Merge pull request #9543 from KasemJaffer:rf64

PiperOrigin-RevId: 408816643

RELEASENOTES.md

@@ -11,6 +11,9 @@
         `buildVideoRenderers()` or `buildAudioRenderers()` can access the codec
         adapter factory and pass it to `MediaCodecRenderer` instances they
         create.
+*   Extractors:
+    *   WAV: Add support for RF64 streams
+        ([#9543](https://github.com/google/ExoPlayer/issues/9543).
 
 ### 2.16.0 (2021-11-04)
 

library/extractor/src/main/java/com/google/android/exoplayer2/audio/WavUtil.java

@@ -30,6 +30,10 @@ public final class WavUtil {
   public static final int FMT_FOURCC = 0x666d7420;
   /** Four character code for "data". */
   public static final int DATA_FOURCC = 0x64617461;
+  /** Four character code for "RF64". */
+  public static final int RF64_FOURCC = 0x52463634;
+  /** Four character code for "ds64". */
+  public static final int DS64_FOURCC = 0x64733634;
 
   /** WAVE type value for integer PCM audio data. */
   public static final int TYPE_PCM = 0x0001;

library/extractor/src/main/java/com/google/android/exoplayer2/extractor/wav/WavExtractor.java

@@ -31,6 +31,7 @@
 import com.google.android.exoplayer2.extractor.PositionHolder;
 import com.google.android.exoplayer2.extractor.TrackOutput;
 import com.google.android.exoplayer2.util.Assertions;
+import com.google.android.exoplayer2.util.Log;
 import com.google.android.exoplayer2.util.MimeTypes;
 import com.google.android.exoplayer2.util.ParsableByteArray;
 import com.google.android.exoplayer2.util.Util;
@@ -47,6 +48,8 @@
 /** Extracts data from WAV byte streams. */
 public final class WavExtractor implements Extractor {
 
+  private static final String TAG = "WavExtractor";
+
   /**
    * When outputting PCM data to a {@link TrackOutput}, we can choose how many frames are grouped
    * into each sample, and hence each sample's duration. This is the target number of samples to
@@ -63,26 +66,30 @@ public final class WavExtractor implements Extractor {
   @Target({ElementType.TYPE_USE})
   @IntDef({
     STATE_READING_FILE_TYPE,
+    STATE_READING_RF64_SAMPLE_DATA_SIZE,
     STATE_READING_FORMAT,
     STATE_SKIPPING_TO_SAMPLE_DATA,
     STATE_READING_SAMPLE_DATA
   })
   private @interface State {}
 
   private static final int STATE_READING_FILE_TYPE = 0;
-  private static final int STATE_READING_FORMAT = 1;
-  private static final int STATE_SKIPPING_TO_SAMPLE_DATA = 2;
-  private static final int STATE_READING_SAMPLE_DATA = 3;
+  private static final int STATE_READING_RF64_SAMPLE_DATA_SIZE = 1;
+  private static final int STATE_READING_FORMAT = 2;
+  private static final int STATE_SKIPPING_TO_SAMPLE_DATA = 3;
+  private static final int STATE_READING_SAMPLE_DATA = 4;
 
   private @MonotonicNonNull ExtractorOutput extractorOutput;
   private @MonotonicNonNull TrackOutput trackOutput;
   private @State int state;
+  private long rf64SampleDataSize;
   private @MonotonicNonNull OutputWriter outputWriter;
   private int dataStartPosition;
   private long dataEndPosition;
 
   public WavExtractor() {
     state = STATE_READING_FILE_TYPE;
+    rf64SampleDataSize = C.LENGTH_UNSET;
     dataStartPosition = C.POSITION_UNSET;
     dataEndPosition = C.POSITION_UNSET;
   }
@@ -120,6 +127,9 @@ public int read(ExtractorInput input, PositionHolder seekPosition) throws IOExce
       case STATE_READING_FILE_TYPE:
         readFileType(input);
         return Extractor.RESULT_CONTINUE;
+      case STATE_READING_RF64_SAMPLE_DATA_SIZE:
+        readRf64SampleDataSize(input);
+        return Extractor.RESULT_CONTINUE;
       case STATE_READING_FORMAT:
         readFormat(input);
         return Extractor.RESULT_CONTINUE;
@@ -152,6 +162,11 @@ private void readFileType(ExtractorInput input) throws IOException {
           "Unsupported or unrecognized wav file type.", /* cause= */ null);
     }
     input.skipFully((int) (input.getPeekPosition() - input.getPosition()));
+    state = STATE_READING_RF64_SAMPLE_DATA_SIZE;
+  }
+
+  private void readRf64SampleDataSize(ExtractorInput input) throws IOException {
+    rf64SampleDataSize = WavHeaderReader.readRf64SampleDataSize(input);
     state = STATE_READING_FORMAT;
   }
 
