Add AC-4 channel configuration support in DASH

libraries/exoplayer_dash/src/main/java/androidx/media3/exoplayer/dash/manifest/DashManifestParser.java

@@ -77,6 +77,17 @@ public class DashManifestParser extends DefaultHandler
   private static final Pattern CEA_708_ACCESSIBILITY_PATTERN =
       Pattern.compile("([1-9]|[1-5][0-9]|6[0-3])=.*");
 
+  /**
+   * Maps the value attribute of an AudioChannelConfiguration with schemeIdUri
+   * "tag:dolby.com,2015:dash:audio_channel_configuration:2015", as defined by ETSI TS 103 190-2
+   * v1.2.1 clause G.3. Table A.27 in ETSI TS 103 190-2 v1.2.1 defines the speaker counts of each
+   * speaker group index, numbers will be counted only when respective indexes present.
+   */
+  private static final int[] DOLBY_AC4_CHANNEL_CONFIGURATION_MAPPING =
+      new int[] {
+          2, 1, 2, 2, 2, 2, 1, 2, 2, 1, 1, 1, 1, 2, 1, 1, 2, 2, 2
+      };
+
   /**
    * Maps the value attribute of an AudioChannelConfiguration with schemeIdUri
    * "urn:mpeg:mpegB:cicp:ChannelConfiguration", as defined by ISO 23001-8 clause 8.1, to a channel
@@ -441,7 +452,7 @@ public class DashManifestParser extends DefaultHandler
       } else if (XmlPullParserUtil.isStartTag(xpp, "Role")) {
         roleDescriptors.add(parseDescriptor(xpp, "Role"));
       } else if (XmlPullParserUtil.isStartTag(xpp, "AudioChannelConfiguration")) {
-        audioChannels = parseAudioChannelConfiguration(xpp);
+        audioChannels = parseAudioChannelConfiguration(xpp, codecs);
       } else if (XmlPullParserUtil.isStartTag(xpp, "Accessibility")) {
         accessibilityDescriptors.add(parseDescriptor(xpp, "Accessibility"));
       } else if (XmlPullParserUtil.isStartTag(xpp, "EssentialProperty")) {
@@ -720,7 +731,7 @@ public class DashManifestParser extends DefaultHandler
         }
         baseUrls.addAll(parseBaseUrl(xpp, parentBaseUrls, dvbProfileDeclared));
       } else if (XmlPullParserUtil.isStartTag(xpp, "AudioChannelConfiguration")) {
-        audioChannels = parseAudioChannelConfiguration(xpp);
+        audioChannels = parseAudioChannelConfiguration(xpp, codecs);
       } else if (XmlPullParserUtil.isStartTag(xpp, "SegmentBase")) {
         segmentBase = parseSegmentBase(xpp, (SingleSegmentBase) segmentBase);
       } else if (XmlPullParserUtil.isStartTag(xpp, "SegmentList")) {
@@ -1487,7 +1498,7 @@ public class DashManifestParser extends DefaultHandler
 
   // AudioChannelConfiguration parsing.
 
-  protected int parseAudioChannelConfiguration(XmlPullParser xpp)
+  protected int parseAudioChannelConfiguration(XmlPullParser xpp, String codecs)
       throws XmlPullParserException, IOException {
     String schemeIdUri = parseString(xpp, "schemeIdUri", null);
     int audioChannels;
@@ -1509,6 +1520,9 @@ public class DashManifestParser extends DefaultHandler
       case "urn:dolby:dash:audio_channel_configuration:2011":
         audioChannels = parseDolbyChannelConfiguration(xpp);
         break;
+      case "tag:dolby.com,2015:dash:audio_channel_configuration:2015":
+        audioChannels = parseDolbyAC4ChannelConfiguration(xpp, codecs);
+        break;
       default:
         audioChannels = Format.NO_VALUE;
         break;
@@ -2043,6 +2057,49 @@ public class DashManifestParser extends DefaultHandler
     }
   }
 
