shaka-player
Version:
DASH/EME video player library
1,417 lines (1,276 loc) • 67.7 kB
JavaScript
/*! @license
* Shaka Player
* Copyright 2016 Google LLC
* SPDX-License-Identifier: Apache-2.0
*/
goog.provide('shaka.media.MediaSourceEngine');
goog.require('goog.asserts');
goog.require('shaka.log');
goog.require('shaka.config.CodecSwitchingStrategy');
goog.require('shaka.media.Capabilities');
goog.require('shaka.media.ContentWorkarounds');
goog.require('shaka.media.ClosedCaptionParser');
goog.require('shaka.media.IClosedCaptionParser');
goog.require('shaka.media.ManifestParser');
goog.require('shaka.media.SegmentReference');
goog.require('shaka.media.TimeRangesUtils');
goog.require('shaka.text.TextEngine');
goog.require('shaka.transmuxer.TransmuxerEngine');
goog.require('shaka.util.BufferUtils');
goog.require('shaka.util.Destroyer');
goog.require('shaka.util.Error');
goog.require('shaka.util.EventManager');
goog.require('shaka.util.Functional');
goog.require('shaka.util.IDestroyable');
goog.require('shaka.util.Id3Utils');
goog.require('shaka.util.ManifestParserUtils');
goog.require('shaka.util.MimeUtils');
goog.require('shaka.util.Mp4BoxParsers');
goog.require('shaka.util.Mp4Parser');
goog.require('shaka.util.Platform');
goog.require('shaka.util.PublicPromise');
goog.require('shaka.util.TsParser');
goog.require('shaka.lcevc.Dec');
/**
* @summary
* MediaSourceEngine wraps all operations on MediaSource and SourceBuffers.
* All asynchronous operations return a Promise, and all operations are
* internally synchronized and serialized as needed. Operations that can
* be done in parallel will be done in parallel.
*
* @implements {shaka.util.IDestroyable}
*/
shaka.media.MediaSourceEngine = class {
/**
* @param {HTMLMediaElement} video The video element, whose source is tied to
* MediaSource during the lifetime of the MediaSourceEngine.
* @param {!shaka.extern.TextDisplayer} textDisplayer
* The text displayer that will be used with the text engine.
* MediaSourceEngine takes ownership of the displayer. When
* MediaSourceEngine is destroyed, it will destroy the displayer.
* @param {!function(!Array.<shaka.extern.ID3Metadata>, number, ?number)=}
* onMetadata
* @param {?shaka.lcevc.Dec} [lcevcDec] Optional - LCEVC Decoder Object
*
*/
constructor(video, textDisplayer, onMetadata, lcevcDec) {
/** @private {HTMLMediaElement} */
this.video_ = video;
/** @private {?shaka.extern.MediaSourceConfiguration} */
this.config_ = null;
/** @private {shaka.extern.TextDisplayer} */
this.textDisplayer_ = textDisplayer;
/** @private {!Object.<shaka.util.ManifestParserUtils.ContentType,
SourceBuffer>} */
this.sourceBuffers_ = {};
/** @private {!Object.<shaka.util.ManifestParserUtils.ContentType,
string>} */
this.sourceBufferTypes_ = {};
/** @private {!Object.<shaka.util.ManifestParserUtils.ContentType,
boolean>} */
this.expectedEncryption_ = {};
/** @private {shaka.text.TextEngine} */
this.textEngine_ = null;
/** @private {boolean} */
this.segmentRelativeVttTiming_ = false;
const onMetadataNoOp = (metadata, timestampOffset, segmentEnd) => {};
/** @private {!function(!Array.<shaka.extern.ID3Metadata>,
number, ?number)} */
this.onMetadata_ = onMetadata || onMetadataNoOp;
/** @private {?shaka.lcevc.Dec} */
this.lcevcDec_ = lcevcDec || null;
/**
* @private {!Object.<string,
* !Array.<shaka.media.MediaSourceEngine.Operation>>}
*/
this.queues_ = {};
/** @private {shaka.util.EventManager} */
this.eventManager_ = new shaka.util.EventManager();
/** @private {!Object.<string, !shaka.extern.Transmuxer>} */
this.transmuxers_ = {};
/** @private {?shaka.media.IClosedCaptionParser} */
this.captionParser_ = null;
/** @private {!shaka.util.PublicPromise} */
this.mediaSourceOpen_ = new shaka.util.PublicPromise();
/** @private {string} */
this.url_ = '';
/** @private {boolean} */
this.playbackHasBegun_ = false;
/** @private {MediaSource} */
this.mediaSource_ = this.createMediaSource(this.mediaSourceOpen_);
/** @private {boolean} */
this.reloadingMediaSource_ = false;
/** @type {!shaka.util.Destroyer} */
this.destroyer_ = new shaka.util.Destroyer(() => this.doDestroy_());
/** @private {boolean} */
this.sequenceMode_ = false;
/** @private {string} */
this.manifestType_ = shaka.media.ManifestParser.UNKNOWN;
/** @private {boolean} */
this.ignoreManifestTimestampsInSegmentsMode_ = false;
/** @private {boolean} */
this.attemptTimestampOffsetCalculation_ = false;
/** @private {!shaka.util.PublicPromise.<number>} */
this.textSequenceModeOffset_ = new shaka.util.PublicPromise();
/** @private {boolean} */
this.needSplitMuxedContent_ = false;
}
/**
* Create a MediaSource object, attach it to the video element, and return it.
* Resolves the given promise when the MediaSource is ready.
*
* Replaced by unit tests.
