mirror of
https://github.com/servo/servo
synced 2026-04-25 17:15:48 +02:00
This change merges http://github.com/servo/media into this repository. It is only used by Servo and version upgrades are complicated by having two repositories. In addition, this avoids the need to refer to individual commit hashes in the Servo `Cargo.toml`. The hope is that merging these two repositories will lead to better code organization / simplification like we have seen with the WebXR support. Initiailly, the idea was that this media support could be shared with the wider Rust ecosystem, but I think that hasn't worked out as planned due to the fact that it is difficult to use the various media packaes outside of the Servo project and the fact that no one seems to be doing this. Some changes were made when importing the code: - The second commit in this PR addresses new clippy warnings from the imported code. - GStreamer Packages are no longer renamed in the media code, so that they are named the same as they are currently in Servo. - Some examples are not ported as they require being run interactively and depend on older version of important libraries like winit. Having these dependencies in the core part of Servo isn't very convenient. Removed examples: - `audio_decoder.rs`: This is meant to be run interactively with a file so isn't very useful for testing. - Depended on winit GUI (`player` subdirectory): - `media_element_source_node.rs` - `play_media_stream.rs` - `simple_player.rs` - `muted_player.rs` - `siple_webrtc.rs`: Depended on `webrtc` library: Testing: This is covered by existing tests. In addition, the job which runs the media examples is added to the unit test workflow. --------- Signed-off-by: Martin Robinson <mrobinson@igalia.com>
440 lines
17 KiB
Rust
440 lines
17 KiB
Rust
/* This Source Code Form is subject to the terms of the Mozilla Public
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
* file, You can obtain one at https://mozilla.org/MPL/2.0/. */
|
|
|
|
use crate::block::{Block, Chunk, FRAMES_PER_BLOCK, Tick};
|
|
use crate::node::{
|
|
AudioNodeEngine, AudioNodeType, AudioScheduledSourceNodeMessage, BlockInfo, ChannelInfo,
|
|
OnEndedCallback, ShouldPlay,
|
|
};
|
|
use crate::param::{Param, ParamType};
|
|
|
|
/// Control messages directed to AudioBufferSourceNodes.
|
|
#[derive(Debug, Clone)]
|
|
pub enum AudioBufferSourceNodeMessage {
|
|
/// Set the data block holding the audio sample data to be played.
|
|
SetBuffer(Option<AudioBuffer>),
|
|
/// Set loop parameter.
|
|
SetLoopEnabled(bool),
|
|
/// Set loop parameter.
|
|
SetLoopEnd(f64),
|
|
/// Set loop parameter.
|
|
SetLoopStart(f64),
|
|
/// Set start parameters (when, offset, duration).
|
|
SetStartParams(f64, Option<f64>, Option<f64>),
|
|
}
|
|
|
|
/// This specifies options for constructing an AudioBufferSourceNode.
|
|
#[derive(Debug, Clone)]
|
|
pub struct AudioBufferSourceNodeOptions {
|
|
/// The audio asset to be played.
|
|
pub buffer: Option<AudioBuffer>,
|
|
/// The initial value for the detune AudioParam.
|
|
pub detune: f32,
|
|
/// The initial value for the loop_enabled attribute.
|
|
pub loop_enabled: bool,
|
|
/// The initial value for the loop_end attribute.
|
|
pub loop_end: Option<f64>,
|
|
/// The initial value for the loop_start attribute.
|
|
pub loop_start: Option<f64>,
|
|
/// The initial value for the playback_rate AudioParam.
|
|
pub playback_rate: f32,
|
|
}
|
|
|
|
impl Default for AudioBufferSourceNodeOptions {
|
|
fn default() -> Self {
|
|
AudioBufferSourceNodeOptions {
|
|
buffer: None,
|
|
detune: 0.,
|
|
loop_enabled: false,
|
|
loop_end: None,
|
|
loop_start: None,
|
|
playback_rate: 1.,
|
|
}
|
|
}
|
|
}
|
|
|
|
/// AudioBufferSourceNode engine.
