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.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  52.    */
  53.   static getBasicInfoFromTs(data) {
  54.     const uint8ArrayData = shaka.util.BufferUtils.toUint8(data);
  55.     const tsParser = new shaka.util.TsParser().parse(uint8ArrayData);
  56.     const tsCodecs = tsParser.getCodecs();
  57.     const videoInfo = tsParser.getVideoInfo();
  58.     const codecs = [];
  59.     let hasAudio = false;
  60.     let hasVideo = false;
  61.     switch (tsCodecs.audio) {
  62.       case 'aac':
  63.       case 'aac-loas':
  64.         codecs.push('mp4a.40.2');
  65.         hasAudio = true;
  66.         break;
  67.       case 'mp3':
  68.         codecs.push('mp4a.40.34');
  69.         hasAudio = true;
  70.         break;
  71.       case 'ac3':
  72.         codecs.push('ac-3');
  73.         hasAudio = true;
  74.         break;
  75.       case 'ec3':
  76.         codecs.push('ec-3');
  77.         hasAudio = true;
  78.         break;
  79.       case 'opus':
  80.         codecs.push('opus');
  81.         hasAudio = true;
  82.         break;
  83.     }
  84.     switch (tsCodecs.video) {
  85.       case 'avc':
  86.         if (videoInfo.codec) {
  87.           codecs.push(videoInfo.codec);
  88.         } else {
  89.           codecs.push('avc1.42E01E');
  90.         }
  91.         hasVideo = true;
  92.         break;
  93.       case 'hvc':
  94.         if (videoInfo.codec) {
  95.           codecs.push(videoInfo.codec);
  96.         } else {
  97.           codecs.push('hvc1.1.6.L93.90');
  98.         }
  99.         hasVideo = true;
  100.         break;
  101.       case 'av1':
  102.         codecs.push('av01.0.01M.08');
  103.         hasVideo = true;
  104.         break;
  105.     }
  106.     if (!codecs.length) {
  107.       return null;
  108.     }
  109.     const onlyAudio = hasAudio && !hasVideo;
  110.     const closedCaptions = new Map();
  111.     if (hasVideo) {
  112.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp2t');
  113.       captionParser.parseFrom(data);
  114.       for (const stream of captionParser.getStreams()) {
  115.         closedCaptions.set(stream, stream);
  116.       }
  117.       captionParser.reset();
  118.     }
  119.     return {
  120.       type: onlyAudio ? 'audio' : 'video',
  121.       mimeType: 'video/mp2t',
  122.       codecs: codecs.join(', '),
  123.       language: null,
  124.       height: videoInfo.height,
  125.       width: videoInfo.width,
  126.       channelCount: null,
  127.       sampleRate: null,
  128.       closedCaptions: closedCaptions,
  129.       videoRange: null,
  130.       colorGamut: null,
  131.       frameRate: videoInfo.frameRate,
  132.     };
  133.   }
  135.   /**
  136.    * @param {?BufferSource} initData
  137.    * @param {!BufferSource} data
  138.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  139.    */
  140.   static getBasicInfoFromMp4(initData, data) {
  141.     const Mp4Parser = shaka.util.Mp4Parser;
  142.     const SegmentUtils = shaka.media.SegmentUtils;
  144.     const audioCodecs = [];
  145.     let videoCodecs = [];
  147.     let hasAudio = false;
  148.     let hasVideo = false;
  150.     const addCodec = (codec) => {
  151.       const codecLC = codec.toLowerCase();
  152.       switch (codecLC) {
  153.         case 'avc1':
  154.         case 'avc3':
  155.           videoCodecs.push(codecLC + '.42E01E');
  156.           hasVideo = true;
  157.           break;
  158.         case 'hev1':
  159.         case 'hvc1':
  160.           videoCodecs.push(codecLC + '.1.6.L93.90');
  161.           hasVideo = true;
  162.           break;
  163.         case 'dvh1':
  164.         case 'dvhe':
  165.           videoCodecs.push(codecLC + '.05.04');
  166.           hasVideo = true;
  167.           break;
  168.         case 'vp09':
  169.           videoCodecs.push(codecLC + '.00.10.08');
  170.           hasVideo = true;
