Add experimentalSetCodecsToParseWithinGopSampleDependencies to HLS

HLS extractors were missing experimentalSetCodecsToParseWithinGopSampleDependencies prior to this patch. PiperOrigin-RevId: 730878460 (cherry picked from commit da402cfd64ef3ae60607491730f01138177a077b)
2025-04-29 22:36:54 +08:00 · 2025-02-25 07:29:54 -08:00 · 2025-02-25 07:29:54 -08:00 · 521c385d4b
commit 521c385d4b
parent a578d43324
6 changed files with 306 additions and 2 deletions
--- a/demos/main/src/main/assets/media.exolist.json
+++ b/demos/main/src/main/assets/media.exolist.json
@ -257,6 +257,10 @@
      {
        "name": "Apple media playlist (AAC)",
        "uri": "https://devstreaming-cdn.apple.com/videos/streaming/examples/bipbop_4x3/gear0/prog_index.m3u8"
+      },
+      {
+        "name": "Bitmovin (FMP4)",
+        "uri": "https://bitdash-a.akamaihd.net/content/MI201109210084_1/m3u8s-fmp4/f08e80da-bf1d-4e3d-8899-f0f6155f6efa.m3u8"
      }
    ]
  },
--- a/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/DefaultHlsExtractorFactory.java
+++ b/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/DefaultHlsExtractorFactory.java
@ -22,6 +22,7 @@ import android.annotation.SuppressLint;
 import android.net.Uri;
 import android.text.TextUtils;
 import androidx.annotation.Nullable;
+import androidx.media3.common.C;
 import androidx.media3.common.FileTypes;
 import androidx.media3.common.Format;
 import androidx.media3.common.Metadata;
@ -71,6 +72,7 @@ public final class DefaultHlsExtractorFactory implements HlsExtractorFactory {

  private SubtitleParser.Factory subtitleParserFactory;
  private boolean parseSubtitlesDuringExtraction;
+  private @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies;

  private final boolean exposeCea608WhenMissingDeclarations;

@ -179,6 +181,14 @@ public final class DefaultHlsExtractorFactory implements HlsExtractorFactory {
    return this;
  }

+  @CanIgnoreReturnValue
+  @Override
+  public DefaultHlsExtractorFactory experimentalSetCodecsToParseWithinGopSampleDependencies(
+      @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies) {
+    this.codecsToParseWithinGopSampleDependencies = codecsToParseWithinGopSampleDependencies;
+    return this;
+  }
+
  /**
   * {@inheritDoc}
   *
@ -242,7 +252,8 @@ public final class DefaultHlsExtractorFactory implements HlsExtractorFactory {
            parseSubtitlesDuringExtraction,
            timestampAdjuster,
            format,
-            muxedCaptionFormats);
+            muxedCaptionFormats,
+            codecsToParseWithinGopSampleDependencies);
      case FileTypes.TS:
        return createTsExtractor(
            payloadReaderFactoryFlags,
@ -313,7 +324,8 @@ public final class DefaultHlsExtractorFactory implements HlsExtractorFactory {
      boolean parseSubtitlesDuringExtraction,
      TimestampAdjuster timestampAdjuster,
      Format format,
-      @Nullable List<Format> muxedCaptionFormats) {
+      @Nullable List<Format> muxedCaptionFormats,
+      @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies) {
    // Only enable the EMSG TrackOutput if this is the 'variant' track (i.e. the main one) to avoid
    // creating a separate EMSG track for every audio track in a video stream.
    @FragmentedMp4Extractor.Flags
@ -322,6 +334,9 @@ public final class DefaultHlsExtractorFactory implements HlsExtractorFactory {
      subtitleParserFactory = SubtitleParser.Factory.UNSUPPORTED;
      flags |= FragmentedMp4Extractor.FLAG_EMIT_RAW_SUBTITLE_DATA;
    }
+    flags |=
+        FragmentedMp4Extractor.codecsToParseWithinGopSampleDependenciesAsFlags(
+            codecsToParseWithinGopSampleDependencies);
    return new FragmentedMp4Extractor(
        subtitleParserFactory,
        flags,
--- a/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/HlsExtractorFactory.java
+++ b/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/HlsExtractorFactory.java
@ -26,6 +26,7 @@ import androidx.media3.exoplayer.analytics.PlayerId;
 import androidx.media3.extractor.Extractor;
 import androidx.media3.extractor.ExtractorInput;
 import androidx.media3.extractor.PositionHolder;
+import androidx.media3.extractor.mp4.FragmentedMp4Extractor;
 import androidx.media3.extractor.text.SubtitleParser;
 import com.google.errorprone.annotations.CanIgnoreReturnValue;
 import java.io.IOException;
@ -101,6 +102,25 @@ public interface HlsExtractorFactory {
    return this;
  }