*
* @param {!shaka.util.PublicPromise} p
* @return {!MediaSource}
*/
createMediaSource(p) {
const mediaSource = new MediaSource();
// Set up MediaSource on the video element.
this.eventManager_.listenOnce(
mediaSource, 'sourceopen', () => this.onSourceOpen_(p));
// Correctly set when playback has begun.
this.eventManager_.listenOnce(this.video_, 'playing', () => {
this.playbackHasBegun_ = true;
});
// Store the object URL for releasing it later.
this.url_ = shaka.media.MediaSourceEngine.createObjectURL(mediaSource);
this.video_.src = this.url_;
return mediaSource;
}
/**
* @param {!shaka.util.PublicPromise} p
* @private
*/
onSourceOpen_(p) {
goog.asserts.assert(this.url_, 'Must have object URL');
// Release the object URL that was previously created, to prevent memory
// leak.
// createObjectURL creates a strong reference to the MediaSource object
// inside the browser. Setting the src of the video then creates another
// reference within the video element. revokeObjectURL will remove the
// strong reference to the MediaSource object, and allow it to be
// garbage-collected later.
URL.revokeObjectURL(this.url_);
p.resolve();
}
/**
* Checks if a certain type is supported.
*
* @param {shaka.extern.Stream} stream
* @return {boolean}
*/
static isStreamSupported(stream) {
const MimeUtils = shaka.util.MimeUtils;
const fullMimeType = MimeUtils.getFullType(stream.mimeType, stream.codecs);
const extendedMimeType = MimeUtils.getExtendedType(stream);
const fullMimeTypeWithAllCodecs = MimeUtils.getFullTypeWithAllCodecs(
stream.mimeType, stream.codecs);
const TransmuxerEngine = shaka.transmuxer.TransmuxerEngine;
return shaka.text.TextEngine.isTypeSupported(fullMimeType) ||
shaka.media.Capabilities.isTypeSupported(extendedMimeType) ||
TransmuxerEngine.isSupported(fullMimeTypeWithAllCodecs, stream.type);
}
/**
* Returns a map of MediaSource support for well-known types.
*
* @return {!Object.<string, boolean>}
*/
static probeSupport() {
const testMimeTypes = [
// MP4 types
'video/mp4; codecs="avc1.42E01E"',
'video/mp4; codecs="avc3.42E01E"',
'video/mp4; codecs="hev1.1.6.L93.90"',
'video/mp4; codecs="hvc1.1.6.L93.90"',
'video/mp4; codecs="hev1.2.4.L153.B0"; eotf="smpte2084"', // HDR HEVC
'video/mp4; codecs="hvc1.2.4.L153.B0"; eotf="smpte2084"', // HDR HEVC
'video/mp4; codecs="vp9"',
'video/mp4; codecs="vp09.00.10.08"',
'video/mp4; codecs="av01.0.01M.08"',
'audio/mp4; codecs="mp4a.40.2"',
'audio/mp4; codecs="ac-3"',
'audio/mp4; codecs="ec-3"',
'audio/mp4; codecs="opus"',
'audio/mp4; codecs="flac"',
'audio/mp4; codecs="dtsc"', // DTS Digital Surround
'audio/mp4; codecs="dtse"', // DTS Express
'audio/mp4; codecs="dtsx"', // DTS:X
// WebM types
'video/webm; codecs="vp8"',
'video/webm; codecs="vp9"',
'video/webm; codecs="vp09.00.10.08"',
'audio/webm; codecs="vorbis"',
'audio/webm; codecs="opus"',
// MPEG2 TS types (video/ is also used for audio: https://bit.ly/TsMse)
'video/mp2t; codecs="avc1.42E01E"',
'video/mp2t; codecs="avc3.42E01E"',
'video/mp2t; codecs="hvc1.1.6.L93.90"',
'video/mp2t; codecs="mp4a.40.2"',
'video/mp2t; codecs="ac-3"',
'video/mp2t; codecs="ec-3"',
// WebVTT types
'text/vtt',
'application/mp4; codecs="wvtt"',
// TTML types
'application/ttml+xml',
'application/mp4; codecs="stpp"',
// Containerless types
...shaka.util.MimeUtils.RAW_FORMATS,
];
const support = {};
for (const type of testMimeTypes) {
if (shaka.util.Platform.supportsMediaSource()) {
// Our TextEngine is only effective for MSE platforms at the moment.
if (shaka.text.TextEngine.isTypeSupported(type)) {
support[type] = true;
} else {
support[type] = shaka.media.Capabilities.isTypeSupported(type) ||
shaka.transmuxer.TransmuxerEngine.isSupported(type);
}
} else {
support[type] = shaka.util.Platform.supportsMediaType(type);
}
const basicType = type.split(';')[0];
support[basicType] = support[basicType] || support[type];
}
return support;
}
/** @override */
destroy() {
return this.destroyer_.destroy();
}
/** @private */
async doDestroy_() {
const Functional = shaka.util.Functional;
const cleanup = [];
for (const contentType in this.queues_) {
// Make a local copy of the queue and the first item.
const q = this.queues_[contentType];
const inProgress = q[0];
// Drop everything else out of the original queue.
this.queues_[contentType] = q.slice(0, 1);
// We will wait for this item to complete/fail.
if (inProgress) {
cleanup.push(inProgress.p.catch(Functional.noop));
}
// The rest will be rejected silently if possible.
for (const item of q.slice(1)) {
item.p.reject(shaka.util.Destroyer.destroyedError());
}
}
if (this.textEngine_) {
cleanup.push(this.textEngine_.destroy());
}
if (this.textDisplayer_) {
cleanup.push(this.textDisplayer_.destroy());
}
for (const contentType in this.transmuxers_) {
cleanup.push(this.transmuxers_[contentType].destroy());
}
await Promise.all(cleanup);
if (this.eventManager_) {
this.eventManager_.release();
this.eventManager_ = null;
}
if (this.video_) {
// "unload" the video element.
