add extjs 6.0.1 sources

[extjs.git] / extjs / build / examples / kitchensink / lib / prettify / prettify.js

// Copyright (C) 2006 Google Inc.\r
//\r
// Licensed under the Apache License, Version 2.0 (the "License");\r
// you may not use this file except in compliance with the License.\r
// You may obtain a copy of the License at\r
//\r
//      http://www.apache.org/licenses/LICENSE-2.0\r
//\r
// Unless required by applicable law or agreed to in writing, software\r
// distributed under the License is distributed on an "AS IS" BASIS,\r
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\r
// See the License for the specific language governing permissions and\r
// limitations under the License.\r
\r
\r
/**\r
 * @fileoverview\r
 * some functions for browser-side pretty printing of code contained in html.\r
 *\r
 * <p>\r
 * For a fairly comprehensive set of languages see the\r
 * <a href="http://google-code-prettify.googlecode.com/svn/trunk/README.html#langs">README</a>\r
 * file that came with this source.  At a minimum, the lexer should work on a\r
 * number of languages including C and friends, Java, Python, Bash, SQL, HTML,\r
 * XML, CSS, Javascript, and Makefiles.  It works passably on Ruby, PHP and Awk\r
 * and a subset of Perl, but, because of commenting conventions, doesn't work on\r
 * Smalltalk, Lisp-like, or CAML-like languages without an explicit lang class.\r
 * <p>\r
 * Usage: <ol>\r
 * <li> include this source file in an html page via\r
 *   {@code <script type="text/javascript" src="/path/to/prettify.js"></script>}\r
 * <li> define style rules.  See the example page for examples.\r
 * <li> mark the {@code <pre>} and {@code <code>} tags in your source with\r
 *    {@code class=prettyprint.}\r
 *    You can also use the (html deprecated) {@code <xmp>} tag, but the pretty\r
 *    printer needs to do more substantial DOM manipulations to support that, so\r
 *    some css styles may not be preserved.\r
 * </ol>\r
 * That's it.  I wanted to keep the API as simple as possible, so there's no\r
 * need to specify which language the code is in, but if you wish, you can add\r
 * another class to the {@code <pre>} or {@code <code>} element to specify the\r
 * language, as in {@code <pre class="prettyprint lang-java">}.  Any class that\r
 * starts with "lang-" followed by a file extension, specifies the file type.\r
 * See the "lang-*.js" files in this directory for code that implements\r
 * per-language file handlers.\r
 * <p>\r
 * Change log:<br>\r
 * cbeust, 2006/08/22\r
 * <blockquote>\r
 *   Java annotations (start with "@") are now captured as literals ("lit")\r
 * </blockquote>\r
 * @requires console\r
 */\r
\r
// JSLint declarations\r
/*global console, document, navigator, setTimeout, window, define */\r
\r
/** @define {boolean} */\r
var IN_GLOBAL_SCOPE = true;\r
\r
/**\r
 * Split {@code prettyPrint} into multiple timeouts so as not to interfere with\r
 * UI events.\r
 * If set to {@code false}, {@code prettyPrint()} is synchronous.\r
 */\r
window['PR_SHOULD_USE_CONTINUATION'] = true;\r
\r
/**\r
 * Pretty print a chunk of code.\r
 * @param {string} sourceCodeHtml The HTML to pretty print.\r
 * @param {string} opt_langExtension The language name to use.\r
 *     Typically, a filename extension like 'cpp' or 'java'.\r
 * @param {number|boolean} opt_numberLines True to number lines,\r
 *     or the 1-indexed number of the first line in sourceCodeHtml.\r
 * @return {string} code as html, but prettier\r
 */\r
var prettyPrintOne;\r
/**\r
 * Find all the {@code <pre>} and {@code <code>} tags in the DOM with\r
 * {@code class=prettyprint} and prettify them.\r
 *\r
 * @param {Function} opt_whenDone called when prettifying is done.\r
 * @param {HTMLElement|HTMLDocument} opt_root an element or document\r
 *   containing all the elements to pretty print.\r
 *   Defaults to {@code document.body}.\r
 */\r
var prettyPrint;\r
\r
\r
(function () {\r
  var win = window;\r
  // Keyword lists for various languages.\r
  // We use things that coerce to strings to make them compact when minified\r
  // and to defeat aggressive optimizers that fold large string constants.\r
  var FLOW_CONTROL_KEYWORDS = ["break,continue,do,else,for,if,return,while"];\r
  var C_KEYWORDS = [FLOW_CONTROL_KEYWORDS,"auto,case,char,const,default," + \r
      "double,enum,extern,float,goto,inline,int,long,register,short,signed," +\r
      "sizeof,static,struct,switch,typedef,union,unsigned,void,volatile"];\r
  var COMMON_KEYWORDS = [C_KEYWORDS,"catch,class,delete,false,import," +\r
      "new,operator,private,protected,public,this,throw,true,try,typeof"];\r
  var CPP_KEYWORDS = [COMMON_KEYWORDS,"alignof,align_union,asm,axiom,bool," +\r
      "concept,concept_map,const_cast,constexpr,decltype,delegate," +\r
      "dynamic_cast,explicit,export,friend,generic,late_check," +\r
      "mutable,namespace,nullptr,property,reinterpret_cast,static_assert," +\r
      "static_cast,template,typeid,typename,using,virtual,where"];\r
  var JAVA_KEYWORDS = [COMMON_KEYWORDS,\r
      "abstract,assert,boolean,byte,extends,final,finally,implements,import," +\r
      "instanceof,interface,null,native,package,strictfp,super,synchronized," +\r
      "throws,transient"];\r
  var CSHARP_KEYWORDS = [JAVA_KEYWORDS,\r
      "as,base,by,checked,decimal,delegate,descending,dynamic,event," +\r
      "fixed,foreach,from,group,implicit,in,internal,into,is,let," +\r
      "lock,object,out,override,orderby,params,partial,readonly,ref,sbyte," +\r
      "sealed,stackalloc,string,select,uint,ulong,unchecked,unsafe,ushort," +\r
      "var,virtual,where"];\r
  var COFFEE_KEYWORDS = "all,and,by,catch,class,else,extends,false,finally," +\r
      "for,if,in,is,isnt,loop,new,no,not,null,of,off,on,or,return,super,then," +\r
      "throw,true,try,unless,until,when,while,yes";\r
  var JSCRIPT_KEYWORDS = [COMMON_KEYWORDS,\r
      "debugger,eval,export,function,get,null,set,undefined,var,with," +\r
      "Infinity,NaN"];\r
  var PERL_KEYWORDS = "caller,delete,die,do,dump,elsif,eval,exit,foreach,for," +\r
      "goto,if,import,last,local,my,next,no,our,print,package,redo,require," +\r
      "sub,undef,unless,until,use,wantarray,while,BEGIN,END";\r
  var PYTHON_KEYWORDS = [FLOW_CONTROL_KEYWORDS, "and,as,assert,class,def,del," +\r
      "elif,except,exec,finally,from,global,import,in,is,lambda," +\r
      "nonlocal,not,or,pass,print,raise,try,with,yield," +\r
      "False,True,None"];\r
  var RUBY_KEYWORDS = [FLOW_CONTROL_KEYWORDS, "alias,and,begin,case,class," +\r
      "def,defined,elsif,end,ensure,false,in,module,next,nil,not,or,redo," +\r
      "rescue,retry,self,super,then,true,undef,unless,until,when,yield," +\r
      "BEGIN,END"];\r
   var RUST_KEYWORDS = [FLOW_CONTROL_KEYWORDS, "as,assert,const,copy,drop," +\r
      "enum,extern,fail,false,fn,impl,let,log,loop,match,mod,move,mut,priv," +\r
      "pub,pure,ref,self,static,struct,true,trait,type,unsafe,use"];\r
  var SH_KEYWORDS = [FLOW_CONTROL_KEYWORDS, "case,done,elif,esac,eval,fi," +\r
      "function,in,local,set,then,until"];\r
  var ALL_KEYWORDS = [\r
      CPP_KEYWORDS, CSHARP_KEYWORDS, JSCRIPT_KEYWORDS, PERL_KEYWORDS,\r
      PYTHON_KEYWORDS, RUBY_KEYWORDS, SH_KEYWORDS];\r
  var C_TYPES = /^(DIR|FILE|vector|(de|priority_)?queue|list|stack|(const_)?iterator|(multi)?(set|map)|bitset|u?(int|float)\d*)\b/;\r
\r
  // token style names.  correspond to css classes\r
  /**\r
   * token style for a string literal\r
   * @const\r
   */\r
  var PR_STRING = 'str';\r
  /**\r
   * token style for a keyword\r
   * @const\r
   */\r
  var PR_KEYWORD = 'kwd';\r
  /**\r
   * token style for a comment\r
   * @const\r
   */\r
  var PR_COMMENT = 'com';\r
  /**\r
   * token style for a type\r
   * @const\r
   */\r
  var PR_TYPE = 'typ';\r
  /**\r
   * token style for a literal value.  e.g. 1, null, true.\r
   * @const\r
   */\r
  var PR_LITERAL = 'lit';\r
  /**\r
   * token style for a punctuation string.\r
   * @const\r
   */\r
  var PR_PUNCTUATION = 'pun';\r
  /**\r
   * token style for plain text.\r
   * @const\r
   */\r
  var PR_PLAIN = 'pln';\r