  171.           break;
  172.         case 'av01':
  173.           videoCodecs.push(codecLC + '.0.01M.08');
  174.           hasVideo = true;
  175.           break;
  176.         case 'mp4a':
  177.           // We assume AAC, but this can be wrong since mp4a supports
  178.           // others codecs
  179.           audioCodecs.push('mp4a.40.2');
  180.           hasAudio = true;
  181.           break;
  182.         case 'ac-3':
  183.         case 'ec-3':
  184.         case 'ac-4':
  185.         case 'opus':
  186.         case 'flac':
  187.           audioCodecs.push(codecLC);
  188.           hasAudio = true;
  189.           break;
  190.       }
  191.     };
  193.     const codecBoxParser = (box) => addCodec(box.name);
  195.     /** @type {?string} */
  196.     let language = null;
  197.     /** @type {?string} */
  198.     let height = null;
  199.     /** @type {?string} */
  200.     let width = null;
  201.     /** @type {?number} */
  202.     let channelCount = null;
  203.     /** @type {?number} */
  204.     let sampleRate = null;
  205.     /** @type {?string} */
  206.     let realVideoRange = null;
  207.     /** @type {?string} */
  208.     let realColorGamut = null;
  209.     /** @type {?string} */
  210.     const realFrameRate = null;
  212.     /** @type {?string} */
  213.     let baseBox;
  215.     const genericAudioBox = (box) => {
  216.       const parsedAudioSampleEntryBox =
  217.           shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  218.       channelCount = parsedAudioSampleEntryBox.channelCount;
  219.       sampleRate = parsedAudioSampleEntryBox.sampleRate;
  220.       codecBoxParser(box);
  221.     };
  223.     const genericVideoBox = (box) => {
  224.       baseBox = box.name;
  225.       const parsedVisualSampleEntryBox =
  226.           shaka.util.Mp4BoxParsers.visualSampleEntry(box.reader);
  227.       width = String(parsedVisualSampleEntryBox.width);
  228.       height = String(parsedVisualSampleEntryBox.height);
  229.       if (box.reader.hasMoreData()) {
  230.         Mp4Parser.children(box);
  231.       }
  232.     };
  234.     new Mp4Parser()
  235.         .box('moov', Mp4Parser.children)
  236.         .box('trak', Mp4Parser.children)
  237.         .box('mdia', Mp4Parser.children)
  238.         .fullBox('mdhd', (box) => {
  239.           goog.asserts.assert(
  240.               box.version != null,
  241.               'MDHD is a full box and should have a valid version.');
  242.           const parsedMDHDBox = shaka.util.Mp4BoxParsers.parseMDHD(
  243.               box.reader, box.version);
  244.           language = parsedMDHDBox.language;
  245.         })
  246.         .box('minf', Mp4Parser.children)
  247.         .box('stbl', Mp4Parser.children)
  248.         .fullBox('stsd', Mp4Parser.sampleDescription)
  250.         // AUDIO
  251.         // These are the various boxes that signal a codec.
  252.         .box('mp4a', (box) => {
  253.           const parsedAudioSampleEntryBox =
  254.               shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  255.           channelCount = parsedAudioSampleEntryBox.channelCount;
  256.           sampleRate = parsedAudioSampleEntryBox.sampleRate;
  257.           if (box.reader.hasMoreData()) {
  258.             Mp4Parser.children(box);
  259.           } else {
  260.             codecBoxParser(box);
  261.           }
  262.         })
  263.         .box('esds', (box) => {
  264.           const parsedESDSBox = shaka.util.Mp4BoxParsers.parseESDS(box.reader);
  265.           audioCodecs.push(parsedESDSBox.codec);
  266.           hasAudio = true;
  267.         })
  268.         .box('ac-3', genericAudioBox)
  269.         .box('ec-3', genericAudioBox)
  270.         .box('ac-4', genericAudioBox)
  271.         .box('Opus', genericAudioBox)
  272.         .box('fLaC', genericAudioBox)
  274.         // VIDEO
  275.         // These are the various boxes that signal a codec.