this.video_.removeAttribute('src');
this.video_.load();
this.video_ = null;
}
this.config_ = null;
this.mediaSource_ = null;
this.textEngine_ = null;
this.textDisplayer_ = null;
this.sourceBuffers_ = {};
this.transmuxers_ = {};
this.captionParser_ = null;
if (goog.DEBUG) {
for (const contentType in this.queues_) {
goog.asserts.assert(
this.queues_[contentType].length == 0,
contentType + ' queue should be empty after destroy!');
}
}
this.queues_ = {};
// This object is owned by Player
this.lcevcDec_ = null;
}
/**
* @return {!Promise} Resolved when MediaSource is open and attached to the
* media element. This process is actually initiated by the constructor.
*/
open() {
return this.mediaSourceOpen_;
}
/**
* Initialize MediaSourceEngine.
*
* Note that it is not valid to call this multiple times, except to add or
* reinitialize text streams.
*
* @param {!Map.<shaka.util.ManifestParserUtils.ContentType,
* shaka.extern.Stream>} streamsByType
* A map of content types to streams. All streams must be supported
* according to MediaSourceEngine.isStreamSupported.
* @param {boolean=} sequenceMode
* If true, the media segments are appended to the SourceBuffer in strict
* sequence.
* @param {string=} manifestType
* Indicates the type of the manifest.
* @param {boolean=} ignoreManifestTimestampsInSegmentsMode
* If true, don't adjust the timestamp offset to account for manifest
* segment durations being out of sync with segment durations. In other
* words, assume that there are no gaps in the segments when appending
* to the SourceBuffer, even if the manifest and segment times disagree.
* Indicates if the manifest has text streams.
*
* @return {!Promise}
*/
async init(streamsByType, sequenceMode=false,
manifestType=shaka.media.ManifestParser.UNKNOWN,
ignoreManifestTimestampsInSegmentsMode=false) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
await this.mediaSourceOpen_;
this.sequenceMode_ = sequenceMode;
this.manifestType_ = manifestType;
this.ignoreManifestTimestampsInSegmentsMode_ =
ignoreManifestTimestampsInSegmentsMode;
this.attemptTimestampOffsetCalculation_ = !this.sequenceMode_ &&
this.manifestType_ == shaka.media.ManifestParser.HLS &&
!this.ignoreManifestTimestampsInSegmentsMode_;
for (const contentType of streamsByType.keys()) {
const stream = streamsByType.get(contentType);
this.initSourceBuffer_(contentType, stream, stream.codecs);
if (this.needSplitMuxedContent_) {
this.queues_[ContentType.AUDIO] = [];
this.queues_[ContentType.VIDEO] = [];
} else {
this.queues_[contentType] = [];
}
}
}
/**
* Initialize a specific SourceBuffer.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {shaka.extern.Stream} stream
* @param {string} codecs
* @private
*/
initSourceBuffer_(contentType, stream, codecs) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
goog.asserts.assert(
shaka.media.MediaSourceEngine.isStreamSupported(stream),
'Type negotiation should happen before MediaSourceEngine.init!');
let mimeType = shaka.util.MimeUtils.getFullType(
stream.mimeType, codecs);
if (contentType == ContentType.TEXT) {
this.reinitText(mimeType, this.sequenceMode_, stream.external);
} else {
let needTransmux = this.config_.forceTransmux;
if (!shaka.media.Capabilities.isTypeSupported(mimeType) ||
(!this.sequenceMode_ &&
shaka.util.MimeUtils.RAW_FORMATS.includes(mimeType))) {
needTransmux = true;
}
const mimeTypeWithAllCodecs =
shaka.util.MimeUtils.getFullTypeWithAllCodecs(
stream.mimeType, codecs);
const TransmuxerEngine = shaka.transmuxer.TransmuxerEngine;
if (needTransmux) {
const audioCodec = shaka.util.ManifestParserUtils.guessCodecsSafe(
ContentType.AUDIO, (codecs || '').split(','));
const videoCodec = shaka.util.ManifestParserUtils.guessCodecsSafe(
ContentType.VIDEO, (codecs || '').split(','));
if (audioCodec && videoCodec) {
this.needSplitMuxedContent_ = true;
this.initSourceBuffer_(ContentType.AUDIO, stream, audioCodec);
this.initSourceBuffer_(ContentType.VIDEO, stream, videoCodec);
return;
}
const transmuxerPlugin =
TransmuxerEngine.findTransmuxer(mimeTypeWithAllCodecs);
if (transmuxerPlugin) {
const transmuxer = transmuxerPlugin();
this.transmuxers_[contentType] = transmuxer;
mimeType =
transmuxer.convertCodecs(contentType, mimeTypeWithAllCodecs);
}
}
const type = mimeType + this.config_.sourceBufferExtraFeatures;
this.destroyer_.ensureNotDestroyed();
let sourceBuffer;
try {
sourceBuffer = this.mediaSource_.addSourceBuffer(type);
} catch (exception) {
throw new shaka.util.Error(
shaka.util.Error.Severity.CRITICAL,
shaka.util.Error.Category.MEDIA,
shaka.util.Error.Code.MEDIA_SOURCE_OPERATION_THREW,
exception,
'The mediaSource_ status was' + this.mediaSource_.readyState +
'expected \'open\'');
}
if (this.sequenceMode_) {
sourceBuffer.mode =
shaka.media.MediaSourceEngine.SourceBufferMode_.SEQUENCE;
}
this.eventManager_.listen(
sourceBuffer, 'error',
() => this.onError_(contentType));
this.eventManager_.listen(
sourceBuffer, 'updateend',
() => this.onUpdateEnd_(contentType));
this.sourceBuffers_[contentType] = sourceBuffer;
this.sourceBufferTypes_[contentType] = mimeType;
this.expectedEncryption_[contentType] = !!stream.drmInfos.length;
}
}
/**
* Called by the Player to provide an updated configuration any time it
* changes. Must be called at least once before init().
