Source: lib/text/vtt_text_parser.js

  1. /*! @license
  2.  * Shaka Player
  3.  * Copyright 2016 Google LLC
  4.  * SPDX-License-Identifier: Apache-2.0
  5.  */
  7. goog.provide('shaka.text.VttTextParser');
  9. goog.require('goog.asserts');
  10. goog.require('shaka.log');
  11. goog.require('shaka.media.ManifestParser');
  12. goog.require('shaka.text.Cue');
  13. goog.require('shaka.text.CueRegion');
  14. goog.require('shaka.text.TextEngine');
  15. goog.require('shaka.util.Error');
  16. goog.require('shaka.util.StringUtils');
  17. goog.require('shaka.util.TextParser');
  18. goog.require('shaka.util.TXml');
  21. /**
  22.  * @implements {shaka.extern.TextParser}
  23.  * @export
  24.  */
  25. shaka.text.VttTextParser = class {
  26.   /** Constructs a VTT parser. */
  27.   constructor() {
  28.     /** @private {boolean} */
  29.     this.sequenceMode_ = false;
  31.     /** @private {string} */
  32.     this.manifestType_ = shaka.media.ManifestParser.UNKNOWN;
  33.   }
  35.   /**
  36.    * @override
  37.    * @export
  38.    */
  39.   parseInit(data) {
  40.     goog.asserts.assert(false, 'VTT does not have init segments');
  41.   }
  43.   /**
  44.    * @override
  45.    * @export
  46.    */
  47.   setSequenceMode(sequenceMode) {
  48.     this.sequenceMode_ = sequenceMode;
  49.   }
  51.   /**
  52.    * @override
  53.    * @export
  54.    */
  55.   setManifestType(manifestType) {
  56.     this.manifestType_ = manifestType;
  57.   }
  59.   /**
  60.    * @override
  61.    * @export
  62.    */
  63.   parseMedia(data, time) {
  64.     const VttTextParser = shaka.text.VttTextParser;
  65.     // Get the input as a string.  Normalize newlines to \n.
  66.     let str = shaka.util.StringUtils.fromUTF8(data);
  67.     str = str.replace(/\r\n|\r(?=[^\n]|$)/gm, '\n');
  68.     const blocks = str.split(/\n{2,}/m);
  70.     if (!/^WEBVTT($|[ \t\n])/m.test(blocks[0])) {
  71.       throw new shaka.util.Error(
  72.           shaka.util.Error.Severity.CRITICAL,
  73.           shaka.util.Error.Category.TEXT,
  74.           shaka.util.Error.Code.INVALID_TEXT_HEADER);
  75.     }
  77.     // Depending on "segmentRelativeVttTiming" configuration,
  78.     // "vttOffset" will correspond to either "periodStart" (default)
  79.     // or "segmentStart", for segmented VTT where timings are relative
  80.     // to the beginning of each segment.
  81.     // NOTE: "periodStart" is the timestamp offset applied via TextEngine.
  82.     // It is no longer closely tied to periods, but the name stuck around.
  83.     // NOTE: This offset and the flag choosing its meaning have no effect on
  84.     // HLS content, which should use X-TIMESTAMP-MAP and periodStart instead.
  85.     let offset = time.vttOffset;
  87.     // Only use 'X-TIMESTAMP-MAP' with HLS.  This overrides offset above.
  88.     if (blocks[0].includes('X-TIMESTAMP-MAP') &&
  89.         this.manifestType_ == shaka.media.ManifestParser.HLS) {
  90.       if (this.sequenceMode_) {
  91.         // Compute a different, rollover-based offset for sequence mode.
  92.         offset = this.computeHlsSequenceModeOffset_(blocks[0], time);
  93.       } else {
  94.         // Calculate the offset from the segment startTime.
  95.         offset = time.segmentStart;
  96.       }
  97.     }
  99.     // Parse VTT regions.
  100.     /* !Array.<!shaka.text.CueRegion> */
  101.     const regions = [];
  102.     for (const line of blocks[0].split('\n')) {
  103.       if (/^Region:/.test(line)) {
  104.         const region = VttTextParser.parseRegion_(line);
  105.         regions.push(region);
  106.       }
  107.     }
  109.     /** @type {!Map.<string, shaka.text.Cue>} */
  110.     const styles = new Map();
  111.     VttTextParser.addDefaultTextColor_(styles);
  113.     // Parse cues.
  114.     const ret = [];
  115.     for (const block of blocks.slice(1)) {
  116.       const lines = block.split('\n');
  117.       VttTextParser.parseStyle_(lines, styles);
  118.       const cue = VttTextParser.parseCue_(lines, offset, regions, styles);
  119.       if (cue) {
  120.         ret.push(cue);
  121.       }
  122.     }
  124.     return ret;
  125.   }
  127.   /**
  128.    * @param {string} headerBlock Contains X-TIMESTAMP-MAP.
