1 /* 2 * Copyright 2012 The WebRTC project authors. All Rights Reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 // This file contains interfaces for MediaStream, MediaTrack and MediaSource. 12 // These interfaces are used for implementing MediaStream and MediaTrack as 13 // defined in http://dev.w3.org/2011/webrtc/editor/webrtc.html#stream-api. These 14 // interfaces must be used only with PeerConnection. 15 16 #ifndef API_MEDIA_STREAM_INTERFACE_H_ 17 #define API_MEDIA_STREAM_INTERFACE_H_ 18 19 #include <stddef.h> 20 21 #include <string> 22 #include <vector> 23 24 #include "absl/types/optional.h" 25 #include "api/audio_options.h" 26 #include "api/scoped_refptr.h" 27 #include "api/video/recordable_encoded_frame.h" 28 #include "api/video/video_frame.h" 29 #include "api/video/video_sink_interface.h" 30 #include "api/video/video_source_interface.h" 31 #include "modules/audio_processing/include/audio_processing_statistics.h" 32 #include "rtc_base/ref_count.h" 33 #include "rtc_base/system/rtc_export.h" 34 35 namespace webrtc { 36 37 // Generic observer interface. 38 class ObserverInterface { 39 public: 40 virtual void OnChanged() = 0; 41 42 protected: ~ObserverInterface()43 virtual ~ObserverInterface() {} 44 }; 45 46 class NotifierInterface { 47 public: 48 virtual void RegisterObserver(ObserverInterface* observer) = 0; 49 virtual void UnregisterObserver(ObserverInterface* observer) = 0; 50 ~NotifierInterface()51 virtual ~NotifierInterface() {} 52 }; 53 54 // Base class for sources. A MediaStreamTrack has an underlying source that 55 // provides media. A source can be shared by multiple tracks. 56 class RTC_EXPORT MediaSourceInterface : public rtc::RefCountInterface, 57 public NotifierInterface { 58 public: 59 enum SourceState { kInitializing, kLive, kEnded, kMuted }; 60 61 virtual SourceState state() const = 0; 62 63 virtual bool remote() const = 0; 64 65 protected: 66 ~MediaSourceInterface() override = default; 67 }; 68 69 // C++ version of MediaStreamTrack. 70 // See: https://www.w3.org/TR/mediacapture-streams/#mediastreamtrack 71 class RTC_EXPORT MediaStreamTrackInterface : public rtc::RefCountInterface, 72 public NotifierInterface { 73 public: 74 enum TrackState { 75 kLive, 76 kEnded, 77 }; 78 79 static const char* const kAudioKind; 80 static const char* const kVideoKind; 81 82 // The kind() method must return kAudioKind only if the object is a 83 // subclass of AudioTrackInterface, and kVideoKind only if the 84 // object is a subclass of VideoTrackInterface. It is typically used 85 // to protect a static_cast<> to the corresponding subclass. 86 virtual std::string kind() const = 0; 87 88 // Track identifier. 89 virtual std::string id() const = 0; 90 91 // A disabled track will produce silence (if audio) or black frames (if 92 // video). Can be disabled and re-enabled. 93 virtual bool enabled() const = 0; 94 virtual bool set_enabled(bool enable) = 0; 95 96 // Live or ended. A track will never be live again after becoming ended. 97 virtual TrackState state() const = 0; 98 99 protected: 100 ~MediaStreamTrackInterface() override = default; 101 }; 102 103 // VideoTrackSourceInterface is a reference counted source used for 104 // VideoTracks. The same source can be used by multiple VideoTracks. 105 // VideoTrackSourceInterface is designed to be invoked on the signaling thread 106 // except for rtc::VideoSourceInterface<VideoFrame> methods that will be invoked 107 // on the worker thread via a VideoTrack. A custom implementation of a source 108 // can inherit AdaptedVideoTrackSource instead of directly implementing this 109 // interface. 110 class VideoTrackSourceInterface : public MediaSourceInterface, 111 public rtc::VideoSourceInterface<VideoFrame> { 112 public: 113 struct Stats { 114 // Original size of captured frame, before video adaptation. 115 int input_width; 116 int input_height; 117 }; 118 119 // Indicates that parameters suitable for screencasts should be automatically 120 // applied to RtpSenders. 121 // TODO(perkj): Remove these once all known applications have moved to 122 // explicitly setting suitable parameters for screencasts and don't need this 123 // implicit behavior. 124 virtual bool is_screencast() const = 0; 125 126 // Indicates that the encoder should denoise video before encoding it. 127 // If it is not set, the default configuration is used which is different 128 // depending on video codec. 129 // TODO(perkj): Remove this once denoising is done by the source, and not by 130 // the encoder. 