*
* @param {shaka.extern.MediaSourceConfiguration} config
*/
configure(config) {
this.config_ = config;
}
/**
* Reinitialize the TextEngine for a new text type.
* @param {string} mimeType
* @param {boolean} sequenceMode
* @param {boolean} external
*/
reinitText(mimeType, sequenceMode, external) {
if (!this.textEngine_) {
this.textEngine_ = new shaka.text.TextEngine(this.textDisplayer_);
}
this.textEngine_.initParser(mimeType, sequenceMode,
external || this.segmentRelativeVttTiming_, this.manifestType_);
}
/**
* @return {boolean} True if the MediaSource is in an "ended" state, or if the
* object has been destroyed.
*/
ended() {
if (this.reloadingMediaSource_) {
return false;
}
return this.mediaSource_ ? this.mediaSource_.readyState == 'ended' : true;
}
/**
* Gets the first timestamp in buffer for the given content type.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @return {?number} The timestamp in seconds, or null if nothing is buffered.
*/
bufferStart(contentType) {
if (this.reloadingMediaSource_ ||
!Object.keys(this.sourceBuffers_).length) {
return null;
}
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
return this.textEngine_.bufferStart();
}
return shaka.media.TimeRangesUtils.bufferStart(
this.getBuffered_(contentType));
}
/**
* Gets the last timestamp in buffer for the given content type.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @return {?number} The timestamp in seconds, or null if nothing is buffered.
*/
bufferEnd(contentType) {
if (this.reloadingMediaSource_) {
return null;
}
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
return this.textEngine_.bufferEnd();
}
return shaka.media.TimeRangesUtils.bufferEnd(
this.getBuffered_(contentType));
}
/**
* Determines if the given time is inside the buffered range of the given
* content type.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} time Playhead time
* @return {boolean}
*/
isBuffered(contentType, time) {
if (this.reloadingMediaSource_) {
return false;
}
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
return this.textEngine_.isBuffered(time);
} else {
const buffered = this.getBuffered_(contentType);
return shaka.media.TimeRangesUtils.isBuffered(buffered, time);
}
}
/**
* Computes how far ahead of the given timestamp is buffered for the given
* content type.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} time
* @return {number} The amount of time buffered ahead in seconds.
*/
bufferedAheadOf(contentType, time) {
if (this.reloadingMediaSource_) {
return 0;
}
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
return this.textEngine_.bufferedAheadOf(time);
} else {
const buffered = this.getBuffered_(contentType);
return shaka.media.TimeRangesUtils.bufferedAheadOf(buffered, time);
}
}
/**
* Returns info about what is currently buffered.
* @return {shaka.extern.BufferedInfo}
*/
getBufferedInfo() {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
const TimeRangesUtils = shaka.media.TimeRangesUtils;
const info = {
total: this.reloadingMediaSource_ ? [] :
TimeRangesUtils.getBufferedInfo(this.video_.buffered),
audio: this.reloadingMediaSource_ ? [] :
TimeRangesUtils.getBufferedInfo(this.getBuffered_(ContentType.AUDIO)),
video: this.reloadingMediaSource_ ? [] :
TimeRangesUtils.getBufferedInfo(this.getBuffered_(ContentType.VIDEO)),
text: [],
};
if (this.textEngine_) {
const start = this.textEngine_.bufferStart();
const end = this.textEngine_.bufferEnd();
if (start != null && end != null) {
info.text.push({start: start, end: end});
}
}
return info;
}
/**
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @return {TimeRanges} The buffered ranges for the given content type, or
* null if the buffered ranges could not be obtained.
* @private
*/
getBuffered_(contentType) {
try {
return this.sourceBuffers_[contentType].buffered;
} catch (exception) {
if (contentType in this.sourceBuffers_) {
// Note: previous MediaSource errors may cause access to |buffered| to
// throw.
shaka.log.error('failed to get buffered range for ' + contentType,
exception);
}
return null;
}
}
/**
* Create a new closed caption parser. This will ONLY be replaced by tests as
* a way to inject fake closed caption parser instances.
