audio_file_{nullptr};
+  const AudioFile *current_audio_file_{nullptr};
   AudioFileType audio_file_type_{AudioFileType::NONE};
   const uint8_t *file_current_{nullptr};
 };

esphome/components/audio_file/__init__.py

@@ -19,7 +19,7 @@ from esphome.const import (
     CONF_URL,
 )
 from esphome.core import CORE, ID, HexInt
-from esphome.cpp_generator import MockObj
+from esphome.cpp_generator import MockObj, ProgmemAssignmentExpression
 from esphome.external_files import download_web_files_in_config
 from esphome.types import ConfigType
 
@@ -151,11 +151,14 @@ TYPED_FILE_SCHEMA = cv.typed_schema(
 )
 
 
+CONF_FILE_DATA_ID = "file_data_id"
+
 MEDIA_FILE_TYPE_SCHEMA = cv.Schema(
     {
         cv.Required(CONF_ID): cv.declare_id(audio.AudioFile),
         cv.Required(CONF_FILE): _file_schema,
         cv.GenerateID(CONF_RAW_DATA_ID): cv.declare_id(cg.uint8),
+        cv.GenerateID(CONF_FILE_DATA_ID): cv.declare_id(audio.AudioFile),
     }
 )
 
@@ -217,9 +220,9 @@ def audio_files_schema() -> cv.All:
 
 
 def generate_audio_file_code(file_config: ConfigType) -> MockObj:
-    """Generate the progmem data, AudioFile struct, and Pvariable for one file.
+    """Generate the progmem data and a flash AudioFile for one file.
 
-    Returns the created Pvariable. Caller is responsible for any further
+    Returns a const pointer to the AudioFile. Caller is responsible for any further
     registration (the audio_file component additionally registers each file in
     its named C++ registry; other consumers may skip that).
     """
@@ -230,17 +233,33 @@ def generate_audio_file_code(file_config: ConfigType) -> MockObj:
     else:
         data, media_file_type = read_audio_file_and_type(file_config)
 
-    rhs = [HexInt(x) for x in data]
-    prog_arr = cg.progmem_array(file_config[CONF_RAW_DATA_ID], rhs)
-
-    media_files_struct = cg.StructInitializer(
+    # Global constants so the AudioFile lives in flash; the id stays a plain pointer
+    # because actions render id arguments as ``::<id>``.
+    data_id = file_config[CONF_RAW_DATA_ID]
+    cg.add_global(
+        ProgmemAssignmentExpression(
+            data_id.type, data_id, cg.safe_exp([HexInt(x) for x in data])
+        )
+    )
+    media_file = cg.StructInitializer(
         audio.AudioFile,
-        ("data", prog_arr),
-        ("length", len(rhs)),
+        ("data", MockObj(data_id, ".")),
+        ("length", len(data)),
         ("file_type", media_file_type),
     )
-
-    return cg.new_Pvariable(file_config[CONF_ID], media_files_struct)
+    file_var_id = file_config[CONF_ID]
+    storage = file_config[CONF_FILE_DATA_ID]
+    cg.add_global(
+        cg.RawStatement(f"static constexpr {storage.type} {storage} = {media_file};")
+    )
+    cg.add_global(
+        cg.RawStatement(
+            f"static const {storage.type} *const {file_var_id} = &{storage};"
+        )
+    )
+    var = MockObj(file_var_id, "->")
+    CORE.register_variable(file_var_id, var)
+    return var
 
 
 CONFIG_SCHEMA = cv.All(

esphome/components/audio_file/audio_file.h

@@ -10,14 +10,14 @@
 namespace esphome::audio_file {
 
 struct NamedAudioFile {
-  audio::AudioFile *file;
+  const audio::AudioFile *file;
   const char *file_id;
 };
 
 inline StaticVector<NamedAudioFile, AUDIO_FILE_MAX_FILES>
     named_audio_files;  // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
 
-inline void add_named_audio_file(audio::AudioFile *file, const char *file_id) {
+inline void add_named_audio_file(const audio::AudioFile *file, const char *file_id) {
   named_audio_files.push_back({file, file_id});
 }
 

esphome/components/audio_file/media_source/audio_file_media_source.h

@@ -46,7 +46,7 @@ class AudioFileMediaSource final : public Component,
  protected:
   std::unique_ptr<micro_decoder::DecoderSource> decoder_;
   audio::AudioStreamInfo stream_info_;
-  audio::AudioFile *current_file_{nullptr};
+  const audio::AudioFile *current_file_{nullptr};
 
   // Written from the main loop in handle_command(), read from the decoder task in
   // on_audio_write(). Must be atomic to avoid a data race.