\r
  /**\r
   * token style for an sgml tag.\r
   * @const\r
   */\r
  var PR_TAG = 'tag';\r
  /**\r
   * token style for a markup declaration such as a DOCTYPE.\r
   * @const\r
   */\r
  var PR_DECLARATION = 'dec';\r
  /**\r
   * token style for embedded source.\r
   * @const\r
   */\r
  var PR_SOURCE = 'src';\r
  /**\r
   * token style for an sgml attribute name.\r
   * @const\r
   */\r
  var PR_ATTRIB_NAME = 'atn';\r
  /**\r
   * token style for an sgml attribute value.\r
   * @const\r
   */\r
  var PR_ATTRIB_VALUE = 'atv';\r
\r
  /**\r
   * A class that indicates a section of markup that is not code, e.g. to allow\r
   * embedding of line numbers within code listings.\r
   * @const\r
   */\r
  var PR_NOCODE = 'nocode';\r
\r
  \r
  \r
  /**\r
   * A set of tokens that can precede a regular expression literal in\r
   * javascript\r
   * http://web.archive.org/web/20070717142515/http://www.mozilla.org/js/language/js20/rationale/syntax.html\r
   * has the full list, but I've removed ones that might be problematic when\r
   * seen in languages that don't support regular expression literals.\r
   *\r
   * <p>Specifically, I've removed any keywords that can't precede a regexp\r
   * literal in a syntactically legal javascript program, and I've removed the\r
   * "in" keyword since it's not a keyword in many languages, and might be used\r
   * as a count of inches.\r
   *\r
   * <p>The link above does not accurately describe EcmaScript rules since\r
   * it fails to distinguish between (a=++/b/i) and (a++/b/i) but it works\r
   * very well in practice.\r
   *\r
   * @private\r
   * @const\r
   */\r
  var REGEXP_PRECEDER_PATTERN = '(?:^^\\.?|[+-]|[!=]=?=?|\\#|%=?|&&?=?|\\(|\\*=?|[+\\-]=|->|\\/=?|::?|<<?=?|>>?>?=?|,|;|\\?|@|\\[|~|{|\\^\\^?=?|\\|\\|?=?|break|case|continue|delete|do|else|finally|instanceof|return|throw|try|typeof)\\s*';\r
  \r
  // CAVEAT: this does not properly handle the case where a regular\r
  // expression immediately follows another since a regular expression may\r
  // have flags for case-sensitivity and the like.  Having regexp tokens\r
  // adjacent is not valid in any language I'm aware of, so I'm punting.\r
  // TODO: maybe style special characters inside a regexp as punctuation.\r
\r
  /**\r
   * Given a group of {@link RegExp}s, returns a {@code RegExp} that globally\r
   * matches the union of the sets of strings matched by the input RegExp.\r
   * Since it matches globally, if the input strings have a start-of-input\r
   * anchor (/^.../), it is ignored for the purposes of unioning.\r
   * @param {Array.<RegExp>} regexs non multiline, non-global regexs.\r
   * @return {RegExp} a global regex.\r
   */\r
  function combinePrefixPatterns(regexs) {\r
    var capturedGroupIndex = 0;\r
  \r
    var needToFoldCase = false;\r
    var ignoreCase = false;\r
    for (var i = 0, n = regexs.length; i < n; ++i) {\r
      var regex = regexs[i];\r
      if (regex.ignoreCase) {\r
        ignoreCase = true;\r
      } else if (/[a-z]/i.test(regex.source.replace(\r
                     /\\u[0-9a-f]{4}|\\x[0-9a-f]{2}|\\[^ux]/gi, ''))) {\r
        needToFoldCase = true;\r
        ignoreCase = false;\r
        break;\r
      }\r
    }\r
  \r
    var escapeCharToCodeUnit = {\r
      'b': 8,\r
      't': 9,\r
      'n': 0xa,\r
      'v': 0xb,\r
      'f': 0xc,\r
      'r': 0xd\r
    };\r
  \r
    function decodeEscape(charsetPart) {\r
      var cc0 = charsetPart.charCodeAt(0);\r
      if (cc0 !== 92 /* \\ */) {\r
        return cc0;\r
      }\r
      var c1 = charsetPart.charAt(1);\r
      cc0 = escapeCharToCodeUnit[c1];\r
      if (cc0) {\r
        return cc0;\r
      } else if ('0' <= c1 && c1 <= '7') {\r
        return parseInt(charsetPart.substring(1), 8);\r
      } else if (c1 === 'u' || c1 === 'x') {\r
        return parseInt(charsetPart.substring(2), 16);\r
      } else {\r
        return charsetPart.charCodeAt(1);\r
      }\r
    }\r
  \r
    function encodeEscape(charCode) {\r
      if (charCode < 0x20) {\r
        return (charCode < 0x10 ? '\\x0' : '\\x') + charCode.toString(16);\r
      }\r
      var ch = String.fromCharCode(charCode);\r
      return (ch === '\\' || ch === '-' || ch === ']' || ch === '^')\r
          ? "\\" + ch : ch;\r
    }\r
  \r
    function caseFoldCharset(charSet) {\r
      var charsetParts = charSet.substring(1, charSet.length - 1).match(\r
          new RegExp(\r
              '\\\\u[0-9A-Fa-f]{4}'\r
              + '|\\\\x[0-9A-Fa-f]{2}'\r
              + '|\\\\[0-3][0-7]{0,2}'\r
              + '|\\\\[0-7]{1,2}'\r
              + '|\\\\[\\s\\S]'\r
              + '|-'\r
              + '|[^-\\\\]',\r
              'g'));\r
      var ranges = [];\r
      var inverse = charsetParts[0] === '^';\r
  \r
      var out = ['['];\r
      if (inverse) { out.push('^'); }\r
  \r
      for (var i = inverse ? 1 : 0, n = charsetParts.length; i < n; ++i) {\r
        var p = charsetParts[i];\r
        if (/\\[bdsw]/i.test(p)) {  // Don't muck with named groups.\r
          out.push(p);\r
        } else {\r
          var start = decodeEscape(p);\r
          var end;\r
          if (i + 2 < n && '-' === charsetParts[i + 1]) {\r
            end = decodeEscape(charsetParts[i + 2]);\r
            i += 2;\r
          } else {\r
            end = start;\r
          }\r
          ranges.push([start, end]);\r
          // If the range might intersect letters, then expand it.\r
          // This case handling is too simplistic.\r
          // It does not deal with non-latin case folding.\r
          // It works for latin source code identifiers though.\r
          if (!(end < 65 || start > 122)) {\r
            if (!(end < 65 || start > 90)) {\r
              ranges.push([Math.max(65, start) | 32, Math.min(end, 90) | 32]);\r
            }\r
            if (!(end < 97 || start > 122)) {\r
              ranges.push([Math.max(97, start) & ~32, Math.min(end, 122) & ~32]);\r
            }\r
          }\r
        }\r
      }\r
  \r
      // [[1, 10], [3, 4], [8, 12], [14, 14], [16, 16], [17, 17]]\r
      // -> [[1, 12], [14, 14], [16, 17]]\r
      ranges.sort(function (a, b) { return (a[0] - b[0]) || (b[1]  - a[1]); });\r
      var consolidatedRanges = [];\r
      var lastRange = [];\r
      for (var i = 0; i < ranges.length; ++i) {\r
        var range = ranges[i];\r
        if (range[0] <= lastRange[1] + 1) {\r
          lastRange[1] = Math.max(lastRange[1], range[1]);\r
        } else {\r
          consolidatedRanges.push(lastRange = range);\r
        }\r
      }\r
  \r
      for (var i = 0; i < consolidatedRanges.length; ++i) {\r
        var range = consolidatedRanges[i];\r
        out.push(encodeEscape(range[0]));\r
        if (range[1] > range[0]) {\r
          if (range[1] + 1 > range[0]) { out.push('-'); }\r
          out.push(encodeEscape(range[1]));\r
        }\r
      }\r
      out.push(']');\r
      return out.join('');\r
    }\r
  \r
    function allowAnywhereFoldCaseAndRenumberGroups(regex) {\r
      // Split into character sets, escape sequences, punctuation strings\r
      // like ('(', '(?:', ')', '^'), and runs of characters that do not\r
      // include any of the above.\r
      var parts = regex.source.match(\r
          new RegExp(\r
              '(?:'\r
              + '\\[(?:[^\\x5C\\x5D]|\\\\[\\s\\S])*\\]'  // a character set\r
              + '|\\\\u[A-Fa-f0-9]{4}'  // a unicode escape\r
              + '|\\\\x[A-Fa-f0-9]{2}'  // a hex escape\r
              + '|\\\\[0-9]+'  // a back-reference or octal escape\r
              + '|\\\\[^ux0-9]'  // other escape sequence\r
              + '|\\(\\?[:!=]'  // start of a non-capturing group\r
              + '|[\\(\\)\\^]'  // start/end of a group, or line start\r
              + '|[^\\x5B\\x5C\\(\\)\\^]+'  // run of other characters\r
              + ')',\r
              'g'));\r
      var n = parts.length;\r
  \r
      // Maps captured group numbers to the number they will occupy in\r
      // the output or to -1 if that has not been determined, or to\r
      // undefined if they need not be capturing in the output.\r
