Merge pull request #973 from jan-varecka-signageos-io:feat/extractTsA…

…udioType

PiperOrigin-RevId: 599547201

RELEASENOTES.md

@@ -19,6 +19,9 @@
         than if we ignore it and assume the file is CBR.
     *   MPEG2-TS: Add DTS, DTS-LBR and DTS:X Profile2 support
         ([#275](https://github.com/androidx/media/pull/275)).
+    *   Extract audio types from TS descriptors and map them to role flags,
+        allowing users to make better-informed audio track selections
+        ([#973](https://github.com/androidx/media/pull/973)).
 *   Audio:
 *   Video:
 *   Text:

libraries/extractor/src/main/java/androidx/media3/extractor/DtsUtil.java

@@ -258,13 +258,15 @@ private DtsHeader(
    * @param frame The DTS Core frame to parse.
    * @param trackId The track identifier to set on the format.
    * @param language The language to set on the format.
+   * @param roleFlags The role flags to set on the format.
    * @param drmInitData {@link DrmInitData} to be included in the format.
    * @return The DTS format parsed from data in the header.
    */
   public static Format parseDtsFormat(
       byte[] frame,
       @Nullable String trackId,
       @Nullable String language,
+      @C.RoleFlags int roleFlags,
       @Nullable DrmInitData drmInitData) {
     ParsableBitArray frameBits = getNormalizedFrame(frame);
     frameBits.skipBits(32 + 1 + 5 + 1 + 7 + 14); // SYNC, FTYPE, SHORT, CPF, NBLKS, FSIZE
@@ -287,6 +289,7 @@ public static Format parseDtsFormat(
         .setSampleRate(sampleRate)
         .setDrmInitData(drmInitData)
         .setLanguage(language)
+        .setRoleFlags(roleFlags)
         .build();
   }
 

libraries/extractor/src/main/java/androidx/media3/extractor/ts/Ac3Reader.java

@@ -60,6 +60,7 @@ public final class Ac3Reader implements ElementaryStreamReader {
   private final ParsableBitArray headerScratchBits;
   private final ParsableByteArray headerScratchBytes;
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
 
   private @MonotonicNonNull String formatId;
   private @MonotonicNonNull TrackOutput output;
@@ -80,20 +81,22 @@ public final class Ac3Reader implements ElementaryStreamReader {
 
   /** Constructs a new reader for (E-)AC-3 elementary streams. */
   public Ac3Reader() {
-    this(null);
+    this(null, /* roleFlags= */ 0);
   }
 
   /**
    * Constructs a new reader for (E-)AC-3 elementary streams.
    *
    * @param language Track language.
+   * @param roleFlags Track role flags.
    */
-  public Ac3Reader(@Nullable String language) {
+  public Ac3Reader(@Nullable String language, @C.RoleFlags int roleFlags) {
     headerScratchBits = new ParsableBitArray(new byte[HEADER_SIZE]);
     headerScratchBytes = new ParsableByteArray(headerScratchBits.data);
     state = STATE_FINDING_SYNC;
     timeUs = C.TIME_UNSET;
     this.language = language;
+    this.roleFlags = roleFlags;
   }
 
   @Override
@@ -216,6 +219,7 @@ private void parseHeader() {
               .setChannelCount(frameInfo.channelCount)
               .setSampleRate(frameInfo.sampleRate)
               .setLanguage(language)
+              .setRoleFlags(roleFlags)
               .setPeakBitrate(frameInfo.bitrate);
       // AC3 has constant bitrate, so averageBitrate = peakBitrate
       if (MimeTypes.AUDIO_AC3.equals(frameInfo.mimeType)) {

libraries/extractor/src/main/java/androidx/media3/extractor/ts/Ac4Reader.java

@@ -57,6 +57,7 @@ public final class Ac4Reader implements ElementaryStreamReader {
   private final ParsableBitArray headerScratchBits;
   private final ParsableByteArray headerScratchBytes;
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
 
   private @MonotonicNonNull String formatId;
   private @MonotonicNonNull TrackOutput output;
@@ -78,15 +79,16 @@ public final class Ac4Reader implements ElementaryStreamReader {
 
   /** Constructs a new reader for AC-4 elementary streams. */
   public Ac4Reader() {
-    this(null);
+    this(null, /* roleFlags= */ 0);
   }
 