*
* @param {string} mimeType
* @return {!shaka.media.IClosedCaptionParser}
*/
getCaptionParser(mimeType) {
return new shaka.media.ClosedCaptionParser(mimeType);
}
/**
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {!BufferSource} data
* @param {?shaka.media.SegmentReference} reference The segment reference
* we are appending, or null for init segments
* @param {!string} mimeType
* @param {!number} timestampOffset
* @return {?number}
* @private
*/
getTimestampAndDispatchMetadata_(contentType, data, reference, mimeType,
timestampOffset) {
let timestamp = null;
const uint8ArrayData = shaka.util.BufferUtils.toUint8(data);
if (shaka.util.MimeUtils.RAW_FORMATS.includes(mimeType)) {
const frames = shaka.util.Id3Utils.getID3Frames(uint8ArrayData);
if (frames.length && reference) {
const metadataTimestamp = frames.find((frame) => {
return frame.description ===
'com.apple.streaming.transportStreamTimestamp';
});
if (metadataTimestamp && metadataTimestamp.data) {
timestamp = Math.round(metadataTimestamp.data) / 1000;
}
/** @private {shaka.extern.ID3Metadata} */
const metadata = {
cueTime: reference.startTime,
data: uint8ArrayData,
frames: frames,
dts: reference.startTime,
pts: reference.startTime,
};
this.onMetadata_([metadata], /* offset= */ 0, reference.endTime);
}
} else if (mimeType.includes('/mp4') &&
reference && reference.timestampOffset == 0 &&
reference.initSegmentReference &&
reference.initSegmentReference.timescale) {
const timescale = reference.initSegmentReference.timescale;
if (!isNaN(timescale)) {
const Mp4Parser = shaka.util.Mp4Parser;
let startTime = 0;
let parsedMedia = false;
new Mp4Parser()
.box('moof', Mp4Parser.children)
.box('traf', Mp4Parser.children)
.fullBox('tfdt', (box) => {
goog.asserts.assert(
box.version == 0 || box.version == 1,
'TFDT version can only be 0 or 1');
const parsed = shaka.util.Mp4BoxParsers.parseTFDTInaccurate(
box.reader, box.version);
startTime = parsed.baseMediaDecodeTime / timescale;
parsedMedia = true;
box.parser.stop();
}).parse(data, /* partialOkay= */ true);
if (parsedMedia) {
timestamp = startTime;
}
}
} else if (!mimeType.includes('/mp4') && !mimeType.includes('/webm') &&
shaka.util.TsParser.probe(uint8ArrayData)) {
const tsParser = new shaka.util.TsParser().parse(uint8ArrayData);
const startTime = tsParser.getStartTime()[contentType];
if (startTime != null) {
timestamp = startTime;
}
const metadata = tsParser.getMetadata();
if (metadata.length) {
this.onMetadata_(metadata, timestampOffset,
reference ? reference.endTime : null);
}
}
return timestamp;
}
/**
* Enqueue an operation to append data to the SourceBuffer.
* Start and end times are needed for TextEngine, but not for MediaSource.
* Start and end times may be null for initialization segments; if present
* they are relative to the presentation timeline.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {!BufferSource} data
* @param {?shaka.media.SegmentReference} reference The segment reference
* we are appending, or null for init segments
* @param {shaka.extern.Stream} stream
* @param {?boolean} hasClosedCaptions True if the buffer contains CEA closed
* captions
* @param {boolean=} seeked True if we just seeked
* @param {boolean=} adaptation True if we just automatically switched active
* variant(s).
* @param {boolean=} isChunkedData True if we add to the buffer from the
* partial read of the segment.
* @return {!Promise}
*/
async appendBuffer(
contentType, data, reference, stream, hasClosedCaptions, seeked = false,
adaptation = false, isChunkedData = false, fromSplit = false) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
if (this.sequenceMode_) {
// This won't be known until the first video segment is appended.
const offset = await this.textSequenceModeOffset_;
this.textEngine_.setTimestampOffset(offset);
}
await this.textEngine_.appendBuffer(
data,
reference ? reference.startTime : null,
reference ? reference.endTime : null);
return;
}
if (!fromSplit && this.needSplitMuxedContent_) {
await this.appendBuffer(ContentType.AUDIO, data, reference, stream,
hasClosedCaptions, seeked, adaptation, isChunkedData,
/* fromSplit= */ true);
await this.appendBuffer(ContentType.VIDEO, data, reference, stream,
hasClosedCaptions, seeked, adaptation, isChunkedData,
/* fromSplit= */ true);
return;
}
if (!this.sourceBuffers_[contentType]) {
shaka.log.warning('Attempted to restore a non-existent source buffer');
return;
}
let timestampOffset = this.sourceBuffers_[contentType].timestampOffset;
let mimeType = this.sourceBufferTypes_[contentType];
if (this.transmuxers_[contentType]) {
mimeType = this.transmuxers_[contentType].getOrginalMimeType();
}
if (reference) {
const timestamp = this.getTimestampAndDispatchMetadata_(
contentType, data, reference, mimeType, timestampOffset);
if (timestamp != null) {
const calculatedTimestampOffset = reference.startTime - timestamp;
const timestampOffsetDifference =
Math.abs(timestampOffset - calculatedTimestampOffset);
if (timestampOffsetDifference >= 0.1 || seeked || adaptation) {
timestampOffset = calculatedTimestampOffset;
if (this.attemptTimestampOffsetCalculation_) {
this.enqueueOperation_(
contentType,
() => this.abort_(contentType));
this.enqueueOperation_(
contentType,
() => this.setTimestampOffset_(contentType, timestampOffset));
}
}
// Timestamps can only be reliably extracted from video, not audio.
// Packed audio formats do not have internal timestamps at all.
// Prefer video for this when available.
const isBestSourceBufferForTimestamps =
contentType == ContentType.VIDEO ||
!(ContentType.VIDEO in this.sourceBuffers_);
if (this.sequenceMode_ && isBestSourceBufferForTimestamps) {
this.textSequenceModeOffset_.resolve(timestampOffset);
}
}
}
if (hasClosedCaptions && contentType == ContentType.VIDEO) {
if (!this.textEngine_) {
this.reinitText(shaka.util.MimeUtils.CEA608_CLOSED_CAPTION_MIMETYPE,
this.sequenceMode_, /* external= */ false);
}
if (!this.captionParser_) {
const basicType = mimeType.split(';', 1)[0];
this.captionParser_ = this.getCaptionParser(basicType);
}
// If it is the init segment for closed captions, initialize the closed
// caption parser.
