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.     };
  46.   }
  48.   /**
  49.    * @param {!BufferSource} data
  50.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  51.    */
  52.   static getBasicInfoFromTs(data) {
  53.     const uint8ArrayData = shaka.util.BufferUtils.toUint8(data);
  54.     const tsParser = new shaka.util.TsParser().parse(uint8ArrayData);
  55.     const tsCodecs = tsParser.getCodecs();
  56.     const videoInfo = tsParser.getVideoInfo();
  57.     const codecs = [];
  58.     let hasAudio = false;
  59.     let hasVideo = false;
  60.     switch (tsCodecs.audio) {
  61.       case 'aac':
  62.       case 'aac-loas':
  63.         codecs.push('mp4a.40.2');
  64.         hasAudio = true;
  65.         break;
  66.       case 'mp3':
  67.         codecs.push('mp4a.40.34');
  68.         hasAudio = true;
  69.         break;
  70.       case 'ac3':
  71.         codecs.push('ac-3');
  72.         hasAudio = true;
  73.         break;
  74.       case 'ec3':
  75.         codecs.push('ec-3');
  76.         hasAudio = true;
  77.         break;
  78.       case 'opus':
  79.         codecs.push('opus');
  80.         hasAudio = true;
  81.         break;
  82.     }
  83.     switch (tsCodecs.video) {
  84.       case 'avc':
  85.         if (videoInfo.codec) {
  86.           codecs.push(videoInfo.codec);
  87.         } else {
  88.           codecs.push('avc1.42E01E');
  89.         }
  90.         hasVideo = true;
  91.         break;
  92.       case 'hvc':
  93.         if (videoInfo.codec) {
  94.           codecs.push(videoInfo.codec);
  95.         } else {
  96.           codecs.push('hvc1.1.6.L93.90');
  97.         }
  98.         hasVideo = true;
  99.         break;
  100.       case 'av1':
  101.         codecs.push('av01.0.01M.08');
  102.         hasVideo = true;
  103.         break;
  104.     }
  105.     if (!codecs.length) {
  106.       return null;
  107.     }
  108.     const onlyAudio = hasAudio && !hasVideo;
  109.     const closedCaptions = new Map();
  110.     if (hasVideo) {
  111.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp2t');
  112.       captionParser.parseFrom(data);
  113.       for (const stream of captionParser.getStreams()) {
  114.         closedCaptions.set(stream, stream);
  115.       }
  116.       captionParser.reset();
  117.     }
  118.     return {
  119.       type: onlyAudio ? 'audio' : 'video',
  120.       mimeType: 'video/mp2t',
  121.       codecs: codecs.join(', '),
  122.       language: null,
  123.       height: videoInfo.height,
  124.       width: videoInfo.width,
  125.       channelCount: null,
  126.       sampleRate: null,
  127.       closedCaptions: closedCaptions,
  128.       videoRange: null,
  129.       colorGamut: null,
  130.     };
  131.   }
  133.   /**
  134.    * @param {?BufferSource} initData
  135.    * @param {!BufferSource} data
  136.    * @return {?shaka.media.SegmentUtils.BasicInfo}
  137.    */
  138.   static getBasicInfoFromMp4(initData, data) {
  139.     const Mp4Parser = shaka.util.Mp4Parser;
  140.     const SegmentUtils = shaka.media.SegmentUtils;
  142.     const audioCodecs = [];
  143.     let videoCodecs = [];
  145.     let hasAudio = false;
  146.     let hasVideo = false;
  148.     const addCodec = (codec) => {
  149.       const codecLC = codec.toLowerCase();
  150.       switch (codecLC) {
  151.         case 'avc1':
  152.         case 'avc3':
  153.           videoCodecs.push(codecLC + '.42E01E');
  154.           hasVideo = true;
  155.           break;
  156.         case 'hev1':
  157.         case 'hvc1':
  158.           videoCodecs.push(codecLC + '.1.6.L93.90');
  159.           hasVideo = true;
  160.           break;
  161.         case 'dvh1':
  162.         case 'dvhe':
  163.           videoCodecs.push(codecLC + '.05.04');
  164.           hasVideo = true;
  165.           break;
  166.         case 'vp09':
  167.           videoCodecs.push(codecLC + '.00.10.08');