  276.         .box('avc1', genericVideoBox)
  277.         .box('avc3', genericVideoBox)
  278.         .box('hev1', genericVideoBox)
  279.         .box('hvc1', genericVideoBox)
  280.         .box('dva1', genericVideoBox)
  281.         .box('dvav', genericVideoBox)
  282.         .box('dvh1', genericVideoBox)
  283.         .box('dvhe', genericVideoBox)
  284.         .box('vp09', genericVideoBox)
  285.         .box('av01', genericVideoBox)
  286.         .box('avcC', (box) => {
  287.           let codecBase = baseBox || '';
  288.           switch (baseBox) {
  289.             case 'dvav':
  290.               codecBase = 'avc3';
  291.               break;
  292.             case 'dva1':
  293.               codecBase = 'avc1';
  294.               break;
  295.           }
  296.           const parsedAVCCBox = shaka.util.Mp4BoxParsers.parseAVCC(
  297.               codecBase, box.reader, box.name);
  298.           videoCodecs.push(parsedAVCCBox.codec);
  299.           hasVideo = true;
  300.         })
  301.         .box('hvcC', (box) => {
  302.           let codecBase = baseBox || '';
  303.           switch (baseBox) {
  304.             case 'dvh1':
  305.               codecBase = 'hvc1';
  306.               break;
  307.             case 'dvhe':
  308.               codecBase = 'hev1';
  309.               break;
  310.           }
  311.           const parsedHVCCBox = shaka.util.Mp4BoxParsers.parseHVCC(
  312.               codecBase, box.reader, box.name);
  313.           videoCodecs.push(parsedHVCCBox.codec);
  314.           hasVideo = true;
  315.         })
  316.         .box('dvcC', (box) => {
  317.           let codecBase = baseBox || '';
  318.           switch (baseBox) {
  319.             case 'hvc1':
  320.               codecBase = 'dvh1';
  321.               break;
  322.             case 'hev1':
  323.               codecBase = 'dvhe';
  324.               break;
  325.             case 'avc1':
  326.               codecBase = 'dva1';
  327.               break;
  328.             case 'avc3':
  329.               codecBase = 'dvav';
  330.               break;
  331.             case 'av01':
  332.               codecBase = 'dav1';
  333.               break;
  334.           }
  335.           const parsedDVCCBox = shaka.util.Mp4BoxParsers.parseDVCC(
  336.               codecBase, box.reader, box.name);
  337.           videoCodecs.push(parsedDVCCBox.codec);
  338.           hasVideo = true;
  339.         })
  340.         .box('dvvC', (box) => {
  341.           let codecBase = baseBox || '';
  342.           switch (baseBox) {
  343.             case 'hvc1':
  344.               codecBase = 'dvh1';
  345.               break;
  346.             case 'hev1':
  347.               codecBase = 'dvhe';
  348.               break;
  349.             case 'avc1':
  350.               codecBase = 'dva1';
  351.               break;
  352.             case 'avc3':
  353.               codecBase = 'dvav';
  354.               break;
  355.             case 'av01':
  356.               codecBase = 'dav1';
  357.               break;
  358.           }
  359.           const parsedDVCCBox = shaka.util.Mp4BoxParsers.parseDVVC(
  360.               codecBase, box.reader, box.name);
  361.           videoCodecs.push(parsedDVCCBox.codec);
  362.           hasVideo = true;
  363.         })
  364.         .fullBox('vpcC', (box) => {
  365.           const codecBase = baseBox || '';
  366.           const parsedVPCCBox = shaka.util.Mp4BoxParsers.parseVPCC(
  367.               codecBase, box.reader, box.name);
  368.           videoCodecs.push(parsedVPCCBox.codec);
  369.           hasVideo = true;
  370.         })
  371.         .box('av1C', (box) => {
  372.           let codecBase = baseBox || '';
  373.           switch (baseBox) {
  374.             case 'dav1':
  375.               codecBase = 'av01';
  376.               break;
  377.           }
  378.           const parsedAV1CBox = shaka.util.Mp4BoxParsers.parseAV1C(
  379.               codecBase, box.reader, box.name);
  380.           videoCodecs.push(parsedAV1CBox.codec);
  381.           hasVideo = true;
  382.         })
  384.         // This signals an encrypted sample, which we can go inside of to
  385.         // find the codec used.
  386.         // Note: If encrypted, you can only have audio or video, not both.
