Plumb MP3 average bitrate from metadata frames into Format

Issue: androidx/media#1081

#minor-release

PiperOrigin-RevId: 619185083

RELEASENOTES.md

@@ -47,6 +47,8 @@
         resolution but a very small number of frames
         ([#1051](https://github.com/androidx/media/issues/1051)).
 *   Extractors:
+    *   MP3: Populate `Format.averageBitrate` from metadata frames such as
+        `XING` and `VBRI`.
 *   Audio:
     *   Allow renderer recovery by disabling offload if audio track fails to
         initialize in offload mode.

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/ConstantBitrateSeeker.java

@@ -24,6 +24,8 @@ import androidx.media3.extractor.MpegAudioUtil;
  */
 /* package */ final class ConstantBitrateSeeker extends ConstantBitrateSeekMap implements Seeker {
 
+  private final int bitrate;
+
   /**
    * @param inputLength The length of the stream in bytes, or {@link C#LENGTH_UNSET} if unknown.
    * @param firstFramePosition The position of the first frame in the stream.
@@ -45,6 +47,7 @@ import androidx.media3.extractor.MpegAudioUtil;
         mpegAudioHeader.bitrate,
         mpegAudioHeader.frameSize,
         allowSeeksIfLengthUnknown);
+    bitrate = mpegAudioHeader.bitrate;
   }
 
   @Override
@@ -56,4 +59,9 @@ import androidx.media3.extractor.MpegAudioUtil;
   public long getDataEndPosition() {
     return C.INDEX_UNSET;
   }
+
+  @Override
+  public int getAverageBitrate() {
+    return bitrate;
+  }
 }

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/IndexSeeker.java

@@ -20,6 +20,7 @@ import androidx.media3.common.C;
 import androidx.media3.common.util.LongArray;
 import androidx.media3.common.util.Util;
 import androidx.media3.extractor.SeekPoint;
+import java.math.RoundingMode;
 
 /** MP3 seeker that builds a time-to-byte mapping as the stream is read. */
 /* package */ final class IndexSeeker implements Seeker {
@@ -30,6 +31,7 @@ import androidx.media3.extractor.SeekPoint;
   private final long dataEndPosition;
   private final LongArray timesUs;
   private final LongArray positions;
+  private final int averageBitrate;
 
   private long durationUs;
 
@@ -40,6 +42,15 @@ import androidx.media3.extractor.SeekPoint;
     positions = new LongArray();
     timesUs.add(0L);
     positions.add(dataStartPosition);
+    if (durationUs != C.TIME_UNSET) {
+      long bitrate =
+          Util.scaleLargeValue(
+              dataStartPosition - dataEndPosition, 8, durationUs, RoundingMode.HALF_UP);
+      this.averageBitrate =
+          bitrate > 0 && bitrate <= Integer.MAX_VALUE ? (int) bitrate : C.RATE_UNSET_INT;
+    } else {
+      this.averageBitrate = C.RATE_UNSET_INT;
+    }
   }
 
   @Override
@@ -79,6 +90,11 @@ import androidx.media3.extractor.SeekPoint;
     }
   }
 
+  @Override
+  public int getAverageBitrate() {
+    return averageBitrate;
+  }
+
   /**
    * Adds a seek point to the index if it is sufficiently distant from the other points.
    *

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/MlltSeeker.java

@@ -127,4 +127,9 @@ import androidx.media3.extractor.metadata.id3.MlltFrame;
   public long getDataEndPosition() {
     return C.INDEX_UNSET;
   }
+
+  @Override
+  public int getAverageBitrate() {
+    return C.RATE_UNSET_INT;
+  }
 }

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/Mp3Extractor.java

@@ -280,7 +280,7 @@ public final class Mp3Extractor implements Extractor {
     if (seeker == null) {
       seeker = computeSeeker(input);
       extractorOutput.seekMap(seeker);
-      currentTrackOutput.format(
+      Format.Builder format =
           new Format.Builder()
               .setSampleMimeType(synchronizedHeader.mimeType)
               .setMaxInputSize(MpegAudioUtil.MAX_FRAME_SIZE_BYTES)
@@ -288,8 +288,11 @@ public final class Mp3Extractor implements Extractor {
               .setSampleRate(synchronizedHeader.sampleRate)
               .setEncoderDelay(gaplessInfoHolder.encoderDelay)
               .setEncoderPadding(gaplessInfoHolder.encoderPadding)
-              .setMetadata((flags & FLAG_DISABLE_ID3_METADATA) != 0 ? null : metadata)
+              .setMetadata((flags & FLAG_DISABLE_ID3_METADATA) != 0 ? null : metadata);
-              .build());
+      if (seeker.getAverageBitrate() != C.RATE_UNSET_INT) {
+        format.setAverageBitrate(seeker.getAverageBitrate());
+      }
+      currentTrackOutput.format(format.build());
       firstSamplePosition = input.getPosition();
     } else if (firstSamplePosition != 0) {
       long inputPosition = input.getPosition();