  168.           hasVideo = true;
  169.           break;
  170.         case 'av01':
  171.           videoCodecs.push(codecLC + '.0.01M.08');
  172.           hasVideo = true;
  173.           break;
  174.         case 'mp4a':
  175.           // We assume AAC, but this can be wrong since mp4a supports
  176.           // others codecs
  177.           audioCodecs.push('mp4a.40.2');
  178.           hasAudio = true;
  179.           break;
  180.         case 'ac-3':
  181.         case 'ec-3':
  182.         case 'ac-4':
  183.         case 'opus':
  184.         case 'flac':
  185.           audioCodecs.push(codecLC);
  186.           hasAudio = true;
  187.           break;
  188.       }
  189.     };
  191.     const codecBoxParser = (box) => addCodec(box.name);
  193.     /** @type {?string} */
  194.     let language = null;
  195.     /** @type {?string} */
  196.     let height = null;
  197.     /** @type {?string} */
  198.     let width = null;
  199.     /** @type {?number} */
  200.     let channelCount = null;
  201.     /** @type {?number} */
  202.     let sampleRate = null;
  203.     /** @type {?string} */
  204.     let realVideoRange = null;
  205.     /** @type {?string} */
  206.     let realColorGamut = null;
  208.     /** @type {?string} */
  209.     let baseBox;
  211.     const genericAudioBox = (box) => {
  212.       const parsedAudioSampleEntryBox =
  213.           shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  214.       channelCount = parsedAudioSampleEntryBox.channelCount;
  215.       sampleRate = parsedAudioSampleEntryBox.sampleRate;
  216.       codecBoxParser(box);
  217.     };
  219.     const genericVideoBox = (box) => {
  220.       baseBox = box.name;
  221.       const parsedVisualSampleEntryBox =
  222.           shaka.util.Mp4BoxParsers.visualSampleEntry(box.reader);
  223.       width = String(parsedVisualSampleEntryBox.width);
  224.       height = String(parsedVisualSampleEntryBox.height);
  225.       if (box.reader.hasMoreData()) {
  226.         Mp4Parser.children(box);
  227.       }
  228.     };
  230.     new Mp4Parser()
  231.         .box('moov', Mp4Parser.children)
  232.         .box('trak', Mp4Parser.children)
  233.         .box('mdia', Mp4Parser.children)
  234.         .fullBox('mdhd', (box) => {
  235.           goog.asserts.assert(
  236.               box.version != null,
  237.               'MDHD is a full box and should have a valid version.');
  238.           const parsedMDHDBox = shaka.util.Mp4BoxParsers.parseMDHD(
  239.               box.reader, box.version);
  240.           language = parsedMDHDBox.language;
  241.         })
  242.         .box('minf', Mp4Parser.children)
  243.         .box('stbl', Mp4Parser.children)
  244.         .fullBox('stsd', Mp4Parser.sampleDescription)
  246.         // AUDIO
  247.         // These are the various boxes that signal a codec.
  248.         .box('mp4a', (box) => {
  249.           const parsedAudioSampleEntryBox =
  250.               shaka.util.Mp4BoxParsers.audioSampleEntry(box.reader);
  251.           channelCount = parsedAudioSampleEntryBox.channelCount;
  252.           sampleRate = parsedAudioSampleEntryBox.sampleRate;
  253.           if (box.reader.hasMoreData()) {
  254.             Mp4Parser.children(box);
  255.           } else {
  256.             codecBoxParser(box);
  257.           }
  258.         })
  259.         .box('esds', (box) => {
  260.           const parsedESDSBox = shaka.util.Mp4BoxParsers.parseESDS(box.reader);
  261.           audioCodecs.push(parsedESDSBox.codec);
  262.           hasAudio = true;
  263.         })
  264.         .box('ac-3', genericAudioBox)
  265.         .box('ec-3', genericAudioBox)
  266.         .box('ac-4', genericAudioBox)
  267.         .box('Opus', genericAudioBox)
  268.         .box('fLaC', genericAudioBox)
  270.         // VIDEO
  271.         // These are the various boxes that signal a codec.