|
|
/// <https://webaudio.github.io/web-audio-api/#AudioBufferSourceNode>
|
|
#[derive(AudioScheduledSourceNode, AudioNodeCommon)]
|
|
#[allow(dead_code)]
|
|
pub(crate) struct AudioBufferSourceNode {
|
|
channel_info: ChannelInfo,
|
|
/// A data block holding the audio sample data to be played.
|
|
buffer: Option<AudioBuffer>,
|
|
/// How many more buffer-frames to output. See buffer_pos for clarification.
|
|
buffer_duration: f64,
|
|
/// "Index" of the next buffer frame to play. "Index" is in quotes because
|
|
/// this variable maps to a playhead position (the offset in seconds can be
|
|
/// obtained by dividing by self.buffer.sample_rate), and therefore has
|
|
/// subsample accuracy; a fractional "index" means interpolation is needed.
|
|
buffer_pos: f64,
|
|
/// AudioParam to modulate the speed at which is rendered the audio stream.
|
|
detune: Param,
|
|
/// Whether we need to compute offsets from scratch.
|
|
initialized_pos: bool,
|
|
/// Indicates if the region of audio data designated by loopStart and loopEnd
|
|
/// should be played continuously in a loop.
|
|
loop_enabled: bool,
|
|
/// An playhead position where looping should end if the loop_enabled
|
|
/// attribute is true.
|
|
loop_end: Option<f64>,
|
|
/// An playhead position where looping should begin if the loop_enabled
|
|
/// attribute is true.
|
|
loop_start: Option<f64>,
|
|
/// The speed at which to render the audio stream. Can be negative if the
|
|
/// audio is to be played backwards. With a negative playback_rate, looping
|
|
/// jumps from loop_start to loop_end instead of the other way around.
|
|
playback_rate: Param,
|
|
/// Time at which the source should start playing.
|
|
start_at: Option<Tick>,
|
|
/// Offset parameter passed to Start().
|
|
start_offset: Option<f64>,
|
|
/// Duration parameter passed to Start().
|
|
start_duration: Option<f64>,
|
|
/// The same as start_at, but with subsample accuracy.
|
|
/// FIXME: AudioScheduledSourceNode should use this as well.
|
|
start_when: f64,
|
|
/// Time at which the source should stop playing.
|
|
stop_at: Option<Tick>,
|
|
/// The ended event callback.
|
|
pub onended_callback: Option<OnEndedCallback>,
|
|
}
|
|
|
|
impl AudioBufferSourceNode {
|
|
pub fn new(options: AudioBufferSourceNodeOptions, channel_info: ChannelInfo) -> Self {
|
|
Self {
|
|
channel_info,
|
|
buffer: options.buffer,
|
|
buffer_pos: 0.,
|
|
detune: Param::new_krate(options.detune),
|
|
initialized_pos: false,
|
|
loop_enabled: options.loop_enabled,
|
|
loop_end: options.loop_end,
|
|
loop_start: options.loop_start,
|
|
playback_rate: Param::new_krate(options.playback_rate),
|
|
buffer_duration: f64::INFINITY,
|
|
start_at: None,
|
|
start_offset: None,
|
|
start_duration: None,
|
|
start_when: 0.,
|
|
stop_at: None,
|
|
onended_callback: None,
|
|
}
|
|
}
|
|
|
|
pub fn handle_message(&mut self, message: AudioBufferSourceNodeMessage, _: f32) {
|
|
match message {
|
|
AudioBufferSourceNodeMessage::SetBuffer(buffer) => {
|
|
self.buffer = buffer;
|
|
},
|
|
// XXX(collares): To fully support dynamically updating loop bounds,
|
|
// Must truncate self.buffer_pos if it is now outside the loop.