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/Seeker.java

@@ -39,6 +39,12 @@ import androidx.media3.extractor.SeekMap;
    */
   long getDataEndPosition();
 
+  /**
+   * Returns the average bitrate (usually derived from the duration and length of the file), or
+   * {@link C#RATE_UNSET_INT} if not known.
+   */
+  int getAverageBitrate();
+
   /** A {@link Seeker} that does not support seeking through audio data. */
   /* package */ class UnseekableSeeker extends SeekMap.Unseekable implements Seeker {
 
@@ -56,5 +62,10 @@ import androidx.media3.extractor.SeekMap;
       // Position unset as we do not know the data end position. Note that returning 0 doesn't work.
       return C.INDEX_UNSET;
     }
+
+    @Override
+    public int getAverageBitrate() {
+      return C.RATE_UNSET_INT;
+    }
   }
 }

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/VbriSeeker.java

@@ -94,19 +94,23 @@ import androidx.media3.extractor.SeekPoint;
     if (inputLength != C.LENGTH_UNSET && inputLength != position) {
       Log.w(TAG, "VBRI data size mismatch: " + inputLength + ", " + position);
     }
-    return new VbriSeeker(timesUs, positions, durationUs, /* dataEndPosition= */ position);
+    return new VbriSeeker(
+        timesUs, positions, durationUs, /* dataEndPosition= */ position, mpegAudioHeader.bitrate);
   }
 
   private final long[] timesUs;
   private final long[] positions;
   private final long durationUs;
   private final long dataEndPosition;
+  private final int bitrate;
 
-  private VbriSeeker(long[] timesUs, long[] positions, long durationUs, long dataEndPosition) {
+  private VbriSeeker(
+      long[] timesUs, long[] positions, long durationUs, long dataEndPosition, int bitrate) {
     this.timesUs = timesUs;
     this.positions = positions;
     this.durationUs = durationUs;
     this.dataEndPosition = dataEndPosition;
+    this.bitrate = bitrate;
   }
 
   @Override
@@ -140,4 +144,9 @@ import androidx.media3.extractor.SeekPoint;
   public long getDataEndPosition() {
     return dataEndPosition;
   }
+
+  @Override
+  public int getAverageBitrate() {
+    return bitrate;
+  }
 }

libraries/extractor/src/main/java/androidx/media3/extractor/mp3/XingSeeker.java

@@ -55,7 +55,8 @@ import androidx.media3.extractor.SeekPoint;
             xingFrame.header.sampleRate);
     if (xingFrame.dataSize == C.LENGTH_UNSET || xingFrame.tableOfContents == null) {
       // If the size in bytes or table of contents is missing, the stream is not seekable.
-      return new XingSeeker(position, xingFrame.header.frameSize, durationUs);
+      return new XingSeeker(
+          position, xingFrame.header.frameSize, durationUs, xingFrame.header.bitrate);
     }
 
     if (inputLength != C.LENGTH_UNSET && inputLength != position + xingFrame.dataSize) {
@@ -66,6 +67,7 @@ import androidx.media3.extractor.SeekPoint;
         position,
         xingFrame.header.frameSize,
         durationUs,
+        xingFrame.header.bitrate,
         xingFrame.dataSize,
         xingFrame.tableOfContents);
   }
@@ -73,6 +75,7 @@ import androidx.media3.extractor.SeekPoint;
   private final long dataStartPosition;
   private final int xingFrameSize;
   private final long durationUs;
+  private final int bitrate;
 
   /** Data size, including the XING frame. */
   private final long dataSize;
@@ -85,11 +88,12 @@ import androidx.media3.extractor.SeekPoint;
    */
   @Nullable private final long[] tableOfContents;
 