  129.    * @param {shaka.extern.TextParser.TimeContext} time
  130.    * @return {number}
  131.    * @private
  132.    */
  133.   computeHlsSequenceModeOffset_(headerBlock, time) {
  134.     // https://bit.ly/2K92l7y
  135.     // The 'X-TIMESTAMP-MAP' header is used in HLS to align text with
  136.     // the rest of the media.
  137.     // The header format is 'X-TIMESTAMP-MAP=MPEGTS:n,LOCAL:m'
  138.     // (the attributes can go in any order)
  139.     // where n is MPEG-2 time and m is cue time it maps to.
  140.     // For example 'X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000'
  141.     // means an offset of 10 seconds
  142.     // 900000/MPEG_TIMESCALE - cue time.
  143.     const cueTimeMatch = headerBlock.match(
  144.         /LOCAL:((?:(\d{1,}):)?(\d{2}):(\d{2})\.(\d{3}))/m);
  145.     const mpegTimeMatch = headerBlock.match(/MPEGTS:(\d+)/m);
  147.     if (!cueTimeMatch || !mpegTimeMatch) {
  148.       throw new shaka.util.Error(
  149.           shaka.util.Error.Severity.CRITICAL,
  150.           shaka.util.Error.Category.TEXT,
  151.           shaka.util.Error.Code.INVALID_TEXT_HEADER);
  152.     }
  154.     const parser = new shaka.util.TextParser(cueTimeMatch[1]);
  155.     const cueTime = shaka.text.VttTextParser.parseTime_(parser);
  156.     if (cueTime == null) {
  157.       throw new shaka.util.Error(
  158.           shaka.util.Error.Severity.CRITICAL,
  159.           shaka.util.Error.Category.TEXT,
  160.           shaka.util.Error.Code.INVALID_TEXT_HEADER);
  161.     }
  163.     const mpegTime = Number(mpegTimeMatch[1]);
  164.     const mpegTimescale = shaka.text.VttTextParser.MPEG_TIMESCALE_;
  166.     return time.periodStart + mpegTime / mpegTimescale - cueTime;
  167.   }
  169.   /**
  170.    * Add default color
  171.    *
  172.    * @param {!Map.<string, shaka.text.Cue>} styles
  173.    * @private
  174.    */
  175.   static addDefaultTextColor_(styles) {
  176.     const textColor = shaka.text.Cue.defaultTextColor;
  177.     for (const [key, value] of Object.entries(textColor)) {
  178.       const cue = new shaka.text.Cue(0, 0, '');
  179.       cue.color = value;
  180.       styles.set('.' + key, cue);
  181.     }
  183.     const bgColor = shaka.text.Cue.defaultTextBackgroundColor;
  184.     for (const [key, value] of Object.entries(bgColor)) {
  185.       const cue = new shaka.text.Cue(0, 0, '');
  186.       cue.backgroundColor = value;
  187.       styles.set('.' + key, cue);
  188.     }
  189.   }
  191.   /**
  192.    * Parses a string into a Region object.
  193.    *
  194.    * @param {string} text
  195.    * @return {!shaka.text.CueRegion}
  196.    * @private
  197.    */
  198.   static parseRegion_(text) {
  199.     const VttTextParser = shaka.text.VttTextParser;
  200.     const parser = new shaka.util.TextParser(text);
  201.     // The region string looks like this:
  202.     // Region: id=fred width=50% lines=3 regionanchor=0%,100%
  203.     //         viewportanchor=10%,90% scroll=up
  204.     const region = new shaka.text.CueRegion();
  206.     // Skip 'Region:'
  207.     parser.readWord();
  208.     parser.skipWhitespace();
  210.     let word = parser.readWord();
  211.     while (word) {
  212.       if (!VttTextParser.parseRegionSetting_(region, word)) {
  213.         shaka.log.warning(
  214.             'VTT parser encountered an invalid VTTRegion setting: ', word,
  215.             ' The setting will be ignored.');
  216.       }
  217.       parser.skipWhitespace();
  218.       word = parser.readWord();
  219.     }
  221.     return region;
  222.   }
  224.   /**
  225.    * Parses a style block into a Cue object.
  226.    *
  227.    * @param {!Array.<string>} text
  228.    * @param {!Map.<string, shaka.text.Cue>} styles
  229.    * @private
  230.    */
  231.   static parseStyle_(text, styles) {
  232.     // Skip empty blocks.
  233.     if (text.length == 1 && !text[0]) {
  234.       return;
  235.     }
  237.     // Skip comment blocks.
  238.     if (/^NOTE($|[ \t])/.test(text[0])) {
  239.       return;
  240.     }
  242.     // Only style block are allowed.
