Source: lib/media/segment_utils.js

  1. /*! @license
  2.  * Shaka Player
  3.  * Copyright 2016 Google LLC
  4.  * SPDX-License-Identifier: Apache-2.0
  5.  */
  7. goog.provide('shaka.media.SegmentUtils');
  9. goog.require('goog.asserts');
  10. goog.require('shaka.log');
  11. goog.require('shaka.media.Capabilities');
  12. goog.require('shaka.media.ClosedCaptionParser');
  13. goog.require('shaka.util.BufferUtils');
  14. goog.require('shaka.util.ManifestParserUtils');
  15. goog.require('shaka.util.MimeUtils');
  16. goog.require('shaka.util.Mp4BoxParsers');
  17. goog.require('shaka.util.Mp4Parser');
  18. goog.require('shaka.util.TsParser');
  21. /**
  22.  * @summary Utility functions for segment parsing.
  23.  */
  24. shaka.media.SegmentUtils = class {
  25.   /**
  26.    * @param {string} mimeType
  27.    * @return {shaka.media.SegmentUtils.BasicInfo}
  28.    */
  29.   static getBasicInfoFromMimeType(mimeType) {
  30.     const baseMimeType = shaka.util.MimeUtils.getBasicType(mimeType);
  31.     const type = baseMimeType.split('/')[0];
  32.     const codecs = shaka.util.MimeUtils.getCodecs(mimeType);
  33.     return {
  34.       type: type,
  35.       mimeType: baseMimeType,
  36.       codecs: codecs,
  37.       language: null,
  38.       height: null,
  39.       width: null,
  40.       channelCount: null,
  41.       sampleRate: null,
  42.       closedCaptions: new Map(),
  43.       videoRange: null,
  44.       colorGamut: null,
  45.       frameRate: null,
  46.     };
  47.   }
  49.   /**
  50.    * @param {!BufferSource} data
  51.    * @param {boolean} disableAudio
  52.    * @param {boolean} disableVideo
  53.    * @param {boolean} disableText
  54.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  55.    */
  56.   static getBasicInfoFromTs(data, disableAudio, disableVideo, disableText) {
  57.     const uint8ArrayData = shaka.util.BufferUtils.toUint8(data);
  58.     const tsParser = new shaka.util.TsParser().parse(uint8ArrayData);
  59.     const tsCodecs = tsParser.getCodecs();
  60.     const videoInfo = tsParser.getVideoInfo();
  61.     const codecs = [];
  62.     let hasAudio = false;
  63.     let hasVideo = false;
  64.     if (!disableAudio) {
  65.       switch (tsCodecs.audio) {
  66.         case 'aac':
  67.         case 'aac-loas':
  68.           if (tsParser.getAudioData().length) {
  69.             codecs.push('mp4a.40.2');
  70.             hasAudio = true;
  71.           }
  72.           break;
  73.         case 'mp3':
  74.           if (tsParser.getAudioData().length) {
  75.             codecs.push('mp4a.40.34');
  76.             hasAudio = true;
  77.           }
  78.           break;
  79.         case 'ac3':
  80.           if (tsParser.getAudioData().length) {
  81.             codecs.push('ac-3');
  82.             hasAudio = true;
  83.           }
  84.           break;
  85.         case 'ec3':
  86.           if (tsParser.getAudioData().length) {
  87.             codecs.push('ec-3');
  88.             hasAudio = true;
  89.           }
  90.           break;
  91.         case 'opus':
  92.           if (tsParser.getAudioData().length) {
  93.             codecs.push('opus');
  94.             hasAudio = true;
  95.           }
  96.           break;
  97.       }
  98.     }
  99.     if (!disableVideo) {
  100.       switch (tsCodecs.video) {
  101.         case 'avc':
  102.           if (videoInfo.codec) {
  103.             codecs.push(videoInfo.codec);
  104.           } else {
  105.             codecs.push('avc1.42E01E');
  106.           }
  107.           hasVideo = true;
  108.           break;
  109.         case 'hvc':
  110.           if (videoInfo.codec) {
  111.             codecs.push(videoInfo.codec);
  112.           } else {
  113.             codecs.push('hvc1.1.6.L93.90');