-  private XingSeeker(long dataStartPosition, int xingFrameSize, long durationUs) {
+  private XingSeeker(long dataStartPosition, int xingFrameSize, long durationUs, int bitrate) {
     this(
         dataStartPosition,
         xingFrameSize,
         durationUs,
+        bitrate,
         /* dataSize= */ C.LENGTH_UNSET,
         /* tableOfContents= */ null);
   }
@@ -98,13 +102,15 @@ import androidx.media3.extractor.SeekPoint;
       long dataStartPosition,
       int xingFrameSize,
       long durationUs,
+      int bitrate,
       long dataSize,
       @Nullable long[] tableOfContents) {
     this.dataStartPosition = dataStartPosition;
     this.xingFrameSize = xingFrameSize;
     this.durationUs = durationUs;
-    this.tableOfContents = tableOfContents;
+    this.bitrate = bitrate;
     this.dataSize = dataSize;
+    this.tableOfContents = tableOfContents;
     dataEndPosition = dataSize == C.LENGTH_UNSET ? C.INDEX_UNSET : dataStartPosition + dataSize;
   }
 
@@ -172,6 +178,11 @@ import androidx.media3.extractor.SeekPoint;
     return dataEndPosition;
   }
 
+  @Override
+  public int getAverageBitrate() {
+    return bitrate;
+  }
+
   /**
    * Returns the time in microseconds for a given table index.
    *

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr-variable-frame-size-no-seek-table.mp3.0.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 45139
   sample count = 108
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr-variable-frame-size-no-seek-table.mp3.1.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 30093
   sample count = 72
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr-variable-frame-size-no-seek-table.mp3.2.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 15046
   sample count = 36
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr-variable-frame-size-no-seek-table.mp3.3.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 0
   sample count = 0
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr-variable-frame-size-no-seek-table.mp3.unknown_length.dump

@@ -7,6 +7,7 @@ track 0:
   total output bytes = 45139
   sample count = 108
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr_cbr-seeking-always-enabled.0.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 45139
   sample count = 108
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr_cbr-seeking-always-enabled.1.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 30093
   sample count = 72
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr_cbr-seeking-always-enabled.2.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 15046
   sample count = 36
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr_cbr-seeking-always-enabled.3.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 0
   sample count = 0
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-cbr_cbr-seeking-always-enabled.unknown_length.dump

@@ -8,6 +8,7 @@ track 0:
   total output bytes = 45139
   sample count = 108
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-vbr-xing-header.mp3.0.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 38160
   sample count = 117
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-vbr-xing-header.mp3.1.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 24384
   sample count = 77
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-vbr-xing-header.mp3.2.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 11328
   sample count = 38
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-vbr-xing-header.mp3.3.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 0
   sample count = 0
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/bear-vbr-xing-header.mp3.unknown_length.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 38160
   sample count = 117
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/play-trimmed.mp3.0.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 418
   sample count = 1
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/play-trimmed.mp3.1.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 418
   sample count = 1
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/play-trimmed.mp3.2.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 418
   sample count = 1
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/play-trimmed.mp3.3.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 418
   sample count = 1
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/play-trimmed.mp3.unknown_length.dump

@@ -7,6 +7,7 @@ track 0:
   total output bytes = 418
   sample count = 1
   format 0:
+    averageBitrate = 128000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 2

libraries/test_data/src/test/assets/extractordumps/mp3/test-cbr-info-header.mp3.0.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 8359
   sample count = 40
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 1

libraries/test_data/src/test/assets/extractordumps/mp3/test-cbr-info-header.mp3.1.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 5643
   sample count = 27
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 1

libraries/test_data/src/test/assets/extractordumps/mp3/test-cbr-info-header.mp3.2.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 2926
   sample count = 14
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 1

libraries/test_data/src/test/assets/extractordumps/mp3/test-cbr-info-header.mp3.3.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 0
   sample count = 0
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 1

libraries/test_data/src/test/assets/extractordumps/mp3/test-cbr-info-header.mp3.unknown_length.dump

@@ -10,6 +10,7 @@ track 0:
   total output bytes = 8359
   sample count = 40
   format 0:
+    averageBitrate = 64000
     sampleMimeType = audio/mpeg
     maxInputSize = 4096
     channelCount = 1

libraries/test_data/src/test/assets/transformerdumps/mp3/test-cbr-info-header.mp3/looping_mixedWith_sample_18byte_nclx_colr.mp4.dump

@@ -1,4 +1,5 @@
 format audio:
+  averageBitrate = 64000
   sampleMimeType = audio/mpeg
   maxInputSize = 4096
   channelCount = 1

libraries/test_data/src/test/assets/transformerdumps/mp4/sample_18byte_nclx_colr.mp4/looping_mixedWith_test-cbr-info-header.mp3.dump

@@ -1,4 +1,5 @@
 format audio:
+  averageBitrate = 64000
   sampleMimeType = audio/mpeg
   maxInputSize = 4096
   channelCount = 1