Simplify HLS MP3 support.

This commit is contained in:
Oliver Woodman 2015-06-26 14:44:10 +01:00
parent 0c577ce2ce
commit 3bcd9ca6c3
9 changed files with 123 additions and 140 deletions

View File

@ -43,8 +43,11 @@ import java.util.Locale;
}
public static final Sample[] YOUTUBE_DASH_MP4 = new Sample[] {
new Sample("XXXXXXXXX",
"http://178.33.229.111/live/mp4:Videolina/playlist.m3u8", PlayerActivity.TYPE_HLS),
new Sample("Google Glass",
"http://www.youtube.com/api/manifest/dash/id/bf5bb2419360daf1/source/youtube?"
+ "as=fmp4_audio_clear,fmp4_sd_hd_clear&sparams=ip,ipbits,expire,source,id,as&ip=0.0.0.0&"
+ "ipbits=0&expire=19000000000&signature=51AF5F39AB0CEC3E5497CD9C900EBFEAECCCB5C7."
+ "8506521BFC350652163895D4C26DEE124209AA9E&key=ik0", PlayerActivity.TYPE_DASH),
new Sample("Google Play",
"http://www.youtube.com/api/manifest/dash/id/3aa39fa2cc27967f/source/youtube?"
+ "as=fmp4_audio_clear,fmp4_sd_hd_clear&sparams=ip,ipbits,expire,source,id,as&ip=0.0.0.0&"

View File

@ -24,7 +24,7 @@ import com.google.android.exoplayer.extractor.ExtractorOutput;
import com.google.android.exoplayer.extractor.PositionHolder;
import com.google.android.exoplayer.extractor.SeekMap;
import com.google.android.exoplayer.extractor.TrackOutput;
import com.google.android.exoplayer.util.MimeTypes;
import com.google.android.exoplayer.util.MpegAudioHeader;
import com.google.android.exoplayer.util.ParsableByteArray;
import com.google.android.exoplayer.util.Util;
@ -43,20 +43,10 @@ public final class Mp3Extractor implements Extractor {
/** Mask that includes the audio header values that must match between frames. */
private static final int HEADER_MASK = 0xFFFE0C00;
private static final int ID3_TAG = Util.getIntegerCodeForString("ID3");
public static final String[] MIME_TYPE_BY_LAYER =
new String[] {MimeTypes.AUDIO_MPEG_L1, MimeTypes.AUDIO_MPEG_L2, MimeTypes.AUDIO_MPEG};
private static final int XING_HEADER = Util.getIntegerCodeForString("Xing");
private static final int INFO_HEADER = Util.getIntegerCodeForString("Info");
private static final int VBRI_HEADER = Util.getIntegerCodeForString("VBRI");
/**
* Theoretical maximum frame size for an MPEG audio stream, which occurs when playing a Layer 2
* MPEG 2.5 audio stream at 16 kb/s (with padding). The size is 1152 sample/frame *
* 160000 bit/s / (8000 sample/s * 8 bit/byte) + 1 padding byte/frame = 2881 byte/frame.
* The next power of two size is 4 KiB.
*/
public static final int MAX_FRAME_SIZE_BYTES = 4096;
private final BufferingInput inputBuffer;
private final ParsableByteArray scratch;
private final MpegAudioHeader synchronizedHeader;
@ -74,7 +64,7 @@ public final class Mp3Extractor implements Extractor {
/** Constructs a new {@link Mp3Extractor}. */
public Mp3Extractor() {
inputBuffer = new BufferingInput(MAX_FRAME_SIZE_BYTES * 3);
inputBuffer = new BufferingInput(MpegAudioHeader.MAX_FRAME_SIZE_BYTES * 3);
scratch = new ParsableByteArray(4);
synchronizedHeader = new MpegAudioHeader();
}
@ -255,10 +245,9 @@ public final class Mp3Extractor implements Extractor {
if (seeker == null) {
setupSeeker(extractorInput, headerPosition);
extractorOutput.seekMap(seeker);
trackOutput.format(MediaFormat.createAudioFormat(
MIME_TYPE_BY_LAYER[synchronizedHeader.layerIndex], MAX_FRAME_SIZE_BYTES,
seeker.getDurationUs(), synchronizedHeader.channels, synchronizedHeader.sampleRate,
Collections.<byte[]>emptyList()));
trackOutput.format(MediaFormat.createAudioFormat(synchronizedHeader.mimeType,
MpegAudioHeader.MAX_FRAME_SIZE_BYTES, seeker.getDurationUs(), synchronizedHeader.channels,
synchronizedHeader.sampleRate, Collections.<byte[]>emptyList()));
}
return headerPosition;

