active_call/media/
processor.rs1use super::INTERNAL_SAMPLERATE;
2use super::track::track_codec::TrackCodec;
3use crate::event::{EventSender, SessionEvent};
4use crate::media::{AudioFrame, Samples, SourcePacket};
5use anyhow::Result;
6use std::any::Any;
7use std::sync::{Arc, Mutex, RwLock};
8use tokio::sync::mpsc;
9
10pub trait Processor: Send + Sync + Any {
11 fn process_frame(&mut self, frame: &mut AudioFrame) -> Result<()>;
12}
13
14pub fn convert_to_mono(samples: &mut Vec<i16>, channels: u16) {
15 if channels != 2 {
16 return;
17 }
18 let mut i = 0;
19 let mut j = 0;
20 while i < samples.len() {
21 let l = samples[i] as i32;
22 let r = samples[i + 1] as i32;
23 samples[j] = ((l + r) / 2) as i16;
24 i += 2;
25 j += 1;
26 }
27 samples.truncate(j);
28}
29
30impl Default for AudioFrame {
31 fn default() -> Self {
32 Self {
33 track_id: "".to_string(),
34 samples: Samples::Empty,
35 timestamp: 0,
36 sample_rate: 16000,
37 channels: 1,
38 src_packet: None,
39 speech_probability: None,
40 }
41 }
42}
43
44impl Samples {
45 pub fn is_empty(&self) -> bool {
46 match self {
47 Samples::PCM { samples } => samples.is_empty(),
48 Samples::RTP { payload, .. } => payload.is_empty(),
49 Samples::Empty => true,
50 }
51 }
52}
53
54#[derive(Clone)]
55pub struct ProcessorChain {
56 processors: Arc<Mutex<Vec<Box<dyn Processor>>>>,
57 pub codec: TrackCodec,
58 sample_rate: u32,
59 pub force_decode: bool,
60 pub raw_tap: Arc<RwLock<Option<mpsc::UnboundedSender<AudioFrame>>>>,
69}
70
71impl ProcessorChain {
72 pub fn new(_sample_rate: u32) -> Self {
73 Self {
74 processors: Arc::new(Mutex::new(Vec::new())),
75 codec: TrackCodec::new(),
76 sample_rate: INTERNAL_SAMPLERATE,
77 force_decode: true,
78 raw_tap: Arc::new(RwLock::new(None)),
79 }
80 }
81 pub fn set_raw_tap(&mut self, tap: Option<mpsc::UnboundedSender<AudioFrame>>) {
82 *self.raw_tap.write().unwrap() = tap;
83 }
84 fn raw_tap(&self) -> Option<mpsc::UnboundedSender<AudioFrame>> {
85 self.raw_tap.read().unwrap().clone()
86 }
87 pub fn insert_processor(&mut self, processor: Box<dyn Processor>) {
88 self.processors.lock().unwrap().insert(0, processor);
89 }
90 pub fn append_processor(&mut self, processor: Box<dyn Processor>) {
91 self.processors.lock().unwrap().push(processor);
92 }
93
94 pub fn has_processor<T: 'static>(&self) -> bool {
95 let processors = self.processors.lock().unwrap();
96 processors
97 .iter()
98 .any(|processor| (processor.as_ref() as &dyn Any).is::<T>())
99 }
100
101 pub fn remove_processor<T: 'static>(&self) {
102 let mut processors = self.processors.lock().unwrap();
103 processors.retain(|processor| !(processor.as_ref() as &dyn Any).is::<T>());
104 }
105
106 pub fn process_frame(&mut self, frame: &mut AudioFrame) -> Result<()> {
107 let mut processors = self.processors.lock().unwrap();
108 if !self.force_decode && processors.is_empty() && self.raw_tap().is_none() {
109 return Ok(());
110 }
111 match &mut frame.samples {
112 Samples::RTP {
113 payload_type,
114 payload,
115 sequence_number,
116 } => {
117 if TrackCodec::is_audio(*payload_type) {
118 let (decoded_sample_rate, channels, samples) =
119 self.codec.decode(*payload_type, &payload);
120 let src_packet = SourcePacket {
121 sequence_number: *sequence_number,
122 payload_type: *payload_type,
123 payload: std::mem::take(payload),
124 };
125 frame.src_packet = Some(src_packet);
126 frame.channels = channels;
127 frame.samples = Samples::PCM { samples };
128 frame.sample_rate = decoded_sample_rate;