  243.     if (text[0] != 'STYLE') {
  244.       return;
  245.     }
  247.     /** @type {!Array.<!Array.<string>>} */
  248.     const styleBlocks = [];
  249.     let lastBlockIndex = -1;
  250.     for (let i = 1; i < text.length; i++) {
  251.       if (text[i].includes('::cue')) {
  252.         styleBlocks.push([]);
  253.         lastBlockIndex = styleBlocks.length - 1;
  254.       }
  255.       if (lastBlockIndex == -1) {
  256.         continue;
  257.       }
  258.       styleBlocks[lastBlockIndex].push(text[i]);
  259.       if (text[i].includes('}')) {
  260.         lastBlockIndex = -1;
  261.       }
  262.     }
  264.     for (const styleBlock of styleBlocks) {
  265.       let styleSelector = 'global';
  266.       // Look for what is within parentheses. For example:
  267.       // <code>:: cue (b) {</code>, what we are looking for is <code>b</code>
  268.       const selector = styleBlock[0].match(/\((.*)\)/);
  269.       if (selector) {
  270.         styleSelector = selector.pop();
  271.       }
  273.       // We start at 1 to avoid '::cue' and end earlier to avoid '}'
  274.       let propertyLines = styleBlock.slice(1, -1);
  275.       if (styleBlock[0].includes('}')) {
  276.         const payload = /\{(.*?)\}/.exec(styleBlock[0]);
  277.         if (payload) {
  278.           propertyLines = payload[1].split(';');
  279.         }
  280.       }
  282.       // Continue styles over multiple selectors if necessary.
  283.       // For example,
  284.       //   ::cue(b) { background: white; } ::cue(b) { color: blue; }
  285.       // should set both the background and foreground of bold tags.
  286.       let cue = styles.get(styleSelector);
  287.       if (!cue) {
  288.         cue = new shaka.text.Cue(0, 0, '');
  289.       }
  291.       let validStyle = false;
  292.       for (let i = 0; i < propertyLines.length; i++) {
  293.         // We look for CSS properties. As a general rule they are separated by
  294.         // <code>:</code>. Eg: <code>color: red;</code>
  295.         const lineParts = /^\s*([^:]+):\s*(.*)/.exec(propertyLines[i]);
  296.         if (lineParts) {
  297.           const name = lineParts[1].trim();
  298.           const value = lineParts[2].trim().replace(';', '');
  299.           switch (name) {
  300.             case 'background-color':
  301.             case 'background':
  302.               validStyle = true;
  303.               cue.backgroundColor = value;
  304.               break;
  305.             case 'color':
  306.               validStyle = true;
  307.               cue.color = value;
  308.               break;
  309.             case 'font-family':
  310.               validStyle = true;
  311.               cue.fontFamily = value;
  312.               break;
  313.             case 'font-size':
  314.               validStyle = true;
  315.               cue.fontSize = value;
  316.               break;
  317.             case 'font-weight':
  318.               if (parseInt(value, 10) >= 700 || value == 'bold') {
  319.                 validStyle = true;
  320.                 cue.fontWeight = shaka.text.Cue.fontWeight.BOLD;
  321.               }
  322.               break;
  323.             case 'font-style':
  324.               switch (value) {
  325.                 case 'normal':
  326.                   validStyle = true;
  327.                   cue.fontStyle = shaka.text.Cue.fontStyle.NORMAL;
  328.                   break;
  329.                 case 'italic':
  330.                   validStyle = true;
  331.                   cue.fontStyle = shaka.text.Cue.fontStyle.ITALIC;
  332.                   break;
  333.                 case 'oblique':
  334.                   validStyle = true;
  335.                   cue.fontStyle = shaka.text.Cue.fontStyle.OBLIQUE;
  336.                   break;
  337.               }
  338.               break;
  339.             case 'opacity':
  340.               validStyle = true;
  341.               cue.opacity = parseFloat(value);
  342.               break;
  343.             case 'text-combine-upright':
  344.               validStyle = true;
  345.               cue.textCombineUpright = value;
  346.               break;
  347.             case 'text-shadow':
  348.               validStyle = true;
  349.               cue.textShadow = value;
  350.               break;
  351.             case 'white-space':
  352.               validStyle = true;
  353.               cue.wrapLine = value != 'noWrap';
  354.               break;
  355.             default:
  356.               shaka.log.warning('VTT parser encountered an unsupported style: ',
  357.                   lineParts);
  358.               break;
  359.           }
  360.         }
  361.       }
  363.       if (validStyle) {
  364.         styles.set(styleSelector, cue);
  365.       }
  366.     }
  367.   }
  369.   /**
  370.    * Parses a text block into a Cue object.
  371.    *
  372.    * @param {!Array.<string>} text
  373.    * @param {number} timeOffset
  374.    * @param {!Array.<!shaka.text.CueRegion>} regions
  375.    * @param {!Map.<string, shaka.text.Cue>} styles
  376.    * @return {shaka.text.Cue}
  377.    * @private
  378.    */
  379.   static parseCue_(text, timeOffset, regions, styles) {
  380.     const VttTextParser = shaka.text.VttTextParser;
  382.     // Skip empty blocks.
  383.     if (text.length == 1 && !text[0]) {
  384.       return null;
  385.     }
  387.     // Skip comment blocks.
  388.     if (/^NOTE($|[ \t])/.test(text[0])) {
  389.       return null;
  390.     }
  392.     // Skip style and region blocks.