      var capturedGroups = [];\r
  \r
      // Walk over and identify back references to build the capturedGroups\r
      // mapping.\r
      for (var i = 0, groupIndex = 0; i < n; ++i) {\r
        var p = parts[i];\r
        if (p === '(') {\r
          // groups are 1-indexed, so max group index is count of '('\r
          ++groupIndex;\r
        } else if ('\\' === p.charAt(0)) {\r
          var decimalValue = +p.substring(1);\r
          if (decimalValue) {\r
            if (decimalValue <= groupIndex) {\r
              capturedGroups[decimalValue] = -1;\r
            } else {\r
              // Replace with an unambiguous escape sequence so that\r
              // an octal escape sequence does not turn into a backreference\r
              // to a capturing group from an earlier regex.\r
              parts[i] = encodeEscape(decimalValue);\r
            }\r
          }\r
        }\r
      }\r
  \r
      // Renumber groups and reduce capturing groups to non-capturing groups\r
      // where possible.\r
      for (var i = 1; i < capturedGroups.length; ++i) {\r
        if (-1 === capturedGroups[i]) {\r
          capturedGroups[i] = ++capturedGroupIndex;\r
        }\r
      }\r
      for (var i = 0, groupIndex = 0; i < n; ++i) {\r
        var p = parts[i];\r
        if (p === '(') {\r
          ++groupIndex;\r
          if (!capturedGroups[groupIndex]) {\r
            parts[i] = '(?:';\r
          }\r
        } else if ('\\' === p.charAt(0)) {\r
          var decimalValue = +p.substring(1);\r
          if (decimalValue && decimalValue <= groupIndex) {\r
            parts[i] = '\\' + capturedGroups[decimalValue];\r
          }\r
        }\r
      }\r
  \r
      // Remove any prefix anchors so that the output will match anywhere.\r
      // ^^ really does mean an anchored match though.\r
      for (var i = 0; i < n; ++i) {\r
        if ('^' === parts[i] && '^' !== parts[i + 1]) { parts[i] = ''; }\r
      }\r
  \r
      // Expand letters to groups to handle mixing of case-sensitive and\r
      // case-insensitive patterns if necessary.\r
      if (regex.ignoreCase && needToFoldCase) {\r
        for (var i = 0; i < n; ++i) {\r
          var p = parts[i];\r
          var ch0 = p.charAt(0);\r
          if (p.length >= 2 && ch0 === '[') {\r
            parts[i] = caseFoldCharset(p);\r
          } else if (ch0 !== '\\') {\r
            // TODO: handle letters in numeric escapes.\r
            parts[i] = p.replace(\r
                /[a-zA-Z]/g,\r
                function (ch) {\r
                  var cc = ch.charCodeAt(0);\r
                  return '[' + String.fromCharCode(cc & ~32, cc | 32) + ']';\r
                });\r
          }\r
        }\r
      }\r
  \r
      return parts.join('');\r
    }\r
  \r
    var rewritten = [];\r
    for (var i = 0, n = regexs.length; i < n; ++i) {\r
      var regex = regexs[i];\r
      if (regex.global || regex.multiline) { throw new Error('' + regex); }\r
      rewritten.push(\r
          '(?:' + allowAnywhereFoldCaseAndRenumberGroups(regex) + ')');\r
    }\r
  \r
    return new RegExp(rewritten.join('|'), ignoreCase ? 'gi' : 'g');\r
  }\r
\r
  /**\r
   * Split markup into a string of source code and an array mapping ranges in\r
   * that string to the text nodes in which they appear.\r
   *\r
   * <p>\r
   * The HTML DOM structure:</p>\r
   * <pre>\r
   * (Element   "p"\r
   *   (Element "b"\r
   *     (Text  "print "))       ; #1\r
   *   (Text    "'Hello '")      ; #2\r
   *   (Element "br")            ; #3\r
   *   (Text    "  + 'World';")) ; #4\r
   * </pre>\r
   * <p>\r
   * corresponds to the HTML\r
   * {@code <p><b>print </b>'Hello '<br>  + 'World';</p>}.</p>\r
   *\r
   * <p>\r
   * It will produce the output:</p>\r
   * <pre>\r
   * {\r
   *   sourceCode: "print 'Hello '\n  + 'World';",\r
   *   //                     1          2\r
   *   //           012345678901234 5678901234567\r
   *   spans: [0, #1, 6, #2, 14, #3, 15, #4]\r
   * }\r
   * </pre>\r
   * <p>\r
   * where #1 is a reference to the {@code "print "} text node above, and so\r
   * on for the other text nodes.\r
   * </p>\r
   *\r
   * <p>\r
   * The {@code} spans array is an array of pairs.  Even elements are the start\r
   * indices of substrings, and odd elements are the text nodes (or BR elements)\r
   * that contain the text for those substrings.\r
   * Substrings continue until the next index or the end of the source.\r
   * </p>\r
   *\r
   * @param {Node} node an HTML DOM subtree containing source-code.\r
   * @param {boolean} isPreformatted true if white-space in text nodes should\r
   *    be considered significant.\r
   * @return {Object} source code and the text nodes in which they occur.\r
   */\r
  function extractSourceSpans(node, isPreformatted) {\r
    var nocode = /(?:^|\s)nocode(?:\s|$)/;\r
  \r
    var chunks = [];\r
    var length = 0;\r
    var spans = [];\r
    var k = 0;\r
  \r
    function walk(node) {\r
      var type = node.nodeType;\r
      if (type == 1) {  // Element\r
        if (nocode.test(node.className)) { return; }\r
        for (var child = node.firstChild; child; child = child.nextSibling) {\r
          walk(child);\r
        }\r
        var nodeName = node.nodeName.toLowerCase();\r
        if ('br' === nodeName || 'li' === nodeName) {\r
          chunks[k] = '\n';\r
          spans[k << 1] = length++;\r
          spans[(k++ << 1) | 1] = node;\r
        }\r
      } else if (type == 3 || type == 4) {  // Text\r
        var text = node.nodeValue;\r
        if (text.length) {\r
          if (!isPreformatted) {\r
            text = text.replace(/[ \t\r\n]+/g, ' ');\r
          } else {\r
            text = text.replace(/\r\n?/g, '\n');  // Normalize newlines.\r
          }\r
          // TODO: handle tabs here?\r
          chunks[k] = text;\r
          spans[k << 1] = length;\r
          length += text.length;\r
          spans[(k++ << 1) | 1] = node;\r
        }\r
      }\r
    }\r
  \r
    walk(node);\r
  \r
    return {\r
      sourceCode: chunks.join('').replace(/\n$/, ''),\r
      spans: spans\r
    };\r
  }\r
\r
  /**\r
   * Apply the given language handler to sourceCode and add the resulting\r
   * decorations to out.\r
   * @param {number} basePos the index of sourceCode within the chunk of source\r
   *    whose decorations are already present on out.\r
   */\r
  function appendDecorations(basePos, sourceCode, langHandler, out) {\r
    if (!sourceCode) { return; }\r
    var job = {\r
      sourceCode: sourceCode,\r
      basePos: basePos\r
    };\r
    langHandler(job);\r
    out.push.apply(out, job.decorations);\r
  }\r
\r
  var notWs = /\S/;\r
\r
  /**\r
   * Given an element, if it contains only one child element and any text nodes\r
   * it contains contain only space characters, return the sole child element.\r
   * Otherwise returns undefined.\r
   * <p>\r
   * This is meant to return the CODE element in {@code <pre><code ...>} when\r
   * there is a single child element that contains all the non-space textual\r
   * content, but not to return anything where there are multiple child elements\r
   * as in {@code <pre><code>...</code><code>...</code></pre>} or when there\r
   * is textual content.\r
   */\r
  function childContentWrapper(element) {\r
    var wrapper = undefined;\r
    for (var c = element.firstChild; c; c = c.nextSibling) {\r
      var type = c.nodeType;\r
      wrapper = (type === 1)  // Element Node\r
          ? (wrapper ? element : c)\r
          : (type === 3)  // Text Node\r
          ? (notWs.test(c.nodeValue) ? element : wrapper)\r
          : wrapper;\r
    }\r
    return wrapper === element ? undefined : wrapper;\r
  }\r
\r
  /** Given triples of [style, pattern, context] returns a lexing function,\r
    * The lexing function interprets the patterns to find token boundaries and\r
    * returns a decoration list of the form\r
    * [index_0, style_0, index_1, style_1, ..., index_n, style_n]\r
    * where index_n is an index into the sourceCode, and style_n is a style\r
    * constant like PR_PLAIN.  index_n-1 <= index_n, and style_n-1 applies to\r