+  /**
+   * Sets the set of video codecs for which within GOP sample dependency information should be
+   * parsed as part of extraction. Defaults to {@code 0} - empty set of codecs.
+   *
+   * <p>Having access to additional sample dependency information can speed up seeking. See {@link
+   * FragmentedMp4Extractor#FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES}.
+   *
+   * <p>This method is experimental and will be renamed or removed in a future release.
+   *
+   * @param codecsToParseWithinGopSampleDependencies The set of codecs for which to parse within GOP
+   *     sample dependency information.
+   * @return This factory, for convenience.
+   */
+  @CanIgnoreReturnValue
+  default HlsExtractorFactory experimentalSetCodecsToParseWithinGopSampleDependencies(
+      @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies) {
+    return this;
+  }
+
  /**
   * Returns the output {@link Format} of emitted {@linkplain C#TRACK_TYPE_TEXT text samples} which
   * were originally in {@code sourceFormat}.
--- a/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/HlsMediaSource.java
+++ b/libraries/exoplayer_hls/src/main/java/androidx/media3/exoplayer/hls/HlsMediaSource.java
@ -109,6 +109,7 @@ public final class HlsMediaSource extends BaseMediaSource
    @Nullable private HlsExtractorFactory extractorFactory;
    @Nullable private SubtitleParser.Factory subtitleParserFactoryOverride;
    private boolean parseSubtitlesDuringExtraction;
+    private @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies;
    private HlsPlaylistParserFactory playlistParserFactory;
    private HlsPlaylistTracker.Factory playlistTrackerFactory;
    private CompositeSequenceableLoaderFactory compositeSequenceableLoaderFactory;
@ -220,6 +221,14 @@ public final class HlsMediaSource extends BaseMediaSource
      return this;
    }

+    @Override
+    @CanIgnoreReturnValue
+    public Factory experimentalSetCodecsToParseWithinGopSampleDependencies(
+        @C.VideoCodecFlags int codecsToParseWithinGopSampleDependencies) {
+      this.codecsToParseWithinGopSampleDependencies = codecsToParseWithinGopSampleDependencies;
+      return this;
+    }
+
    /**
     * Sets the factory from which playlist parsers will be obtained.
     *
@ -396,6 +405,8 @@ public final class HlsMediaSource extends BaseMediaSource
        extractorFactory.setSubtitleParserFactory(subtitleParserFactoryOverride);
      }
      extractorFactory.experimentalParseSubtitlesDuringExtraction(parseSubtitlesDuringExtraction);
+      extractorFactory.experimentalSetCodecsToParseWithinGopSampleDependencies(
+          codecsToParseWithinGopSampleDependencies);
      HlsExtractorFactory extractorFactory = this.extractorFactory;
      HlsPlaylistParserFactory playlistParserFactory = this.playlistParserFactory;
      List<StreamKey> streamKeys = mediaItem.localConfiguration.streamKeys;
--- a/libraries/exoplayer_hls/src/test/java/androidx/media3/exoplayer/hls/e2etest/HlsPlaybackTest.java
+++ b/libraries/exoplayer_hls/src/test/java/androidx/media3/exoplayer/hls/e2etest/HlsPlaybackTest.java
@ -29,6 +29,7 @@ import android.graphics.SurfaceTexture;
 import android.net.Uri;
 import android.view.Surface;
 import androidx.annotation.Nullable;
+import androidx.media3.common.C;
 import androidx.media3.common.MediaItem;
 import androidx.media3.common.Player;
 import androidx.media3.datasource.DefaultDataSource;
@ -42,6 +43,7 @@ import androidx.media3.exoplayer.source.DefaultMediaSourceFactory;
 import androidx.media3.exoplayer.source.LoadEventInfo;
 import androidx.media3.exoplayer.source.MediaLoadData;
 import androidx.media3.exoplayer.upstream.CmcdConfiguration;
+import androidx.media3.extractor.DefaultExtractorsFactory;
 import androidx.media3.test.utils.CapturingRenderersFactory;
 import androidx.media3.test.utils.DumpFileAsserts;
 import androidx.media3.test.utils.FakeClock;
@ -464,6 +466,39 @@ public final class HlsPlaybackTest {
    assertThat(loadCompletedDataSpecUris).containsExactlyElementsIn(loadStartedUris);
  }