   /**
    * Constructs a new reader for AC-4 elementary streams.
    *
    * @param language Track language.
+   * @param roleFlags Track role flags.
    */
-  public Ac4Reader(@Nullable String language) {
+  public Ac4Reader(@Nullable String language, @C.RoleFlags int roleFlags) {
     headerScratchBits = new ParsableBitArray(new byte[Ac4Util.HEADER_SIZE_FOR_PARSER]);
     headerScratchBytes = new ParsableByteArray(headerScratchBits.data);
     state = STATE_FINDING_SYNC;
@@ -95,6 +97,7 @@ public Ac4Reader(@Nullable String language) {
     hasCRC = false;
     timeUs = C.TIME_UNSET;
     this.language = language;
+    this.roleFlags = roleFlags;
   }
 
   @Override
@@ -217,6 +220,7 @@ private void parseHeader() {
               .setChannelCount(frameInfo.channelCount)
               .setSampleRate(frameInfo.sampleRate)
               .setLanguage(language)
+              .setRoleFlags(roleFlags)
               .build();
       output.format(format);
     }

libraries/extractor/src/main/java/androidx/media3/extractor/ts/AdtsReader.java

@@ -71,6 +71,7 @@ public final class AdtsReader implements ElementaryStreamReader {
   private final ParsableBitArray adtsScratch;
   private final ParsableByteArray id3HeaderBuffer;
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
 
   private @MonotonicNonNull String formatId;
   private @MonotonicNonNull TrackOutput output;
@@ -105,14 +106,15 @@ public final class AdtsReader implements ElementaryStreamReader {
    * @param exposeId3 True if the reader should expose ID3 information.
    */
   public AdtsReader(boolean exposeId3) {
-    this(exposeId3, null);
+    this(exposeId3, null, /* roleFlags= */ 0);
   }
 
   /**
    * @param exposeId3 True if the reader should expose ID3 information.
    * @param language Track language.
+   * @param roleFlags Track role flags.
    */
-  public AdtsReader(boolean exposeId3, @Nullable String language) {
+  public AdtsReader(boolean exposeId3, @Nullable String language, @C.RoleFlags int roleFlags) {
     adtsScratch = new ParsableBitArray(new byte[HEADER_SIZE + CRC_SIZE]);
     id3HeaderBuffer = new ParsableByteArray(Arrays.copyOf(ID3_IDENTIFIER, ID3_HEADER_SIZE));
     setFindingSampleState();
@@ -122,6 +124,7 @@ public AdtsReader(boolean exposeId3, @Nullable String language) {
     timeUs = C.TIME_UNSET;
     this.exposeId3 = exposeId3;
     this.language = language;
+    this.roleFlags = roleFlags;
   }
 
   /** Returns whether an integer matches an ADTS SYNC word. */
@@ -510,6 +513,7 @@ private void parseAdtsHeader() throws ParserException {
               .setSampleRate(aacConfig.sampleRateHz)
               .setInitializationData(Collections.singletonList(audioSpecificConfig))
               .setLanguage(language)
+              .setRoleFlags(roleFlags)
               .build();
       // In this class a sample is an access unit, but the MediaFormat sample rate specifies the
       // number of PCM audio samples per second.

libraries/extractor/src/main/java/androidx/media3/extractor/ts/DefaultTsPayloadReaderFactory.java

@@ -151,30 +151,32 @@ public TsPayloadReader createPayloadReader(int streamType, EsInfo esInfo) {
     switch (streamType) {
       case TsExtractor.TS_STREAM_TYPE_MPA:
       case TsExtractor.TS_STREAM_TYPE_MPA_LSF:
-        return new PesReader(new MpegAudioReader(esInfo.language));
+        return new PesReader(new MpegAudioReader(esInfo.language, esInfo.getRoleFlags()));
       case TsExtractor.TS_STREAM_TYPE_AAC_ADTS:
         return isSet(FLAG_IGNORE_AAC_STREAM)
             ? null
-            : new PesReader(new AdtsReader(false, esInfo.language));
+            : new PesReader(new AdtsReader(false, esInfo.language, esInfo.getRoleFlags()));
       case TsExtractor.TS_STREAM_TYPE_AAC_LATM:
         return isSet(FLAG_IGNORE_AAC_STREAM)
             ? null
-            : new PesReader(new LatmReader(esInfo.language));
+            : new PesReader(new LatmReader(esInfo.language, esInfo.getRoleFlags()));
       case TsExtractor.TS_STREAM_TYPE_AC3:
       case TsExtractor.TS_STREAM_TYPE_E_AC3:
-        return new PesReader(new Ac3Reader(esInfo.language));
+        return new PesReader(new Ac3Reader(esInfo.language, esInfo.getRoleFlags()));
       case TsExtractor.TS_STREAM_TYPE_AC4:
-        return new PesReader(new Ac4Reader(esInfo.language));
+        return new PesReader(new Ac4Reader(esInfo.language, esInfo.getRoleFlags()));
       case TsExtractor.TS_STREAM_TYPE_HDMV_DTS:
         if (!isSet(FLAG_ENABLE_HDMV_DTS_AUDIO_STREAMS)) {
           return null;
         }
         // Fall through.
       case TsExtractor.TS_STREAM_TYPE_DTS:
       case TsExtractor.TS_STREAM_TYPE_DTS_HD:
-        return new PesReader(new DtsReader(esInfo.language, DtsReader.EXTSS_HEADER_SIZE_MAX));
+        return new PesReader(
+            new DtsReader(esInfo.language, esInfo.getRoleFlags(), DtsReader.EXTSS_HEADER_SIZE_MAX));
       case TsExtractor.TS_STREAM_TYPE_DTS_UHD:
-        return new PesReader(new DtsReader(esInfo.language, DtsReader.FTOC_MAX_HEADER_SIZE));
+        return new PesReader(
+            new DtsReader(esInfo.language, esInfo.getRoleFlags(), DtsReader.FTOC_MAX_HEADER_SIZE));
       case TsExtractor.TS_STREAM_TYPE_H262:
       case TsExtractor.TS_STREAM_TYPE_DC2_H262:
         return new PesReader(new H262Reader(buildUserDataReader(esInfo)));