    * all characters in sourceCode[index_n-1:index_n].\r
    *\r
    * The stylePatterns is a list whose elements have the form\r
    * [style : string, pattern : RegExp, DEPRECATED, shortcut : string].\r
    *\r
    * Style is a style constant like PR_PLAIN, or can be a string of the\r
    * form 'lang-FOO', where FOO is a language extension describing the\r
    * language of the portion of the token in $1 after pattern executes.\r
    * E.g., if style is 'lang-lisp', and group 1 contains the text\r
    * '(hello (world))', then that portion of the token will be passed to the\r
    * registered lisp handler for formatting.\r
    * The text before and after group 1 will be restyled using this decorator\r
    * so decorators should take care that this doesn't result in infinite\r
    * recursion.  For example, the HTML lexer rule for SCRIPT elements looks\r
    * something like ['lang-js', /<[s]cript>(.+?)<\/script>/].  This may match\r
    * '<script>foo()<\/script>', which would cause the current decorator to\r
    * be called with '<script>' which would not match the same rule since\r
    * group 1 must not be empty, so it would be instead styled as PR_TAG by\r
    * the generic tag rule.  The handler registered for the 'js' extension would\r
    * then be called with 'foo()', and finally, the current decorator would\r
    * be called with '<\/script>' which would not match the original rule and\r
    * so the generic tag rule would identify it as a tag.\r
    *\r
    * Pattern must only match prefixes, and if it matches a prefix, then that\r
    * match is considered a token with the same style.\r
    *\r
    * Context is applied to the last non-whitespace, non-comment token\r
    * recognized.\r
    *\r
    * Shortcut is an optional string of characters, any of which, if the first\r
    * character, gurantee that this pattern and only this pattern matches.\r
    *\r
    * @param {Array} shortcutStylePatterns patterns that always start with\r
    *   a known character.  Must have a shortcut string.\r
    * @param {Array} fallthroughStylePatterns patterns that will be tried in\r
    *   order if the shortcut ones fail.  May have shortcuts.\r
    *\r
    * @return {function (Object)} a\r
    *   function that takes source code and returns a list of decorations.\r
    */\r
  function createSimpleLexer(shortcutStylePatterns, fallthroughStylePatterns) {\r
    var shortcuts = {};\r
    var tokenizer;\r
    (function () {\r
      var allPatterns = shortcutStylePatterns.concat(fallthroughStylePatterns);\r
      var allRegexs = [];\r
      var regexKeys = {};\r
      for (var i = 0, n = allPatterns.length; i < n; ++i) {\r
        var patternParts = allPatterns[i];\r
        var shortcutChars = patternParts[3];\r
        if (shortcutChars) {\r
          for (var c = shortcutChars.length; --c >= 0;) {\r
            shortcuts[shortcutChars.charAt(c)] = patternParts;\r
          }\r
        }\r
        var regex = patternParts[1];\r
        var k = '' + regex;\r
        if (!regexKeys.hasOwnProperty(k)) {\r
          allRegexs.push(regex);\r
          regexKeys[k] = null;\r
        }\r
      }\r
      allRegexs.push(/[\0-\uffff]/);\r
      tokenizer = combinePrefixPatterns(allRegexs);\r
    })();\r
\r
    var nPatterns = fallthroughStylePatterns.length;\r
\r
    /**\r
     * Lexes job.sourceCode and produces an output array job.decorations of\r
     * style classes preceded by the position at which they start in\r
     * job.sourceCode in order.\r
     *\r
     * @param {Object} job an object like <pre>{\r
     *    sourceCode: {string} sourceText plain text,\r
     *    basePos: {int} position of job.sourceCode in the larger chunk of\r
     *        sourceCode.\r
     * }</pre>\r
     */\r
    var decorate = function (job) {\r
      var sourceCode = job.sourceCode, basePos = job.basePos;\r
      /** Even entries are positions in source in ascending order.  Odd enties\r
        * are style markers (e.g., PR_COMMENT) that run from that position until\r
        * the end.\r
        * @type {Array.<number|string>}\r
        */\r
      var decorations = [basePos, PR_PLAIN];\r
      var pos = 0;  // index into sourceCode\r
      var tokens = sourceCode.match(tokenizer) || [];\r
      var styleCache = {};\r
\r
      for (var ti = 0, nTokens = tokens.length; ti < nTokens; ++ti) {\r
        var token = tokens[ti];\r
        var style = styleCache[token];\r
        var match = void 0;\r
\r
        var isEmbedded;\r
        if (typeof style === 'string') {\r
          isEmbedded = false;\r
        } else {\r
          var patternParts = shortcuts[token.charAt(0)];\r
          if (patternParts) {\r
            match = token.match(patternParts[1]);\r
            style = patternParts[0];\r
          } else {\r
            for (var i = 0; i < nPatterns; ++i) {\r
              patternParts = fallthroughStylePatterns[i];\r
              match = token.match(patternParts[1]);\r
              if (match) {\r
                style = patternParts[0];\r
                break;\r
              }\r
            }\r
\r
            if (!match) {  // make sure that we make progress\r
              style = PR_PLAIN;\r
            }\r
          }\r
\r
          isEmbedded = style.length >= 5 && 'lang-' === style.substring(0, 5);\r
          if (isEmbedded && !(match && typeof match[1] === 'string')) {\r
            isEmbedded = false;\r
            style = PR_SOURCE;\r
          }\r
\r
          if (!isEmbedded) { styleCache[token] = style; }\r
        }\r
\r
        var tokenStart = pos;\r
        pos += token.length;\r
\r
        if (!isEmbedded) {\r
          decorations.push(basePos + tokenStart, style);\r
        } else {  // Treat group 1 as an embedded block of source code.\r
          var embeddedSource = match[1];\r
          var embeddedSourceStart = token.indexOf(embeddedSource);\r
          var embeddedSourceEnd = embeddedSourceStart + embeddedSource.length;\r
          if (match[2]) {\r
            // If embeddedSource can be blank, then it would match at the\r
            // beginning which would cause us to infinitely recurse on the\r
            // entire token, so we catch the right context in match[2].\r
            embeddedSourceEnd = token.length - match[2].length;\r
            embeddedSourceStart = embeddedSourceEnd - embeddedSource.length;\r
          }\r
          var lang = style.substring(5);\r
          // Decorate the left of the embedded source\r
          appendDecorations(\r
              basePos + tokenStart,\r
              token.substring(0, embeddedSourceStart),\r
              decorate, decorations);\r
          // Decorate the embedded source\r
          appendDecorations(\r
              basePos + tokenStart + embeddedSourceStart,\r
              embeddedSource,\r
              langHandlerForExtension(lang, embeddedSource),\r
              decorations);\r
          // Decorate the right of the embedded section\r
          appendDecorations(\r
              basePos + tokenStart + embeddedSourceEnd,\r
              token.substring(embeddedSourceEnd),\r
              decorate, decorations);\r
        }\r
      }\r
      job.decorations = decorations;\r
    };\r
    return decorate;\r
  }\r
\r
  /** returns a function that produces a list of decorations from source text.\r
    *\r
    * This code treats ", ', and ` as string delimiters, and \ as a string\r
    * escape.  It does not recognize perl's qq() style strings.\r
    * It has no special handling for double delimiter escapes as in basic, or\r
    * the tripled delimiters used in python, but should work on those regardless\r
    * although in those cases a single string literal may be broken up into\r
    * multiple adjacent string literals.\r
    *\r
    * It recognizes C, C++, and shell style comments.\r
    *\r
    * @param {Object} options a set of optional parameters.\r
    * @return {function (Object)} a function that examines the source code\r
    *     in the input job and builds the decoration list.\r
    */\r
