Add support for extracting Vorbis audio in WebM Extractor.

24726372 · Oliver Woodman · 6a544da2 · 24726372 · 24726372 · 24726372
Commit 24726372 authored Nov 18, 2014 by Oliver Woodman
Showing with 178 additions and 40 deletions
library/src/main/java/com/google/android/exoplayer/dash/DashChunkSource.java
library/src/main/java/com/google/android/exoplayer/parser/webm/DefaultEbmlReader.java
library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlEventHandler.java
library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlReader.java
library/src/main/java/com/google/android/exoplayer/parser/webm/WebmExtractor.java
library/src/main/java/com/google/android/exoplayer/util/MimeTypes.java
--- a/library/src/main/java/com/google/android/exoplayer/dash/DashChunkSource.java
+++ b/library/src/main/java/com/google/android/exoplayer/dash/DashChunkSource.java
@@ -87,7 +87,7 @@ public class DashChunkSource implements ChunkSource {
      formats[i] = representations[i].format;
      maxWidth = Math.max(formats[i].width, maxWidth);
      maxHeight = Math.max(formats[i].height, maxHeight);
-      Extractor extractor = formats[i].mimeType.startsWith(MimeTypes.VIDEO_WEBM)
+      Extractor extractor = mimeTypeIsWebm(formats[i].mimeType)
          ? new WebmExtractor() : new FragmentedMp4Extractor();
      extractors.put(formats[i].id, extractor);
      this.representations.put(formats[i].id, representations[i]);
@@ -197,6 +197,10 @@ public class DashChunkSource implements ChunkSource {
    // Do nothing.
  }
+  private boolean mimeTypeIsWebm(String mimeType) {
+    return mimeType.startsWith(MimeTypes.VIDEO_WEBM) || mimeType.startsWith(MimeTypes.AUDIO_WEBM);
+  }
  private Chunk newInitializationChunk(RangedUri initializationUri, RangedUri indexUri,
      Representation representation, Extractor extractor, DataSource dataSource,
      int trigger) {

--- a/library/src/main/java/com/google/android/exoplayer/parser/webm/DefaultEbmlReader.java
+++ b/library/src/main/java/com/google/android/exoplayer/parser/webm/DefaultEbmlReader.java
@@ -16,6 +16,7 @@
 package com.google.android.exoplayer.parser.webm;
 import com.google.android.exoplayer.C;
+import com.google.android.exoplayer.ParserException;
 import com.google.android.exoplayer.upstream.NonBlockingInputStream;
 import com.google.android.exoplayer.util.Assertions;
@@ -134,7 +135,7 @@ import java.util.Stack;
  }
  @Override
-  public int read(NonBlockingInputStream inputStream) {
+  public int read(NonBlockingInputStream inputStream) throws ParserException {
    Assertions.checkState(eventHandler != null);
    while (true) {
      while (!masterElementsStack.isEmpty()

--- a/library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlEventHandler.java
+++ b/library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlEventHandler.java
@@ -15,6 +15,7 @@
 */
 package com.google.android.exoplayer.parser.webm;
+import com.google.android.exoplayer.ParserException;
 import com.google.android.exoplayer.upstream.NonBlockingInputStream;
 import java.nio.ByteBuffer;
@@ -46,41 +47,47 @@ import java.nio.ByteBuffer;
   * @param elementOffsetBytes The byte offset where this element starts
   * @param headerSizeBytes The byte length of this element's ID and size header
   * @param contentsSizeBytes The byte length of this element's children
+   * @throws ParserException If a parsing error occurs.
   */
  public void onMasterElementStart(
-      int id, long elementOffsetBytes, int headerSizeBytes, long contentsSizeBytes);
+      int id, long elementOffsetBytes, int headerSizeBytes,
+      long contentsSizeBytes) throws ParserException;
  /**
   * Called when a master element has finished reading in all of its children from the
   * {@link NonBlockingInputStream}.
   *
   * @param id The integer ID of this element
+   * @throws ParserException If a parsing error occurs.
   */
-  public void onMasterElementEnd(int id);
+  public void onMasterElementEnd(int id) throws ParserException;
  /**
   * Called when an integer element is encountered in the {@link NonBlockingInputStream}.