libraries/extractor/src/main/java/androidx/media3/extractor/ts/DtsReader.java

@@ -68,6 +68,7 @@ public final class DtsReader implements ElementaryStreamReader {
   private final AtomicInteger uhdAudioChunkId;
 
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
 
   private @MonotonicNonNull String formatId;
   private @MonotonicNonNull TrackOutput output;
@@ -93,16 +94,18 @@ public final class DtsReader implements ElementaryStreamReader {
    * Constructs a new reader for DTS elementary streams.
    *
    * @param language Track language.
+   * @param roleFlags Track role flags.
    * @param maxHeaderSize Maximum size of the header in a frame.
    */
-  public DtsReader(@Nullable String language, int maxHeaderSize) {
+  public DtsReader(@Nullable String language, @C.RoleFlags int roleFlags, int maxHeaderSize) {
     headerScratchBytes = new ParsableByteArray(new byte[maxHeaderSize]);
     state = STATE_FINDING_SYNC;
     timeUs = C.TIME_UNSET;
     uhdAudioChunkId = new AtomicInteger();
     extensionSubstreamHeaderSize = C.LENGTH_UNSET;
     uhdHeaderSize = C.LENGTH_UNSET;
     this.language = language;
+    this.roleFlags = roleFlags;
   }
 
   @Override
@@ -263,7 +266,7 @@ private boolean skipToNextSyncWord(ParsableByteArray pesBuffer) {
   private void parseCoreHeader() {
     byte[] frameData = headerScratchBytes.getData();
     if (format == null) {
-      format = DtsUtil.parseDtsFormat(frameData, formatId, language, null);
+      format = DtsUtil.parseDtsFormat(frameData, formatId, language, roleFlags, null);
       output.format(format);
     }
     sampleSize = DtsUtil.getDtsFrameSize(frameData);
@@ -314,6 +317,7 @@ private void updateFormatWithDtsHeaderInfo(DtsUtil.DtsHeader dtsHeader) {
               .setChannelCount(dtsHeader.channelCount)
               .setSampleRate(dtsHeader.sampleRate)
               .setLanguage(language)
+              .setRoleFlags(roleFlags)
               .build();
       output.format(format);
     }

libraries/extractor/src/main/java/androidx/media3/extractor/ts/LatmReader.java

@@ -49,6 +49,7 @@ public final class LatmReader implements ElementaryStreamReader {
   private static final int SYNC_BYTE_SECOND = 0xE0;
 
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
   private final ParsableByteArray sampleDataBuffer;
   private final ParsableBitArray sampleBitArray;
 
@@ -78,9 +79,11 @@ public final class LatmReader implements ElementaryStreamReader {
 
   /**
    * @param language Track language.
+   * @param roleFlags Track role flags.
    */
-  public LatmReader(@Nullable String language) {
+  public LatmReader(@Nullable String language, @C.RoleFlags int roleFlags) {
     this.language = language;
+    this.roleFlags = roleFlags;
     sampleDataBuffer = new ParsableByteArray(INITIAL_BUFFER_SIZE);
     sampleBitArray = new ParsableBitArray(sampleDataBuffer.getData());
     timeUs = C.TIME_UNSET;
@@ -217,6 +220,7 @@ private void parseStreamMuxConfig(ParsableBitArray data) throws ParserException
                 .setSampleRate(sampleRateHz)
                 .setInitializationData(Collections.singletonList(initData))
                 .setLanguage(language)
+                .setRoleFlags(roleFlags)
                 .build();
         if (!format.equals(this.format)) {
           this.format = format;

libraries/extractor/src/main/java/androidx/media3/extractor/ts/MpegAudioReader.java

