Files
pmomusic/pmoaudio/src/nodes/file_source.rs
Eric Coissac 8924552696 Add StreamType for multi-client support and improved UPnP control
- Introduce `Streamtype` enum (Continuous vs Finite) to distinguish radio streams from finite tracks
- Enrich `TrackBoundary` sync marker with stream type for proper pause behavior per mode (silence vs backpressure)
- Update all sources and sinks to pass `StreamType` when creating track boundaries
  - Radio Paradise, HTTP source → Continuous (infinite)
- Improve UPnP control architecture: pause sends silence for radio, blocks pipeline via backpressure for tracks
- Prepare groundwork for multi-client DSP architecture with shared source and per-DSP pipelines
2026-04-04 14:37:32 +02:00

542 lines
20 KiB
Rust
Executable File

use crate::{
nodes::{AudioError, TypedAudioNode, DEFAULT_CHUNK_DURATION_MS},
pipeline::{send_to_children, AudioPipelineNode, Node, NodeLogic},
type_constraints::TypeRequirement,
AudioChunk, AudioChunkData, AudioSegment, I24, StreamType,
};
use pmoflac::{decode_audio_stream, AudioFileMetadata, StreamInfo};
use pmometadata::{MemoryTrackMetadata, TrackMetadata};
use std::{path::PathBuf, sync::Arc, time::Duration};
use tokio::{fs::File, io::AsyncReadExt, sync::mpsc};
use tokio_util::sync::CancellationToken;
use tracing;
// ═══════════════════════════════════════════════════════════════════════════
// NOUVELLE ARCHITECTURE - FileSourceLogic
// ═══════════════════════════════════════════════════════════════════════════
/// Logique pure de lecture de fichier audio
///
/// Contient seulement la logique de décodage et d'envoi des segments,
/// sans la plomberie d'orchestration (gérée par Node<FileSourceLogic>).
pub struct FileSourceLogic {
path: PathBuf,
chunk_frames: usize,
}
impl FileSourceLogic {
pub fn new<P: Into<PathBuf>>(path: P, chunk_frames: usize) -> Self {
Self {
path: path.into(),
chunk_frames,
}
}
}
#[async_trait::async_trait]
impl NodeLogic for FileSourceLogic {
async fn process(
&mut self,
_input: Option<mpsc::Receiver<Arc<AudioSegment>>>,
output: Vec<mpsc::Sender<Arc<AudioSegment>>>,
stop_token: CancellationToken,
) -> Result<(), AudioError> {
tracing::debug!(
"FileSourceLogic::process started, path={:?}, {} children",
self.path,
output.len()
);
// Ouvrir le fichier
let file = File::open(&self.path)
.await
.map_err(|e| AudioError::IoError(format!("Failed to open {:?}: {}", self.path, e)))?;
// Décoder le flux audio
let mut stream = decode_audio_stream(file)
.await
.map_err(|e| AudioError::ProcessingError(format!("Decode error: {}", e)))?;
let stream_info = stream.info().clone();
validate_stream(&stream_info)?;
// Calculer la taille des chunks si non spécifiée (0 = auto)
let chunk_frames = if self.chunk_frames == 0 {
let frames =
(stream_info.sample_rate as f64 * DEFAULT_CHUNK_DURATION_MS / 1000.0) as usize;
frames.next_power_of_two().max(256)
} else {
self.chunk_frames.max(1)
};
// Émettre TopZeroSync
let top_zero = AudioSegment::new_top_zero_sync();
send_to_children(std::any::type_name::<Self>(), &output, top_zero).await?;
// Extraire et émettre les métadonnées du fichier
if let Ok(file_metadata) = AudioFileMetadata::from_file(&self.path) {
let mut metadata = MemoryTrackMetadata::new();
if let Some(title) = file_metadata.title {
let _ = metadata.set_title(Some(title)).await;
}
if let Some(artist) = file_metadata.artist {
let _ = metadata.set_artist(Some(artist)).await;
}
if let Some(album) = file_metadata.album {
let _ = metadata.set_album(Some(album)).await;
}
if let Some(year) = file_metadata.year {
let _ = metadata.set_year(Some(year)).await;
}
if let Some(duration_secs) = file_metadata.duration_secs {
let _ = metadata
.set_duration(Some(Duration::from_secs(duration_secs)))
.await;
}
let track_boundary = AudioSegment::new_track_boundary(
0,
0.0,
Arc::new(tokio::sync::RwLock::new(metadata)),
StreamType::Finite,
);
send_to_children(std::any::type_name::<Self>(), &output, track_boundary).await?;