   *
   * @param id The integer ID of this element
   * @param value The integer value this element contains
+   * @throws ParserException If a parsing error occurs.
   */
-  public void onIntegerElement(int id, long value);
+  public void onIntegerElement(int id, long value) throws ParserException;
  /**
   * Called when a float element is encountered in the {@link NonBlockingInputStream}.
   *
   * @param id The integer ID of this element
   * @param value The float value this element contains
+   * @throws ParserException If a parsing error occurs.
   */
-  public void onFloatElement(int id, double value);
+  public void onFloatElement(int id, double value) throws ParserException;
  /**
   * Called when a string element is encountered in the {@link NonBlockingInputStream}.
   *
   * @param id The integer ID of this element
   * @param value The string value this element contains
+   * @throws ParserException If a parsing error occurs.
   */
-  public void onStringElement(int id, String value);
+  public void onStringElement(int id, String value) throws ParserException;
  /**
   * Called when a binary element is encountered in the {@link NonBlockingInputStream}.
@@ -109,9 +116,10 @@ import java.nio.ByteBuffer;
   * @param inputStream The {@link NonBlockingInputStream} from which this
   *        element's contents should be read
   * @return True if the element was read. False otherwise.
+   * @throws ParserException If a parsing error occurs.
   */
  public boolean onBinaryElement(
      int id, long elementOffsetBytes, int headerSizeBytes, int contentsSizeBytes,
-      NonBlockingInputStream inputStream);
+      NonBlockingInputStream inputStream) throws ParserException;
 }
--- a/library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlReader.java
+++ b/library/src/main/java/com/google/android/exoplayer/parser/webm/EbmlReader.java
@@ -15,6 +15,7 @@
 */
 package com.google.android.exoplayer.parser.webm;
+import com.google.android.exoplayer.ParserException;
 import com.google.android.exoplayer.upstream.NonBlockingInputStream;
 import java.nio.ByteBuffer;
@@ -53,8 +54,9 @@ import java.nio.ByteBuffer;
   *
   * @param inputStream The input stream from which data should be read
   * @return One of the {@code RESULT_*} flags defined in this interface
+   * @throws ParserException If parsing fails.
   */
-  public int read(NonBlockingInputStream inputStream);
+  public int read(NonBlockingInputStream inputStream) throws ParserException;
  /**
   * The total number of bytes consumed by the reader since first created or last {@link #reset()}.

--- a/library/src/main/java/com/google/android/exoplayer/parser/webm/WebmExtractor.java
+++ b/library/src/main/java/com/google/android/exoplayer/parser/webm/WebmExtractor.java
@@ -16,6 +16,7 @@
 package com.google.android.exoplayer.parser.webm;
 import com.google.android.exoplayer.MediaFormat;
+import com.google.android.exoplayer.ParserException;
 import com.google.android.exoplayer.SampleHolder;
 import com.google.android.exoplayer.parser.Extractor;
 import com.google.android.exoplayer.parser.SegmentIndex;
@@ -27,6 +28,7 @@ import android.annotation.TargetApi;
 import android.media.MediaExtractor;
 import java.nio.ByteBuffer;
+import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.Map;
 import java.util.UUID;
@@ -44,6 +46,8 @@ public final class WebmExtractor implements Extractor {
  private static final String DOC_TYPE_WEBM = "webm";
  private static final String CODEC_ID_VP9 = "V_VP9";
+  private static final String CODEC_ID_VORBIS = "A_VORBIS";
+  private static final int VORBIS_MAX_INPUT_SIZE = 8192;
  private static final int UNKNOWN = -1;
  // Element IDs
@@ -65,9 +69,13 @@ public final class WebmExtractor implements Extractor {
  private static final int ID_TRACKS = 0x1654AE6B;