  393.     if (text[0] == 'STYLE' || text[0] == 'REGION') {
  394.       return null;
  395.     }
  397.     let id = null;
  398.     if (!text[0].includes('-->')) {
  399.       id = text[0];
  400.       text.splice(0, 1);
  401.     }
  403.     // Parse the times.
  404.     const parser = new shaka.util.TextParser(text[0]);
  405.     let start = VttTextParser.parseTime_(parser);
  406.     const expect = parser.readRegex(/[ \t]+-->[ \t]+/g);
  407.     let end = VttTextParser.parseTime_(parser);
  409.     if (start == null || expect == null || end == null) {
  410.       shaka.log.alwaysWarn(
  411.           'Failed to parse VTT time code. Cue skipped:', id, text);
  412.       return null;
  413.     }
  415.     start += timeOffset;
  416.     end += timeOffset;
  418.     // Get the payload.
  419.     const payload = text.slice(1).join('\n').trim();
  421.     let cue = null;
  422.     if (styles.has('global')) {
  423.       cue = styles.get('global').clone();
  424.       cue.startTime = start;
  425.       cue.endTime = end;
  426.       cue.payload = '';
  427.     } else {
  428.       cue = new shaka.text.Cue(start, end, '');
  429.     }
  431.     // Parse optional settings.
  432.     parser.skipWhitespace();
  433.     let word = parser.readWord();
  434.     while (word) {
  435.       if (!VttTextParser.parseCueSetting(cue, word, regions)) {
  436.         shaka.log.warning('VTT parser encountered an invalid VTT setting: ',
  437.             word,
  438.             ' The setting will be ignored.');
  439.       }
  440.       parser.skipWhitespace();
  441.       word = parser.readWord();
  442.     }
  444.     VttTextParser.parseCueStyles(payload, cue, styles);
  446.     if (id != null) {
  447.       cue.id = id;
  448.     }
  449.     return cue;
  450.   }
  452.   /**
  453.    * Parses a WebVTT styles from the given payload.
  454.    *
  455.    * @param {string} payload
  456.    * @param {!shaka.text.Cue} rootCue
  457.    * @param {!Map.<string, shaka.text.Cue>} styles
  458.    */
  459.   static parseCueStyles(payload, rootCue, styles) {
  460.     const VttTextParser = shaka.text.VttTextParser;
  461.     const StringUtils = shaka.util.StringUtils;
  462.     const TXml = shaka.util.TXml;
  463.     // Optimization for unstyled payloads.
  464.     if (!payload.includes('<')) {
  465.       rootCue.payload = StringUtils.htmlUnescape(payload);
  466.       return;
  467.     }
  468.     if (styles.size === 0) {
  469.       VttTextParser.addDefaultTextColor_(styles);
  470.     }
  472.     payload = VttTextParser.replaceKaraokeStylePayload_(payload);
  473.     payload = VttTextParser.replaceVoiceStylePayload_(payload);
  474.     payload = VttTextParser.escapeInvalidChevrons_(payload);
  475.     const xmlPayload = '<span>' + payload + '</span>';
  476.     let element;
  477.     try {
  478.       element = TXml.parseXmlString(xmlPayload, 'span');
  479.     } catch (e) {
  480.       shaka.log.warning('cue parse fail: ', e);
  481.     }
  483.     if (element) {
  484.       const childNodes = element.children;
  485.       if (childNodes.length == 1) {
  486.         const childNode = childNodes[0];
  487.         if (!TXml.isNode(childNode)) {
  488.           rootCue.payload = StringUtils.htmlUnescape(payload);
  489.           return;
  490.         }
  491.       }
  492.       for (const childNode of childNodes) {
  493.         VttTextParser.generateCueFromElement_(childNode, rootCue, styles);
  494.       }
  495.     } else {
  496.       shaka.log.warning('The cue\'s markup could not be parsed: ', payload);
  497.       rootCue.payload = StringUtils.htmlUnescape(payload);
  498.     }
  499.   }
  501.   /**
  502.    * This method converts invalid > chevrons to HTML entities.
  503.    * It also removes < chevrons as per spec.
  504.    *
  505.    * @param {!string} input
  506.    * @return {string}
  507.    * @private
  508.    */
  509.   static escapeInvalidChevrons_(input) {
  510.     // Used to map HTML entities to characters.
  511.     const htmlEscapes = {
  512.       '< ': '',
  513.       ' >': ' &gt;',
  514.     };
  516.     const reEscapedHtml = /(< +>|<\s|\s>)/g;
  517.     const reHasEscapedHtml = RegExp(reEscapedHtml.source);
  518.     // This check is an optimization, since replace always makes a copy
  519.     if (input && reHasEscapedHtml.test(input)) {
  520.       return input.replace(reEscapedHtml, (entity) => {
  521.         return htmlEscapes[entity] || '';
  522.       });
  523.     }
  524.     return input || '';
  525.   }
  527.   /**
  528.    * Converts voice style tag to be valid for xml parsing
  529.    * For example,
  530.    * input: <v Shaka>Test
  531.    * output: <v.voice-Shaka>Test</v.voice-Shaka>
  532.    *
  533.    * @param {string} payload
  534.    * @return {string} processed payload
  535.    * @private
  536.    */
  537.   static replaceVoiceStylePayload_(payload) {
  538.     const voiceTag = 'v';
  539.     const names = [];
  540.     let nameStart = -1;
  541.     let newPayload = '';
  542.     let hasVoiceEndTag = false;
  543.     for (let i = 0; i < payload.length; i++) {
  544.       // This condition is used to manage tags that have end tags.