@@ -194,7 +209,18 @@ private void readFormat(ExtractorInput input) throws IOException {
   private void skipToSampleData(ExtractorInput input) throws IOException {
     Pair<Long, Long> dataBounds = WavHeaderReader.skipToSampleData(input);
     dataStartPosition = dataBounds.first.intValue();
-    dataEndPosition = dataBounds.second;
+    long dataSize = dataBounds.second;
+    if (rf64SampleDataSize != C.LENGTH_UNSET && dataSize == 0xFFFFFFFFL) {
+      // Following EBU - Tech 3306-2007, the data size indicated in the ds64 chunk should only be
+      // used if the size of the data chunk is unset.
+      dataSize = rf64SampleDataSize;
+    }
+    dataEndPosition = dataStartPosition + dataSize;
+    long inputLength = input.getLength();
+    if (inputLength != C.LENGTH_UNSET && dataEndPosition > inputLength) {
+      Log.w(TAG, "Data exceeds input length: " + dataEndPosition + ", " + inputLength);
+      dataEndPosition = inputLength;
+    }
     Assertions.checkNotNull(outputWriter).init(dataStartPosition, dataEndPosition);
     state = STATE_READING_SAMPLE_DATA;
   }

library/extractor/src/main/java/com/google/android/exoplayer2/extractor/wav/WavHeaderReader.java

@@ -32,20 +32,19 @@
   private static final String TAG = "WavHeaderReader";
 
   /**
-   * Returns whether the given {@code input} starts with a RIFF chunk header, followed by a WAVE
-   * tag.
+   * Returns whether the given {@code input} starts with a RIFF or RF64 chunk header, followed by a
+   * WAVE tag.
    *
    * @param input The input stream to peek from. The position should point to the start of the
    *     stream.
-   * @return Whether the given {@code input} starts with a RIFF chunk header, followed by a WAVE
-   *     tag.
+   * @return Whether the given {@code input} starts with a RIFF or RF64 chunk header, followed by a
+   *     WAVE tag.
    * @throws IOException If peeking from the input fails.
    */
   public static boolean checkFileType(ExtractorInput input) throws IOException {
     ParsableByteArray scratch = new ParsableByteArray(ChunkHeader.SIZE_IN_BYTES);
-    // Attempt to read the RIFF chunk.
     ChunkHeader chunkHeader = ChunkHeader.peek(input, scratch);
-    if (chunkHeader.id != WavUtil.RIFF_FOURCC) {
+    if (chunkHeader.id != WavUtil.RIFF_FOURCC && chunkHeader.id != WavUtil.RF64_FOURCC) {
       return false;
     }
 
@@ -60,11 +59,36 @@ public static boolean checkFileType(ExtractorInput input) throws IOException {
     return true;
   }
 