  private static final int ID_TRACK_ENTRY = 0xAE;
  private static final int ID_CODEC_ID = 0x86;
+  private static final int ID_CODEC_PRIVATE = 0x63A2;
  private static final int ID_VIDEO = 0xE0;
  private static final int ID_PIXEL_WIDTH = 0xB0;
  private static final int ID_PIXEL_HEIGHT = 0xBA;
+  private static final int ID_AUDIO = 0xE1;
+  private static final int ID_CHANNELS = 0x9F;
+  private static final int ID_SAMPLING_FREQUENCY = 0xB5;
  private static final int ID_CUES = 0x1C53BB6B;
  private static final int ID_CUE_POINT = 0xBB;
@@ -96,6 +104,10 @@ public final class WebmExtractor implements Extractor {
  private long durationUs = UNKNOWN;
  private int pixelWidth = UNKNOWN;
  private int pixelHeight = UNKNOWN;
+  private int channelCount = UNKNOWN;
+  private int sampleRate = UNKNOWN;
+  private byte[] codecPrivate;
+  private boolean seenAudioTrack;
  private long cuesSizeBytes = UNKNOWN;
  private long clusterTimecodeUs = UNKNOWN;
  private long simpleBlockTimecodeUs = UNKNOWN;
@@ -114,7 +126,8 @@ public final class WebmExtractor implements Extractor {
  }
  @Override
-  public int read(NonBlockingInputStream inputStream, SampleHolder sampleHolder) {
+  public int read(
+      NonBlockingInputStream inputStream, SampleHolder sampleHolder) throws ParserException {
    this.sampleHolder = sampleHolder;
    this.readResults = 0;
    while ((readResults & READ_TERMINATING_RESULTS) == 0) {
@@ -176,6 +189,7 @@ public final class WebmExtractor implements Extractor {
      case ID_CLUSTER:
      case ID_TRACKS:
      case ID_TRACK_ENTRY:
+      case ID_AUDIO:
      case ID_VIDEO:
      case ID_CUES:
      case ID_CUE_POINT:
@@ -187,6 +201,7 @@ public final class WebmExtractor implements Extractor {
      case ID_TIME_CODE:
      case ID_PIXEL_WIDTH:
      case ID_PIXEL_HEIGHT:
+      case ID_CHANNELS:
      case ID_CUE_TIME:
      case ID_CUE_CLUSTER_POSITION:
        return EbmlReader.TYPE_UNSIGNED_INT;
@@ -194,8 +209,10 @@ public final class WebmExtractor implements Extractor {
      case ID_CODEC_ID:
        return EbmlReader.TYPE_STRING;
      case ID_SIMPLE_BLOCK:
+      case ID_CODEC_PRIVATE:
        return EbmlReader.TYPE_BINARY;
      case ID_DURATION:
+      case ID_SAMPLING_FREQUENCY:
        return EbmlReader.TYPE_FLOAT;
      default:
        return EbmlReader.TYPE_UNKNOWN;
@@ -203,11 +220,12 @@ public final class WebmExtractor implements Extractor {
  }
  /* package */ boolean onMasterElementStart(
-      int id, long elementOffsetBytes, int headerSizeBytes, long contentsSizeBytes) {
+      int id, long elementOffsetBytes, int headerSizeBytes,
+      long contentsSizeBytes) throws ParserException {
    switch (id) {
      case ID_SEGMENT:
        if (segmentStartOffsetBytes != UNKNOWN || segmentEndOffsetBytes != UNKNOWN) {
-          throw new IllegalStateException("Multiple Segment elements not supported");
+          throw new ParserException("Multiple Segment elements not supported");
        }
        segmentStartOffsetBytes = elementOffsetBytes + headerSizeBytes;
        segmentEndOffsetBytes = elementOffsetBytes + headerSizeBytes + contentsSizeBytes;
@@ -223,31 +241,41 @@ public final class WebmExtractor implements Extractor {
    return true;
  }
-  /* package */ boolean onMasterElementEnd(int id) {
+  /* package */ boolean onMasterElementEnd(int id) throws ParserException {
    switch (id) {
      case ID_CUES:
        buildCues();
        return false;
      case ID_VIDEO:
-        buildFormat();
+        buildVideoFormat();
+        return true;
+      case ID_AUDIO:
+        seenAudioTrack = true;
+        return true;
+      case ID_TRACK_ENTRY:
+        if (seenAudioTrack) {
+          // Audio format has to be built here since codec private may not be available at the end
+          // of ID_AUDIO.