  545.       if (payload[i] === '/') {
  546.         const end = payload.indexOf('>', i);
  547.         if (end === -1) {
  548.           return payload;
  549.         }
  550.         const tagEnd = payload.substring(i + 1, end);
  551.         if (!tagEnd || tagEnd != voiceTag) {
  552.           newPayload += payload[i];
  553.           continue;
  554.         }
  555.         hasVoiceEndTag = true;
  556.         let tagStart = null;
  557.         if (names.length) {
  558.           tagStart = names[names.length -1];
  559.         }
  560.         if (!tagStart) {
  561.           newPayload += payload[i];
  562.         } else if (tagStart === tagEnd) {
  563.           newPayload += '/' + tagEnd + '>';
  564.           i += tagEnd.length + 1;
  565.         } else {
  566.           if (!tagStart.startsWith(voiceTag)) {
  567.             newPayload += payload[i];
  568.             continue;
  569.           }
  570.           newPayload += '/' + tagStart + '>';
  571.           i += tagEnd.length + 1;
  572.         }
  573.       } else {
  574.         // Here we only want the tag name, not any other payload.
  575.         if (payload[i] === '<') {
  576.           nameStart = i + 1;
  577.           if (payload[nameStart] != voiceTag) {
  578.             nameStart = -1;
  579.           }
  580.         } else if (payload[i] === '>') {
  581.           if (nameStart > 0) {
  582.             names.push(payload.substr(nameStart, i - nameStart));
  583.             nameStart = -1;
  584.           }
  585.         }
  586.         newPayload += payload[i];
  587.       }
  588.     }
  589.     for (const name of names) {
  590.       const newName = name.replace(' ', '.voice-');
  591.       newPayload = newPayload.replace(`<${name}>`, `<${newName}>`);
  592.       newPayload = newPayload.replace(`</${name}>`, `</${newName}>`);
  593.       if (!hasVoiceEndTag) {
  594.         newPayload += `</${newName}>`;
  595.       }
  596.     }
  597.     return newPayload;
  598.   }
  600.   /**
  601.    * Converts karaoke style tag to be valid for xml parsing
  602.    * For example,
  603.    * input: Text <00:00:00.450> time <00:00:01.450> 1
  604.    * output: Text <div time="00:00:00.450"> time
  605.    *         <div time="00:00:01.450"> 1</div></div>
  606.    *
  607.    * @param {string} payload
  608.    * @return {string} processed payload
  609.    * @private
  610.    */
  611.   static replaceKaraokeStylePayload_(payload) {
  612.     const names = [];
  613.     let nameStart = -1;
  614.     for (let i = 0; i < payload.length; i++) {
  615.       if (payload[i] === '<') {
  616.         nameStart = i + 1;
  617.       } else if (payload[i] === '>') {
  618.         if (nameStart > 0) {
  619.           const name = payload.substr(nameStart, i - nameStart);
  620.           if (name.match(shaka.text.VttTextParser.timeFormat_)) {
  621.             names.push(name);
  622.           }
  623.           nameStart = -1;
  624.         }
  625.       }
  626.     }
  627.     let newPayload = payload;
  628.     for (const name of names) {
  629.       const replaceTag = '<' + name + '>';
  630.       const startTag = '<div time="' + name + '">';
  631.       const endTag = '</div>';
  632.       newPayload = newPayload.replace(replaceTag, startTag);
  633.       newPayload += endTag;
  634.     }
  635.     return newPayload;
  636.   }
  638.   /**
  639.    * @param {string} value
  640.    * @param {string} defaultValue
  641.    * @private
  642.    */
  643.   static getOrDefault_(value, defaultValue) {
  644.     if (value && value.length > 0) {
  645.       return value;
  646.     }
  647.     return defaultValue;
  648.   }
  650.   /**
  651.    * Merges values created in parseStyle_
  652.    * @param {!shaka.text.Cue} cue
  653.    * @param {shaka.text.Cue} refCue
  654.    * @private
  655.    */
  656.   static mergeStyle_(cue, refCue) {
  657.     if (!refCue) {
  658.       return;
  659.     }
  661.     const VttTextParser = shaka.text.VttTextParser;
  662.     // Overwrites if new value string length > 0
  663.     cue.backgroundColor = VttTextParser.getOrDefault_(
  664.         refCue.backgroundColor, cue.backgroundColor);
  665.     cue.color = VttTextParser.getOrDefault_(
  666.         refCue.color, cue.color);
  667.     cue.fontFamily = VttTextParser.getOrDefault_(
  668.         refCue.fontFamily, cue.fontFamily);
  669.     cue.fontSize = VttTextParser.getOrDefault_(
  670.         refCue.fontSize, cue.fontSize);
  671.     cue.textShadow = VttTextParser.getOrDefault_(
  672.         refCue.textShadow, cue.textShadow);
  674.     // Overwrite with new values as unable to determine
  675.     // if new value is set or not
  676.     cue.fontWeight = refCue.fontWeight;
  677.     cue.fontStyle = refCue.fontStyle;
  678.     cue.opacity = refCue.opacity;
  679.     cue.rubyTag = refCue.rubyTag;
  680.     cue.textCombineUpright = refCue.textCombineUpright;
  681.     cue.wrapLine = refCue.wrapLine;
  682.   }
  684.   /**
  685.    * @param {!shaka.extern.xml.Node} element
  686.    * @param {!shaka.text.Cue} rootCue
  687.    * @param {!Map.<string, shaka.text.Cue>} styles
  688.    * @private
  689.    */
  690.   static generateCueFromElement_(element, rootCue, styles) {
  691.     const VttTextParser = shaka.text.VttTextParser;
  692.     const TXml = shaka.util.TXml;
  693.     const nestedCue = rootCue.clone();
  694.     // We don't want propagate some properties.