  114.           }
  115.           hasVideo = true;
  116.           break;
  117.         case 'av1':
  118.           codecs.push('av01.0.01M.08');
  119.           hasVideo = true;
  120.           break;
  121.       }
  122.     }
  123.     if (!codecs.length) {
  124.       return null;
  125.     }
  126.     const onlyAudio = hasAudio && !hasVideo;
  127.     const closedCaptions = new Map();
  128.     if (hasVideo && !disableText) {
  129.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp2t');
  130.       captionParser.parseFrom(data);
  131.       for (const stream of captionParser.getStreams()) {
  132.         closedCaptions.set(stream, stream);
  133.       }
  134.       captionParser.reset();
  135.     }
  136.     return {
  137.       type: onlyAudio ? 'audio' : 'video',
  138.       mimeType: 'video/mp2t',
  139.       codecs: codecs.join(', '),
  140.       language: null,
  141.       height: videoInfo.height,
  142.       width: videoInfo.width,
  143.       channelCount: null,
  144.       sampleRate: null,
  145.       closedCaptions: closedCaptions,
  146.       videoRange: null,
  147.       colorGamut: null,
  148.       frameRate: videoInfo.frameRate,
  149.     };
  150.   }
  152.   /**
  153.    * @param {?BufferSource} initData
  154.    * @param {!BufferSource} data
  155.    * @param {boolean} disableText
  156.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  157.    */
  158.   static getBasicInfoFromMp4(initData, data, disableText) {
  159.     const Mp4Parser = shaka.util.Mp4Parser;
  160.     const SegmentUtils = shaka.media.SegmentUtils;
  162.     const audioCodecs = [];
  163.     let videoCodecs = [];
  165.     let hasAudio = false;
  166.     let hasVideo = false;
  168.     const addCodec = (codec) => {
  169.       const codecLC = codec.toLowerCase();
  170.       switch (codecLC) {
  171.         case 'avc1':
  172.         case 'avc3':
  173.           videoCodecs.push(codecLC + '.42E01E');
  174.           hasVideo = true;
  175.           break;
  176.         case 'hev1':
  177.         case 'hvc1':
  178.           videoCodecs.push(codecLC + '.1.6.L93.90');
  179.           hasVideo = true;
  180.           break;
  181.         case 'dvh1':
  182.         case 'dvhe':
  183.           videoCodecs.push(codecLC + '.05.04');
  184.           hasVideo = true;
  185.           break;
  186.         case 'vp09':
  187.           videoCodecs.push(codecLC + '.00.10.08');
  188.           hasVideo = true;
  189.           break;
  190.         case 'av01':
  191.           videoCodecs.push(codecLC + '.0.01M.08');
  192.           hasVideo = true;
  193.           break;
  194.         case 'mp4a':
  195.           // We assume AAC, but this can be wrong since mp4a supports
  196.           // others codecs
  197.           audioCodecs.push('mp4a.40.2');
  198.           hasAudio = true;
  199.           break;
  200.         case 'ac-3':
  201.         case 'ec-3':
  202.         case 'ac-4':
  203.         case 'opus':
  204.         case 'flac':
  205.           audioCodecs.push(codecLC);
  206.           hasAudio = true;
  207.           break;
  208.         case 'apac':
  209.           audioCodecs.push('apac.31.00');
  210.           hasAudio = true;
  211.           break;
  212.       }
  213.     };
  215.     const codecBoxParser = (box) => addCodec(box.name);
  217.     /** @type {?string} */
  218.     let language = null;
  219.     /** @type {?string} */
  220.     let height = null;
  221.     /** @type {?string} */
  222.     let width = null;
  223.     /** @type {?number} */
  224.     let channelCount = null;
  225.     /** @type {?number} */
  226.     let sampleRate = null;
  227.     /** @type {?string} */
  228.     let realVideoRange = null;
  229.     /** @type {?string} */