+          buildAudioFormat();
+        }
        return true;
      default:
        return true;
    }
  }
-  /* package */ boolean onIntegerElement(int id, long value) {
+  /* package */ boolean onIntegerElement(int id, long value) throws ParserException {
    switch (id) {
      case ID_EBML_READ_VERSION:
        // Validate that EBMLReadVersion is supported. This extractor only supports v1.
        if (value != 1) {
-          throw new IllegalArgumentException("EBMLReadVersion " + value + " not supported");
+          throw new ParserException("EBMLReadVersion " + value + " not supported");
        }
        break;
      case ID_DOC_TYPE_READ_VERSION:
        // Validate that DocTypeReadVersion is supported. This extractor only supports up to v2.
        if (value < 1 || value > 2) {
-          throw new IllegalArgumentException("DocTypeReadVersion " + value + " not supported");
+          throw new ParserException("DocTypeReadVersion " + value + " not supported");
        }
        break;
      case ID_TIMECODE_SCALE:
@@ -259,6 +287,9 @@ public final class WebmExtractor implements Extractor {
      case ID_PIXEL_HEIGHT:
        pixelHeight = (int) value;
        break;
+      case ID_CHANNELS:
+        channelCount = (int) value;
+        break;
      case ID_CUE_TIME:
        cueTimesUs.add(scaleTimecodeToUs(value));
        break;
@@ -275,24 +306,31 @@ public final class WebmExtractor implements Extractor {
  }
  /* package */ boolean onFloatElement(int id, double value) {
-    if (id == ID_DURATION) {
+    switch (id) {
+      case ID_DURATION:
        durationUs = scaleTimecodeToUs((long) value);
+        break;
+      case ID_SAMPLING_FREQUENCY:
+        sampleRate = (int) value;
+        break;
+      default:
+        // pass
    }
    return true;
  }
-  /* package */ boolean onStringElement(int id, String value) {
+  /* package */ boolean onStringElement(int id, String value) throws ParserException {
    switch (id) {
      case ID_DOC_TYPE:
        // Validate that DocType is supported. This extractor only supports "webm".
        if (!DOC_TYPE_WEBM.equals(value)) {
-          throw new IllegalArgumentException("DocType " + value + " not supported");
+          throw new ParserException("DocType " + value + " not supported");
        }
        break;
      case ID_CODEC_ID:
-        // Validate that CodecID is supported. This extractor only supports "V_VP9".
+        // Validate that CodecID is supported. This extractor only supports "V_VP9" and "A_VORBIS".
-        if (!CODEC_ID_VP9.equals(value)) {
+        if (!CODEC_ID_VP9.equals(value) && !CODEC_ID_VORBIS.equals(value)) {
-          throw new IllegalArgumentException("CodecID " + value + " not supported");
+          throw new ParserException("CodecID " + value + " not supported");
        }
        break;
      default:
@@ -303,8 +341,9 @@ public final class WebmExtractor implements Extractor {
  /* package */ boolean onBinaryElement(
      int id, long elementOffsetBytes, int headerSizeBytes, int contentsSizeBytes,
-      NonBlockingInputStream inputStream) {
+      NonBlockingInputStream inputStream) throws ParserException {
-    if (id == ID_SIMPLE_BLOCK) {
+    switch (id) {
+      case ID_SIMPLE_BLOCK:
        // Please refer to http://www.matroska.org/technical/specs/index.html#simpleblock_structure
        // for info about how data is organized in a SimpleBlock element.
@@ -344,7 +383,7 @@ public final class WebmExtractor implements Extractor {
          case LACING_FIXED:
          case LACING_XIPH:
          default:
-          throw new IllegalStateException("Lacing mode " + lacing + " not supported");
+            throw new ParserException("Lacing mode " + lacing + " not supported");
        }
        if (sampleHolder.data == null || sampleHolder.data.capacity() < sampleHolder.size) {
@@ -359,6 +398,13 @@ public final class WebmExtractor implements Extractor {
          reader.readBytes(inputStream, outputData, sampleHolder.size);
        }
        readResults |= RESULT_READ_SAMPLE;
+        break;
+      case ID_CODEC_PRIVATE:
+        codecPrivate = new byte[contentsSizeBytes];
+        reader.readBytes(inputStream, codecPrivate, contentsSizeBytes);
+        break;
+      default:
+        // pass
    }
    return true;
  }
@@ -372,16 +418,38 @@ public final class WebmExtractor implements Extractor {
   *
   * <p>Replaces the previous {@link #format} only if video width/height have changed.