View File

@ -15,6 +15,7 @@
*/
package com.google.android.exoplayer.extractor.mp3;
import com.google.android.exoplayer.util.MpegAudioHeader;
import com.google.android.exoplayer.util.ParsableByteArray;
import com.google.android.exoplayer.util.Util;

View File

@ -16,6 +16,7 @@
package com.google.android.exoplayer.extractor.mp3;
import com.google.android.exoplayer.C;
import com.google.android.exoplayer.util.MpegAudioHeader;
import com.google.android.exoplayer.util.ParsableByteArray;
import com.google.android.exoplayer.util.Util;

View File

@ -130,7 +130,7 @@ import java.util.Collections;
* Locates the next sync word, advancing the position to the byte that immediately follows it.
* If a sync word was not located, the position is advanced to the limit.
*
* @param pesBuffer The buffer in which to search for the sync word.
* @param pesBuffer The buffer whose position should be advanced.
* @return True if a sync word position was found. False otherwise.
*/
private boolean skipToNextSync(ParsableByteArray pesBuffer) {

View File

@ -18,8 +18,7 @@ package com.google.android.exoplayer.extractor.ts;
import com.google.android.exoplayer.C;
import com.google.android.exoplayer.MediaFormat;
import com.google.android.exoplayer.extractor.TrackOutput;
import com.google.android.exoplayer.extractor.mp3.Mp3Extractor;
import com.google.android.exoplayer.extractor.mp3.MpegAudioHeader;
import com.google.android.exoplayer.util.MpegAudioHeader;
import com.google.android.exoplayer.util.ParsableByteArray;
import java.util.Collections;
@ -27,7 +26,7 @@ import java.util.Collections;
/**
* Parses a continuous MPEG Audio byte stream and extracts individual frames.
*/
/* package */ public class MpaReader extends ElementaryStreamReader {
/* package */ class MpegAudioReader extends ElementaryStreamReader {
private static final int STATE_FINDING_HEADER = 0;
private static final int STATE_READING_HEADER = 1;
@ -36,33 +35,35 @@ import java.util.Collections;
private static final int HEADER_SIZE = 4;
private final ParsableByteArray headerScratch;
private final MpegAudioHeader header;
private int state;
private int bytesRead;
private int frameBytesRead;
private boolean hasOutputFormat;
// Used to find the header.
// Used when finding the frame header.
private boolean lastByteWasFF;
// Used when parsing the header.
private boolean hasOutputFormat;
// Parsed from the frame header.
private long frameDurationUs;
private int sampleSize;
private int frameSize;
// Used when reading the samples.
// The timestamp to attach to the next sample in the current packet.
private long timeUs;
public MpaReader(TrackOutput output) {
public MpegAudioReader(TrackOutput output) {
super(output);
state = STATE_FINDING_HEADER;
// The first byte of an MPEG Audio frame header is always 0xFF.
headerScratch = new ParsableByteArray(4);
headerScratch.data[0] = (byte) 0xFF;
header = new MpegAudioHeader();
}
@Override
public void seek() {
state = STATE_FINDING_HEADER;
bytesRead = 0;
frameBytesRead = 0;
lastByteWasFF = false;
}
@ -74,19 +75,13 @@ import java.util.Collections;
while (data.bytesLeft() > 0) {
switch (state) {
case STATE_FINDING_HEADER:
if (findHeader(data)) {
state = STATE_READING_HEADER;
}
findHeader(data);
break;
case STATE_READING_HEADER:
if (readHeaderRemainder(data)) {
state = STATE_READING_FRAME;
}
readHeaderRemainder(data);
break;
case STATE_READING_FRAME:
if (readFrame(data)) {
state = STATE_FINDING_HEADER;
}
readFrameRemainder(data);
break;
}
}
@ -100,78 +95,83 @@ import java.util.Collections;
/**
* Attempts to locate the start of the next frame header.
* <p>
* If a frame header is located then true is returned. The first two bytes of the header will have
* been written into {@link #headerScratch}, and the position of the source will have been
* advanced to the byte that immediately follows these two bytes.
* If a frame header is located then the state is changed to {@link #STATE_READING_HEADER}, the
* first two bytes of the header are written into {@link #headerScratch}, and the position of the
* source is advanced to the byte that immediately follows these two bytes.
* <p>
* If a frame header is not located then the position of the source will have been advanced to the
* limit, and the method should be called again with the next source to continue the search.
* If a frame header is not located then the position of the source is advanced to the limit, and
* the method should be called again with the next source to continue the search.
*
* @param source The source from which to read.
* @return True if the frame header was located. False otherwise.
*/
private boolean findHeader(ParsableByteArray source) {
byte[] mpaData = source.data;
private void findHeader(ParsableByteArray source) {
byte[] data = source.data;
int startOffset = source.getPosition();
int endOffset = source.limit();
for (int i = startOffset; i < endOffset; i++) {