+  /**
+   * Parses the number of channels from the value attribute of an AudioChannelConfiguration with
+   * schemeIdUri "tag:dolby.com,2015:dash:audio_channel_configuration:2015" as defined by table A.27
+   * in ETSI TS 103 190-2 v1.2.1.
+   *
+   * @param xpp The parser from which to read.
+   * @return The parsed number of channels, or {@link Format#NO_VALUE} if the channel count could
+   *     not be parsed.
+   */
+  protected static int parseDolbyAC4ChannelConfiguration(XmlPullParser xpp, String codecs) {
+    @Nullable String value = xpp.getAttributeValue(null, "value");
+    // The value attribute must be set to a six-digit uppercase hexadecimal
+    if (value == null || value.length() != 6) {
+      return Format.NO_VALUE;
+    }
+    final int DOLBY_AC4_OBJECT_BASED_AUDIO_STREAM = 0x800000;
+    final int ac4ChannelMask = Integer.parseInt(value, /* radix= */ 16);
+    if ((ac4ChannelMask & DOLBY_AC4_OBJECT_BASED_AUDIO_STREAM) ==
+        DOLBY_AC4_OBJECT_BASED_AUDIO_STREAM) {
+      // object-based audio content
+      String[] codecList = Util.splitCodecs(codecs);
+      // The AC-4 codec string format is "ac-4.xx.yy.zz", where zz is presentation level
+      final String AC4_PRESENTATION_LEVEL3 = "03";
+      final String AC4_PRESENTATION_LEVEL4 = "04";
+      String[] parts = Ascii.toLowerCase(codecList[0].trim()).split("\\.");
+      if (parts.length == 4 && parts[0].equals("ac-4")) {
+        if (parts[3].equals(AC4_PRESENTATION_LEVEL3)) {
+          return 18;  // AC-4 Level 3 object-based content is mapped to 17.1 channels
+        } else if (parts[3].equals(AC4_PRESENTATION_LEVEL4)) {
+          return 21;  // AC-4 Level 4 object-based content is mapped to 20.1 channels
+        }
+      }
+      return Format.NO_VALUE;
+    }
+    // channel-based audio content
+    // bits 0...18 indicate the presence of individual channel groups
+    int channelCount = 0;
+    for (int i = 0; i < DOLBY_AC4_CHANNEL_CONFIGURATION_MAPPING.length; i++) {
+      channelCount += ((ac4ChannelMask >> i) & 0x1) * DOLBY_AC4_CHANNEL_CONFIGURATION_MAPPING[i];
+    }
+    return channelCount == 0 ? Format.NO_VALUE : channelCount;
+  }
+
   protected static long parseLastSegmentNumberSupplementalProperty(
       List<Descriptor> supplementalProperties) {
     for (int i = 0; i < supplementalProperties.size(); i++) {

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/FragmentedMp4ExtractorParameterizedTest.java

@@ -110,6 +110,15 @@ public final class FragmentedMp4ExtractorParameterizedTest {
         /* closedCaptionFormats= */ ImmutableList.of(), "media/mp4/sample_ac4_fragmented.mp4");
   }
 
+  @Test
+  public void sampleWithAc4Level4Track() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        getExtractorFactory(
+            /* closedCaptionFormats= */ ImmutableList.of(), subtitlesParsedDuringExtraction),
+        "media/mp4/sample_ac4_level4_fragmented.mp4",
+        simulationConfig);
+  }
+
   @Test
   public void sampleWithProtectedAc4Track() throws Exception {
     assertExtractorBehavior(

libraries/test_data/src/test/assets/extractordumps/mp4/sample_ac4_level4_fragmented.mp4.0.dump

@@ -0,0 +1,98 @@
+seekMap:
+  isSeekable = true
+  duration = 853333
+  getPosition(0) = [[timeUs=0, position=669]]
+  getPosition(1) = [[timeUs=0, position=669]]
+  getPosition(426666) = [[timeUs=0, position=669]]
+  getPosition(853333) = [[timeUs=0, position=669]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 162700
+  sample count = 20
+  format 0:
+    id = 1
+    sampleMimeType = audio/ac4
+    channelCount = 21
+    sampleRate = 48000
+    language = und
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 8135, hash B524F88E
+  sample 1:
+    time = 42666
+    flags = 0
+    data = length 8135, hash FB80C2FB
+  sample 2:
+    time = 85333
+    flags = 0
+    data = length 8135, hash 907C0C31
+  sample 3:
+    time = 128000
+    flags = 0
+    data = length 8135, hash FDFBD32B
+  sample 4:
+    time = 170666
+    flags = 0
+    data = length 8135, hash 6CAF0549
+  sample 5:
+    time = 213333
+    flags = 0
+    data = length 8135, hash F5CA1C9A
+  sample 6:
+    time = 256000
+    flags = 0
+    data = length 8135, hash B1B5160D
+  sample 7:
+    time = 298666
+    flags = 0
+    data = length 8135, hash 9E923B3F
+  sample 8:
+    time = 341333
+    flags = 0
+    data = length 8135, hash B1C0BB1F
+  sample 9:
+    time = 384000
+    flags = 0
+    data = length 8135, hash 56F65A03
+  sample 10:
+    time = 426666
+    flags = 1
+    data = length 8135, hash D07FA9A1
+  sample 11:
+    time = 469333
+    flags = 0
+    data = length 8135, hash EF26FDDE
+  sample 12:
+    time = 512000
+    flags = 0
+    data = length 8135, hash 8946EEEB
+  sample 13:
+    time = 554666
+    flags = 0
+    data = length 8135, hash AC2E4C99
+  sample 14:
+    time = 597333
+    flags = 0
+    data = length 8135, hash B63A1D8
+  sample 15:
+    time = 640000
+    flags = 0
+    data = length 8135, hash 23119F0F
+  sample 16:
+    time = 682666
+    flags = 0
+    data = length 8135, hash 507972CA
+  sample 17:
+    time = 725333
+    flags = 0
+    data = length 8135, hash E574BC00
+  sample 18:
+    time = 768000
+    flags = 0
+    data = length 8135, hash 52F482FA
+  sample 19:
+    time = 810666
+    flags = 0
+    data = length 8135, hash C1A7B518
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_ac4_level4_fragmented.mp4.1.dump