   * {@link #format} is guaranteed to not be null after calling this method. In
-   * the event that it can't be built, an {@link IllegalStateException} will be thrown.
+   * the event that it can't be built, an {@link ParserException} will be thrown.
   */
-  private void buildFormat() {
+  private void buildVideoFormat() throws ParserException {
    if (pixelWidth != UNKNOWN && pixelHeight != UNKNOWN
        && (format == null || format.width != pixelWidth || format.height != pixelHeight)) {
      format = MediaFormat.createVideoFormat(
          MimeTypes.VIDEO_VP9, MediaFormat.NO_VALUE, pixelWidth, pixelHeight, null);
      readResults |= RESULT_READ_INIT;
    } else if (format == null) {
-      throw new IllegalStateException("Unable to build format");
+      throw new ParserException("Unable to build format");
+    }
+  }
+  /**
+   * Build an audio {@link MediaFormat} containing recently gathered Audio information, if needed.
+   *
+   * <p>Replaces the previous {@link #format} only if audio channel count/sample rate have changed.
+   * {@link #format} is guaranteed to not be null after calling this method.
+   *
+   * @throws ParserException If an error occurs when parsing codec's private data or if the format
+   *    can't be built.
+   */
+  private void buildAudioFormat() throws ParserException {
+    if (channelCount != UNKNOWN && sampleRate != UNKNOWN
+        && (format == null || format.channelCount != channelCount
+            || format.sampleRate != sampleRate)) {
+      format = MediaFormat.createAudioFormat(
+          MimeTypes.AUDIO_VORBIS, VORBIS_MAX_INPUT_SIZE,
+          sampleRate, channelCount, parseVorbisCodecPrivate());
+      readResults |= RESULT_READ_INIT;
+    } else if (format == null) {
+      throw new ParserException("Unable to build format");
    }
  }
@@ -389,18 +457,18 @@ public final class WebmExtractor implements Extractor {
   * Build a {@link SegmentIndex} containing recently gathered Cues information.
   *
   * <p>{@link #cues} is guaranteed to not be null after calling this method. In
-   * the event that it can't be built, an {@link IllegalStateException} will be thrown.
+   * the event that it can't be built, an {@link ParserException} will be thrown.
   */
-  private void buildCues() {
+  private void buildCues() throws ParserException {
    if (segmentStartOffsetBytes == UNKNOWN) {
-      throw new IllegalStateException("Segment start/end offsets unknown");
+      throw new ParserException("Segment start/end offsets unknown");
    } else if (durationUs == UNKNOWN) {
-      throw new IllegalStateException("Duration unknown");
+      throw new ParserException("Duration unknown");
    } else if (cuesSizeBytes == UNKNOWN) {
-      throw new IllegalStateException("Cues size unknown");
+      throw new ParserException("Cues size unknown");
    } else if (cueTimesUs == null || cueClusterPositions == null
        || cueTimesUs.size() == 0 || cueTimesUs.size() != cueClusterPositions.size()) {
-      throw new IllegalStateException("Invalid/missing cue points");
+      throw new ParserException("Invalid/missing cue points");
    }
    int cuePointsSize = cueTimesUs.size();
    int[] sizes = new int[cuePointsSize];
@@ -424,6 +492,58 @@ public final class WebmExtractor implements Extractor {
  }
  /**
+   * Parses Vorbis Codec Private data and adds it as initialization data to the {@link #format}.
+   * WebM Vorbis Codec Private data specification can be found
+   * <a href="http://matroska.org/technical/specs/codecid/index.html">here</a>.
+   *
+   * @return ArrayList of byte arrays containing the initialization data on success.
+   * @throws ParserException If parsing codec private data fails.