esphome/components/speaker/media_player/__init__.py

@@ -251,7 +251,7 @@ automation.register_apply_action(
             # so a trigger arg with the same name cannot shadow it.
             (
                 CONF_MEDIA_FILE,
-                audio.AudioFile.operator("ptr"),
+                audio.AudioFile.operator("const").operator("ptr"),
                 lambda _, value: f"::{value}",
             ),
             (CONF_ANNOUNCEMENT, cg.bool_),

esphome/components/speaker/media_player/audio_pipeline.cpp

@@ -66,7 +66,7 @@ void AudioPipeline::start_url(const std::string &uri) {
   this->pending_url_ = true;
 }
 
-void AudioPipeline::start_file(audio::AudioFile *audio_file) {
+void AudioPipeline::start_file(const audio::AudioFile *audio_file) {
   if (this->is_playing_) {
     xEventGroupSetBits(this->event_group_, PIPELINE_COMMAND_STOP);
   }

esphome/components/speaker/media_player/audio_pipeline.h

@@ -74,7 +74,7 @@ class AudioPipeline final {
   /// @brief Starts an audio pipeline given a AudioFile pointer
   /// @param audio_file pointer to an AudioFile object
   /// @return ESP_OK if successful or an appropriate error if not
-  void start_file(audio::AudioFile *audio_file);
+  void start_file(const audio::AudioFile *audio_file);
 
   /// @brief Stops the pipeline. Sends a stop signal to each task (if running) and clears the ring buffers.
   /// @return ESP_OK if successful or ESP_ERR_TIMEOUT if the tasks did not indicate they stopped
@@ -121,7 +121,7 @@ class AudioPipeline final {
   speaker::Speaker *speaker_{nullptr};
 
   std::string current_uri_{};
-  audio::AudioFile *current_audio_file_{nullptr};
+  const audio::AudioFile *current_audio_file_{nullptr};
 
   audio::AudioFileType current_audio_file_type_;
   audio::AudioStreamInfo current_audio_stream_info_;

esphome/components/speaker/media_player/speaker_media_player.cpp

@@ -493,7 +493,7 @@ void SpeakerMediaPlayer::loop() {
 #endif
 }
 
-void SpeakerMediaPlayer::play_file(audio::AudioFile *media_file, bool announcement, bool enqueue) {
+void SpeakerMediaPlayer::play_file(const audio::AudioFile *media_file, bool announcement, bool enqueue) {
   if (!this->is_ready()) {
     // Ignore any commands sent before the media player is setup
     return;

esphome/components/speaker/media_player/speaker_media_player.h

@@ -28,13 +28,13 @@ struct MediaCallCommand {
   optional<float> volume;
   optional<bool> announce;
   optional<std::string *> url;  // Must be manually deleted after receiving this struct from a queue
-  optional<audio::AudioFile *> file;
+  optional<const audio::AudioFile *> file;
   optional<bool> enqueue;
 };
 
 struct PlaylistItem {
   optional<std::string> url;
-  optional<audio::AudioFile *> file;
+  optional<const audio::AudioFile *> file;
 };
 
 struct VolumeRestoreState {
@@ -87,7 +87,7 @@ class SpeakerMediaPlayer final : public Component,
   Trigger<> *get_unmute_trigger() { return &this->unmute_trigger_; }
   Trigger<float> *get_volume_trigger() { return &this->volume_trigger_; }
 
-  void play_file(audio::AudioFile *media_file, bool announcement, bool enqueue);
+  void play_file(const audio::AudioFile *media_file, bool announcement, bool enqueue);
 
   void set_playlist_delay_ms(AudioPipelineType pipeline_type, uint32_t delay_ms);
 

tests/component_tests/audio_file/config/audio_file.yaml

@@ -0,0 +1,14 @@
+---
+esphome:
+  name: test
+
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf
+
+audio_file:
+  - id: chime
+    file:
+      type: local
+      path: ../../../components/audio_file/test.wav

tests/component_tests/audio_file/test_audio_file.py

@@ -0,0 +1,22 @@
+"""Tests for the audio_file component codegen."""
+
+from collections.abc import Callable
+from pathlib import Path
+
+
+def test_audio_file_is_constant_in_flash(
+    generate_main: Callable[[str | Path], str],
+    component_config_path: Callable[[str], Path],
+) -> None:
+    """The AudioFile is a global constant behind a const pointer, not a placement new."""
+    main_cpp = generate_main(component_config_path("audio_file.yaml"))
+
+    assert (
+        "static constexpr audio::AudioFile audio_audiofile_id = audio::AudioFile{"
+        in main_cpp
+    )
+    assert (
+        "static const audio::AudioFile *const chime = &audio_audiofile_id;" in main_cpp
+    )
+    assert "new(chime)" not in main_cpp
+    assert 'audio_file::add_named_audio_file(chime, "chime");' in main_cpp