  272.         .box('avc1', genericVideoBox)
  273.         .box('avc3', genericVideoBox)
  274.         .box('hev1', genericVideoBox)
  275.         .box('hvc1', genericVideoBox)
  276.         .box('dva1', genericVideoBox)
  277.         .box('dvav', genericVideoBox)
  278.         .box('dvh1', genericVideoBox)
  279.         .box('dvhe', genericVideoBox)
  280.         .box('vp09', genericVideoBox)
  281.         .box('av01', genericVideoBox)
  282.         .box('avcC', (box) => {
  283.           let codecBase = baseBox || '';
  284.           switch (baseBox) {
  285.             case 'dvav':
  286.               codecBase = 'avc3';
  287.               break;
  288.             case 'dva1':
  289.               codecBase = 'avc1';
  290.               break;
  291.           }
  292.           const parsedAVCCBox = shaka.util.Mp4BoxParsers.parseAVCC(
  293.               codecBase, box.reader, box.name);
  294.           videoCodecs.push(parsedAVCCBox.codec);
  295.           hasVideo = true;
  296.         })
  297.         .box('hvcC', (box) => {
  298.           let codecBase = baseBox || '';
  299.           switch (baseBox) {
  300.             case 'dvh1':
  301.               codecBase = 'hvc1';
  302.               break;
  303.             case 'dvhe':
  304.               codecBase = 'hev1';
  305.               break;
  306.           }
  307.           const parsedHVCCBox = shaka.util.Mp4BoxParsers.parseHVCC(
  308.               codecBase, box.reader, box.name);
  309.           videoCodecs.push(parsedHVCCBox.codec);
  310.           hasVideo = true;
  311.         })
  312.         .box('dvcC', (box) => {
  313.           let codecBase = baseBox || '';
  314.           switch (baseBox) {
  315.             case 'hvc1':
  316.               codecBase = 'dvh1';
  317.               break;
  318.             case 'hev1':
  319.               codecBase = 'dvhe';
  320.               break;
  321.             case 'avc1':
  322.               codecBase = 'dva1';
  323.               break;
  324.             case 'avc3':
  325.               codecBase = 'dvav';
  326.               break;
  327.           }
  328.           const parsedDVCCBox = shaka.util.Mp4BoxParsers.parseDVCC(
  329.               codecBase, box.reader, box.name);
  330.           videoCodecs.push(parsedDVCCBox.codec);
  331.           hasVideo = true;
  332.         })
  333.         .fullBox('vpcC', (box) => {
  334.           const codecBase = baseBox || '';
  335.           const parsedVPCCBox = shaka.util.Mp4BoxParsers.parseVPCC(
  336.               codecBase, box.reader, box.name);
  337.           videoCodecs.push(parsedVPCCBox.codec);
  338.           hasVideo = true;
  339.         })
  340.         .box('av1C', (box) => {
  341.           const codecBase = baseBox || '';
  342.           const parsedAV1CBox = shaka.util.Mp4BoxParsers.parseAV1C(
  343.               codecBase, box.reader, box.name);
  344.           videoCodecs.push(parsedAV1CBox.codec);
  345.           hasVideo = true;
  346.         })
  348.         // This signals an encrypted sample, which we can go inside of to
  349.         // find the codec used.
  350.         // Note: If encrypted, you can only have audio or video, not both.
  351.         .box('enca', Mp4Parser.audioSampleEntry)
  352.         .box('encv', Mp4Parser.visualSampleEntry)
  353.         .box('sinf', Mp4Parser.children)
  354.         .box('frma', (box) => {
  355.           const {codec} = shaka.util.Mp4BoxParsers.parseFRMA(box.reader);
  356.           addCodec(codec);
  357.         })
  359.         .box('colr', (box) => {
  360.           videoCodecs = videoCodecs.map((codec) => {
  361.             if (codec.startsWith('av01.')) {
  362.               return shaka.util.Mp4BoxParsers.updateAV1CodecWithCOLRBox(
  363.                   codec, box.reader);
  364.             }
  365.             return codec;
  366.           });
  367.           const {videoRange, colorGamut} =
  368.               shaka.util.Mp4BoxParsers.parseCOLR(box.reader);
  369.           realVideoRange = videoRange;
  370.           realColorGamut = colorGamut;
  371.         })
  373.         .parse(initData || data, /* partialOkay= */ true);
  374.     if (!audioCodecs.length && !videoCodecs.length) {
  375.       return null;
  376.     }
  377.     const onlyAudio = hasAudio && !hasVideo;
  378.     const closedCaptions = new Map();
  379.     if (hasVideo) {
  380.       const captionParser = new shaka.media.ClosedCaptionParser('video/mp4');
  381.       if (initData) {
  382.         captionParser.init(initData);
  383.       }
  384.       captionParser.parseFrom(data);
  385.       for (const stream of captionParser.getStreams()) {
  386.         closedCaptions.set(stream, stream);
  387.       }
  388.       captionParser.reset();
  389.     }
  390.     const codecs = audioCodecs.concat(videoCodecs);
  391.     return {
  392.       type: onlyAudio ? 'audio' : 'video',
  393.       mimeType: onlyAudio ? 'audio/mp4' : 'video/mp4',
  394.       codecs: SegmentUtils.codecsFiltering(codecs).join(', '),
  395.       language: language,
  396.       height: height,
  397.       width: width,
  398.       channelCount: channelCount,
  399.       sampleRate: sampleRate,
  400.       closedCaptions: closedCaptions,
  401.       videoRange: realVideoRange,
  402.       colorGamut: realColorGamut,
  403.     };
  404.   }
  406.   /**
  407.    * @param {!Array.<string>} codecs
  408.    * @return {!Array.<string>} codecs
  409.    */
  410.   static codecsFiltering(codecs) {
  411.     const ContentType = shaka.util.ManifestParserUtils.ContentType;
  412.     const ManifestParserUtils = shaka.util.ManifestParserUtils;
  413.     const SegmentUtils = shaka.media.SegmentUtils;
  414.     const allCodecs = SegmentUtils.filterDuplicateCodecs_(codecs);
  415.     const audioCodecs =
  416.         ManifestParserUtils.guessAllCodecsSafe(ContentType.AUDIO, allCodecs);
  417.     const videoCodecs =
  418.         ManifestParserUtils.guessAllCodecsSafe(ContentType.VIDEO, allCodecs);
  419.     const textCodecs =
  420.         ManifestParserUtils.guessAllCodecsSafe(ContentType.TEXT, allCodecs);
  421.     const validVideoCodecs = SegmentUtils.chooseBetterCodecs_(videoCodecs);
  422.     const finalCodecs =
  423.         audioCodecs.concat(validVideoCodecs).concat(textCodecs);
  424.     if (allCodecs.length && !finalCodecs.length) {
  425.       return allCodecs;
  426.     }
  427.     return finalCodecs;
  428.   }
  430.   /**
  431.    * @param {!Array.<string>} codecs
  432.    * @return {!Array.<string>} codecs
  433.    * @private
  434.    */
  435.   static filterDuplicateCodecs_(codecs) {
  436.     // Filter out duplicate codecs.