@@ -0,0 +1,98 @@
+seekMap:
+  isSeekable = true
+  duration = 853333
+  getPosition(0) = [[timeUs=0, position=669]]
+  getPosition(1) = [[timeUs=0, position=669]]
+  getPosition(426666) = [[timeUs=0, position=669]]
+  getPosition(853333) = [[timeUs=0, position=669]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 162700
+  sample count = 20
+  format 0:
+    id = 1
+    sampleMimeType = audio/ac4
+    channelCount = 21
+    sampleRate = 48000
+    language = und
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 8135, hash B524F88E
+  sample 1:
+    time = 42666
+    flags = 0
+    data = length 8135, hash FB80C2FB
+  sample 2:
+    time = 85333
+    flags = 0
+    data = length 8135, hash 907C0C31
+  sample 3:
+    time = 128000
+    flags = 0
+    data = length 8135, hash FDFBD32B
+  sample 4:
+    time = 170666
+    flags = 0
+    data = length 8135, hash 6CAF0549
+  sample 5:
+    time = 213333
+    flags = 0
+    data = length 8135, hash F5CA1C9A
+  sample 6:
+    time = 256000
+    flags = 0
+    data = length 8135, hash B1B5160D
+  sample 7:
+    time = 298666
+    flags = 0
+    data = length 8135, hash 9E923B3F
+  sample 8:
+    time = 341333
+    flags = 0
+    data = length 8135, hash B1C0BB1F
+  sample 9:
+    time = 384000
+    flags = 0
+    data = length 8135, hash 56F65A03
+  sample 10:
+    time = 426666
+    flags = 1
+    data = length 8135, hash D07FA9A1
+  sample 11:
+    time = 469333
+    flags = 0
+    data = length 8135, hash EF26FDDE
+  sample 12:
+    time = 512000
+    flags = 0
+    data = length 8135, hash 8946EEEB
+  sample 13:
+    time = 554666
+    flags = 0
+    data = length 8135, hash AC2E4C99
+  sample 14:
+    time = 597333
+    flags = 0
+    data = length 8135, hash B63A1D8
+  sample 15:
+    time = 640000
+    flags = 0
+    data = length 8135, hash 23119F0F
+  sample 16:
+    time = 682666
+    flags = 0
+    data = length 8135, hash 507972CA
+  sample 17:
+    time = 725333
+    flags = 0
+    data = length 8135, hash E574BC00
+  sample 18:
+    time = 768000
+    flags = 0
+    data = length 8135, hash 52F482FA
+  sample 19:
+    time = 810666
+    flags = 0
+    data = length 8135, hash C1A7B518
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_ac4_level4_fragmented.mp4.2.dump

@@ -0,0 +1,58 @@
+seekMap:
+  isSeekable = true
+  duration = 853333
+  getPosition(0) = [[timeUs=0, position=669]]
+  getPosition(1) = [[timeUs=0, position=669]]
+  getPosition(426666) = [[timeUs=0, position=669]]
+  getPosition(853333) = [[timeUs=0, position=669]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 81350
+  sample count = 10
+  format 0:
+    id = 1
+    sampleMimeType = audio/ac4
+    channelCount = 21
+    sampleRate = 48000
+    language = und
+  sample 0:
+    time = 426666
+    flags = 1
+    data = length 8135, hash D07FA9A1
+  sample 1:
+    time = 469333
+    flags = 0
+    data = length 8135, hash EF26FDDE
+  sample 2:
+    time = 512000
+    flags = 0
+    data = length 8135, hash 8946EEEB
+  sample 3:
+    time = 554666
+    flags = 0
+    data = length 8135, hash AC2E4C99
+  sample 4:
+    time = 597333
+    flags = 0
+    data = length 8135, hash B63A1D8
+  sample 5:
+    time = 640000
+    flags = 0
+    data = length 8135, hash 23119F0F
+  sample 6:
+    time = 682666
+    flags = 0
+    data = length 8135, hash 507972CA
+  sample 7:
+    time = 725333
+    flags = 0
+    data = length 8135, hash E574BC00
+  sample 8:
+    time = 768000
+    flags = 0
+    data = length 8135, hash 52F482FA
+  sample 9:
+    time = 810666
+    flags = 0
+    data = length 8135, hash C1A7B518
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_ac4_level4_fragmented.mp4.3.dump