}
// Préparer la lecture des chunks audio
let frame_bytes = stream_info.bytes_per_sample() * stream_info.channels as usize;
let chunk_byte_len = chunk_frames * frame_bytes;
let mut pending = Vec::new();
let mut read_buf = vec![0u8; frame_bytes * 512.max(chunk_frames)];
let mut chunk_index = 0u64;
let mut total_frames = 0u64;
// Lire et émettre les chunks audio
loop {
tokio::select! {
_ = stop_token.cancelled() => {
tracing::info!("FileSourceLogic: stop requested");
break;
}
read_result = stream.read(&mut read_buf) => {
// Remplir le buffer
if pending.len() < chunk_byte_len {
let read = read_result.map_err(|e| {
AudioError::IoError(format!("I/O error while decoding: {}", e))
})?;
if read == 0 && pending.is_empty() {
break;
}
if read > 0 {
pending.extend_from_slice(&read_buf[..read]);
}
}
if pending.is_empty() {
break;
}
// Extraire un chunk
let frames_in_pending = pending.len() / frame_bytes;
let frames_to_emit = frames_in_pending.min(chunk_frames);
if frames_to_emit == 0 {
break;
}
let take_bytes = frames_to_emit * frame_bytes;
let chunk_bytes = pending.drain(..take_bytes).collect::<Vec<u8>>();
// Calculer le timestamp
let timestamp_sec = total_frames as f64 / stream_info.sample_rate as f64;
// Créer et envoyer le segment audio
let segment = bytes_to_segment(
&chunk_bytes,
&stream_info,
frames_to_emit,
chunk_index,
timestamp_sec,
)?;
send_to_children(std::any::type_name::<Self>(), &output, segment).await?;
chunk_index += 1;
total_frames += frames_to_emit as u64;
}
}
}
// Traiter le reste éventuel (moins qu'un chunk complet)
if !pending.is_empty() {
let frames = pending.len() / frame_bytes;
if frames > 0 {
let timestamp_sec = total_frames as f64 / stream_info.sample_rate as f64;
let segment =
bytes_to_segment(&pending, &stream_info, frames, chunk_index, timestamp_sec)?;
send_to_children(std::any::type_name::<Self>(), &output, segment).await?;
total_frames += frames as u64;
chunk_index += 1;
}
}
// Émettre EndOfStream
let final_timestamp = total_frames as f64 / stream_info.sample_rate as f64;
let eos = AudioSegment::new_end_of_stream(chunk_index, final_timestamp);
send_to_children(std::any::type_name::<Self>(), &output, eos).await?;
// Attendre la fin du décodage
stream
.wait()
.await
.map_err(|e| AudioError::ProcessingError(format!("Decode task failed: {}", e)))?;
Ok(())
}
}
// ═════════════════════════════════════════════════════════════════════════════
// WRAPPER FileSource - Délègue à Node<FileSourceLogic>
// ═══════════════════════════════════════════════════════════════════════════════
/// FileSource - Lit un fichier audio et publie des `AudioSegment`
///
/// Cette source utilise `pmoflac` pour décoder le fichier (FLAC/MP3/OGG/WAV/AIFF)
/// puis transforme les échantillons PCM en `AudioSegment` stéréo avec le type approprié
/// (I16, I24, ou I32) selon la profondeur de bit du fichier source.
///
/// Le node émet trois types de syncmarkers :
/// - `TopZeroSync` au début du flux
/// - `TrackBoundary` avec les métadonnées du fichier
/// - `EndOfStream` à la fin du flux
///
/// # Architecture
///
/// Utilise la nouvelle architecture avec `Node<FileSourceLogic>` pour séparer
/// la logique métier (décodage) de la plomberie (spawning, monitoring).
pub struct FileSource {
inner: Node<FileSourceLogic>,
}
impl FileSource {
/// Crée une nouvelle source de fichier avec calcul automatique de la taille des chunks.
///
/// La taille des chunks sera calculée automatiquement pour obtenir environ 50ms
/// de latence par chunk, en fonction du sample rate du fichier.
///
/// * `path` - chemin du fichier audio à lire
pub fn new<P: Into<PathBuf>>(path: P) -> Self {
let logic = FileSourceLogic::new(path, 0);
Self { inner: Node::new_source(logic) }
}
pub fn make<P: Into<PathBuf>>(path: P) -> Box<dyn AudioPipelineNode> {
Self::new(path).boxed()
}
/// Crée une nouvelle source de fichier avec une taille de chunk spécifique.