131 virtual absl::optional<bool> needs_denoising() const = 0; 132 133 // Returns false if no stats are available, e.g, for a remote source, or a 134 // source which has not seen its first frame yet. 135 // 136 // Implementation should avoid blocking. 137 virtual bool GetStats(Stats* stats) = 0; 138 139 // Returns true if encoded output can be enabled in the source. 140 virtual bool SupportsEncodedOutput() const = 0; 141 142 // Reliably cause a key frame to be generated in encoded output. 143 // TODO(bugs.webrtc.org/11115): find optimal naming. 144 virtual void GenerateKeyFrame() = 0; 145 146 // Add an encoded video sink to the source and additionally cause 147 // a key frame to be generated from the source. The sink will be 148 // invoked from a decoder queue. 149 // TODO(bugs.webrtc.org/11114): make pure virtual once downstream project 150 // adapts. 151 virtual void AddEncodedSink( 152 rtc::VideoSinkInterface<RecordableEncodedFrame>* sink) = 0; 153 154 // Removes an encoded video sink from the source. 155 virtual void RemoveEncodedSink( 156 rtc::VideoSinkInterface<RecordableEncodedFrame>* sink) = 0; 157 158 protected: 159 ~VideoTrackSourceInterface() override = default; 160 }; 161 162 // VideoTrackInterface is designed to be invoked on the signaling thread except 163 // for rtc::VideoSourceInterface<VideoFrame> methods that must be invoked 164 // on the worker thread. 165 // PeerConnectionFactory::CreateVideoTrack can be used for creating a VideoTrack 166 // that ensures thread safety and that all methods are called on the right 167 // thread. 168 class RTC_EXPORT VideoTrackInterface 169 : public MediaStreamTrackInterface, 170 public rtc::VideoSourceInterface<VideoFrame> { 171 public: 172 // Video track content hint, used to override the source is_screencast 173 // property. 174 // See https://crbug.com/653531 and https://w3c.github.io/mst-content-hint. 175 enum class ContentHint { kNone, kFluid, kDetailed, kText }; 176 177 // Register a video sink for this track. Used to connect the track to the 178 // underlying video engine. AddOrUpdateSink(rtc::VideoSinkInterface<VideoFrame> * sink,const rtc::VideoSinkWants & wants)179 void AddOrUpdateSink(rtc::VideoSinkInterface<VideoFrame>* sink, 180 const rtc::VideoSinkWants& wants) override {} RemoveSink(rtc::VideoSinkInterface<VideoFrame> * sink)181 void RemoveSink(rtc::VideoSinkInterface<VideoFrame>* sink) override {} 182 183 virtual VideoTrackSourceInterface* GetSource() const = 0; 184 185 virtual ContentHint content_hint() const; set_content_hint(ContentHint hint)186 virtual void set_content_hint(ContentHint hint) {} 187 188 protected: 189 ~VideoTrackInterface() override = default; 190 }; 191 192 // Interface for receiving audio data from a AudioTrack. 193 class AudioTrackSinkInterface { 194 public: OnData(const void * audio_data,int bits_per_sample,int sample_rate,size_t number_of_channels,size_t number_of_frames)195 virtual void OnData(const void* audio_data, 196 int bits_per_sample, 197 int sample_rate, 198 size_t number_of_channels, 199 size_t number_of_frames) { 200 RTC_NOTREACHED() << "This method must be overridden, or not used."; 201 } 202 203 // In this method, |absolute_capture_timestamp_ms|, when available, is 204 // supposed to deliver the timestamp when this audio frame was originally 205 // captured. This timestamp MUST be based on the same clock as 206 // rtc::TimeMillis(). OnData(const void * audio_data,int bits_per_sample,int sample_rate,size_t number_of_channels,size_t number_of_frames,absl::optional<int64_t> absolute_capture_timestamp_ms)207 virtual void OnData(const void* audio_data, 208 int bits_per_sample, 209 int sample_rate, 210 size_t number_of_channels, 211 size_t number_of_frames, 212 absl::optional<int64_t> absolute_capture_timestamp_ms) { 213 // TODO(bugs.webrtc.org/10739): Deprecate the old OnData and make this one 214 // pure virtual. 215 return OnData(audio_data, bits_per_sample, sample_rate, number_of_channels, 216 number_of_frames); 217 } 218 219 protected: ~AudioTrackSinkInterface()220 virtual ~AudioTrackSinkInterface() {} 221 }; 222 223 // AudioSourceInterface is a reference counted source used for AudioTracks. 224 // The same source can be used by multiple AudioTracks. 225 class RTC_EXPORT AudioSourceInterface : public MediaSourceInterface { 226 public: 227 class AudioObserver { 228 public: 229 virtual void OnSetVolume(double volume) = 0; 230 231 protected: ~AudioObserver()232 virtual ~AudioObserver() {} 233 }; 234 235 // TODO(deadbeef): Makes all the interfaces pure virtual after they're 236 // implemented in chromium. 