  695.     nestedCue.nestedCues = [];
  696.     nestedCue.payload = '';
  697.     nestedCue.rubyTag = '';
  698.     // We don't want propagate some position settings
  699.     nestedCue.line = null;
  700.     nestedCue.region = new shaka.text.CueRegion();
  701.     nestedCue.position = null;
  702.     nestedCue.size = 0;
  704.     if (shaka.util.TXml.isNode(element)) {
  705.       const bold = shaka.text.Cue.fontWeight.BOLD;
  706.       const italic = shaka.text.Cue.fontStyle.ITALIC;
  707.       const underline = shaka.text.Cue.textDecoration.UNDERLINE;
  708.       const tags = element.tagName.split(/(?=[ .])+/g);
  709.       for (const tag of tags) {
  710.         let styleTag = tag;
  711.         // White blanks at start indicate that the style is a voice
  712.         if (styleTag.startsWith('.voice-')) {
  713.           const voice = styleTag.split('-').pop();
  714.           styleTag = `v[voice="${voice}"]`;
  715.           // The specification allows to have quotes and not, so we check to
  716.           // see which one is being used.
  717.           if (!styles.has(styleTag)) {
  718.             styleTag = `v[voice=${voice}]`;
  719.           }
  720.         }
  721.         if (styles.has(styleTag)) {
  722.           VttTextParser.mergeStyle_(nestedCue, styles.get(styleTag));
  723.         }
  724.         switch (tag) {
  725.           case 'br': {
  726.             const lineBreakCue = shaka.text.Cue.lineBreak(
  727.                 nestedCue.startTime, nestedCue.endTime);
  728.             rootCue.nestedCues.push(lineBreakCue);
  729.             return;
  730.           }
  731.           case 'b':
  732.             nestedCue.fontWeight = bold;
  733.             break;
  734.           case 'i':
  735.             nestedCue.fontStyle = italic;
  736.             break;
  737.           case 'u':
  738.             nestedCue.textDecoration.push(underline);
  739.             break;
  740.           case 'font': {
  741.             const color = element.attributes['color'];
  742.             if (color) {
  743.               nestedCue.color = color;
  744.             }
  745.             break;
  746.           }
  747.           case 'div': {
  748.             const time = element.attributes['time'];
  749.             if (!time) {
  750.               break;
  751.             }
  752.             const parser = new shaka.util.TextParser(time);
  753.             const cueTime = shaka.text.VttTextParser.parseTime_(parser);
  754.             if (cueTime) {
  755.               nestedCue.startTime = cueTime;
  756.             }
  757.             break;
  758.           }
  759.           case 'ruby':
  760.           case 'rp':
  761.           case 'rt':
  762.             nestedCue.rubyTag = tag;
  763.             break;
  764.           default:
  765.             break;
  766.         }
  767.       }
  768.     }
  770.     const isTextNode = (item) => shaka.util.TXml.isText(item);
  771.     const childNodes = element.children;
  772.     if (isTextNode(element) ||
  773.         (childNodes.length == 1 && isTextNode(childNodes[0]))) {
  774.       // Trailing line breaks may lost when convert cue to HTML tag
  775.       // Need to insert line break cue to preserve line breaks
  776.       const textArr = TXml.getTextContents(element).split('\n');
  777.       let isFirst = true;
  778.       for (const text of textArr) {
  779.         if (!isFirst) {
  780.           const lineBreakCue = shaka.text.Cue.lineBreak(
  781.               nestedCue.startTime, nestedCue.endTime);
  782.           rootCue.nestedCues.push(lineBreakCue);
  783.         }
  784.         if (text.length > 0) {
  785.           const textCue = nestedCue.clone();
  786.           textCue.payload = shaka.util.StringUtils.htmlUnescape(text);
  787.           rootCue.nestedCues.push(textCue);
  788.         }
  789.         isFirst = false;
  790.       }
  791.     } else {
  792.       rootCue.nestedCues.push(nestedCue);
  793.       for (const childNode of childNodes) {
  794.         VttTextParser.generateCueFromElement_(childNode, nestedCue, styles);
  795.       }
  796.     }
  797.   }
  799.   /**
  800.    * Parses a WebVTT setting from the given word.