+  @Test
+  public void playVideo_usingWithinGopSampleDependencies_withSeek() throws Exception {
+    Context applicationContext = ApplicationProvider.getApplicationContext();
+    CapturingRenderersFactory capturingRenderersFactory =
+        new CapturingRenderersFactory(applicationContext);
+    DefaultMediaSourceFactory defaultMediaSourceFactory =
+        new DefaultMediaSourceFactory(applicationContext, new DefaultExtractorsFactory());
+    defaultMediaSourceFactory.experimentalSetCodecsToParseWithinGopSampleDependencies(
+        C.VIDEO_CODEC_FLAG_H264);
+    ExoPlayer player =
+        new ExoPlayer.Builder(
+                applicationContext, capturingRenderersFactory, defaultMediaSourceFactory)
+            .setClock(new FakeClock(/* isAutoAdvancing= */ true))
+            .build();
+    Surface surface = new Surface(new SurfaceTexture(/* texName= */ 1));
+    player.setVideoSurface(surface);
+    PlaybackOutput playbackOutput = PlaybackOutput.register(player, capturingRenderersFactory);
+
+    player.setMediaItem(
+        MediaItem.fromUri("asset:///media/hls/standalone-webvtt/multivariant_playlist.m3u8"));
+    player.seekTo(500L);
+    player.prepare();
+    player.play();
+    advance(player).untilState(Player.STATE_ENDED);
+    player.release();
+    surface.release();
+
+    DumpFileAsserts.assertOutput(
+        applicationContext,
+        playbackOutput,
+        "playbackdumps/hls/standalone-webvtt-optimized-seek.dump");
+  }
+
  private static class AnalyticsListenerImpl implements AnalyticsListener {