  437.     const seen = new Set();
  438.     const ret = [];
  439.     for (const codec of codecs) {
  440.       const shortCodec = shaka.util.MimeUtils.getCodecBase(codec);
  441.       if (!seen.has(shortCodec)) {
  442.         ret.push(codec);
  443.         seen.add(shortCodec);
  444.       } else {
  445.         shaka.log.debug('Ignoring duplicate codec');
  446.       }
  447.     }
  448.     return ret;
  449.   }
  451.   /**
  452.    * Prioritizes Dolby Vision if supported. This is necessary because with
  453.    * Dolby Vision we could have hvcC and dvcC boxes at the same time.
  454.    *
  455.    * @param {!Array.<string>} codecs
  456.    * @return {!Array.<string>} codecs
  457.    * @private
  458.    */
  459.   static chooseBetterCodecs_(codecs) {
  460.     if (codecs.length <= 1) {
  461.       return codecs;
  462.     }
  463.     const dolbyVision = codecs.find((codec) => {
  464.       return codec.startsWith('dvh1.') ||
  465.           codec.startsWith('dvhe.') ||
  466.           codec.startsWith('dav1.');
  467.     });
  468.     if (!dolbyVision) {
  469.       return codecs;
  470.     }
  471.     const type = `video/mp4; codecs="${dolbyVision}"`;
  472.     if (shaka.media.Capabilities.isTypeSupported(type)) {
  473.       return [dolbyVision];
  474.     }
  475.     return codecs.filter((codec) => codec != dolbyVision);
  476.   }
  478.   /**
  479.    * @param {!BufferSource} data
  480.    * @return {?string}
  481.    */
  482.   static getDefaultKID(data) {
  483.     const Mp4Parser = shaka.util.Mp4Parser;
  485.     let defaultKID = null;
  486.     new Mp4Parser()
  487.         .box('moov', Mp4Parser.children)
  488.         .box('trak', Mp4Parser.children)
  489.         .box('mdia', Mp4Parser.children)
  490.         .box('minf', Mp4Parser.children)
  491.         .box('stbl', Mp4Parser.children)
  492.         .fullBox('stsd', Mp4Parser.sampleDescription)
  493.         .box('encv', Mp4Parser.visualSampleEntry)
  494.         .box('enca', Mp4Parser.audioSampleEntry)
  495.         .box('sinf', Mp4Parser.children)
  496.         .box('schi', Mp4Parser.children)
  497.         .fullBox('tenc', (box) => {
  498.           const parsedTENCBox = shaka.util.Mp4BoxParsers.parseTENC(box.reader);
  499.           defaultKID = parsedTENCBox.defaultKID;
  500.         })
  502.         .parse(data, /* partialOkay= */ true);
  503.     return defaultKID;
  504.   }
  505. };
  508. /**
  509.  * @typedef {{
  510.  *   type: string,
  511.  *   mimeType: string,
  512.  *   codecs: string,
  513.  *   language: ?string,
  514.  *   height: ?string,
  515.  *   width: ?string,
  516.  *   channelCount: ?number,
  517.  *   sampleRate: ?number,
  518.  *   closedCaptions: Map.<string, string>,
  519.  *   videoRange: ?string,
  520.  *   colorGamut: ?string
  521.  * }}
  522.  *
  523.  * @property {string} type
  524.  * @property {string} mimeType
  525.  * @property {string} codecs
  526.  * @property {?string} language
  527.  * @property {?string} height
  528.  * @property {?string} width
  529.  * @property {?number} channelCount
  530.  * @property {?number} sampleRate
  531.  * @property {Map.<string, string>} closedCaptions
  532.  * @property {?string} videoRange
  533.  * @property {?string} colorGamut
  534.  */
  535. shaka.media.SegmentUtils.BasicInfo;