|
|
AudioBufferSourceNodeMessage::SetLoopEnabled(loop_enabled) => {
|
|
self.loop_enabled = loop_enabled
|
|
},
|
|
AudioBufferSourceNodeMessage::SetLoopEnd(loop_end) => self.loop_end = Some(loop_end),
|
|
AudioBufferSourceNodeMessage::SetLoopStart(loop_start) => {
|
|
self.loop_start = Some(loop_start)
|
|
},
|
|
AudioBufferSourceNodeMessage::SetStartParams(when, offset, duration) => {
|
|
self.start_when = when;
|
|
self.start_offset = offset;
|
|
self.start_duration = duration;
|
|
},
|
|
}
|
|
}
|
|
}
|
|
|
|
impl AudioNodeEngine for AudioBufferSourceNode {
|
|
fn node_type(&self) -> AudioNodeType {
|
|
AudioNodeType::AudioBufferSourceNode
|
|
}
|
|
|
|
fn input_count(&self) -> u32 {
|
|
0
|
|
}
|
|
|
|
fn process(&mut self, mut inputs: Chunk, info: &BlockInfo) -> Chunk {
|
|
debug_assert!(inputs.is_empty());
|
|
|
|
if self.buffer.is_none() {
|
|
inputs.blocks.push(Default::default());
|
|
return inputs;
|
|
}
|
|
|
|
let (start_at, stop_at) = match self.should_play_at(info.frame) {
|
|
ShouldPlay::No => {
|
|
inputs.blocks.push(Default::default());
|
|
return inputs;
|
|
},
|
|
ShouldPlay::Between(start, end) => (start.0 as usize, end.0 as usize),
|
|
};
|
|
|
|
let buffer = self.buffer.as_ref().unwrap();
|
|
|
|
let (mut actual_loop_start, mut actual_loop_end) = (0., buffer.len() as f64);
|
|
if self.loop_enabled {
|
|
let loop_start = self.loop_start.unwrap_or(0.);
|
|
let loop_end = self.loop_end.unwrap_or(0.);
|
|
|
|
if loop_start >= 0. && loop_end > loop_start {
|
|
actual_loop_start = loop_start * (buffer.sample_rate as f64);
|
|
actual_loop_end = loop_end * (buffer.sample_rate as f64);
|
|
}
|
|
}
|
|
|
|
// https://webaudio.github.io/web-audio-api/#computedplaybackrate
|
|
self.playback_rate.update(info, Tick(0));
|
|
self.detune.update(info, Tick(0));
|
|
// computed_playback_rate can be negative or zero.
|
|
let computed_playback_rate =
|
|
self.playback_rate.value() as f64 * (2.0_f64).powf(self.detune.value() as f64 / 1200.);
|
|
let forward = computed_playback_rate >= 0.;
|
|
|
|
if !self.initialized_pos {
|
|
self.initialized_pos = true;
|
|
|
|
// Apply the offset and duration parameters passed to start. We handle
|
|
// this here because the buffer may be set after Start() gets called, so
|
|
// this might be the first time we know the buffer's sample rate.
|
|
if let Some(start_offset) = self.start_offset {
|
|
self.buffer_pos = start_offset * (buffer.sample_rate as f64);
|
|
if self.buffer_pos < 0. {
|
|
self.buffer_pos = 0.
|
|
} else if self.buffer_pos > buffer.len() as f64 {
|
|
self.buffer_pos = buffer.len() as f64;
|
|
}
|
|
}
|
|
|
|
if self.loop_enabled {
|
|
if forward && self.buffer_pos >= actual_loop_end {
|
|
self.buffer_pos = actual_loop_start;
|
|
}
|
|
// https://github.com/WebAudio/web-audio-api/issues/2031
|
|
if !forward && self.buffer_pos < actual_loop_start {
|
|
self.buffer_pos = actual_loop_end;
|
|
}
|
|
}
|
|
|
|
if let Some(start_duration) = self.start_duration {
|
|
self.buffer_duration = start_duration * (buffer.sample_rate as f64);
|
|
}
|
|
|
|
// start_when can be subsample accurate. Correct buffer_pos.
|
|
//
|
|
// XXX(collares): What happens to "start_when" if the buffer gets
|
|
// set after Start()?
|
|
// XXX(collares): Need a better way to distingush between Start()
|
|
// being called with "when" in the past (in which case "when" must
|
|
// be ignored) and Start() being called with "when" in the future.
|
|
// This can now make a difference if "when" shouldn't be ignored
|
|
// but falls after the last frame of the previous quantum.