237 238 // Sets the volume of the source. |volume| is in the range of [0, 10]. 239 // TODO(tommi): This method should be on the track and ideally volume should 240 // be applied in the track in a way that does not affect clones of the track. SetVolume(double volume)241 virtual void SetVolume(double volume) {} 242 243 // Registers/unregisters observers to the audio source. RegisterAudioObserver(AudioObserver * observer)244 virtual void RegisterAudioObserver(AudioObserver* observer) {} UnregisterAudioObserver(AudioObserver * observer)245 virtual void UnregisterAudioObserver(AudioObserver* observer) {} 246 247 // TODO(tommi): Make pure virtual. AddSink(AudioTrackSinkInterface * sink)248 virtual void AddSink(AudioTrackSinkInterface* sink) {} RemoveSink(AudioTrackSinkInterface * sink)249 virtual void RemoveSink(AudioTrackSinkInterface* sink) {} 250 251 // Returns options for the AudioSource. 252 // (for some of the settings this approach is broken, e.g. setting 253 // audio network adaptation on the source is the wrong layer of abstraction). 254 virtual const cricket::AudioOptions options() const; 255 }; 256 257 // Interface of the audio processor used by the audio track to collect 258 // statistics. 259 class AudioProcessorInterface : public rtc::RefCountInterface { 260 public: 261 struct AudioProcessorStatistics { 262 bool typing_noise_detected = false; 263 AudioProcessingStats apm_statistics; 264 }; 265 266 // Get audio processor statistics. The |has_remote_tracks| argument should be 267 // set if there are active remote tracks (this would usually be true during 268 // a call). If there are no remote tracks some of the stats will not be set by 269 // the AudioProcessor, because they only make sense if there is at least one 270 // remote track. 271 virtual AudioProcessorStatistics GetStats(bool has_remote_tracks) = 0; 272 273 protected: 274 ~AudioProcessorInterface() override = default; 275 }; 276 277 class RTC_EXPORT AudioTrackInterface : public MediaStreamTrackInterface { 278 public: 279 // TODO(deadbeef): Figure out if the following interface should be const or 280 // not. 281 virtual AudioSourceInterface* GetSource() const = 0; 282 283 // Add/Remove a sink that will receive the audio data from the track. 284 virtual void AddSink(AudioTrackSinkInterface* sink) = 0; 285 virtual void RemoveSink(AudioTrackSinkInterface* sink) = 0; 286 287 // Get the signal level from the audio track. 288 // Return true on success, otherwise false. 289 // TODO(deadbeef): Change the interface to int GetSignalLevel() and pure 290 // virtual after it's implemented in chromium. 291 virtual bool GetSignalLevel(int* level); 292 293 // Get the audio processor used by the audio track. Return null if the track 294 // does not have any processor. 295 // TODO(deadbeef): Make the interface pure virtual. 296 virtual rtc::scoped_refptr<AudioProcessorInterface> GetAudioProcessor(); 297 298 protected: 299 ~AudioTrackInterface() override = default; 300 }; 301 302 typedef std::vector<rtc::scoped_refptr<AudioTrackInterface> > AudioTrackVector; 303 typedef std::vector<rtc::scoped_refptr<VideoTrackInterface> > VideoTrackVector; 304 305 // C++ version of https://www.w3.org/TR/mediacapture-streams/#mediastream. 306 // 307 // A major difference is that remote audio/video tracks (received by a 308 // PeerConnection/RtpReceiver) are not synchronized simply by adding them to 309 // the same stream; a session description with the correct "a=msid" attributes 310 // must be pushed down. 311 // 312 // Thus, this interface acts as simply a container for tracks. 313 class MediaStreamInterface : public rtc::RefCountInterface, 314 public NotifierInterface { 315 public: 316 virtual std::string id() const = 0; 317 318 virtual AudioTrackVector GetAudioTracks() = 0; 319 virtual VideoTrackVector GetVideoTracks() = 0; 320 virtual rtc::scoped_refptr<AudioTrackInterface> FindAudioTrack( 321 const std::string& track_id) = 0; 322 virtual rtc::scoped_refptr<VideoTrackInterface> FindVideoTrack( 323 const std::string& track_id) = 0; 324 325 virtual bool AddTrack(AudioTrackInterface* track) = 0; 326 virtual bool AddTrack(VideoTrackInterface* track) = 0; 327 virtual bool RemoveTrack(AudioTrackInterface* track) = 0; 328 virtual bool RemoveTrack(VideoTrackInterface* track) = 0; 329 330 protected: 331 ~MediaStreamInterface() override = default; 332 }; 333 334 } // namespace webrtc 335 336 #endif // API_MEDIA_STREAM_INTERFACE_H_ 337