  function sourceDecorator(options) {\r
    var shortcutStylePatterns = [], fallthroughStylePatterns = [];\r
    if (options['tripleQuotedStrings']) {\r
      // '''multi-line-string''', 'single-line-string', and double-quoted\r
      shortcutStylePatterns.push(\r
          [PR_STRING,  /^(?:\'\'\'(?:[^\'\\]|\\[\s\S]|\'{1,2}(?=[^\']))*(?:\'\'\'|$)|\"\"\"(?:[^\"\\]|\\[\s\S]|\"{1,2}(?=[^\"]))*(?:\"\"\"|$)|\'(?:[^\\\']|\\[\s\S])*(?:\'|$)|\"(?:[^\\\"]|\\[\s\S])*(?:\"|$))/,\r
           null, '\'"']);\r
    } else if (options['multiLineStrings']) {\r
      // 'multi-line-string', "multi-line-string"\r
      shortcutStylePatterns.push(\r
          [PR_STRING,  /^(?:\'(?:[^\\\']|\\[\s\S])*(?:\'|$)|\"(?:[^\\\"]|\\[\s\S])*(?:\"|$)|\`(?:[^\\\`]|\\[\s\S])*(?:\`|$))/,\r
           null, '\'"`']);\r
    } else {\r
      // 'single-line-string', "single-line-string"\r
      shortcutStylePatterns.push(\r
          [PR_STRING,\r
           /^(?:\'(?:[^\\\'\r\n]|\\.)*(?:\'|$)|\"(?:[^\\\"\r\n]|\\.)*(?:\"|$))/,\r
           null, '"\'']);\r
    }\r
    if (options['verbatimStrings']) {\r
      // verbatim-string-literal production from the C# grammar.  See issue 93.\r
      fallthroughStylePatterns.push(\r
          [PR_STRING, /^@\"(?:[^\"]|\"\")*(?:\"|$)/, null]);\r
    }\r
    var hc = options['hashComments'];\r
    if (hc) {\r
      if (options['cStyleComments']) {\r
        if (hc > 1) {  // multiline hash comments\r
          shortcutStylePatterns.push(\r
              [PR_COMMENT, /^#(?:##(?:[^#]|#(?!##))*(?:###|$)|.*)/, null, '#']);\r
        } else {\r
          // Stop C preprocessor declarations at an unclosed open comment\r
          shortcutStylePatterns.push(\r
              [PR_COMMENT, /^#(?:(?:define|e(?:l|nd)if|else|error|ifn?def|include|line|pragma|undef|warning)\b|[^\r\n]*)/,\r
               null, '#']);\r
        }\r
        // #include <stdio.h>\r
        fallthroughStylePatterns.push(\r
            [PR_STRING,\r
             /^<(?:(?:(?:\.\.\/)*|\/?)(?:[\w-]+(?:\/[\w-]+)+)?[\w-]+\.h(?:h|pp|\+\+)?|[a-z]\w*)>/,\r
             null]);\r
      } else {\r
        shortcutStylePatterns.push([PR_COMMENT, /^#[^\r\n]*/, null, '#']);\r
      }\r
    }\r
    if (options['cStyleComments']) {\r
      fallthroughStylePatterns.push([PR_COMMENT, /^\/\/[^\r\n]*/, null]);\r
      fallthroughStylePatterns.push(\r
          [PR_COMMENT, /^\/\*[\s\S]*?(?:\*\/|$)/, null]);\r
    }\r
    var regexLiterals = options['regexLiterals'];\r
    if (regexLiterals) {\r
      /**\r
       * @const\r
       */\r
      var regexExcls = regexLiterals > 1\r
        ? ''  // Multiline regex literals\r
        : '\n\r';\r
      /**\r
       * @const\r
       */\r
      var regexAny = regexExcls ? '.' : '[\\S\\s]';\r
      /**\r
       * @const\r
       */\r
      var REGEX_LITERAL = (\r
          // A regular expression literal starts with a slash that is\r
          // not followed by * or / so that it is not confused with\r
          // comments.\r
          '/(?=[^/*' + regexExcls + '])'\r
          // and then contains any number of raw characters,\r
          + '(?:[^/\\x5B\\x5C' + regexExcls + ']'\r
          // escape sequences (\x5C),\r
          +    '|\\x5C' + regexAny\r
          // or non-nesting character sets (\x5B\x5D);\r
          +    '|\\x5B(?:[^\\x5C\\x5D' + regexExcls + ']'\r
          +             '|\\x5C' + regexAny + ')*(?:\\x5D|$))+'\r
          // finally closed by a /.\r
          + '/');\r
      fallthroughStylePatterns.push(\r
          ['lang-regex',\r
           RegExp('^' + REGEXP_PRECEDER_PATTERN + '(' + REGEX_LITERAL + ')')\r
           ]);\r
    }\r
\r
    var types = options['types'];\r
    if (types) {\r
      fallthroughStylePatterns.push([PR_TYPE, types]);\r
    }\r
\r
    var keywords = ("" + options['keywords']).replace(/^ | $/g, '');\r
    if (keywords.length) {\r
      fallthroughStylePatterns.push(\r
          [PR_KEYWORD,\r
           new RegExp('^(?:' + keywords.replace(/[\s,]+/g, '|') + ')\\b'),\r
           null]);\r
    }\r
\r
    shortcutStylePatterns.push([PR_PLAIN,       /^\s+/, null, ' \r\n\t\xA0']);\r
\r
    var punctuation =\r
      // The Bash man page says\r
\r
      // A word is a sequence of characters considered as a single\r
      // unit by GRUB. Words are separated by metacharacters,\r
      // which are the following plus space, tab, and newline: { }\r
      // | & $ ; < >\r
      // ...\r
      \r
      // A word beginning with # causes that word and all remaining\r
      // characters on that line to be ignored.\r
\r
      // which means that only a '#' after /(?:^|[{}|&$;<>\s])/ starts a\r
      // comment but empirically\r
      // $ echo {#}\r
      // {#}\r
      // $ echo \$#\r
      // $#\r
      // $ echo }#\r
      // }#\r
\r
      // so /(?:^|[|&;<>\s])/ is more appropriate.\r
\r
      // http://gcc.gnu.org/onlinedocs/gcc-2.95.3/cpp_1.html#SEC3\r
      // suggests that this definition is compatible with a\r
      // default mode that tries to use a single token definition\r
      // to recognize both bash/python style comments and C\r
      // preprocessor directives.\r
\r
      // This definition of punctuation does not include # in the list of\r
      // follow-on exclusions, so # will not be broken before if preceeded\r
      // by a punctuation character.  We could try to exclude # after\r
      // [|&;<>] but that doesn't seem to cause many major problems.\r
      // If that does turn out to be a problem, we should change the below\r
      // when hc is truthy to include # in the run of punctuation characters\r
      // only when not followint [|&;<>].\r
      '^.[^\\s\\w.$@\'"`/\\\\]*';\r
    if (options['regexLiterals']) {\r
      punctuation += '(?!\s*\/)';\r
    }\r
\r
    fallthroughStylePatterns.push(\r
        // TODO(mikesamuel): recognize non-latin letters and numerals in idents\r
        [PR_LITERAL,     /^@[a-z_$][a-z_$@0-9]*/i, null],\r
        [PR_TYPE,        /^(?:[@_]?[A-Z]+[a-z][A-Za-z_$@0-9]*|\w+_t\b)/, null],\r
        [PR_PLAIN,       /^[a-z_$][a-z_$@0-9]*/i, null],\r
        [PR_LITERAL,\r
         new RegExp(\r
             '^(?:'\r
             // A hex number\r
             + '0x[a-f0-9]+'\r
             // or an octal or decimal number,\r
             + '|(?:\\d(?:_\\d+)*\\d*(?:\\.\\d*)?|\\.\\d\\+)'\r
             // possibly in scientific notation\r
             + '(?:e[+\\-]?\\d+)?'\r
             + ')'\r
             // with an optional modifier like UL for unsigned long\r
             + '[a-z]*', 'i'),\r
         null, '0123456789'],\r
        // Don't treat escaped quotes in bash as starting strings.\r
        // See issue 144.\r
        [PR_PLAIN,       /^\\[\s\S]?/, null],\r
        [PR_PUNCTUATION, new RegExp(punctuation), null]);\r
\r
    return createSimpleLexer(shortcutStylePatterns, fallthroughStylePatterns);\r
  }\r
\r
  var decorateSource = sourceDecorator({\r
        'keywords': ALL_KEYWORDS,\r
        'hashComments': true,\r
        'cStyleComments': true,\r
        'multiLineStrings': true,\r
        'regexLiterals': true\r
      });\r
\r
  /**\r
   * Given a DOM subtree, wraps it in a list, and puts each line into its own\r
   * list item.\r
   *\r
   * @param {Node} node modified in place.  Its content is pulled into an\r
   *     HTMLOListElement, and each line is moved into a separate list item.\r
   *     This requires cloning elements, so the input might not have unique\r
   *     IDs after numbering.\r
   * @param {boolean} isPreformatted true iff white-space in text nodes should\r
   *     be treated as significant.\r
   */\r
  function numberLines(node, opt_startLineNum, isPreformatted) {\r
    var nocode = /(?:^|\s)nocode(?:\s|$)/;\r
    var lineBreak = /\r\n?|\n/;\r
  \r
    var document = node.ownerDocument;\r
  \r
    var li = document.createElement('li');\r
    while (node.firstChild) {\r
      li.appendChild(node.firstChild);\r
    }\r
    // An array of lines.  We split below, so this is initialized to one\r
    // un-split line.\r
    var listItems = [li];\r
  \r
    function walk(node) {\r
      var type = node.nodeType;\r