+   */
+  private ArrayList<byte[]> parseVorbisCodecPrivate() throws ParserException {
+    try {
+      if (codecPrivate[0] != 0x02) {
+        throw new ParserException("Error parsing vorbis codec private");
+      }
+      int offset = 1;
+      int vorbisInfoLength = 0;
+      while (codecPrivate[offset] == (byte) 0xFF) {
+        vorbisInfoLength += 0xFF;
+        offset++;
+      }
+      vorbisInfoLength += codecPrivate[offset++];
+      int vorbisSkipLength = 0;
+      while (codecPrivate[offset] == (byte) 0xFF) {
+        vorbisSkipLength += 0xFF;
+        offset++;
+      }
+      vorbisSkipLength += codecPrivate[offset++];
+      if (codecPrivate[offset] != 0x01) {
+        throw new ParserException("Error parsing vorbis codec private");
+      }
+      byte[] vorbisInfo = new byte[vorbisInfoLength];
+      System.arraycopy(codecPrivate, offset, vorbisInfo, 0, vorbisInfoLength);
+      offset += vorbisInfoLength;
+      if (codecPrivate[offset] != 0x03) {
+        throw new ParserException("Error parsing vorbis codec private");
+      }
+      offset += vorbisSkipLength;
+      if (codecPrivate[offset] != 0x05) {
+        throw new ParserException("Error parsing vorbis codec private");
+      }
+      byte[] vorbisBooks = new byte[codecPrivate.length - offset];
+      System.arraycopy(codecPrivate, offset, vorbisBooks, 0, codecPrivate.length - offset);
+      ArrayList<byte[]> initializationData = new ArrayList<byte[]>(2);
+      initializationData.add(vorbisInfo);
+      initializationData.add(vorbisBooks);
+      return initializationData;
+    } catch (ArrayIndexOutOfBoundsException e) {
+      throw new ParserException("Error parsing vorbis codec private");
+    }
+  }
+  /**
   * Passes events through to {@link WebmExtractor} as
   * callbacks from {@link EbmlReader} are received.
   */
@@ -436,18 +556,19 @@ public final class WebmExtractor implements Extractor {
    @Override
    public void onMasterElementStart(
-        int id, long elementOffsetBytes, int headerSizeBytes, long contentsSizeBytes) {
+        int id, long elementOffsetBytes, int headerSizeBytes,
+        long contentsSizeBytes) throws ParserException {
      WebmExtractor.this.onMasterElementStart(
          id, elementOffsetBytes, headerSizeBytes, contentsSizeBytes);
    }
    @Override
-    public void onMasterElementEnd(int id) {
+    public void onMasterElementEnd(int id) throws ParserException {
      WebmExtractor.this.onMasterElementEnd(id);
    }
    @Override
-    public void onIntegerElement(int id, long value) {
+    public void onIntegerElement(int id, long value) throws ParserException {
      WebmExtractor.this.onIntegerElement(id, value);
    }
@@ -457,14 +578,14 @@ public final class WebmExtractor implements Extractor {
    }
    @Override
-    public void onStringElement(int id, String value) {
+    public void onStringElement(int id, String value) throws ParserException {
      WebmExtractor.this.onStringElement(id, value);
    }
    @Override
    public boolean onBinaryElement(
        int id, long elementOffsetBytes, int headerSizeBytes, int contentsSizeBytes,
-        NonBlockingInputStream inputStream) {
+        NonBlockingInputStream inputStream) throws ParserException {
      return WebmExtractor.this.onBinaryElement(
          id, elementOffsetBytes, headerSizeBytes, contentsSizeBytes, inputStream);
    }

--- a/library/src/main/java/com/google/android/exoplayer/util/MimeTypes.java
+++ b/library/src/main/java/com/google/android/exoplayer/util/MimeTypes.java
@@ -34,6 +34,8 @@ public class MimeTypes {
  public static final String AUDIO_AAC = BASE_TYPE_AUDIO + "/mp4a-latm";
  public static final String AUDIO_AC3 = BASE_TYPE_AUDIO + "/ac3";
  public static final String AUDIO_EC3 = BASE_TYPE_AUDIO + "/eac3";
+  public static final String AUDIO_WEBM = BASE_TYPE_AUDIO + "/webm";
+  public static final String AUDIO_VORBIS = BASE_TYPE_AUDIO + "/vorbis";
  public static final String TEXT_VTT = BASE_TYPE_TEXT + "/vtt";