+  /**
+   * Reads the ds64 chunk defined in EBU - TECH 3306-2007, if present. If there is no such chunk,
+   * the input's position is left unchanged.
+   *
+   * @param input Input stream to read from. The position should point to the byte following the
+   *     WAVE tag.
+   * @throws IOException If reading from the input fails.
+   * @return The value of the data size field in the ds64 chunk, or {@link C#LENGTH_UNSET} if there
+   *     is no such chunk.
+   */
+  public static long readRf64SampleDataSize(ExtractorInput input) throws IOException {
+    ParsableByteArray scratch = new ParsableByteArray(ChunkHeader.SIZE_IN_BYTES);
+    ChunkHeader chunkHeader = ChunkHeader.peek(input, scratch);
+    if (chunkHeader.id != WavUtil.DS64_FOURCC) {
+      input.resetPeekPosition();
+      return C.LENGTH_UNSET;
+    }
+    input.advancePeekPosition(8); // RIFF size
+    scratch.setPosition(0);
+    input.peekFully(scratch.getData(), 0, 8);
+    long sampleDataSize = scratch.readLittleEndianLong();
+    input.skipFully(ChunkHeader.SIZE_IN_BYTES + (int) chunkHeader.size);
+    return sampleDataSize;
+  }
+
   /**
    * Reads and returns a {@code WavFormat}.
    *
    * @param input Input stream to read the WAV format from. The position should point to the byte
-   *     following the WAVE tag.
+   *     following the ds64 chunk if present, or to the byte following the WAVE tag otherwise.
    * @throws IOException If reading from the input fails.
    * @return A new {@code WavFormat} read from {@code input}.
    */
@@ -104,13 +128,14 @@ public static WavFormat readFormat(ExtractorInput input) throws IOException {
   }
 
   /**
-   * Skips to the data in the given WAV input stream, and returns its bounds. After calling, the
-   * input stream's position will point to the start of sample data in the WAV. If an exception is
-   * thrown, the input position will be left pointing to a chunk header (that may not be the data
-   * chunk header).
+   * Skips to the data in the given WAV input stream, and returns its start position and size. After
+   * calling, the input stream's position will point to the start of sample data in the WAV. If an
+   * exception is thrown, the input position will be left pointing to a chunk header (that may not
+   * be the data chunk header).
    *
    * @param input The input stream, whose read position must be pointing to a valid chunk header.
-   * @return The byte positions at which the data starts (inclusive) and ends (exclusive).
+   * @return The byte positions at which the data starts (inclusive) and the size of the data, in
+   *     bytes.
    * @throws ParserException If an error occurs parsing chunks.
    * @throws IOException If reading from the input fails.
    */
@@ -125,13 +150,7 @@ public static Pair<Long, Long> skipToSampleData(ExtractorInput input) throws IOE
     input.skipFully(ChunkHeader.SIZE_IN_BYTES);
 
     long dataStartPosition = input.getPosition();
-    long dataEndPosition = dataStartPosition + chunkHeader.size;
-    long inputLength = input.getLength();
-    if (inputLength != C.LENGTH_UNSET && dataEndPosition > inputLength) {
-      Log.w(TAG, "Data exceeds input length: " + dataEndPosition + ", " + inputLength);
-      dataEndPosition = inputLength;
-    }
-    return Pair.create(dataStartPosition, dataEndPosition);
+    return Pair.create(dataStartPosition, chunkHeader.size);
   }
 
   /**

library/extractor/src/test/java/com/google/android/exoplayer2/extractor/wav/WavExtractorTest.java

@@ -53,4 +53,10 @@ public void sample_imaAdpcm() throws Exception {
     ExtractorAsserts.assertBehavior(
         WavExtractor::new, "media/wav/sample_ima_adpcm.wav", simulationConfig);
   }
+
+  @Test
+  public void sample_rf64() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        WavExtractor::new, "media/wav/sample_rf64.wav", simulationConfig);
+  }
 }

testdata/src/test/assets/extractordumps/wav/sample_rf64.wav.0.dump

@@ -0,0 +1,36 @@
+seekMap:
+  isSeekable = true
+  duration = 348625
+  getPosition(0) = [[timeUs=0, position=80]]
+  getPosition(1) = [[timeUs=0, position=80], [timeUs=20, position=84]]
+  getPosition(174312) = [[timeUs=174291, position=33544], [timeUs=174312, position=33548]]
+  getPosition(348625) = [[timeUs=348604, position=67012]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 66936
+  sample count = 4
+  format 0:
+    averageBitrate = 1536000
+    peakBitrate = 1536000
+    sampleMimeType = audio/raw
+    maxInputSize = 19200
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 2
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 19200, hash EF6C7C27
+  sample 1:
+    time = 100000
+    flags = 1
+    data = length 19200, hash 5AB97AFC
+  sample 2:
+    time = 200000
+    flags = 1
+    data = length 19200, hash 37920F33
+  sample 3:
+    time = 300000
+    flags = 1
+    data = length 9336, hash 135F1C30
+tracksEnded = true