///
/// * `path` - chemin du fichier audio à lire
/// * `chunk_frames` - nombre d'échantillons par canal par chunk (0 = auto)
pub fn with_chunk_size<P: Into<PathBuf>>(path: P, chunk_frames: usize) -> Box<dyn AudioPipelineNode> {
let logic = FileSourceLogic::new(path, chunk_frames);
Self { inner: Node::new_source(logic) }.boxed()
}
}
#[async_trait::async_trait]
impl AudioPipelineNode for FileSource {
fn get_tx(&self) -> Option<mpsc::Sender<Arc<AudioSegment>>> {
self.inner.get_tx()
}
fn register(&mut self, child: Box<dyn AudioPipelineNode>) {
self.inner.register(child)
}
async fn run(self: Box<Self>, stop_token: CancellationToken) -> Result<(), AudioError> {
Box::new(self.inner).run(stop_token).await
}
}
fn validate_stream(info: &StreamInfo) -> Result<(), AudioError> {
if !(1..=2).contains(&info.channels) {
return Err(AudioError::ProcessingError(format!(
"Unsupported channel count: {}",
info.channels
)));
}
match info.bits_per_sample {
8 | 16 | 24 | 32 => Ok(()),
other => Err(AudioError::ProcessingError(format!(
"Unsupported bit depth: {}",
other
))),
}
}
/// Convertit des bytes PCM en AudioSegment avec le type approprié
fn bytes_to_segment(
chunk_bytes: &[u8],
info: &StreamInfo,
frames: usize,
order: u64,
timestamp_sec: f64,
) -> Result<Arc<AudioSegment>, AudioError> {
let bytes_per_sample = info.bytes_per_sample();
let channels = info.channels as usize;
let frame_bytes = bytes_per_sample * channels;
// Créer le chunk du bon type selon la profondeur de bit
let chunk = match info.bits_per_sample {
16 => {
// Type I16
let mut stereo = Vec::with_capacity(frames);
for frame_idx in 0..frames {
let base = frame_idx * frame_bytes;
let l = i16::from_le_bytes(
chunk_bytes[base..base + bytes_per_sample]
.try_into()
.unwrap(),
);
let r = if channels == 1 {
l
} else {
i16::from_le_bytes(
chunk_bytes[base + bytes_per_sample..base + 2 * bytes_per_sample]
.try_into()
.unwrap(),
)
};
stereo.push([l, r]);
}
let chunk_data = AudioChunkData::new(stereo, info.sample_rate, 0.0);
AudioChunk::I16(chunk_data)
}
24 => {
// Type I24
let mut stereo = Vec::with_capacity(frames);
for frame_idx in 0..frames {
let base = frame_idx * frame_bytes;
let l_i32 = {
let mut buf = [0u8; 4];
buf[..3].copy_from_slice(&chunk_bytes[base..base + 3]);
// Sign extend
if chunk_bytes[base + 2] & 0x80 != 0 {
buf[3] = 0xFF;
}
i32::from_le_bytes(buf)
};
let l = I24::new(l_i32).ok_or_else(|| {
AudioError::ProcessingError(format!("Invalid I24 value: {}", l_i32))
})?;
let r = if channels == 1 {
l
} else {
let r_i32 = {
let mut buf = [0u8; 4];
buf[..3].copy_from_slice(
&chunk_bytes[base + bytes_per_sample..base + bytes_per_sample + 3],
);
// Sign extend
if chunk_bytes[base + bytes_per_sample + 2] & 0x80 != 0 {
buf[3] = 0xFF;
}
i32::from_le_bytes(buf)
};
I24::new(r_i32).ok_or_else(|| {
AudioError::ProcessingError(format!("Invalid I24 value: {}", r_i32))
})?