|
|
if self.start_when > info.time - 1. / info.sample_rate as f64 {
|
|
let first_time = info.time + start_at as f64 / info.sample_rate as f64;
|
|
if self.start_when <= first_time {
|
|
let subsample_offset = (first_time - self.start_when) *
|
|
(buffer.sample_rate as f64) *
|
|
computed_playback_rate;
|
|
self.buffer_pos += subsample_offset;
|
|
self.buffer_duration -= subsample_offset.abs();
|
|
}
|
|
}
|
|
}
|
|
|
|
let mut buffer_offset_per_tick =
|
|
computed_playback_rate * (buffer.sample_rate as f64 / info.sample_rate as f64);
|
|
|
|
// WebAudio §1.9.5: "Setting the loop attribute to true causes playback of
|
|
// the region of the buffer defined by the endpoints loopStart and loopEnd
|
|
// to continue indefinitely, once any part of the looped region has been
|
|
// played. While loop remains true, looped playback will continue until one
|
|
// of the following occurs:
|
|
// * stop() is called,
|
|
// * the scheduled stop time has been reached,
|
|
// * the duration has been exceeded, if start() was called with a duration value."
|
|
// Even with extreme playback rates we must stay inside the loop body, so wrap
|
|
// the per-tick delta instead of bailing.
|
|
if self.loop_enabled && actual_loop_end > actual_loop_start {
|
|
let loop_length = actual_loop_end - actual_loop_start;
|
|
if loop_length > 0. {
|
|
let step = buffer_offset_per_tick.abs();
|
|
if step >= loop_length {
|
|
let mut wrapped = step.rem_euclid(loop_length);
|
|
if wrapped == 0. {
|
|
wrapped = loop_length;
|
|
}
|
|
buffer_offset_per_tick = wrapped.copysign(buffer_offset_per_tick);
|
|
}
|
|
}
|
|
}
|
|
|
|
// We will output at most this many frames (fewer if we run out of data).
|
|
let frames_to_output = stop_at - start_at;
|
|
|
|
// Fast path for the case where we can just copy FRAMES_PER_BLOCK
|
|
// frames straight from the buffer.
|
|
if frames_to_output == FRAMES_PER_BLOCK.0 as usize &&
|
|
forward &&
|
|
buffer_offset_per_tick == 1. &&
|
|
self.buffer_pos.trunc() == self.buffer_pos &&
|
|
self.buffer_pos + (FRAMES_PER_BLOCK.0 as f64) <= actual_loop_end &&
|
|
FRAMES_PER_BLOCK.0 as f64 <= self.buffer_duration
|
|
{
|
|
let mut block = Block::empty();
|
|
let pos = self.buffer_pos as usize;
|
|
|
|
for chan in 0..buffer.chans() {
|
|
block.push_chan(&buffer.buffers[chan as usize][pos..(pos + frames_to_output)]);
|
|
}
|
|
|
|
inputs.blocks.push(block);
|
|
self.buffer_pos += FRAMES_PER_BLOCK.0 as f64;
|
|
self.buffer_duration -= FRAMES_PER_BLOCK.0 as f64;
|
|
} else {
|
|
// Slow path, with interpolation.
|
|
let mut block = Block::default();
|
|
block.repeat(buffer.chans());
|
|
block.explicit_repeat();
|
|
|
|
debug_assert!(buffer.chans() > 0);
|
|
|
|
for chan in 0..buffer.chans() {
|
|
let data = block.data_chan_mut(chan);
|
|
let (_, data) = data.split_at_mut(start_at);
|
|
let (data, _) = data.split_at_mut(frames_to_output);
|
|
|
|
let mut pos = self.buffer_pos;
|
|
let mut duration = self.buffer_duration;
|
|
|
|
for sample in data {
|
|
if duration <= 0. {
|
|
break;
|
|
}
|
|
|
|
if self.loop_enabled {
|
|
if forward && pos >= actual_loop_end {
|
|
pos -= actual_loop_end - actual_loop_start;
|
|
} else if !forward && pos < actual_loop_start {
|
|
pos += actual_loop_end - actual_loop_start;
|
|
}
|
|
} else if pos < 0. || pos >= buffer.len() as f64 {
|
|
break;
|
|
}
|
|
|
|
*sample = buffer.interpolate(chan, pos);
|
|
pos += buffer_offset_per_tick;
|
|
duration -= buffer_offset_per_tick.abs();
|
|
}
|
|
|
|
// This is the last channel, update parameters.