testdata/src/test/assets/extractordumps/wav/sample_rf64.wav.1.dump

@@ -0,0 +1,32 @@
+seekMap:
+  isSeekable = true
+  duration = 348625
+  getPosition(0) = [[timeUs=0, position=80]]
+  getPosition(1) = [[timeUs=0, position=80], [timeUs=20, position=84]]
+  getPosition(174312) = [[timeUs=174291, position=33544], [timeUs=174312, position=33548]]
+  getPosition(348625) = [[timeUs=348604, position=67012]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 44628
+  sample count = 3
+  format 0:
+    averageBitrate = 1536000
+    peakBitrate = 1536000
+    sampleMimeType = audio/raw
+    maxInputSize = 19200
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 2
+  sample 0:
+    time = 116208
+    flags = 1
+    data = length 19200, hash E4B962ED
+  sample 1:
+    time = 216208
+    flags = 1
+    data = length 19200, hash 4F13D6CF
+  sample 2:
+    time = 316208
+    flags = 1
+    data = length 6228, hash 3FB5F446
+tracksEnded = true

testdata/src/test/assets/extractordumps/wav/sample_rf64.wav.2.dump

@@ -0,0 +1,28 @@
+seekMap:
+  isSeekable = true
+  duration = 348625
+  getPosition(0) = [[timeUs=0, position=80]]
+  getPosition(1) = [[timeUs=0, position=80], [timeUs=20, position=84]]
+  getPosition(174312) = [[timeUs=174291, position=33544], [timeUs=174312, position=33548]]
+  getPosition(348625) = [[timeUs=348604, position=67012]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 22316
+  sample count = 2
+  format 0:
+    averageBitrate = 1536000
+    peakBitrate = 1536000
+    sampleMimeType = audio/raw
+    maxInputSize = 19200
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 2
+  sample 0:
+    time = 232416
+    flags = 1
+    data = length 19200, hash F82E494B
+  sample 1:
+    time = 332416
+    flags = 1
+    data = length 3116, hash 93C99CFD
+tracksEnded = true

testdata/src/test/assets/extractordumps/wav/sample_rf64.wav.3.dump

@@ -0,0 +1,24 @@
+seekMap:
+  isSeekable = true
+  duration = 348625
+  getPosition(0) = [[timeUs=0, position=80]]
+  getPosition(1) = [[timeUs=0, position=80], [timeUs=20, position=84]]
+  getPosition(174312) = [[timeUs=174291, position=33544], [timeUs=174312, position=33548]]
+  getPosition(348625) = [[timeUs=348604, position=67012]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 4
+  sample count = 1
+  format 0:
+    averageBitrate = 1536000
+    peakBitrate = 1536000
+    sampleMimeType = audio/raw
+    maxInputSize = 19200
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 2
+  sample 0:
+    time = 348625
+    flags = 1
+    data = length 4, hash FFD4C53F
+tracksEnded = true

testdata/src/test/assets/extractordumps/wav/sample_rf64.wav.unknown_length.dump

@@ -0,0 +1,36 @@
+seekMap:
+  isSeekable = true
+  duration = 348625
+  getPosition(0) = [[timeUs=0, position=80]]
+  getPosition(1) = [[timeUs=0, position=80], [timeUs=20, position=84]]
+  getPosition(174312) = [[timeUs=174291, position=33544], [timeUs=174312, position=33548]]
+  getPosition(348625) = [[timeUs=348604, position=67012]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 66936
+  sample count = 4
+  format 0:
+    averageBitrate = 1536000
+    peakBitrate = 1536000
+    sampleMimeType = audio/raw
+    maxInputSize = 19200
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 2
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 19200, hash EF6C7C27
+  sample 1:
+    time = 100000
+    flags = 1
+    data = length 19200, hash 5AB97AFC
+  sample 2:
+    time = 200000
+    flags = 1
+    data = length 19200, hash 37920F33
+  sample 3:
+    time = 300000
+    flags = 1
+    data = length 9336, hash 135F1C30
+tracksEnded = true