  230.     let realColorGamut = null;
  231.     /** @type {?string} */
  232.     const realFrameRate = null;
  234.     /** @type {?string} */
  235.     let baseBox;
  237.     const genericAudioBox = (box) => {
  238.       const parsedAudioSampleEntryBox =
  239.           shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  240.       channelCount = parsedAudioSampleEntryBox.channelCount;
  241.       sampleRate = parsedAudioSampleEntryBox.sampleRate;
  242.       codecBoxParser(box);
  243.     };
  245.     const genericVideoBox = (box) => {
  246.       baseBox = box.name;
  247.       const parsedVisualSampleEntryBox =
  248.           shaka.util.Mp4BoxParsers.visualSampleEntry(box.reader);
  249.       width = String(parsedVisualSampleEntryBox.width);
  250.       height = String(parsedVisualSampleEntryBox.height);
  251.       if (box.reader.hasMoreData()) {
  252.         Mp4Parser.children(box);
  253.       }
  254.     };
  256.     new Mp4Parser()
  257.         .box('moov', Mp4Parser.children)
  258.         .box('trak', Mp4Parser.children)
  259.         .box('mdia', Mp4Parser.children)
  260.         .fullBox('mdhd', (box) => {
  261.           goog.asserts.assert(
  262.               box.version != null,
  263.               'MDHD is a full box and should have a valid version.');
  264.           const parsedMDHDBox = shaka.util.Mp4BoxParsers.parseMDHD(
  265.               box.reader, box.version);
  266.           language = parsedMDHDBox.language;
  267.         })
  268.         .box('minf', Mp4Parser.children)
  269.         .box('stbl', Mp4Parser.children)
  270.         .fullBox('stsd', Mp4Parser.sampleDescription)
  272.         // AUDIO
  273.         // These are the various boxes that signal a codec.
  274.         .box('mp4a', (box) => {
  275.           const parsedAudioSampleEntryBox =
  276.               shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  277.           channelCount = parsedAudioSampleEntryBox.channelCount;
  278.           sampleRate = parsedAudioSampleEntryBox.sampleRate;
  279.           if (box.reader.hasMoreData()) {
  280.             Mp4Parser.children(box);
  281.           } else {
  282.             codecBoxParser(box);
  283.           }
  284.         })
  285.         .box('esds', (box) => {
  286.           const parsedESDSBox = shaka.util.Mp4BoxParsers.parseESDS(box.reader);
  287.           audioCodecs.push(parsedESDSBox.codec);
  288.           hasAudio = true;
  289.         })
  290.         .box('ac-3', genericAudioBox)
  291.         .box('ec-3', genericAudioBox)
  292.         .box('ac-4', genericAudioBox)
  293.         .box('Opus', genericAudioBox)
  294.         .box('fLaC', genericAudioBox)
  295.         .box('apac', genericAudioBox)
  297.         // VIDEO
  298.         // These are the various boxes that signal a codec.
  299.         .box('avc1', genericVideoBox)
  300.         .box('avc3', genericVideoBox)
  301.         .box('hev1', genericVideoBox)
  302.         .box('hvc1', genericVideoBox)
  303.         .box('dva1', genericVideoBox)
  304.         .box('dvav', genericVideoBox)
  305.         .box('dvh1', genericVideoBox)
  306.         .box('dvhe', genericVideoBox)
  307.         .box('vp09', genericVideoBox)
  308.         .box('av01', genericVideoBox)
  309.         .box('avcC', (box) => {
  310.           let codecBase = baseBox || '';
  311.           switch (baseBox) {
  312.             case 'dvav':
  313.               codecBase = 'avc3';
  314.               break;
  315.             case 'dva1':
  316.               codecBase = 'avc1';
  317.               break;
  318.           }
  319.           const parsedAVCCBox = shaka.util.Mp4BoxParsers.parseAVCC(
  320.               codecBase, box.reader, box.name);
  321.           videoCodecs.push(parsedAVCCBox.codec);