boolean byteIsFF = (mpaData[i] & 0xFF) == 0xFF;
boolean found = lastByteWasFF && (mpaData[i] & 0xF0) == 0xF0;
boolean byteIsFF = (data[i] & 0xFF) == 0xFF;
boolean found = lastByteWasFF && (data[i] & 0xE0) == 0xE0;
lastByteWasFF = byteIsFF;
if (found) {
source.setPosition(i + 1);
// Reset lastByteWasFF for next time.
lastByteWasFF = false;
headerScratch.data[0] = (byte) 0xFF;
headerScratch.data[1] = mpaData[i];
bytesRead = 2;
return true;
headerScratch.data[1] = data[i];
frameBytesRead = 2;
state = STATE_READING_HEADER;
return;
}
}
source.setPosition(endOffset);
return false;
}
/**
* Attempts to read the remaining two bytes of the frame header.
* <p>
* If a frame header is read in full then true is returned. The media format will have been output
* if this has not previously occurred, the four header bytes will have been output as sample
* data, and the position of the source will have been advanced to the byte that immediately
* If a frame header is read in full then the state is changed to {@link #STATE_READING_FRAME},
* the media format is output if this has not previously occurred, the four header bytes are
* output as sample data, and the position of the source is advanced to the byte that immediately
* follows the header.
* <p>
* If a frame header is not read in full then the position of the source will have been advanced
* to the limit, and the method should be called again with the next source to continue the read.
* If a frame header is read in full but cannot be parsed then the state is changed to
* {@link #STATE_READING_HEADER}.
* <p>
* If a frame header is not read in full then the position of the source is advanced to the limit,
* and the method should be called again with the next source to continue the read.
*
* @param source The source from which to read.
* @return True if the frame header was read in full. False otherwise.
*/
private boolean readHeaderRemainder(ParsableByteArray source) {
int bytesToRead = Math.min(source.bytesLeft(), HEADER_SIZE - bytesRead);
source.readBytes(headerScratch.data, bytesRead, bytesToRead);
bytesRead += bytesToRead;
if (bytesRead < HEADER_SIZE) {
return false;
private void readHeaderRemainder(ParsableByteArray source) {
int bytesToRead = Math.min(source.bytesLeft(), HEADER_SIZE - frameBytesRead);
source.readBytes(headerScratch.data, frameBytesRead, bytesToRead);
frameBytesRead += bytesToRead;
if (frameBytesRead < HEADER_SIZE) {
// We haven't read the whole header yet.
return;
}
headerScratch.setPosition(0);
boolean parsedHeader = MpegAudioHeader.populateHeader(headerScratch.readInt(), header);
if (!parsedHeader) {
// We thought we'd located a frame header, but we hadn't.
frameBytesRead = 0;
state = STATE_READING_HEADER;
return;
}
frameSize = header.frameSize;
if (!hasOutputFormat) {
headerScratch.setPosition(0);
int headerInt = headerScratch.readInt();
MpegAudioHeader synchronizedHeader = new MpegAudioHeader();
MpegAudioHeader.populateHeader(headerInt, synchronizedHeader);
MediaFormat mediaFormat = MediaFormat.createAudioFormat(
Mp3Extractor.MIME_TYPE_BY_LAYER[synchronizedHeader.layerIndex], Mp3Extractor.MAX_FRAME_SIZE_BYTES,
C.UNKNOWN_TIME_US, synchronizedHeader.channels, synchronizedHeader.sampleRate,
Collections.<byte[]>emptyList());
frameDurationUs = (C.MICROS_PER_SECOND * header.samplesPerFrame) / header.sampleRate;
MediaFormat mediaFormat = MediaFormat.createAudioFormat(header.mimeType,
MpegAudioHeader.MAX_FRAME_SIZE_BYTES, C.UNKNOWN_TIME_US, header.channels,
header.sampleRate, Collections.<byte[]>emptyList());
output.format(mediaFormat);
hasOutputFormat = true;
frameDurationUs = (C.MICROS_PER_SECOND * synchronizedHeader.samplesPerFrame) / mediaFormat.sampleRate;
sampleSize = synchronizedHeader.frameSize;
}
headerScratch.setPosition(0);
output.sampleData(headerScratch, HEADER_SIZE);
return true;
state = STATE_READING_FRAME;
}
/**
@ -185,20 +185,20 @@ import java.util.Collections;
* limit, and the method should be called again with the next source to continue the read.
*
* @param source The source from which to read.
* @return True if the frame was read in full. False otherwise.
*/
private boolean readFrame(ParsableByteArray source) {
int bytesToRead = Math.min(source.bytesLeft(), sampleSize - bytesRead);
private void readFrameRemainder(ParsableByteArray source) {
int bytesToRead = Math.min(source.bytesLeft(), frameSize - frameBytesRead);
output.sampleData(source, bytesToRead);
bytesRead += bytesToRead;
if (bytesRead < sampleSize) {
return false;
frameBytesRead += bytesToRead;
if (frameBytesRead < frameSize) {
// We haven't read the whole of the frame yet.
return;
}
output.sampleMetadata(timeUs, C.SAMPLE_FLAG_SYNC, sampleSize, 0, null);
output.sampleMetadata(timeUs, C.SAMPLE_FLAG_SYNC, frameSize, 0, null);
timeUs += frameDurationUs;
bytesRead = 0;
return true;
frameBytesRead = 0;
state = STATE_FINDING_HEADER;
}
}