|
|
if chan == buffer.chans() - 1 {
|
|
self.buffer_pos = pos;
|
|
self.buffer_duration = duration;
|
|
}
|
|
}
|
|
|
|
inputs.blocks.push(block);
|
|
}
|
|
|
|
if !self.loop_enabled && (self.buffer_pos < 0. || self.buffer_pos >= buffer.len() as f64) ||
|
|
self.buffer_duration <= 0.
|
|
{
|
|
self.maybe_trigger_onended_callback();
|
|
}
|
|
|
|
inputs
|
|
}
|
|
|
|
fn get_param(&mut self, id: ParamType) -> &mut Param {
|
|
match id {
|
|
ParamType::PlaybackRate => &mut self.playback_rate,
|
|
ParamType::Detune => &mut self.detune,
|
|
_ => panic!("Unknown param {:?} for AudioBufferSourceNode", id),
|
|
}
|
|
}
|
|
|
|
make_message_handler!(
|
|
AudioBufferSourceNode: handle_message,
|
|
AudioScheduledSourceNode: handle_source_node_message
|
|
);
|
|
}
|
|
|
|
#[derive(Debug, Clone)]
|
|
pub struct AudioBuffer {
|
|
/// Invariant: all buffers must be of the same length
|
|
pub buffers: Vec<Vec<f32>>,
|
|
pub sample_rate: f32,
|
|
}
|
|
|
|
impl AudioBuffer {
|
|
pub fn new(chan: u8, len: usize, sample_rate: f32) -> Self {
|
|
assert!(chan > 0);
|
|
let mut buffers = Vec::with_capacity(chan as usize);
|
|
let single = vec![0.; len];
|
|
buffers.resize(chan as usize, single);
|
|
AudioBuffer {
|
|
buffers,
|
|
sample_rate,
|
|
}
|
|
}
|
|
|
|
pub fn from_buffers(buffers: Vec<Vec<f32>>, sample_rate: f32) -> Self {
|
|
for buf in &buffers {
|
|
assert_eq!(buf.len(), buffers[0].len())
|
|
}
|
|
|
|
Self {
|
|
buffers,
|
|
sample_rate,
|
|
}
|
|
}
|
|
|
|
pub fn from_buffer(buffer: Vec<f32>, sample_rate: f32) -> Self {
|
|
AudioBuffer::from_buffers(vec![buffer], sample_rate)
|
|
}
|
|
|
|
pub fn len(&self) -> usize {
|
|
self.buffers[0].len()
|
|
}
|
|
|
|
pub fn is_empty(&self) -> bool {
|
|
self.len() == 0
|
|
}
|
|
|
|
pub fn chans(&self) -> u8 {
|
|
self.buffers.len() as u8
|
|
}
|
|
|
|
// XXX(collares): There are better fast interpolation algorithms.
|
|
// Firefox uses (via Speex's resampler) the algorithm described in
|
|
// https://ccrma.stanford.edu/~jos/resample/resample.pdf
|
|
// There are Rust bindings: https://github.com/rust-av/speexdsp-rs
|
|
pub fn interpolate(&self, chan: u8, pos: f64) -> f32 {
|
|
debug_assert!(pos >= 0. && pos < self.len() as f64);
|
|
|
|
let prev = pos.floor() as usize;
|
|
let offset = pos - pos.floor();
|
|
match self.buffers[chan as usize].get(prev + 1) {
|
|
Some(next_sample) => {
|
|
((1. - offset) * (self.buffers[chan as usize][prev] as f64) +
|
|
offset * (*next_sample as f64)) as f32
|
|
},
|
|
_ => {
|
|
// linear extrapolation of two prev samples if there are two
|
|
if prev > 0 {
|
|
((1. + offset) * (self.buffers[chan as usize][prev] as f64) -
|
|
offset * (self.buffers[chan as usize][prev - 1] as f64))
|
|
as f32
|
|
} else {
|
|
self.buffers[chan as usize][prev]
|
|
}
|
|
},
|
|
}
|
|
}
|
|
|
|
pub fn data_chan_mut(&mut self, chan: u8) -> &mut [f32] {
|
|
&mut self.buffers[chan as usize]
|
|
}
|
|
}
|