  322.           hasVideo = true;
  323.         })
  324.         .box('hvcC', (box) => {
  325.           let codecBase = baseBox || '';
  326.           switch (baseBox) {
  327.             case 'dvh1':
  328.               codecBase = 'hvc1';
  329.               break;
  330.             case 'dvhe':
  331.               codecBase = 'hev1';
  332.               break;
  333.           }
  334.           const parsedHVCCBox = shaka.util.Mp4BoxParsers.parseHVCC(
  335.               codecBase, box.reader, box.name);
  336.           videoCodecs.push(parsedHVCCBox.codec);
  337.           hasVideo = true;
  338.         })
  339.         .box('dvcC', (box) => {
  340.           let codecBase = baseBox || '';
  341.           switch (baseBox) {
  342.             case 'hvc1':
  343.               codecBase = 'dvh1';
  344.               break;
  345.             case 'hev1':
  346.               codecBase = 'dvhe';
  347.               break;
  348.             case 'avc1':
  349.               codecBase = 'dva1';
  350.               break;
  351.             case 'avc3':
  352.               codecBase = 'dvav';
  353.               break;
  354.             case 'av01':
  355.               codecBase = 'dav1';
  356.               break;
  357.           }
  358.           const parsedDVCCBox = shaka.util.Mp4BoxParsers.parseDVCC(
  359.               codecBase, box.reader, box.name);
  360.           videoCodecs.push(parsedDVCCBox.codec);
  361.           hasVideo = true;
  362.         })
  363.         .box('dvvC', (box) => {
  364.           let codecBase = baseBox || '';
  365.           switch (baseBox) {
  366.             case 'hvc1':
  367.               codecBase = 'dvh1';
  368.               break;
  369.             case 'hev1':
  370.               codecBase = 'dvhe';
  371.               break;
  372.             case 'avc1':
  373.               codecBase = 'dva1';
  374.               break;
  375.             case 'avc3':
  376.               codecBase = 'dvav';
  377.               break;
  378.             case 'av01':
  379.               codecBase = 'dav1';
  380.               break;
  381.           }
  382.           const parsedDVCCBox = shaka.util.Mp4BoxParsers.parseDVVC(
  383.               codecBase, box.reader, box.name);
  384.           videoCodecs.push(parsedDVCCBox.codec);
  385.           hasVideo = true;
  386.         })
  387.         .fullBox('vpcC', (box) => {
  388.           const codecBase = baseBox || '';
  389.           const parsedVPCCBox = shaka.util.Mp4BoxParsers.parseVPCC(
  390.               codecBase, box.reader, box.name);
  391.           videoCodecs.push(parsedVPCCBox.codec);
  392.           hasVideo = true;
  393.         })
  394.         .box('av1C', (box) => {
  395.           let codecBase = baseBox || '';
  396.           switch (baseBox) {
  397.             case 'dav1':
  398.               codecBase = 'av01';
  399.               break;
  400.           }
  401.           const parsedAV1CBox = shaka.util.Mp4BoxParsers.parseAV1C(
  402.               codecBase, box.reader, box.name);
  403.           videoCodecs.push(parsedAV1CBox.codec);
  404.           hasVideo = true;
  405.         })
  407.         // This signals an encrypted sample, which we can go inside of to
  408.         // find the codec used.
  409.         // Note: If encrypted, you can only have audio or video, not both.
  410.         .box('enca', Mp4Parser.audioSampleEntry)
  411.         .box('encv', Mp4Parser.visualSampleEntry)
  412.         .box('sinf', Mp4Parser.children)
  413.         .box('frma', (box) => {
  414.           const {codec} = shaka.util.Mp4BoxParsers.parseFRMA(box.reader);
  415.           addCodec(codec);
  416.         })
  418.         .box('colr', (box) => {
  419.           videoCodecs = videoCodecs.map((codec) => {