View File

@ -354,10 +354,10 @@ public final class TsExtractor implements Extractor, SeekMap {
ElementaryStreamReader pesPayloadReader = null;
switch (streamType) {
case TS_STREAM_TYPE_MPA:
pesPayloadReader = new MpaReader(output.track(TS_STREAM_TYPE_MPA));
pesPayloadReader = new MpegAudioReader(output.track(TS_STREAM_TYPE_MPA));
break;
case TS_STREAM_TYPE_MPA_LSF:
pesPayloadReader = new MpaReader(output.track(TS_STREAM_TYPE_MPA_LSF));
pesPayloadReader = new MpegAudioReader(output.track(TS_STREAM_TYPE_MPA_LSF));
break;
case TS_STREAM_TYPE_AAC:
pesPayloadReader = new AdtsReader(output.track(TS_STREAM_TYPE_AAC));

View File

@ -41,22 +41,6 @@ public final class CodecSpecificDataUtil {
private CodecSpecificDataUtil() {}
/**
* Gets the sample rate index.
*
* @param sampleRate The sample rate in Hz.
* @return The sample rate index.
*/
public static int getSampleRateIndex(int sampleRate) {
int sampleRateIndex = 0;
for (; sampleRateIndex < AUDIO_SPECIFIC_CONFIG_SAMPLING_RATE_TABLE.length; sampleRateIndex++) {
if (AUDIO_SPECIFIC_CONFIG_SAMPLING_RATE_TABLE[sampleRateIndex] == sampleRate) {
return sampleRateIndex;
}
}
return -1;
}
/**
* Parses an AudioSpecificConfig, as defined in ISO 14496-3 1.6.2.1
*
@ -65,24 +49,13 @@ public final class CodecSpecificDataUtil {
*/
public static Pair<Integer, Integer> parseAacAudioSpecificConfig(byte[] audioSpecificConfig) {
int audioObjectType = (audioSpecificConfig[0] >> 3) & 0x1F;
if (audioObjectType < 31) {
int byteOffset = audioObjectType == 5 || audioObjectType == 29 ? 1 : 0;
int frequencyIndex = (audioSpecificConfig[byteOffset] & 0x7) << 1 |
((audioSpecificConfig[byteOffset + 1] >> 7) & 0x1);
Assertions.checkState(frequencyIndex < 13);
int sampleRate = AUDIO_SPECIFIC_CONFIG_SAMPLING_RATE_TABLE[frequencyIndex];
int channelCount = AUDIO_SPECIFIC_CONFIG_CHANNEL_COUNT_TABLE[
(audioSpecificConfig[byteOffset + 1] >> 3) & 0xF];
return Pair.create(sampleRate, channelCount);
} else {
int frequencyIndex = (audioSpecificConfig[1] & 0x1E) >> 1;
Assertions.checkState(frequencyIndex < 13);
int sampleRate = AUDIO_SPECIFIC_CONFIG_SAMPLING_RATE_TABLE[frequencyIndex];
int channelCount = AUDIO_SPECIFIC_CONFIG_CHANNEL_COUNT_TABLE[
(audioSpecificConfig[1] & 0x01) << 3 |
((audioSpecificConfig[2] >> 5) & 0x07)];
return Pair.create(sampleRate, channelCount);
}
int byteOffset = audioObjectType == 5 || audioObjectType == 29 ? 1 : 0;
int frequencyIndex = (audioSpecificConfig[byteOffset] & 0x7) << 1
| ((audioSpecificConfig[byteOffset + 1] >> 7) & 0x1);
Assertions.checkState(frequencyIndex < 13);
int sampleRate = AUDIO_SPECIFIC_CONFIG_SAMPLING_RATE_TABLE[frequencyIndex];
int channelCount = (audioSpecificConfig[byteOffset + 1] >> 3) & 0xF;
return Pair.create(sampleRate, channelCount);
}
/**