  801.    *
  802.    * @param {!shaka.text.Cue} cue
  803.    * @param {string} word
  804.    * @param {!Array.<!shaka.text.CueRegion>} regions
  805.    * @return {boolean} True on success.
  806.    */
  807.   static parseCueSetting(cue, word, regions) {
  808.     const VttTextParser = shaka.text.VttTextParser;
  809.     let results = null;
  810.     if ((results = /^align:(start|middle|center|end|left|right)$/.exec(word))) {
  811.       VttTextParser.setTextAlign_(cue, results[1]);
  812.     } else if ((results = /^vertical:(lr|rl)$/.exec(word))) {
  813.       VttTextParser.setVerticalWritingMode_(cue, results[1]);
  814.     } else if ((results = /^size:([\d.]+)%$/.exec(word))) {
  815.       cue.size = Number(results[1]);
  816.     } else if ((results =
  817.         // eslint-disable-next-line max-len
  818.         /^position:([\d.]+)%(?:,(line-left|line-right|middle|center|start|end|auto))?$/
  819.             .exec(word))) {
  820.       cue.position = Number(results[1]);
  821.       if (results[2]) {
  822.         VttTextParser.setPositionAlign_(cue, results[2]);
  823.       }
  824.     } else if ((results = /^region:(.*)$/.exec(word))) {
  825.       const region = VttTextParser.getRegionById_(regions, results[1]);
  826.       if (region) {
  827.         cue.region = region;
  828.       }
  829.     } else {
  830.       return VttTextParser.parsedLineValueAndInterpretation_(cue, word);
  831.     }
  833.     return true;
  834.   }
  836.   /**
  837.    *
  838.    * @param {!Array.<!shaka.text.CueRegion>} regions
  839.    * @param {string} id
  840.    * @return {?shaka.text.CueRegion}
  841.    * @private
  842.    */
  843.   static getRegionById_(regions, id) {
  844.     const regionsWithId = regions.filter((region) => {
  845.       return region.id == id;
  846.     });
  847.     if (!regionsWithId.length) {
  848.       shaka.log.warning('VTT parser could not find a region with id: ',
  849.           id,
  850.           ' The region will be ignored.');
  851.       return null;
  852.     }
  853.     goog.asserts.assert(regionsWithId.length == 1,
  854.         'VTTRegion ids should be unique!');
  856.     return regionsWithId[0];
  857.   }
  859.   /**
  860.    * Parses a WebVTTRegion setting from the given word.
  861.    *
  862.    * @param {!shaka.text.CueRegion} region
  863.    * @param {string} word
  864.    * @return {boolean} True on success.
  865.    * @private
  866.    */
  867.   static parseRegionSetting_(region, word) {
  868.     let results = null;
  869.     if ((results = /^id=(.*)$/.exec(word))) {
  870.       region.id = results[1];
  871.     } else if ((results = /^width=(\d{1,2}|100)%$/.exec(word))) {
  872.       region.width = Number(results[1]);
  873.     } else if ((results = /^lines=(\d+)$/.exec(word))) {
  874.       region.height = Number(results[1]);
  875.       region.heightUnits = shaka.text.CueRegion.units.LINES;
  876.     } else if ((results = /^regionanchor=(\d{1,2}|100)%,(\d{1,2}|100)%$/
  877.         .exec(word))) {
  878.       region.regionAnchorX = Number(results[1]);
  879.       region.regionAnchorY = Number(results[2]);
  880.     } else if ((results = /^viewportanchor=(\d{1,2}|100)%,(\d{1,2}|100)%$/
  881.         .exec(word))) {
  882.       region.viewportAnchorX = Number(results[1]);
  883.       region.viewportAnchorY = Number(results[2]);
  884.     } else if ((results = /^scroll=up$/.exec(word))) {
  885.       region.scroll = shaka.text.CueRegion.scrollMode.UP;
  886.     } else {
  887.       return false;
  888.     }
  890.     return true;
  891.   }
  893.   /**
  894.    * @param {!shaka.text.Cue} cue
  895.    * @param {string} align
  896.    * @private
  897.    */
  898.   static setTextAlign_(cue, align) {
  899.     const Cue = shaka.text.Cue;
  900.     if (align == 'middle') {
  901.       cue.textAlign = Cue.textAlign.CENTER;
  902.     } else {
  903.       goog.asserts.assert(align.toUpperCase() in Cue.textAlign,
  904.           align.toUpperCase() +
  905.                           ' Should be in Cue.textAlign values!');
  907.       cue.textAlign = Cue.textAlign[align.toUpperCase()];
  908.     }