  420.             if (codec.startsWith('av01.')) {
  421.               return shaka.util.Mp4BoxParsers.updateAV1CodecWithCOLRBox(
  422.                   codec, box.reader);
  423.             }
  424.             return codec;
  425.           });
  426.           const {videoRange, colorGamut} =
  427.               shaka.util.Mp4BoxParsers.parseCOLR(box.reader);
  428.           realVideoRange = videoRange;
  429.           realColorGamut = colorGamut;
  430.         })
  432.         .parse(initData || data, /* partialOkay= */ true);
  433.     if (!audioCodecs.length && !videoCodecs.length) {
  434.       return null;
  435.     }
  436.     const onlyAudio = hasAudio && !hasVideo;
  437.     const closedCaptions = new Map();
  438.     if (hasVideo && !disableText) {
  439.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp4');
  440.       if (initData) {
  441.         captionParser.init(initData);
  442.       }
  443.       try {
  444.         captionParser.parseFrom(data);
  445.         for (const stream of captionParser.getStreams()) {
  446.           closedCaptions.set(stream, stream);
  447.         }
  448.       } catch (e) {
  449.         shaka.log.debug('Error detecting CC streams', e);
  450.       }
  451.       captionParser.reset();
  452.     }
  453.     const codecs = audioCodecs.concat(videoCodecs);
  454.     return {
  455.       type: onlyAudio ? 'audio' : 'video',
  456.       mimeType: onlyAudio ? 'audio/mp4' : 'video/mp4',
  457.       codecs: SegmentUtils.codecsFiltering(codecs).join(', '),
  458.       language: language,
  459.       height: height,
  460.       width: width,
  461.       channelCount: channelCount,
  462.       sampleRate: sampleRate,
  463.       closedCaptions: closedCaptions,
  464.       videoRange: realVideoRange,
  465.       colorGamut: realColorGamut,
  466.       frameRate: realFrameRate,
  467.     };
  468.   }
  470.   /**
  471.    * @param {!Array<string>} codecs
  472.    * @return {!Array<string>} codecs
  473.    */
  474.   static codecsFiltering(codecs) {
  475.     const ContentType = shaka.util.ManifestParserUtils.ContentType;
  476.     const ManifestParserUtils = shaka.util.ManifestParserUtils;
  477.     const SegmentUtils = shaka.media.SegmentUtils;
  478.     const allCodecs = SegmentUtils.filterDuplicateCodecs_(codecs);
  479.     const audioCodecs =
  480.         ManifestParserUtils.guessAllCodecsSafe(ContentType.AUDIO, allCodecs);
  481.     const videoCodecs =
  482.         ManifestParserUtils.guessAllCodecsSafe(ContentType.VIDEO, allCodecs);
  483.     const textCodecs =
  484.         ManifestParserUtils.guessAllCodecsSafe(ContentType.TEXT, allCodecs);
  485.     const validVideoCodecs = SegmentUtils.chooseBetterCodecs_(videoCodecs);
  486.     const finalCodecs =
  487.         audioCodecs.concat(validVideoCodecs).concat(textCodecs);
  488.     if (allCodecs.length && !finalCodecs.length) {
  489.       return allCodecs;
  490.     }
  491.     return finalCodecs;
  492.   }
  494.   /**
  495.    * @param {!Array<string>} codecs
  496.    * @return {!Array<string>} codecs
  497.    * @private
  498.    */
  499.   static filterDuplicateCodecs_(codecs) {
  500.     // Filter out duplicate codecs.
  501.     const seen = new Set();
  502.     const ret = [];
  503.     for (const codec of codecs) {
  504.       const shortCodec = shaka.util.MimeUtils.getCodecBase(codec);
  505.       if (!seen.has(shortCodec)) {
  506.         ret.push(codec);
  507.         seen.add(shortCodec);
  508.       } else {
  509.         shaka.log.debug('Ignoring duplicate codec');
  510.       }
  511.     }
  512.     return ret;
  513.   }
  515.   /**
  516.    * Prioritizes Dolby Vision if supported. This is necessary because with
  517.    * Dolby Vision we could have hvcC and dvcC boxes at the same time.