if (!reference) {
this.captionParser_.init(data);
} else {
const closedCaptions = this.captionParser_.parseFrom(data);
if (closedCaptions.length) {
this.textEngine_.storeAndAppendClosedCaptions(
closedCaptions,
reference.startTime,
reference.endTime,
timestampOffset);
}
}
}
if (this.transmuxers_[contentType]) {
data = await this.transmuxers_[contentType].transmux(
data, stream, reference, this.mediaSource_.duration, contentType);
}
data = this.workAroundBrokenPlatforms_(
data, reference ? reference.startTime : null, contentType);
if (reference && this.sequenceMode_ && contentType != ContentType.TEXT) {
// In sequence mode, for non-text streams, if we just cleared the buffer
// and are either performing an unbuffered seek or handling an automatic
// adaptation, we need to set a new timestampOffset on the sourceBuffer.
if (seeked || adaptation) {
const timestampOffset = reference.startTime;
// The logic to call abort() before setting the timestampOffset is
// extended during unbuffered seeks or automatic adaptations; it is
// possible for the append state to be PARSING_MEDIA_SEGMENT from the
// previous SourceBuffer#appendBuffer() call.
this.enqueueOperation_(contentType, () => this.abort_(contentType));
this.enqueueOperation_(
contentType,
() => this.setTimestampOffset_(contentType, timestampOffset));
}
}
let bufferedBefore = null;
await this.enqueueOperation_(contentType, () => {
if (goog.DEBUG && reference && !reference.isPreload() && !isChunkedData) {
bufferedBefore = this.getBuffered_(contentType);
}
this.append_(contentType, data, timestampOffset);
});
if (goog.DEBUG && reference && !reference.isPreload() && !isChunkedData) {
const bufferedAfter = this.getBuffered_(contentType);
const newBuffered = shaka.media.TimeRangesUtils.computeAddedRange(
bufferedBefore, bufferedAfter);
if (newBuffered) {
const segmentDuration = reference.endTime - reference.startTime;
// Check end times instead of start times. We may be overwriting a
// buffer and only the end changes, and that would be fine.
// Also, exclude tiny segments. Sometimes alignment segments as small
// as 33ms are seen in Google DAI content. For such tiny segments,
// half a segment duration would be no issue.
const offset = Math.abs(newBuffered.end - reference.endTime);
if (segmentDuration > 0.100 && offset > segmentDuration / 2) {
shaka.log.error('Possible encoding problem detected!',
'Unexpected buffered range for reference', reference,
'from URIs', reference.getUris(),
'should be', {start: reference.startTime, end: reference.endTime},
'but got', newBuffered);
}
}
}
}
/**
* Set the selected closed captions Id and language.
*
* @param {string} id
*/
setSelectedClosedCaptionId(id) {
const VIDEO = shaka.util.ManifestParserUtils.ContentType.VIDEO;
const videoBufferEndTime = this.bufferEnd(VIDEO) || 0;
this.textEngine_.setSelectedClosedCaptionId(id, videoBufferEndTime);
}
/** Disable embedded closed captions. */
clearSelectedClosedCaptionId() {
if (this.textEngine_) {
this.textEngine_.setSelectedClosedCaptionId('', 0);
}
}
/**
* Enqueue an operation to remove data from the SourceBuffer.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} startTime relative to the start of the presentation
* @param {number} endTime relative to the start of the presentation
* @return {!Promise}
*/
async remove(contentType, startTime, endTime) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
await this.textEngine_.remove(startTime, endTime);
} else {
await this.enqueueOperation_(
contentType,
() => this.remove_(contentType, startTime, endTime));
if (this.needSplitMuxedContent_) {
await this.enqueueOperation_(
ContentType.AUDIO,
() => this.remove_(ContentType.AUDIO, startTime, endTime));
}
}
}
/**
* Enqueue an operation to clear the SourceBuffer.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @return {!Promise}
*/
async clear(contentType) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
if (!this.textEngine_) {
return;
}
await this.textEngine_.remove(0, Infinity);
} else {
// Note that not all platforms allow clearing to Infinity.
await this.enqueueOperation_(
contentType,
() => this.remove_(contentType, 0, this.mediaSource_.duration));
if (this.needSplitMuxedContent_) {
await this.enqueueOperation_(
ContentType.AUDIO,
() => this.remove_(
ContentType.AUDIO, 0, this.mediaSource_.duration));
}
}
}
/**
* Fully reset the state of the caption parser owned by MediaSourceEngine.
*/
resetCaptionParser() {
if (this.captionParser_) {
this.captionParser_.reset();
}
}
/**
* Enqueue an operation to flush the SourceBuffer.
* This is a workaround for what we believe is a Chromecast bug.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @return {!Promise}
*/
async flush(contentType) {
// Flush the pipeline. Necessary on Chromecast, even though we have removed
// everything.
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
// Nothing to flush for text.
return;
}
await this.enqueueOperation_(
contentType,
() => this.flush_(contentType));
if (this.needSplitMuxedContent_) {
await this.enqueueOperation_(
ContentType.AUDIO,
() => this.flush_(ContentType.AUDIO));
}
}
/**
* Sets the timestamp offset and append window end for the given content type.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} timestampOffset The timestamp offset. Segments which start
* at time t will be inserted at time t + timestampOffset instead. This
* value does not affect segments which have already been inserted.
* @param {number} appendWindowStart The timestamp to set the append window
* start to. For future appends, frames/samples with timestamps less than
* this value will be dropped.
* @param {number} appendWindowEnd The timestamp to set the append window end
* to. For future appends, frames/samples with timestamps greater than this
* value will be dropped.
* @param {boolean} ignoreTimestampOffset If true, the timestampOffset will
* not be applied in this step.
* @param {shaka.extern.Stream} stream The current stream.