@@ -44,6 +44,7 @@ public final class MpegAudioReader implements ElementaryStreamReader {
   private final ParsableByteArray headerScratch;
   private final MpegAudioUtil.Header header;
   @Nullable private final String language;
+  private final @C.RoleFlags int roleFlags;
 
   private @MonotonicNonNull TrackOutput output;
   private @MonotonicNonNull String formatId;
@@ -63,17 +64,18 @@ public final class MpegAudioReader implements ElementaryStreamReader {
   private long timeUs;
 
   public MpegAudioReader() {
-    this(null);
+    this(null, /* roleFlags= */ 0);
   }
 
-  public MpegAudioReader(@Nullable String language) {
+  public MpegAudioReader(@Nullable String language, @C.RoleFlags int roleFlags) {
     state = STATE_FINDING_HEADER;
     // The first byte of an MPEG Audio frame header is always 0xFF.
     headerScratch = new ParsableByteArray(4);
     headerScratch.getData()[0] = (byte) 0xFF;
     header = new MpegAudioUtil.Header();
     timeUs = C.TIME_UNSET;
     this.language = language;
+    this.roleFlags = roleFlags;
   }
 
   @Override
@@ -200,6 +202,7 @@ private void readHeaderRemainder(ParsableByteArray source) {
               .setChannelCount(header.channels)
               .setSampleRate(header.sampleRate)
               .setLanguage(language)
+              .setRoleFlags(roleFlags)
               .build();
       output.format(format);
       hasOutputFormat = true;

libraries/extractor/src/main/java/androidx/media3/extractor/ts/TsExtractor.java

@@ -15,6 +15,7 @@
  */
 package androidx.media3.extractor.ts;
 
+import static androidx.media3.extractor.ts.TsPayloadReader.EsInfo.AUDIO_TYPE_UNDEFINED;
 import static androidx.media3.extractor.ts.TsPayloadReader.FLAG_PAYLOAD_UNIT_START_INDICATOR;
 import static java.lang.annotation.ElementType.TYPE_USE;
 
@@ -752,7 +753,8 @@ public void consume(ParsableByteArray sectionData) {
       if (mode == MODE_HLS && id3Reader == null) {
         // Setup an ID3 track regardless of whether there's a corresponding entry, in case one
         // appears intermittently during playback. See [Internal: b/20261500].
-        EsInfo id3EsInfo = new EsInfo(TS_STREAM_TYPE_ID3, null, null, Util.EMPTY_BYTE_ARRAY);
+        EsInfo id3EsInfo =
+            new EsInfo(TS_STREAM_TYPE_ID3, null, AUDIO_TYPE_UNDEFINED, null, Util.EMPTY_BYTE_ARRAY);
         id3Reader = payloadReaderFactory.createPayloadReader(TS_STREAM_TYPE_ID3, id3EsInfo);
         if (id3Reader != null) {
           id3Reader.init(
@@ -842,6 +844,7 @@ private EsInfo readEsInfo(ParsableByteArray data, int length) {
       int descriptorsEndPosition = descriptorsStartPosition + length;
       int streamType = -1;
       String language = null;
+      @EsInfo.AudioType int audioType = AUDIO_TYPE_UNDEFINED;
       List<DvbSubtitleInfo> dvbSubtitleInfos = null;
       while (data.getPosition() < descriptorsEndPosition) {
         int descriptorTag = data.readUnsignedByte();
@@ -883,7 +886,7 @@ private EsInfo readEsInfo(ParsableByteArray data, int length) {
           streamType = TS_STREAM_TYPE_DTS;
         } else if (descriptorTag == TS_PMT_DESC_ISO639_LANG) {
           language = data.readString(3).trim();
-          // Audio type is ignored.
+          audioType = data.readUnsignedByte();
         } else if (descriptorTag == TS_PMT_DESC_DVBSUBS) {
           streamType = TS_STREAM_TYPE_DVBSUBS;
           dvbSubtitleInfos = new ArrayList<>();
@@ -905,6 +908,7 @@ private EsInfo readEsInfo(ParsableByteArray data, int length) {
       return new EsInfo(
           streamType,
           language,
+          audioType,
           dvbSubtitleInfos,
           Arrays.copyOfRange(data.getData(), descriptorsStartPosition, descriptorsEndPosition));
     }