  909.   }
  911.   /**
  912.    * @param {!shaka.text.Cue} cue
  913.    * @param {string} align
  914.    * @private
  915.    */
  916.   static setPositionAlign_(cue, align) {
  917.     const Cue = shaka.text.Cue;
  918.     if (align == 'line-left' || align == 'start') {
  919.       cue.positionAlign = Cue.positionAlign.LEFT;
  920.     } else if (align == 'line-right' || align == 'end') {
  921.       cue.positionAlign = Cue.positionAlign.RIGHT;
  922.     } else if (align == 'center' || align == 'middle') {
  923.       cue.positionAlign = Cue.positionAlign.CENTER;
  924.     } else {
  925.       cue.positionAlign = Cue.positionAlign.AUTO;
  926.     }
  927.   }
  929.   /**
  930.    * @param {!shaka.text.Cue} cue
  931.    * @param {string} value
  932.    * @private
  933.    */
  934.   static setVerticalWritingMode_(cue, value) {
  935.     const Cue = shaka.text.Cue;
  936.     if (value == 'lr') {
  937.       cue.writingMode = Cue.writingMode.VERTICAL_LEFT_TO_RIGHT;
  938.     } else {
  939.       cue.writingMode = Cue.writingMode.VERTICAL_RIGHT_TO_LEFT;
  940.     }
  941.   }
  943.   /**
  944.    * @param {!shaka.text.Cue} cue
  945.    * @param {string} word
  946.    * @return {boolean}
  947.    * @private
  948.    */
  949.   static parsedLineValueAndInterpretation_(cue, word) {
  950.     const Cue = shaka.text.Cue;
  951.     let results = null;
  952.     if ((results = /^line:([\d.]+)%(?:,(start|end|center))?$/.exec(word))) {
  953.       cue.lineInterpretation = Cue.lineInterpretation.PERCENTAGE;
  954.       cue.line = Number(results[1]);
  955.       if (results[2]) {
  956.         goog.asserts.assert(
  957.             results[2].toUpperCase() in Cue.lineAlign,
  958.             results[2].toUpperCase() + ' Should be in Cue.lineAlign values!');
  959.         cue.lineAlign = Cue.lineAlign[results[2].toUpperCase()];
  960.       }
  961.     } else if ((results =
  962.                     /^line:(-?\d+)(?:,(start|end|center))?$/.exec(word))) {
  963.       cue.lineInterpretation = Cue.lineInterpretation.LINE_NUMBER;
  964.       cue.line = Number(results[1]);
  965.       if (results[2]) {
  966.         goog.asserts.assert(
  967.             results[2].toUpperCase() in Cue.lineAlign,
  968.             results[2].toUpperCase() + ' Should be in Cue.lineAlign values!');
  969.         cue.lineAlign = Cue.lineAlign[results[2].toUpperCase()];
  970.       }
  971.     } else {
  972.       return false;
  973.     }
  975.     return true;
  976.   }
  978.   /**
  979.    * Parses a WebVTT time from the given parser.
  980.    *
  981.    * @param {!shaka.util.TextParser} parser
  982.    * @return {?number}
  983.    * @private
  984.    */
  985.   static parseTime_(parser) {
  986.     const results = parser.readRegex(shaka.text.VttTextParser.timeFormat_);
  987.     if (results == null) {
  988.       return null;
  989.     }
  990.     // This capture is optional, but will still be in the array as undefined,
  991.     // in which case it is 0.
  992.     const hours = Number(results[1]) || 0;
  993.     const minutes = Number(results[2]);
  994.     const seconds = Number(results[3]);
  995.     const milliseconds = Number(results[4]);
  996.     if (minutes > 59 || seconds > 59) {
  997.       return null;
  998.     }
  1000.     return (milliseconds / 1000) + seconds + (minutes * 60) + (hours * 3600);
  1001.   }
  1002. };
  1004. /**
  1005.  * @const {number}
  1006.  * @private
  1007.  */
  1008. shaka.text.VttTextParser.MPEG_TIMESCALE_ = 90000;
  1010. /**
  1011.  * @const
  1012.  * @private {!RegExp}
  1013.  * @example 00:00.000 or 00:00:00.000 or 0:00:00.000 or
  1014.  * 00:00.00 or 00:00:00.00 or 0:00:00.00
  1015.  */
  1016. shaka.text.VttTextParser.timeFormat_ =
  1017.     /(?:(\d{1,}):)?(\d{2}):(\d{2})\.(\d{2,3})/g;
  1019. shaka.text.TextEngine.registerParser(
  1020.     'text/vtt', () => new shaka.text.VttTextParser());
  1022. shaka.text.TextEngine.registerParser(
  1023.     'text/vtt; codecs="vtt"', () => new shaka.text.VttTextParser());
  1025. shaka.text.TextEngine.registerParser(
  1026.     'text/vtt; codecs="wvtt"', () => new shaka.text.VttTextParser());