129 }
130 }
131 _ => {}
132 }
133
134 if let Some(tap) = self.raw_tap()
137 && let Samples::PCM { samples } = &frame.samples
138 && !samples.is_empty()
139 && frame.sample_rate > 0
140 {
141 let mut raw = frame.clone();
142 raw.src_packet = None;
143 let mono = match &mut raw.samples {
144 Samples::PCM { samples } => samples,
145 _ => unreachable!("checked PCM above"),
146 };
147 if raw.channels == 2 {
148 convert_to_mono(mono, 2);
149 raw.channels = 1;
150 }
151 let _ = tap.send(raw);
152 }
153
154 if let Samples::PCM { samples } = &mut frame.samples {
155 if frame.sample_rate != self.sample_rate {
156 let new_samples = self.codec.resample(
157 std::mem::take(samples),
158 frame.sample_rate,
159 self.sample_rate,
160 );
161 *samples = new_samples;
162 frame.sample_rate = self.sample_rate;
163 }
164 if frame.channels == 2 {
165 convert_to_mono(samples, 2);
166 frame.channels = 1;
167 }
168 }
169 for processor in processors.iter_mut() {
171 processor.process_frame(frame)?;
172 }
173 Ok(())
174 }
175}
176
177pub struct SubscribeProcessor {
178 event_sender: EventSender,
179 track_id: String,
180 track_index: u8, }
182
183impl SubscribeProcessor {
184 pub fn new(event_sender: EventSender, track_id: String, track_index: u8) -> Self {
185 Self {
186 event_sender,
187 track_id,
188 track_index,
189 }
190 }
191}
192
193impl Processor for SubscribeProcessor {
194 fn process_frame(&mut self, frame: &mut AudioFrame) -> Result<()> {
195 if let Samples::PCM { samples } = &frame.samples {
196 if !samples.is_empty() {
197 let pcm_data = audio_codec::samples_to_bytes(samples);
198 let mut data = Vec::with_capacity(pcm_data.len() + 1);
199 data.push(self.track_index);
200 data.extend_from_slice(&pcm_data);
201
202 let event = SessionEvent::Binary {
203 track_id: self.track_id.clone(),
204 timestamp: frame.timestamp,
205 data,
206 };
207 self.event_sender.send(event).ok();
208 }
209 }
210 Ok(())
211 }
212}
213
214#[cfg(test)]
215mod tests {
216 use super::*;
217
218 #[test]
224 fn raw_tap_visible_to_clones_taken_before_attach() {
225 let mut chain = ProcessorChain::new(INTERNAL_SAMPLERATE);
226 let mut worker = chain.clone();
227
228 let (tx, mut rx) = mpsc::unbounded_channel();
229 chain.set_raw_tap(Some(tx));
230
231 let mut frame = AudioFrame {
232 track_id: "track".to_string(),
233 samples: Samples::PCM {
234 samples: vec![100i16; 160],
235 },
236 sample_rate: 8000,
237 channels: 1,
238 ..Default::default()
239 };
240 worker.process_frame(&mut frame).unwrap();
241
242 let raw = rx
243 .try_recv()
244 .expect("raw tap should receive a frame from a pre-attach clone");
245 assert_eq!(raw.sample_rate, 8000);
246 assert_eq!(raw.channels, 1);
247 match raw.samples {
248 Samples::PCM { samples } => assert_eq!(samples.len(), 160),
249 _ => panic!("expected PCM samples on the raw tap"),
250 }
251 assert_eq!(frame.sample_rate, INTERNAL_SAMPLERATE);
253 }
254
255 #[test]
258 fn raw_tap_detach_visible_to_clones() {
259 let mut chain = ProcessorChain::new(INTERNAL_SAMPLERATE);
260 let (tx, _rx) = mpsc::unbounded_channel();
261 chain.set_raw_tap(Some(tx));
262 let mut worker = chain.clone();
263 chain.set_raw_tap(None);
264
265 let mut frame = AudioFrame {
266 track_id: "track".to_string(),
267 samples: Samples::PCM {
268 samples: vec![100i16; 160],
269 },
270 sample_rate: 8000,
271 channels: 1,
272 ..Default::default()
273 };
274 worker.process_frame(&mut frame).unwrap();
275 assert_eq!(frame.sample_rate, INTERNAL_SAMPLERATE);
276 }
277}