      if (type == 1 && !nocode.test(node.className)) {  // Element\r
        if ('br' === node.nodeName) {\r
          breakAfter(node);\r
          // Discard the <BR> since it is now flush against a </LI>.\r
          if (node.parentNode) {\r
            node.parentNode.removeChild(node);\r
          }\r
        } else {\r
          for (var child = node.firstChild; child; child = child.nextSibling) {\r
            walk(child);\r
          }\r
        }\r
      } else if ((type == 3 || type == 4) && isPreformatted) {  // Text\r
        var text = node.nodeValue;\r
        var match = text.match(lineBreak);\r
        if (match) {\r
          var firstLine = text.substring(0, match.index);\r
          node.nodeValue = firstLine;\r
          var tail = text.substring(match.index + match[0].length);\r
          if (tail) {\r
            var parent = node.parentNode;\r
            parent.insertBefore(\r
              document.createTextNode(tail), node.nextSibling);\r
          }\r
          breakAfter(node);\r
          if (!firstLine) {\r
            // Don't leave blank text nodes in the DOM.\r
            node.parentNode.removeChild(node);\r
          }\r
        }\r
      }\r
    }\r
  \r
    // Split a line after the given node.\r
    function breakAfter(lineEndNode) {\r
      // If there's nothing to the right, then we can skip ending the line\r
      // here, and move root-wards since splitting just before an end-tag\r
      // would require us to create a bunch of empty copies.\r
      while (!lineEndNode.nextSibling) {\r
        lineEndNode = lineEndNode.parentNode;\r
        if (!lineEndNode) { return; }\r
      }\r
  \r
      function breakLeftOf(limit, copy) {\r
        // Clone shallowly if this node needs to be on both sides of the break.\r
        var rightSide = copy ? limit.cloneNode(false) : limit;\r
        var parent = limit.parentNode;\r
        if (parent) {\r
          // We clone the parent chain.\r
          // This helps us resurrect important styling elements that cross lines.\r
          // E.g. in <i>Foo<br>Bar</i>\r
          // should be rewritten to <li><i>Foo</i></li><li><i>Bar</i></li>.\r
          var parentClone = breakLeftOf(parent, 1);\r
          // Move the clone and everything to the right of the original\r
          // onto the cloned parent.\r
          var next = limit.nextSibling;\r
          parentClone.appendChild(rightSide);\r
          for (var sibling = next; sibling; sibling = next) {\r
            next = sibling.nextSibling;\r
            parentClone.appendChild(sibling);\r
          }\r
        }\r
        return rightSide;\r
      }\r
  \r
      var copiedListItem = breakLeftOf(lineEndNode.nextSibling, 0);\r
  \r
      // Walk the parent chain until we reach an unattached LI.\r
      for (var parent;\r
           // Check nodeType since IE invents document fragments.\r
           (parent = copiedListItem.parentNode) && parent.nodeType === 1;) {\r
        copiedListItem = parent;\r
      }\r
      // Put it on the list of lines for later processing.\r
      listItems.push(copiedListItem);\r
    }\r
  \r
    // Split lines while there are lines left to split.\r
    for (var i = 0;  // Number of lines that have been split so far.\r
         i < listItems.length;  // length updated by breakAfter calls.\r
         ++i) {\r
      walk(listItems[i]);\r
    }\r
  \r
    // Make sure numeric indices show correctly.\r
    if (opt_startLineNum === (opt_startLineNum|0)) {\r
      listItems[0].setAttribute('value', opt_startLineNum);\r
    }\r
  \r
    var ol = document.createElement('ol');\r
    ol.className = 'linenums';\r
    var offset = Math.max(0, ((opt_startLineNum - 1 /* zero index */)) | 0) || 0;\r
    for (var i = 0, n = listItems.length; i < n; ++i) {\r
      li = listItems[i];\r
      // Stick a class on the LIs so that stylesheets can\r
      // color odd/even rows, or any other row pattern that\r
      // is co-prime with 10.\r
      li.className = 'L' + ((i + offset) % 10);\r
      if (!li.firstChild) {\r
        li.appendChild(document.createTextNode('\xA0'));\r
      }\r
      ol.appendChild(li);\r
    }\r
  \r
    node.appendChild(ol);\r
  }\r
  /**\r
   * Breaks {@code job.sourceCode} around style boundaries in\r
   * {@code job.decorations} and modifies {@code job.sourceNode} in place.\r
   * @param {Object} job like <pre>{\r
   *    sourceCode: {string} source as plain text,\r
   *    sourceNode: {HTMLElement} the element containing the source,\r
   *    spans: {Array.<number|Node>} alternating span start indices into source\r
   *       and the text node or element (e.g. {@code <BR>}) corresponding to that\r
   *       span.\r
   *    decorations: {Array.<number|string} an array of style classes preceded\r
   *       by the position at which they start in job.sourceCode in order\r
   * }</pre>\r
   * @private\r
   */\r
  function recombineTagsAndDecorations(job) {\r
    var isIE8OrEarlier = /\bMSIE\s(\d+)/.exec(navigator.userAgent);\r
    isIE8OrEarlier = isIE8OrEarlier && +isIE8OrEarlier[1] <= 8;\r
    var newlineRe = /\n/g;\r
  \r
    var source = job.sourceCode;\r
    var sourceLength = source.length;\r
    // Index into source after the last code-unit recombined.\r
    var sourceIndex = 0;\r
  \r
    var spans = job.spans;\r
    var nSpans = spans.length;\r
    // Index into spans after the last span which ends at or before sourceIndex.\r
    var spanIndex = 0;\r
  \r
    var decorations = job.decorations;\r
    var nDecorations = decorations.length;\r
    // Index into decorations after the last decoration which ends at or before\r
    // sourceIndex.\r
    var decorationIndex = 0;\r
  \r
    // Remove all zero-length decorations.\r
    decorations[nDecorations] = sourceLength;\r
    var decPos, i;\r
    for (i = decPos = 0; i < nDecorations;) {\r
      if (decorations[i] !== decorations[i + 2]) {\r
        decorations[decPos++] = decorations[i++];\r
        decorations[decPos++] = decorations[i++];\r
      } else {\r
        i += 2;\r
      }\r
    }\r
    nDecorations = decPos;\r
  \r
    // Simplify decorations.\r
    for (i = decPos = 0; i < nDecorations;) {\r
      var startPos = decorations[i];\r
      // Conflate all adjacent decorations that use the same style.\r
      var startDec = decorations[i + 1];\r
      var end = i + 2;\r
      while (end + 2 <= nDecorations && decorations[end + 1] === startDec) {\r
        end += 2;\r
      }\r
      decorations[decPos++] = startPos;\r
      decorations[decPos++] = startDec;\r
      i = end;\r
    }\r
  \r
    nDecorations = decorations.length = decPos;\r
  \r
    var sourceNode = job.sourceNode;\r
    var oldDisplay;\r
    if (sourceNode) {\r
      oldDisplay = sourceNode.style.display;\r
      sourceNode.style.display = 'none';\r
    }\r
    try {\r
      var decoration = null;\r
      while (spanIndex < nSpans) {\r
        var spanStart = spans[spanIndex];\r
        var spanEnd = spans[spanIndex + 2] || sourceLength;\r
  \r
        var decEnd = decorations[decorationIndex + 2] || sourceLength;\r
  \r
        var end = Math.min(spanEnd, decEnd);\r
  \r
        var textNode = spans[spanIndex + 1];\r
        var styledText;\r
        if (textNode.nodeType !== 1  // Don't muck with <BR>s or <LI>s\r
            // Don't introduce spans around empty text nodes.\r
            && (styledText = source.substring(sourceIndex, end))) {\r
          // This may seem bizarre, and it is.  Emitting LF on IE causes the\r
          // code to display with spaces instead of line breaks.\r
          // Emitting Windows standard issue linebreaks (CRLF) causes a blank\r
          // space to appear at the beginning of every line but the first.\r
          // Emitting an old Mac OS 9 line separator makes everything spiffy.\r
          if (isIE8OrEarlier) {\r
            styledText = styledText.replace(newlineRe, '\r');\r
          }\r
          textNode.nodeValue = styledText;\r
          var document = textNode.ownerDocument;\r
          var span = document.createElement('span');\r