* @param {!Map.<shaka.util.ManifestParserUtils.ContentType,
* shaka.extern.Stream>} streamsByType
* A map of content types to streams. All streams must be supported
* according to MediaSourceEngine.isStreamSupported.
*
* @return {!Promise}
*/
async setStreamProperties(
contentType, timestampOffset, appendWindowStart, appendWindowEnd,
ignoreTimestampOffset, stream, streamsByType) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
if (!ignoreTimestampOffset) {
this.textEngine_.setTimestampOffset(timestampOffset);
}
this.textEngine_.setAppendWindow(appendWindowStart, appendWindowEnd);
return;
}
const operations = [];
const hasChangedCodecs =
await this.codecSwitchIfNecessary_(contentType, stream, streamsByType);
if (!hasChangedCodecs) {
// Queue an abort() to help MSE splice together overlapping segments.
// We set appendWindowEnd when we change periods in DASH content, and the
// period transition may result in overlap.
//
// An abort() also helps with MPEG2-TS. When we append a TS segment, we
// always enter a PARSING_MEDIA_SEGMENT state and we can't change the
// timestamp offset. By calling abort(), we reset the state so we can
// set it.
operations.push(this.enqueueOperation_(
contentType,
() => this.abort_(contentType)));
if (this.needSplitMuxedContent_) {
operations.push(this.enqueueOperation_(
ContentType.AUDIO,
() => this.abort_(ContentType.AUDIO)));
}
}
if (!ignoreTimestampOffset) {
operations.push(this.enqueueOperation_(
contentType,
() => this.setTimestampOffset_(contentType, timestampOffset)));
if (this.needSplitMuxedContent_) {
operations.push(this.enqueueOperation_(
ContentType.AUDIO,
() => this.setTimestampOffset_(
ContentType.AUDIO, timestampOffset)));
}
}
operations.push(this.enqueueOperation_(
contentType,
() => this.setAppendWindow_(
contentType, appendWindowStart, appendWindowEnd)));
if (this.needSplitMuxedContent_) {
operations.push(this.enqueueOperation_(
ContentType.AUDIO,
() => this.setAppendWindow_(
ContentType.AUDIO, appendWindowStart, appendWindowEnd)));
}
await Promise.all(operations);
}
/**
* Adjust timestamp offset to maintain AV sync across discontinuities.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} timestampOffset
* @return {!Promise}
*/
async resync(contentType, timestampOffset) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
if (contentType == ContentType.TEXT) {
// This operation is for audio and video only.
return;
}
// Queue an abort() to help MSE splice together overlapping segments.
// We set appendWindowEnd when we change periods in DASH content, and the
// period transition may result in overlap.
//
// An abort() also helps with MPEG2-TS. When we append a TS segment, we
// always enter a PARSING_MEDIA_SEGMENT state and we can't change the
// timestamp offset. By calling abort(), we reset the state so we can
// set it.
this.enqueueOperation_(
contentType,
() => this.abort_(contentType));
if (this.needSplitMuxedContent_) {
this.enqueueOperation_(
ContentType.AUDIO,
() => this.abort_(ContentType.AUDIO));
}
await this.enqueueOperation_(
contentType,
() => this.setTimestampOffset_(contentType, timestampOffset));
if (this.needSplitMuxedContent_) {
await this.enqueueOperation_(
ContentType.AUDIO,
() => this.setTimestampOffset_(ContentType.AUDIO, timestampOffset));
}
}
/**
* @param {string=} reason Valid reasons are 'network' and 'decode'.
* @return {!Promise}
* @see http://w3c.github.io/media-source/#idl-def-EndOfStreamError
*/
async endOfStream(reason) {
await this.enqueueBlockingOperation_(() => {
// If endOfStream() has already been called on the media source,
// don't call it again. Also do not call if readyState is
// 'closed' (not attached to video element) since it is not a
// valid operation.
if (this.ended() || this.mediaSource_.readyState === 'closed') {
return;
}
// Tizen won't let us pass undefined, but it will let us omit the
// argument.
if (reason) {
this.mediaSource_.endOfStream(reason);
} else {
this.mediaSource_.endOfStream();
}
});
}
/**
* @param {number} duration
* @return {!Promise}
*/
async setDuration(duration) {
await this.enqueueBlockingOperation_(() => {
// Reducing the duration causes the MSE removal algorithm to run, which
// triggers an 'updateend' event to fire. To handle this scenario, we
// have to insert a dummy operation into the beginning of each queue,
// which the 'updateend' handler will remove.
if (duration < this.mediaSource_.duration) {
for (const contentType in this.sourceBuffers_) {
const dummyOperation = {
start: () => {},
p: new shaka.util.PublicPromise(),
};
this.queues_[contentType].unshift(dummyOperation);
}
}
this.mediaSource_.duration = duration;
});
}
/**
* Get the current MediaSource duration.
*
* @return {number}
*/
getDuration() {
return this.mediaSource_.duration;
}
/**
* Append data to the SourceBuffer.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {BufferSource} data
* @param {number} timestampOffset
* @private
*/
append_(contentType, data, timestampOffset) {
const ContentType = shaka.util.ManifestParserUtils.ContentType;
// Append only video data to the LCEVC Dec.
if (contentType == ContentType.VIDEO && this.lcevcDec_) {
// Append video buffers to the LCEVC Dec for parsing and storing
// of LCEVC data.
this.lcevcDec_.appendBuffer(data, timestampOffset);
}
// This will trigger an 'updateend' event.
this.sourceBuffers_[contentType].appendBuffer(data);
}
/**
* Remove data from the SourceBuffer.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} startTime relative to the start of the presentation
* @param {number} endTime relative to the start of the presentation
* @private
*/
remove_(contentType, startTime, endTime) {
if (endTime <= startTime) {
// Ignore removal of inverted or empty ranges.