  387.         .box('enca', Mp4Parser.audioSampleEntry)
  388.         .box('encv', Mp4Parser.visualSampleEntry)
  389.         .box('sinf', Mp4Parser.children)
  390.         .box('frma', (box) => {
  391.           const {codec} = shaka.util.Mp4BoxParsers.parseFRMA(box.reader);
  392.           addCodec(codec);
  393.         })
  395.         .box('colr', (box) => {
  396.           videoCodecs = videoCodecs.map((codec) => {
  397.             if (codec.startsWith('av01.')) {
  398.               return shaka.util.Mp4BoxParsers.updateAV1CodecWithCOLRBox(
  399.                   codec, box.reader);
  400.             }
  401.             return codec;
  402.           });
  403.           const {videoRange, colorGamut} =
  404.               shaka.util.Mp4BoxParsers.parseCOLR(box.reader);
  405.           realVideoRange = videoRange;
  406.           realColorGamut = colorGamut;
  407.         })
  409.         .parse(initData || data, /* partialOkay= */ true);
  410.     if (!audioCodecs.length && !videoCodecs.length) {
  411.       return null;
  412.     }
  413.     const onlyAudio = hasAudio && !hasVideo;
  414.     const closedCaptions = new Map();
  415.     if (hasVideo) {
  416.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp4');
  417.       if (initData) {
  418.         captionParser.init(initData);
  419.       }
  420.       captionParser.parseFrom(data);
  421.       for (const stream of captionParser.getStreams()) {
  422.         closedCaptions.set(stream, stream);
  423.       }
  424.       captionParser.reset();
  425.     }
  426.     const codecs = audioCodecs.concat(videoCodecs);
  427.     return {
  428.       type: onlyAudio ? 'audio' : 'video',
  429.       mimeType: onlyAudio ? 'audio/mp4' : 'video/mp4',
  430.       codecs: SegmentUtils.codecsFiltering(codecs).join(', '),
  431.       language: language,
  432.       height: height,
  433.       width: width,
  434.       channelCount: channelCount,
  435.       sampleRate: sampleRate,
  436.       closedCaptions: closedCaptions,
  437.       videoRange: realVideoRange,
  438.       colorGamut: realColorGamut,
  439.       frameRate: realFrameRate,
  440.     };
  441.   }
  443.   /**
  444.    * @param {!Array.<string>} codecs
  445.    * @return {!Array.<string>} codecs
  446.    */
  447.   static codecsFiltering(codecs) {
  448.     const ContentType = shaka.util.ManifestParserUtils.ContentType;
  449.     const ManifestParserUtils = shaka.util.ManifestParserUtils;
  450.     const SegmentUtils = shaka.media.SegmentUtils;
  451.     const allCodecs = SegmentUtils.filterDuplicateCodecs_(codecs);
  452.     const audioCodecs =
  453.         ManifestParserUtils.guessAllCodecsSafe(ContentType.AUDIO, allCodecs);
  454.     const videoCodecs =
  455.         ManifestParserUtils.guessAllCodecsSafe(ContentType.VIDEO, allCodecs);
  456.     const textCodecs =
  457.         ManifestParserUtils.guessAllCodecsSafe(ContentType.TEXT, allCodecs);
  458.     const validVideoCodecs = SegmentUtils.chooseBetterCodecs_(videoCodecs);
  459.     const finalCodecs =
  460.         audioCodecs.concat(validVideoCodecs).concat(textCodecs);
  461.     if (allCodecs.length && !finalCodecs.length) {
  462.       return allCodecs;
  463.     }
  464.     return finalCodecs;
  465.   }
  467.   /**
  468.    * @param {!Array.<string>} codecs
  469.    * @return {!Array.<string>} codecs
  470.    * @private
  471.    */
  472.   static filterDuplicateCodecs_(codecs) {
  473.     // Filter out duplicate codecs.
  474.     const seen = new Set();
  475.     const ret = [];
  476.     for (const codec of codecs) {
  477.       const shortCodec = shaka.util.MimeUtils.getCodecBase(codec);
  478.       if (!seen.has(shortCodec)) {
  479.         ret.push(codec);
  480.         seen.add(shortCodec);
  481.       } else {
  482.         shaka.log.debug('Ignoring duplicate codec');
  483.       }
  484.     }
  485.     return ret;
  486.   }
  488.   /**
  489.    * Prioritizes Dolby Vision if supported. This is necessary because with
  490.    * Dolby Vision we could have hvcC and dvcC boxes at the same time.