  518.    *
  519.    * @param {!Array<string>} codecs
  520.    * @return {!Array<string>} codecs
  521.    * @private
  522.    */
  523.   static chooseBetterCodecs_(codecs) {
  524.     if (codecs.length <= 1) {
  525.       return codecs;
  526.     }
  527.     const dolbyVision = codecs.find((codec) => {
  528.       return codec.startsWith('dvav.') ||
  529.           codec.startsWith('dva1.') ||
  530.           codec.startsWith('dvh1.') ||
  531.           codec.startsWith('dvhe.') ||
  532.           codec.startsWith('dav1.') ||
  533.           codec.startsWith('dvc1.') ||
  534.           codec.startsWith('dvi1.');
  535.     });
  536.     if (!dolbyVision) {
  537.       return codecs;
  538.     }
  539.     const type = `video/mp4; codecs="${dolbyVision}"`;
  540.     if (shaka.media.Capabilities.isTypeSupported(type)) {
  541.       return [dolbyVision];
  542.     }
  543.     return codecs.filter((codec) => codec != dolbyVision);
  544.   }
  546.   /**
  547.    * @param {!BufferSource} data
  548.    * @return {?string}
  549.    */
  550.   static getDefaultKID(data) {
  551.     const Mp4Parser = shaka.util.Mp4Parser;
  553.     let defaultKID = null;
  554.     new Mp4Parser()
  555.         .box('moov', Mp4Parser.children)
  556.         .box('trak', Mp4Parser.children)
  557.         .box('mdia', Mp4Parser.children)
  558.         .box('minf', Mp4Parser.children)
  559.         .box('stbl', Mp4Parser.children)
  560.         .fullBox('stsd', Mp4Parser.sampleDescription)
  561.         .box('encv', Mp4Parser.visualSampleEntry)
  562.         .box('enca', Mp4Parser.audioSampleEntry)
  563.         .box('sinf', Mp4Parser.children)
  564.         .box('schi', Mp4Parser.children)
  565.         .fullBox('tenc', (box) => {
  566.           const parsedTENCBox = shaka.util.Mp4BoxParsers.parseTENC(box.reader);
  567.           defaultKID = parsedTENCBox.defaultKID;
  568.         })
  570.         .parse(data, /* partialOkay= */ true);
  571.     return defaultKID;
  572.   }
  574.   /**
  575.    * @param {!BufferSource} rawResult
  576.    * @param {shaka.extern.aesKey} aesKey
  577.    * @param {number} position
  578.    * @return {!Promise<!BufferSource>}
  579.    */
  580.   static async aesDecrypt(rawResult, aesKey, position) {
  581.     const key = aesKey;
  582.     if (!key.cryptoKey) {
  583.       goog.asserts.assert(key.fetchKey, 'If AES cryptoKey was not ' +
  584.           'preloaded, fetchKey function should be provided');
  585.       await key.fetchKey();
  586.       goog.asserts.assert(key.cryptoKey, 'AES cryptoKey should now be set');
  587.     }
  588.     let iv = key.iv;
  589.     if (!iv) {
  590.       iv = shaka.util.BufferUtils.toUint8(new ArrayBuffer(16));
  591.       let sequence = key.firstMediaSequenceNumber + position;
  592.       for (let i = iv.byteLength - 1; i >= 0; i--) {
  593.         iv[i] = sequence & 0xff;
  594.         sequence >>= 8;
  595.       }
  596.     }
  597.     let algorithm;
  598.     if (aesKey.blockCipherMode == 'CBC') {
  599.       algorithm = {
  600.         name: 'AES-CBC',
  601.         iv,
  602.       };
  603.     } else {
  604.       algorithm = {
  605.         name: 'AES-CTR',
  606.         counter: iv,
  607.         // NIST SP800-38A standard suggests that the counter should occupy half
  608.         // of the counter block
  609.         length: 64,
  610.       };
  611.     }
  612.     return window.crypto.subtle.decrypt(algorithm, key.cryptoKey, rawResult);
  613.   }
  614. };
  617. /**
  618.  * @typedef {{
  619.  *   type: string,
  620.  *   mimeType: string,
  621.  *   codecs: string,
  622.  *   language: ?string,
  623.  *   height: ?string,
  624.  *   width: ?string,
  625.  *   channelCount: ?number,
  626.  *   sampleRate: ?number,
  627.  *   closedCaptions: Map<string, string>,
  628.  *   videoRange: ?string,
  629.  *   colorGamut: ?string,
  630.  *   frameRate: ?string
  631.  * }}
  632.  *
  633.  * @property {string} type
  634.  * @property {string} mimeType
  635.  * @property {string} codecs
  636.  * @property {?string} language
  637.  * @property {?string} height
  638.  * @property {?string} width
  639.  * @property {?number} channelCount
  640.  * @property {?number} sampleRate
  641.  * @property {Map<string, string>} closedCaptions
  642.  * @property {?string} videoRange
  643.  * @property {?string} colorGamut
  644.  * @property {?string} frameRate
  645.  */
  646. shaka.media.SegmentUtils.BasicInfo;