// Fake 'updateend' event to resolve the operation.
this.onUpdateEnd_(contentType);
return;
}
// This will trigger an 'updateend' event.
this.sourceBuffers_[contentType].remove(startTime, endTime);
}
/**
* Call abort() on the SourceBuffer.
* This resets MSE's last_decode_timestamp on all track buffers, which should
* trigger the splicing logic for overlapping segments.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @private
*/
abort_(contentType) {
// Save the append window, which is reset on abort().
const appendWindowStart =
this.sourceBuffers_[contentType].appendWindowStart;
const appendWindowEnd = this.sourceBuffers_[contentType].appendWindowEnd;
// This will not trigger an 'updateend' event, since nothing is happening.
// This is only to reset MSE internals, not to abort an actual operation.
this.sourceBuffers_[contentType].abort();
// Restore the append window.
this.sourceBuffers_[contentType].appendWindowStart = appendWindowStart;
this.sourceBuffers_[contentType].appendWindowEnd = appendWindowEnd;
// Fake an 'updateend' event to resolve the operation.
this.onUpdateEnd_(contentType);
}
/**
* Nudge the playhead to force the media pipeline to be flushed.
* This seems to be necessary on Chromecast to get new content to replace old
* content.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @private
*/
flush_(contentType) {
// Never use flush_ if there's data. It causes a hiccup in playback.
goog.asserts.assert(
this.video_.buffered.length == 0, 'MediaSourceEngine.flush_ should ' +
'only be used after clearing all data!');
// Seeking forces the pipeline to be flushed.
this.video_.currentTime -= 0.001;
// Fake an 'updateend' event to resolve the operation.
this.onUpdateEnd_(contentType);
}
/**
* Set the SourceBuffer's timestamp offset.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} timestampOffset
* @private
*/
setTimestampOffset_(contentType, timestampOffset) {
// Work around for
// https://github.com/shaka-project/shaka-player/issues/1281:
// TODO(https://bit.ly/2ttKiBU): follow up when this is fixed in Edge
if (timestampOffset < 0) {
// Try to prevent rounding errors in Edge from removing the first
// keyframe.
timestampOffset += 0.001;
}
this.sourceBuffers_[contentType].timestampOffset = timestampOffset;
// Fake an 'updateend' event to resolve the operation.
this.onUpdateEnd_(contentType);
}
/**
* Set the SourceBuffer's append window end.
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {number} appendWindowStart
* @param {number} appendWindowEnd
* @private
*/
setAppendWindow_(contentType, appendWindowStart, appendWindowEnd) {
// You can't set start > end, so first set start to 0, then set the new
// end, then set the new start. That way, there are no intermediate
// states which are invalid.
this.sourceBuffers_[contentType].appendWindowStart = 0;
this.sourceBuffers_[contentType].appendWindowEnd = appendWindowEnd;
this.sourceBuffers_[contentType].appendWindowStart = appendWindowStart;
// Fake an 'updateend' event to resolve the operation.
this.onUpdateEnd_(contentType);
}
/**
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @private
*/
onError_(contentType) {
const operation = this.queues_[contentType][0];
goog.asserts.assert(operation, 'Spurious error event!');
goog.asserts.assert(!this.sourceBuffers_[contentType].updating,
'SourceBuffer should not be updating on error!');
const code = this.video_.error ? this.video_.error.code : 0;
operation.p.reject(new shaka.util.Error(
shaka.util.Error.Severity.CRITICAL,
shaka.util.Error.Category.MEDIA,
shaka.util.Error.Code.MEDIA_SOURCE_OPERATION_FAILED,
code));
// Do not pop from queue. An 'updateend' event will fire next, and to
// avoid synchronizing these two event handlers, we will allow that one to
// pop from the queue as normal. Note that because the operation has
// already been rejected, the call to resolve() in the 'updateend' handler
// will have no effect.
}
/**
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @private
*/
onUpdateEnd_(contentType) {
if (this.reloadingMediaSource_) {
return;
}
const operation = this.queues_[contentType][0];
goog.asserts.assert(operation, 'Spurious updateend event!');
if (!operation) {
return;
}
goog.asserts.assert(!this.sourceBuffers_[contentType].updating,
'SourceBuffer should not be updating on updateend!');
operation.p.resolve();
this.popFromQueue_(contentType);
}
/**
* Enqueue an operation and start it if appropriate.
*
* @param {shaka.util.ManifestParserUtils.ContentType} contentType
* @param {function()} start
* @return {!Promise}
* @private
*/
enqueueOperation_(contentType, start) {
this.destroyer_.ensureNotDestroyed();
const operation = {
start: start,
p: new shaka.util.PublicPromise(),
};
this.queues_[contentType].push(operation);
if (this.queues_[contentType].length == 1) {
this.startOperation_(contentType);
}
return operation.p;
}
/**
* Enqueue an operation which must block all other operations on all
* SourceBuffers.
*
* @param {function():(Promise|undefined)} run
* @return {!Promise}
* @private
*/
async enqueueBlockingOperation_(run) {
this.destroyer_.ensureNotDestroyed();
/** @type {!Array.<!shaka.util.PublicPromise>} */
const allWaiters = [];
// Enqueue a 'wait' operation onto each queue.
// This operation signals its readiness when it starts.
// When all wait operations are ready, the real operation takes place.
for (const contentType in this.sourceBuffers_) {
const ready = new shaka.util.PublicPromise();
const operation = {