  491.    *
  492.    * @param {!Array.<string>} codecs
  493.    * @return {!Array.<string>} codecs
  494.    * @private
  495.    */
  496.   static chooseBetterCodecs_(codecs) {
  497.     if (codecs.length <= 1) {
  498.       return codecs;
  499.     }
  500.     const dolbyVision = codecs.find((codec) => {
  501.       return codec.startsWith('dvav.') ||
  502.           codec.startsWith('dva1.') ||
  503.           codec.startsWith('dvh1.') ||
  504.           codec.startsWith('dvhe.') ||
  505.           codec.startsWith('dav1.') ||
  506.           codec.startsWith('dvc1.') ||
  507.           codec.startsWith('dvi1.');
  508.     });
  509.     if (!dolbyVision) {
  510.       return codecs;
  511.     }
  512.     const type = `video/mp4; codecs="${dolbyVision}"`;
  513.     if (shaka.media.Capabilities.isTypeSupported(type)) {
  514.       return [dolbyVision];
  515.     }
  516.     return codecs.filter((codec) => codec != dolbyVision);
  517.   }
  519.   /**
  520.    * @param {!BufferSource} data
  521.    * @return {?string}
  522.    */
  523.   static getDefaultKID(data) {
  524.     const Mp4Parser = shaka.util.Mp4Parser;
  526.     let defaultKID = null;
  527.     new Mp4Parser()
  528.         .box('moov', Mp4Parser.children)
  529.         .box('trak', Mp4Parser.children)
  530.         .box('mdia', Mp4Parser.children)
  531.         .box('minf', Mp4Parser.children)
  532.         .box('stbl', Mp4Parser.children)
  533.         .fullBox('stsd', Mp4Parser.sampleDescription)
  534.         .box('encv', Mp4Parser.visualSampleEntry)
  535.         .box('enca', Mp4Parser.audioSampleEntry)
  536.         .box('sinf', Mp4Parser.children)
  537.         .box('schi', Mp4Parser.children)
  538.         .fullBox('tenc', (box) => {
  539.           const parsedTENCBox = shaka.util.Mp4BoxParsers.parseTENC(box.reader);
  540.           defaultKID = parsedTENCBox.defaultKID;
  541.         })
  543.         .parse(data, /* partialOkay= */ true);
  544.     return defaultKID;
  545.   }
  547.   /**
  548.    * @param {!BufferSource} rawResult
  549.    * @param {shaka.extern.aesKey} aesKey
  550.    * @param {number} position
  551.    * @return {!Promise.<!BufferSource>}
  552.    */
  553.   static async aesDecrypt(rawResult, aesKey, position) {
  554.     const key = aesKey;
  555.     if (!key.cryptoKey) {
  556.       goog.asserts.assert(key.fetchKey, 'If AES cryptoKey was not ' +
  557.           'preloaded, fetchKey function should be provided');
  558.       await key.fetchKey();
  559.       goog.asserts.assert(key.cryptoKey, 'AES cryptoKey should now be set');
  560.     }
  561.     let iv = key.iv;
  562.     if (!iv) {
  563.       iv = shaka.util.BufferUtils.toUint8(new ArrayBuffer(16));
  564.       let sequence = key.firstMediaSequenceNumber + position;
  565.       for (let i = iv.byteLength - 1; i >= 0; i--) {
  566.         iv[i] = sequence & 0xff;
  567.         sequence >>= 8;
  568.       }
  569.     }
  570.     let algorithm;
  571.     if (aesKey.blockCipherMode == 'CBC') {
  572.       algorithm = {
  573.         name: 'AES-CBC',
  574.         iv,
  575.       };
  576.     } else {
  577.       algorithm = {
  578.         name: 'AES-CTR',
  579.         counter: iv,
  580.         // NIST SP800-38A standard suggests that the counter should occupy half
  581.         // of the counter block
  582.         length: 64,
  583.       };
  584.     }
  585.     return window.crypto.subtle.decrypt(algorithm, key.cryptoKey, rawResult);
  586.   }
  587. };
  590. /**
  591.  * @typedef {{
  592.  *   type: string,
  593.  *   mimeType: string,
  594.  *   codecs: string,
  595.  *   language: ?string,
  596.  *   height: ?string,
  597.  *   width: ?string,
  598.  *   channelCount: ?number,
  599.  *   sampleRate: ?number,
  600.  *   closedCaptions: Map.<string, string>,
  601.  *   videoRange: ?string,
  602.  *   colorGamut: ?string,
  603.  *   frameRate: ?string
  604.  * }}
  605.  *
  606.  * @property {string} type
  607.  * @property {string} mimeType
  608.  * @property {string} codecs
  609.  * @property {?string} language
  610.  * @property {?string} height
  611.  * @property {?string} width
  612.  * @property {?number} channelCount
  613.  * @property {?number} sampleRate
  614.  * @property {Map.<string, string>} closedCaptions
  615.  * @property {?string} videoRange
  616.  * @property {?string} colorGamut
  617.  * @property {?string} frameRate
  618.  */
  619. shaka.media.SegmentUtils.BasicInfo;