          span.className = decorations[decorationIndex + 1];\r
          var parentNode = textNode.parentNode;\r
          parentNode.replaceChild(span, textNode);\r
          span.appendChild(textNode);\r
          if (sourceIndex < spanEnd) {  // Split off a text node.\r
            spans[spanIndex + 1] = textNode\r
                // TODO: Possibly optimize by using '' if there's no flicker.\r
                = document.createTextNode(source.substring(end, spanEnd));\r
            parentNode.insertBefore(textNode, span.nextSibling);\r
          }\r
        }\r
  \r
        sourceIndex = end;\r
  \r
        if (sourceIndex >= spanEnd) {\r
          spanIndex += 2;\r
        }\r
        if (sourceIndex >= decEnd) {\r
          decorationIndex += 2;\r
        }\r
      }\r
    } finally {\r
      if (sourceNode) {\r
        sourceNode.style.display = oldDisplay;\r
      }\r
    }\r
  }\r
\r
  /** Maps language-specific file extensions to handlers. */\r
  var langHandlerRegistry = {};\r
  /** Register a language handler for the given file extensions.\r
    * @param {function (Object)} handler a function from source code to a list\r
    *      of decorations.  Takes a single argument job which describes the\r
    *      state of the computation.   The single parameter has the form\r
    *      {@code {\r
    *        sourceCode: {string} as plain text.\r
    *        decorations: {Array.<number|string>} an array of style classes\r
    *                     preceded by the position at which they start in\r
    *                     job.sourceCode in order.\r
    *                     The language handler should assigned this field.\r
    *        basePos: {int} the position of source in the larger source chunk.\r
    *                 All positions in the output decorations array are relative\r
    *                 to the larger source chunk.\r
    *      } }\r
    * @param {Array.<string>} fileExtensions\r
    */\r
  function registerLangHandler(handler, fileExtensions) {\r
    for (var i = fileExtensions.length; --i >= 0;) {\r
      var ext = fileExtensions[i];\r
      if (!langHandlerRegistry.hasOwnProperty(ext)) {\r
        langHandlerRegistry[ext] = handler;\r
      } else if (win['console']) {\r
        console['warn']('cannot override language handler %s', ext);\r
      }\r
    }\r
  }\r
  function langHandlerForExtension(extension, source) {\r
    if (!(extension && langHandlerRegistry.hasOwnProperty(extension))) {\r
      // Treat it as markup if the first non whitespace character is a < and\r
      // the last non-whitespace character is a >.\r
      extension = /^\s*</.test(source)\r
          ? 'default-markup'\r
          : 'default-code';\r
    }\r
    return langHandlerRegistry[extension];\r
  }\r
  registerLangHandler(decorateSource, ['default-code']);\r
  registerLangHandler(\r
      createSimpleLexer(\r
          [],\r
          [\r
           [PR_PLAIN,       /^[^<?]+/],\r
           [PR_DECLARATION, /^<!\w[^>]*(?:>|$)/],\r
           [PR_COMMENT,     /^<\!--[\s\S]*?(?:-\->|$)/],\r
           // Unescaped content in an unknown language\r
           ['lang-',        /^<\?([\s\S]+?)(?:\?>|$)/],\r
           ['lang-',        /^<%([\s\S]+?)(?:%>|$)/],\r
           [PR_PUNCTUATION, /^(?:<[%?]|[%?]>)/],\r
           ['lang-',        /^<xmp\b[^>]*>([\s\S]+?)<\/xmp\b[^>]*>/i],\r
           // Unescaped content in javascript.  (Or possibly vbscript).\r
           ['lang-js',      /^<script\b[^>]*>([\s\S]*?)(<\/script\b[^>]*>)/i],\r
           // Contains unescaped stylesheet content\r
           ['lang-css',     /^<style\b[^>]*>([\s\S]*?)(<\/style\b[^>]*>)/i],\r
           ['lang-in.tag',  /^(<\/?[a-z][^<>]*>)/i]\r
          ]),\r
      ['default-markup', 'htm', 'html', 'mxml', 'xhtml', 'xml', 'xsl']);\r
  registerLangHandler(\r
      createSimpleLexer(\r
          [\r
           [PR_PLAIN,        /^[\s]+/, null, ' \t\r\n'],\r
           [PR_ATTRIB_VALUE, /^(?:\"[^\"]*\"?|\'[^\']*\'?)/, null, '\"\'']\r
           ],\r
          [\r
           [PR_TAG,          /^^<\/?[a-z](?:[\w.:-]*\w)?|\/?>$/i],\r
           [PR_ATTRIB_NAME,  /^(?!style[\s=]|on)[a-z](?:[\w:-]*\w)?/i],\r
           ['lang-uq.val',   /^=\s*([^>\'\"\s]*(?:[^>\'\"\s\/]|\/(?=\s)))/],\r
           [PR_PUNCTUATION,  /^[=<>\/]+/],\r
           ['lang-js',       /^on\w+\s*=\s*\"([^\"]+)\"/i],\r
           ['lang-js',       /^on\w+\s*=\s*\'([^\']+)\'/i],\r
           ['lang-js',       /^on\w+\s*=\s*([^\"\'>\s]+)/i],\r
           ['lang-css',      /^style\s*=\s*\"([^\"]+)\"/i],\r
           ['lang-css',      /^style\s*=\s*\'([^\']+)\'/i],\r
           ['lang-css',      /^style\s*=\s*([^\"\'>\s]+)/i]\r
           ]),\r
      ['in.tag']);\r
  registerLangHandler(\r
      createSimpleLexer([], [[PR_ATTRIB_VALUE, /^[\s\S]+/]]), ['uq.val']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': CPP_KEYWORDS,\r
          'hashComments': true,\r
          'cStyleComments': true,\r
          'types': C_TYPES\r
        }), ['c', 'cc', 'cpp', 'cxx', 'cyc', 'm']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': 'null,true,false'\r
        }), ['json']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': CSHARP_KEYWORDS,\r
          'hashComments': true,\r
          'cStyleComments': true,\r
          'verbatimStrings': true,\r
          'types': C_TYPES\r
        }), ['cs']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': JAVA_KEYWORDS,\r
          'cStyleComments': true\r
        }), ['java']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': SH_KEYWORDS,\r
          'hashComments': true,\r
          'multiLineStrings': true\r
        }), ['bash', 'bsh', 'csh', 'sh']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': PYTHON_KEYWORDS,\r
          'hashComments': true,\r
          'multiLineStrings': true,\r
          'tripleQuotedStrings': true\r
        }), ['cv', 'py', 'python']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': PERL_KEYWORDS,\r
          'hashComments': true,\r
          'multiLineStrings': true,\r
          'regexLiterals': 2  // multiline regex literals\r
        }), ['perl', 'pl', 'pm']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': RUBY_KEYWORDS,\r
          'hashComments': true,\r
          'multiLineStrings': true,\r
          'regexLiterals': true\r
        }), ['rb', 'ruby']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': JSCRIPT_KEYWORDS,\r
          'cStyleComments': true,\r
          'regexLiterals': true\r
        }), ['javascript', 'js']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': COFFEE_KEYWORDS,\r
          'hashComments': 3,  // ### style block comments\r
          'cStyleComments': true,\r
          'multilineStrings': true,\r
          'tripleQuotedStrings': true,\r
          'regexLiterals': true\r
        }), ['coffee']);\r
  registerLangHandler(sourceDecorator({\r
          'keywords': RUST_KEYWORDS,\r
          'cStyleComments': true,\r
          'multilineStrings': true\r
        }), ['rc', 'rs', 'rust']);\r
  registerLangHandler(\r
      createSimpleLexer([], [[PR_STRING, /^[\s\S]+/]]), ['regex']);\r
\r
  function applyDecorator(job) {\r
    var opt_langExtension = job.langExtension;\r
\r
    try {\r
      // Extract tags, and convert the source code to plain text.\r
      var sourceAndSpans = extractSourceSpans(job.sourceNode, job.pre);\r
      /** Plain text. @type {string} */\r
      var source = sourceAndSpans.sourceCode;\r
      job.sourceCode = source;\r
      job.spans = sourceAndSpans.spans;\r
      job.basePos = 0;\r
\r
      // Apply the appropriate language handler\r
      langHandlerForExtension(opt_langExtension, source)(job);\r
\r
      // Integrate the decorations and tags back into the source code,\r
      // modifying the sourceNode in place.\r
      recombineTagsAndDecorations(job);\r
    } catch (e) {\r
      if (win['console']) {\r
        console['log'](e && e['stack'] || e);\r
      }\r
    }\r
  }\r
\r
  /**\r
   * Pretty print a chunk of code.\r
   * @param sourceCodeHtml {string} The HTML to pretty print.\r