    @Nullable private LoadEventInfo loadErrorEventInfo;
--- a/libraries/test_data/src/test/assets/playbackdumps/hls/standalone-webvtt-optimized-seek.dump
+++ b/libraries/test_data/src/test/assets/playbackdumps/hls/standalone-webvtt-optimized-seek.dump
@ -0,0 +1,219 @@
+MediaCodecAdapter (exotest.video.avc):
+  inputBuffers:
+    count = 24
+    input buffer #0:
+      timeUs = 1000000000000
+      contents = length 36692, hash D216076E
+    input buffer #1:
+      timeUs = 1000000066733
+      contents = length 5312, hash D45D3CA0
+    input buffer #2:
+      timeUs = 1000000200200
+      contents = length 7735, hash 4490F110
+    input buffer #3:
+      timeUs = 1000000133466
+      contents = length 987, hash 560B5036
+    input buffer #4:
+      timeUs = 1000000333666
+      contents = length 6061, hash 736C72B2
+    input buffer #5:
+      timeUs = 1000000266933
+      contents = length 992, hash FE132F23
+    input buffer #6:
+      timeUs = 1000000433766
+      contents = length 4899, hash F72F86A1
+    input buffer #7:
+      timeUs = 1000000400400
+      contents = length 568, hash 519A8E50
+    input buffer #8:
+      timeUs = 1000000567233
+      contents = length 5450, hash F06EC4AA
+    input buffer #9:
+      timeUs = 1000000500500
+      contents = length 1051, hash 92DFA63A
+    input buffer #10:
+      timeUs = 1000000533866
+      contents = length 781, hash 36BE495B
+    input buffer #11:
+      timeUs = 1000000700700
+      contents = length 4725, hash AC0C8CD3
+    input buffer #12:
+      timeUs = 1000000633966
+      contents = length 1022, hash 5D8BFF34
+    input buffer #13:
+      timeUs = 1000000600600
+      contents = length 790, hash 99413A99
+    input buffer #14:
+      timeUs = 1000000667333
+      contents = length 610, hash 5E129290
+    input buffer #15:
+      timeUs = 1000000834166
+      contents = length 2751, hash 769974CB
+    input buffer #16:
+      timeUs = 1000000767433
+      contents = length 745, hash B78A477A
+    input buffer #17:
+      timeUs = 1000000734066
+      contents = length 621, hash CF741E7A
+    input buffer #18:
+      timeUs = 1000000800800
+      contents = length 505, hash 1DB4894E
+    input buffer #19:
+      timeUs = 1000000967633
+      contents = length 1268, hash C15348DC
+    input buffer #20:
+      timeUs = 1000000900900
+      contents = length 880, hash C2DE85D0
+    input buffer #21:
+      timeUs = 1000000867533
+      contents = length 530, hash C98BC6A8
+    input buffer #22:
+      timeUs = 1000000934266
+      contents = length 568, hash 4FE5C8EA
+    input buffer #23:
+      timeUs = 0
+      flags = 4
+      contents = length 0, hash 1
+  outputBuffers:
+    count = 23
+    output buffer #0:
+      timeUs = 1000000000000
+      size = 36692
+      rendered = false
+    output buffer #1:
+      timeUs = 1000000066733
+      size = 5312
+      rendered = false
+    output buffer #2:
+      timeUs = 1000000200200
+      size = 7735
+      rendered = false
+    output buffer #3:
+      timeUs = 1000000133466
+      size = 987
+      rendered = false
+    output buffer #4:
+      timeUs = 1000000333666
+      size = 6061
+      rendered = false
+    output buffer #5:
+      timeUs = 1000000266933
+      size = 992
+      rendered = false
+    output buffer #6:
+      timeUs = 1000000433766
+      size = 4899
+      rendered = false
+    output buffer #7:
+      timeUs = 1000000400400
+      size = 568
+      rendered = false
+    output buffer #8:
+      timeUs = 1000000567233
+      size = 5450
+      rendered = true
+    output buffer #9:
+      timeUs = 1000000500500
+      size = 1051
+      rendered = true
+    output buffer #10:
+      timeUs = 1000000533866
+      size = 781
+      rendered = true
+    output buffer #11:
+      timeUs = 1000000700700
+      size = 4725
+      rendered = true
+    output buffer #12:
+      timeUs = 1000000633966
+      size = 1022
+      rendered = true
+    output buffer #13:
+      timeUs = 1000000600600
+      size = 790
+      rendered = true
+    output buffer #14:
+      timeUs = 1000000667333
+      size = 610
+      rendered = true
+    output buffer #15:
+      timeUs = 1000000834166
+      size = 2751
+      rendered = true
+    output buffer #16:
+      timeUs = 1000000767433
+      size = 745
+      rendered = true
+    output buffer #17:
+      timeUs = 1000000734066
+      size = 621
+      rendered = true
+    output buffer #18:
+      timeUs = 1000000800800
+      size = 505
+      rendered = true
+    output buffer #19:
+      timeUs = 1000000967633
+      size = 1268
+      rendered = true
+    output buffer #20:
+      timeUs = 1000000900900
+      size = 880
+      rendered = true
+    output buffer #21:
+      timeUs = 1000000867533
+      size = 530
+      rendered = true
+    output buffer #22:
+      timeUs = 1000000934266
+      size = 568
+      rendered = true
+TextOutput:
+  Subtitle[0]:
+    presentationTimeUs = 500000
+    Cues = []
+  Subtitle[1]:
+    presentationTimeUs = 567000
+    Cue[0]:
+      text = This is the first subtitle.
+      textAlignment = ALIGN_CENTER
+      line = -1.0
+      lineType = 1
+      lineAnchor = 0
+      position = 0.5
+      positionAnchor = 1
+      size = 1.0
+  Subtitle[2]:
+    presentationTimeUs = 667000
+    Cue[0]:
+      text = This is the first subtitle.
+      textAlignment = ALIGN_CENTER
+      line = -1.0
+      lineType = 1
+      lineAnchor = 0
+      position = 0.5
+      positionAnchor = 1
+      size = 1.0
+    Cue[1]:
+      text = This is the second subtitle.
+      textAlignment = ALIGN_CENTER
+      line = -2.0
+      lineType = 1
+      lineAnchor = 0
+      position = 0.5
+      positionAnchor = 1
+      size = 1.0
+  Subtitle[3]:
+    presentationTimeUs = 801000
+    Cue[0]:
+      text = This is the second subtitle.
+      textAlignment = ALIGN_CENTER
+      line = -1.0
+      lineType = 1
+      lineAnchor = 0
+      position = 0.5
+      positionAnchor = 1
+      size = 1.0
+  Subtitle[4]:
+    presentationTimeUs = 951000
+    Cues = []