@@ -0,0 +1,58 @@
+seekMap:
+  isSeekable = true
+  duration = 853333
+  getPosition(0) = [[timeUs=0, position=669]]
+  getPosition(1) = [[timeUs=0, position=669]]
+  getPosition(426666) = [[timeUs=0, position=669]]
+  getPosition(853333) = [[timeUs=0, position=669]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 81350
+  sample count = 10
+  format 0:
+    id = 1
+    sampleMimeType = audio/ac4
+    channelCount = 21
+    sampleRate = 48000
+    language = und
+  sample 0:
+    time = 426666
+    flags = 1
+    data = length 8135, hash D07FA9A1
+  sample 1:
+    time = 469333
+    flags = 0
+    data = length 8135, hash EF26FDDE
+  sample 2:
+    time = 512000
+    flags = 0
+    data = length 8135, hash 8946EEEB
+  sample 3:
+    time = 554666
+    flags = 0
+    data = length 8135, hash AC2E4C99
+  sample 4:
+    time = 597333
+    flags = 0
+    data = length 8135, hash B63A1D8
+  sample 5:
+    time = 640000
+    flags = 0
+    data = length 8135, hash 23119F0F
+  sample 6:
+    time = 682666
+    flags = 0
+    data = length 8135, hash 507972CA
+  sample 7:
+    time = 725333
+    flags = 0
+    data = length 8135, hash E574BC00
+  sample 8:
+    time = 768000
+    flags = 0
+    data = length 8135, hash 52F482FA
+  sample 9:
+    time = 810666
+    flags = 0
+    data = length 8135, hash C1A7B518
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_ac4_level4_fragmented.mp4.unknown_length.dump

@@ -0,0 +1,98 @@
+seekMap:
+  isSeekable = true
+  duration = 853333
+  getPosition(0) = [[timeUs=0, position=669]]
+  getPosition(1) = [[timeUs=0, position=669]]
+  getPosition(426666) = [[timeUs=0, position=669]]
+  getPosition(853333) = [[timeUs=0, position=669]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 162700
+  sample count = 20
+  format 0:
+    id = 1
+    sampleMimeType = audio/ac4
+    channelCount = 21
+    sampleRate = 48000
+    language = und
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 8135, hash B524F88E
+  sample 1:
+    time = 42666
+    flags = 0
+    data = length 8135, hash FB80C2FB
+  sample 2:
+    time = 85333
+    flags = 0
+    data = length 8135, hash 907C0C31
+  sample 3:
+    time = 128000
+    flags = 0
+    data = length 8135, hash FDFBD32B
+  sample 4:
+    time = 170666
+    flags = 0
+    data = length 8135, hash 6CAF0549
+  sample 5:
+    time = 213333
+    flags = 0
+    data = length 8135, hash F5CA1C9A
+  sample 6:
+    time = 256000
+    flags = 0
+    data = length 8135, hash B1B5160D
+  sample 7:
+    time = 298666
+    flags = 0
+    data = length 8135, hash 9E923B3F
+  sample 8:
+    time = 341333
+    flags = 0
+    data = length 8135, hash B1C0BB1F
+  sample 9:
+    time = 384000
+    flags = 0
+    data = length 8135, hash 56F65A03
+  sample 10:
+    time = 426666
+    flags = 1
+    data = length 8135, hash D07FA9A1
+  sample 11:
+    time = 469333
+    flags = 0
+    data = length 8135, hash EF26FDDE
+  sample 12:
+    time = 512000
+    flags = 0
+    data = length 8135, hash 8946EEEB
+  sample 13:
+    time = 554666
+    flags = 0
+    data = length 8135, hash AC2E4C99
+  sample 14:
+    time = 597333
+    flags = 0
+    data = length 8135, hash B63A1D8
+  sample 15:
+    time = 640000
+    flags = 0
+    data = length 8135, hash 23119F0F
+  sample 16:
+    time = 682666
+    flags = 0
+    data = length 8135, hash 507972CA
+  sample 17:
+    time = 725333
+    flags = 0
+    data = length 8135, hash E574BC00
+  sample 18:
+    time = 768000
+    flags = 0
+    data = length 8135, hash 52F482FA
+  sample 19:
+    time = 810666
+    flags = 0
+    data = length 8135, hash C1A7B518
+tracksEnded = true