};
stereo.push([l, r]);
}
let chunk_data = AudioChunkData::new(stereo, info.sample_rate, 0.0);
AudioChunk::I24(chunk_data)
}
32 => {
// Type I32
let mut stereo = Vec::with_capacity(frames);
for frame_idx in 0..frames {
let base = frame_idx * frame_bytes;
let l = i32::from_le_bytes(
chunk_bytes[base..base + bytes_per_sample]
.try_into()
.unwrap(),
);
let r = if channels == 1 {
l
} else {
i32::from_le_bytes(
chunk_bytes[base + bytes_per_sample..base + 2 * bytes_per_sample]
.try_into()
.unwrap(),
)
};
stereo.push([l, r]);
}
let chunk_data = AudioChunkData::new(stereo, info.sample_rate, 0.0);
AudioChunk::I32(chunk_data)
}
other => {
return Err(AudioError::ProcessingError(format!(
"Unsupported bit depth: {}",
other
)))
}
};
// Créer le segment audio
Ok(Arc::new(AudioSegment {
order,
timestamp_sec,
segment: crate::_AudioSegment::Chunk(Arc::new(chunk)),
}))
}
impl TypedAudioNode for FileSource {
fn input_type(&self) -> Option<TypeRequirement> {
// FileSource est une source, elle ne consomme pas d'audio
None
}
fn output_type(&self) -> Option<TypeRequirement> {
// FileSource peut produire n'importe quel type entier (I16, I24, I32)
// selon la profondeur de bit du fichier source
Some(TypeRequirement::any_integer())
}
}
#[cfg(test)]
mod tests {
use super::*;
use pmoflac::{encode_flac_stream, EncoderOptions, PcmFormat};
use std::io::Cursor;
use tokio::io::AsyncWriteExt;
use tokio::sync::mpsc;
#[tokio::test]
async fn test_file_source_decodes_flac() {
let temp_dir = tempfile::tempdir().unwrap();
let flac_path = temp_dir.path().join("test.flac");
let sample_rate = 48_000;
let frames = 256;
let mut pcm = Vec::with_capacity(frames * 4);
for i in 0..frames {
let sample = ((i % 32) as f32 / 31.0 * 2.0 - 1.0) * 0.5; // simple ramp
let sample_i16 = (sample * 32767.0) as i16;
pcm.extend_from_slice(&sample_i16.to_le_bytes());
pcm.extend_from_slice(&sample_i16.to_le_bytes());
}
let format = PcmFormat {
sample_rate,
channels: 2,
bits_per_sample: 16,
};
let mut flac_stream =
encode_flac_stream(Cursor::new(pcm.clone()), format, EncoderOptions::default())
.await
.unwrap();
let mut file = File::create(&flac_path).await.expect("create flac file");
tokio::io::copy(&mut flac_stream, &mut file)
.await
.expect("write flac");
file.flush().await.expect("flush file");
flac_stream.wait().await.unwrap();
// Créer un collecteur simple qui transmet les segments à un channel de test
struct TestCollectorNode {
tx: mpsc::Sender<Arc<AudioSegment>>,
rx: mpsc::Receiver<Arc<AudioSegment>>,
test_tx: mpsc::Sender<Arc<AudioSegment>>,
}
impl TestCollectorNode {
fn new(test_tx: mpsc::Sender<Arc<AudioSegment>>) -> Self {
let (tx, rx) = mpsc::channel(16);
Self { tx, rx, test_tx }
}
}
#[async_trait::async_trait]
impl AudioPipelineNode for TestCollectorNode {
fn get_tx(&self) -> Option<mpsc::Sender<Arc<AudioSegment>>> {
Some(self.tx.clone())
}
fn register(&mut self, _child: Box<dyn AudioPipelineNode>) {
panic!("TestCollectorNode is a terminal node");
}
async fn run(
mut self: Box<Self>,
stop_token: CancellationToken,
) -> Result<(), AudioError> {
// Transférer tous les segments au test
loop {
tokio::select! {
_ = stop_token.cancelled() => {
break;
}
segment = self.rx.recv() => {
match segment {
Some(seg) => {
if self.test_tx.send(seg).await.is_err() {
break;
}
}
None => break,
}
}
}
}
Ok(())
}
}
let (test_tx, mut rx) = mpsc::channel(1024);
let mut source = FileSource::with_chunk_size(&flac_path, 64);
let collector = TestCollectorNode::new(test_tx);
source.register(Box::new(collector));
tokio::spawn(async move {
let token = CancellationToken::new();
Box::new(source).run(token).await.unwrap();
});
let mut received_frames = 0usize;
let mut received_syncmarkers = 0usize;
let mut seen_top_zero = false;
let mut seen_eos = false;
while let Some(segment) = rx.recv().await {
if segment.is_audio_chunk() {
if let Some(chunk) = segment.as_chunk() {
received_frames += chunk.len();
assert_eq!(chunk.sample_rate(), sample_rate);
}
} else {
received_syncmarkers += 1;
if let Some(marker) = segment.as_sync_marker() {
match **marker {
crate::SyncMarker::TopZeroSync => seen_top_zero = true,
crate::SyncMarker::EndOfStream => seen_eos = true,
crate::SyncMarker::TrackBoundary { .. } => {}
_ => {}
}
}
}
}
// Vérifier que tous les frames ont été reçus
assert_eq!(received_frames, frames);
// Vérifier qu'on a bien reçu des syncmarkers
assert!(received_syncmarkers >= 2); // Au moins TopZeroSync et EndOfStream
assert!(seen_top_zero, "Should have received TopZeroSync");
assert!(seen_eos, "Should have received EndOfStream");
}
}