media_stream: add frame size

This will help once audio streaming is implemented, as the encoder will
need to know the number of samples per frame in the decoded file.

Change-Id: I6d6375b46d74eec6618c0f1e7ee1f07fd86b6e45

src/media/audio/audio_rtp_session.cpp

@@ -161,7 +161,7 @@ AudioSender::update(Observable<std::shared_ptr<ring::AudioFrame>>* /*obs*/, cons
 {
     auto frame = framePtr->pointer();
     auto ms = MediaStream("a:local", frame->format, rational<int>(1, frame->sample_rate),
-                          frame->sample_rate, frame->channels);
+                          frame->sample_rate, frame->channels, frame->nb_samples);
     frame->pts = sent_samples;
     ms.firstTimestamp = frame->pts;
     sent_samples += frame->nb_samples;

src/media/media_stream.h

@@ -40,6 +40,7 @@ struct MediaStream {
     rational<int> frameRate;
     int sampleRate {0};
     int nbChannels {0};
+    int frameSize {0};
 
     MediaStream()
     {}
@@ -56,13 +57,14 @@ struct MediaStream {
         , frameRate(fr)
     {}
 
-    MediaStream(const std::string& streamName, int fmt, rational<int> tb, int sr, int channels)
+    MediaStream(const std::string& streamName, int fmt, rational<int> tb, int sr, int channels, int size)
         : name(streamName)
         , format(fmt)
         , isVideo(false)
         , timeBase(tb)
         , sampleRate(sr)
         , nbChannels(channels)
+        , frameSize(size)
     {}
 
     MediaStream(const std::string& streamName, AudioFormat fmt)
@@ -72,6 +74,7 @@ struct MediaStream {
         , timeBase(1, fmt.sample_rate)
         , sampleRate(fmt.sample_rate)
         , nbChannels(fmt.nb_channels)
+        , frameSize(fmt.sample_rate / 50) // standard frame size for our encoder is 20 ms
     {}
 
     MediaStream(const std::string& streamName, AVCodecContext* c)
@@ -97,6 +100,7 @@ struct MediaStream {
             isVideo = false;
             sampleRate = c->sample_rate;
             nbChannels = c->channels;
+            frameSize = c->frame_size;
             break;
         default:
             break;
@@ -127,6 +131,8 @@ struct MediaStream {
             sampleRate = f->sample_rate;
             nbChannels = f->channels;
             timeBase = rational<int>(1, f->sample_rate);
+            if (!frameSize)
+                frameSize = f->nb_samples;
         }
     }
 };
@@ -142,7 +148,8 @@ inline std::ostream& operator<<(std::ostream& os, const MediaStream& ms)
         os << (ms.name.empty() ? "(null)" : ms.name) << ": "
             << av_get_sample_fmt_name(static_cast<AVSampleFormat>(ms.format)) << " audio, "
             << ms.nbChannels << " channel(s), "
-            << ms.sampleRate << " Hz (" << ms.timeBase << ")";
+            << ms.sampleRate << " Hz (" << ms.timeBase << "), "
+            << ms.frameSize << " samples per frame";
     }
     if (ms.firstTimestamp > 0)
         os << ", start: " << ms.firstTimestamp;

test/unitTest/media/test_media_filter.cpp

@@ -117,7 +117,7 @@ fillAudioFrameProps(AVFrame* frame, const MediaStream& ms)
 {
     frame->format = ms.format;
     frame->channel_layout = av_get_default_channel_layout(ms.nbChannels);
-    frame->nb_samples = 960;
+    frame->nb_samples = ms.frameSize;
     frame->sample_rate = ms.sampleRate;
     frame->channels = ms.nbChannels;
     CPPUNIT_ASSERT(frame->format > AV_SAMPLE_FMT_NONE);
@@ -146,7 +146,7 @@ MediaFilterTest::testAudioFilter()
     frame->channels = av_get_channel_layout_nb_channels(channelLayout);
 
     // construct the filter parameters
-    auto params = MediaStream("in1", format, rational<int>(1, sampleRate), sampleRate, frame->channels);
+    auto params = MediaStream("in1", format, rational<int>(1, sampleRate), sampleRate, frame->channels, nbSamples);
 
     // allocate and fill frame buffers
     CPPUNIT_ASSERT(av_frame_get_buffer(frame, 0) >= 0);
@@ -180,9 +180,9 @@ MediaFilterTest::testAudioMixing()
     auto frame3 = af3.pointer();
 
     std::vector<MediaStream> vec;
-    vec.emplace_back("a1", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2);
-    vec.emplace_back("a2", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2);
-    vec.emplace_back("a3", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2);
+    vec.emplace_back("a1", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2, 960);
+    vec.emplace_back("a2", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2, 960);
+    vec.emplace_back("a3", AV_SAMPLE_FMT_S16, rational<int>(1, 48000), 48000, 2, 960);
     CPPUNIT_ASSERT(filter_->initialize(filterSpec, vec) >= 0);
 
     float t1 = 0, t2 = 0, t3 = 0;
@@ -324,7 +324,7 @@ MediaFilterTest::testReinit()
     frame->channels = 2;
 
     // construct the filter parameters with different sample rate
-    auto params = MediaStream("in1", frame->format, rational<int>(1, 16000), 16000, frame->channels);
+    auto params = MediaStream("in1", frame->format, rational<int>(1, 16000), 16000, frame->channels, frame->nb_samples);
 
     // allocate and fill frame buffers
     CPPUNIT_ASSERT(av_frame_get_buffer(frame, 0) >= 0);