View File

@ -13,11 +13,23 @@
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.google.android.exoplayer.extractor.mp3;
package com.google.android.exoplayer.util;
/** Parsed MPEG audio frame header. */
/**
* Representation of an MPEG audio frame header.
*/
public final class MpegAudioHeader {
/**
* Theoretical maximum frame size for an MPEG audio stream, which occurs when playing a Layer 2
* MPEG 2.5 audio stream at 16 kb/s (with padding). The size is 1152 sample/frame *
* 160000 bit/s / (8000 sample/s * 8 bit/byte) + 1 padding byte/frame = 2881 byte/frame.
* The next power of two size is 4 KiB.
*/
public static final int MAX_FRAME_SIZE_BYTES = 4096;
private static final String[] MIME_TYPE_BY_LAYER =
new String[] {MimeTypes.AUDIO_MPEG_L1, MimeTypes.AUDIO_MPEG_L2, MimeTypes.AUDIO_MPEG};
private static final int[] SAMPLING_RATE_V1 = {44100, 48000, 32000};
private static final int[] BITRATE_V1_L1 =
{32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448};
@ -30,7 +42,9 @@ public final class MpegAudioHeader {
private static final int[] BITRATE_V2 =
{8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160};
/** Returns the size of the frame associated with {@code header}, or -1 if it is invalid. */
/**
* Returns the size of the frame associated with {@code header}, or -1 if it is invalid.
*/
public static int getFrameSize(int header) {
if ((header & 0xFFE00000) != 0xFFE00000) {
return -1;
@ -92,35 +106,37 @@ public final class MpegAudioHeader {
}
/**
* Returns the header represented by {@code header}, if it is valid; {@code null} otherwise.
* Parses {@code headerData}, populating {@code header} with the parsed data.
*
* @param headerData Header data to parse.
* @param header Header to populate with data from {@code headerData}.
* @return True if the header was populated. False otherwise, indicating that {@code headerData}
* is not a valid MPEG audio header.
*/
public static void populateHeader(int headerData, MpegAudioHeader header) {
public static boolean populateHeader(int headerData, MpegAudioHeader header) {
if ((headerData & 0xFFE00000) != 0xFFE00000) {
return;
return false;
}
int version = (headerData >>> 19) & 3;
if (version == 1) {
return;
return false;
}
int layer = (headerData >>> 17) & 3;
if (layer == 0) {
return;
return false;
}
int bitrateIndex = (headerData >>> 12) & 15;
if (bitrateIndex == 0 || bitrateIndex == 0xF) {
// Disallow "free" bitrate.
return;
return false;
}
int samplingRateIndex = (headerData >>> 10) & 3;
if (samplingRateIndex == 3) {
return;
return false;
}
int sampleRate = SAMPLING_RATE_V1[samplingRateIndex];
@ -154,16 +170,16 @@ public final class MpegAudioHeader {
}
}
String mimeType = MIME_TYPE_BY_LAYER[3 - layer];
int channels = ((headerData >> 6) & 3) == 3 ? 1 : 2;
int layerIndex = 3 - layer;
header.setValues(
version, layerIndex, frameSize, sampleRate, channels, bitrate, samplesPerFrame);
header.setValues(version, mimeType, frameSize, sampleRate, channels, bitrate, samplesPerFrame);
return true;
}
/** MPEG audio header version. */
public int version;
/** MPEG audio layer index, starting at zero. */
public int layerIndex;
/** The mime type. */
public String mimeType;
/** Size of the frame associated with this header, in bytes. */
public int frameSize;
/** Sample rate in samples per second. */
@ -175,10 +191,10 @@ public final class MpegAudioHeader {
/** Number of samples stored in the frame. */
public int samplesPerFrame;
private void setValues(int version, int layerIndex, int frameSize, int sampleRate, int channels,
int bitrate, int samplesPerFrame) {
private void setValues(int version, String mimeType, int frameSize,
int sampleRate, int channels, int bitrate, int samplesPerFrame) {
this.version = version;
this.layerIndex = layerIndex;
this.mimeType = mimeType;
this.frameSize = frameSize;
this.sampleRate = sampleRate;
this.channels = channels;