   * @param opt_langExtension {string} The language name to use.\r
   *     Typically, a filename extension like 'cpp' or 'java'.\r
   * @param opt_numberLines {number|boolean} True to number lines,\r
   *     or the 1-indexed number of the first line in sourceCodeHtml.\r
   */\r
  function $prettyPrintOne(sourceCodeHtml, opt_langExtension, opt_numberLines) {\r
    var container = document.createElement('div');\r
    // This could cause images to load and onload listeners to fire.\r
    // E.g. <img onerror="alert(1337)" src="nosuchimage.png">.\r
    // We assume that the inner HTML is from a trusted source.\r
    // The pre-tag is required for IE8 which strips newlines from innerHTML\r
    // when it is injected into a <pre> tag.\r
    // http://stackoverflow.com/questions/451486/pre-tag-loses-line-breaks-when-setting-innerhtml-in-ie\r
    // http://stackoverflow.com/questions/195363/inserting-a-newline-into-a-pre-tag-ie-javascript\r
    container.innerHTML = '<pre>' + sourceCodeHtml + '</pre>';\r
    container = container.firstChild;\r
    if (opt_numberLines) {\r
      numberLines(container, opt_numberLines, true);\r
    }\r
\r
    var job = {\r
      langExtension: opt_langExtension,\r
      numberLines: opt_numberLines,\r
      sourceNode: container,\r
      pre: 1\r
    };\r
    applyDecorator(job);\r
    return container.innerHTML;\r
  }\r
\r
   /**\r
    * Find all the {@code <pre>} and {@code <code>} tags in the DOM with\r
    * {@code class=prettyprint} and prettify them.\r
    *\r
    * @param {Function} opt_whenDone called when prettifying is done.\r
    * @param {HTMLElement|HTMLDocument} opt_root an element or document\r
    *   containing all the elements to pretty print.\r
    *   Defaults to {@code document.body}.\r
    */\r
  function $prettyPrint(opt_whenDone, opt_root) {\r
    var root = opt_root || document.body;\r
    var doc = root.ownerDocument || document;\r
    function byTagName(tn) { return root.getElementsByTagName(tn); }\r
    // fetch a list of nodes to rewrite\r
    var codeSegments = [byTagName('pre'), byTagName('code'), byTagName('xmp')];\r
    var elements = [];\r
    for (var i = 0; i < codeSegments.length; ++i) {\r
      for (var j = 0, n = codeSegments[i].length; j < n; ++j) {\r
        elements.push(codeSegments[i][j]);\r
      }\r
    }\r
    codeSegments = null;\r
\r
    var clock = Date;\r
    if (!clock['now']) {\r
      clock = { 'now': function () { return +(new Date); } };\r
    }\r
\r
    // The loop is broken into a series of continuations to make sure that we\r
    // don't make the browser unresponsive when rewriting a large page.\r
    var k = 0;\r
    var prettyPrintingJob;\r
\r
    var langExtensionRe = /\blang(?:uage)?-([\w.]+)(?!\S)/;\r
    var prettyPrintRe = /\bprettyprint\b/;\r
    var prettyPrintedRe = /\bprettyprinted\b/;\r
    var preformattedTagNameRe = /pre|xmp/i;\r
    var codeRe = /^code$/i;\r
    var preCodeXmpRe = /^(?:pre|code|xmp)$/i;\r
    var EMPTY = {};\r
\r
    function doWork() {\r
      var endTime = (win['PR_SHOULD_USE_CONTINUATION'] ?\r
                     clock['now']() + 250 /* ms */ :\r
                     Infinity);\r
      for (; k < elements.length && clock['now']() < endTime; k++) {\r
        var cs = elements[k];\r
\r
        // Look for a preceding comment like\r
        // <?prettify lang="..." linenums="..."?>\r
        var attrs = EMPTY;\r
        {\r
          for (var preceder = cs; (preceder = preceder.previousSibling);) {\r
            var nt = preceder.nodeType;\r
            // <?foo?> is parsed by HTML 5 to a comment node (8)\r
            // like <!--?foo?-->, but in XML is a processing instruction\r
            var value = (nt === 7 || nt === 8) && preceder.nodeValue;\r
            if (value\r
                ? !/^\??prettify\b/.test(value)\r
                : (nt !== 3 || /\S/.test(preceder.nodeValue))) {\r
              // Skip over white-space text nodes but not others.\r
              break;\r
            }\r
            if (value) {\r
              attrs = {};\r
              value.replace(\r
                  /\b(\w+)=([\w:.%+-]+)/g,\r
                function (_, name, value) { attrs[name] = value; });\r
              break;\r
            }\r
          }\r
        }\r
\r
        var className = cs.className;\r
        if ((attrs !== EMPTY || prettyPrintRe.test(className))\r
            // Don't redo this if we've already done it.\r
            // This allows recalling pretty print to just prettyprint elements\r
            // that have been added to the page since last call.\r
            && !prettyPrintedRe.test(className)) {\r
\r
          // make sure this is not nested in an already prettified element\r
          var nested = false;\r
          for (var p = cs.parentNode; p; p = p.parentNode) {\r
            var tn = p.tagName;\r
            if (preCodeXmpRe.test(tn)\r
                && p.className && prettyPrintRe.test(p.className)) {\r
              nested = true;\r
              break;\r
            }\r
          }\r
          if (!nested) {\r
            // Mark done.  If we fail to prettyprint for whatever reason,\r
            // we shouldn't try again.\r
            cs.className += ' prettyprinted';\r
\r
            // If the classes includes a language extensions, use it.\r
            // Language extensions can be specified like\r
            //     <pre class="prettyprint lang-cpp">\r
            // the language extension "cpp" is used to find a language handler\r
            // as passed to PR.registerLangHandler.\r
            // HTML5 recommends that a language be specified using "language-"\r
            // as the prefix instead.  Google Code Prettify supports both.\r
            // http://dev.w3.org/html5/spec-author-view/the-code-element.html\r
            var langExtension = attrs['lang'];\r
            if (!langExtension) {\r
              langExtension = className.match(langExtensionRe);\r
              // Support <pre class="prettyprint"><code class="language-c">\r
              var wrapper;\r
              if (!langExtension && (wrapper = childContentWrapper(cs))\r
                  && codeRe.test(wrapper.tagName)) {\r
                langExtension = wrapper.className.match(langExtensionRe);\r
              }\r
\r
              if (langExtension) { langExtension = langExtension[1]; }\r
            }\r
\r
            var preformatted;\r
            if (preformattedTagNameRe.test(cs.tagName)) {\r
              preformatted = 1;\r
            } else {\r
              var currentStyle = cs['currentStyle'];\r
              var defaultView = doc.defaultView;\r
              var whitespace = (\r
                  currentStyle\r
                  ? currentStyle['whiteSpace']\r
                  : (defaultView\r
                     && defaultView.getComputedStyle)\r
                  ? defaultView.getComputedStyle(cs, null)\r
                  .getPropertyValue('white-space')\r
                  : 0);\r
              preformatted = whitespace\r
                  && 'pre' === whitespace.substring(0, 3);\r
            }\r
\r
            // Look for a class like linenums or linenums:<n> where <n> is the\r
            // 1-indexed number of the first line.\r
            var lineNums = attrs['linenums'];\r
            if (!(lineNums = lineNums === 'true' || +lineNums)) {\r
              lineNums = className.match(/\blinenums\b(?::(\d+))?/);\r
              lineNums =\r
                lineNums\r
                ? lineNums[1] && lineNums[1].length\r
                  ? +lineNums[1] : true\r
                : false;\r
            }\r
            if (lineNums) { numberLines(cs, lineNums, preformatted); }\r
\r
            // do the pretty printing\r
            prettyPrintingJob = {\r
              langExtension: langExtension,\r
              sourceNode: cs,\r
              numberLines: lineNums,\r
              pre: preformatted\r
            };\r
            applyDecorator(prettyPrintingJob);\r
          }\r
        }\r
      }\r
      if (k < elements.length) {\r
        // finish up in a continuation\r
        setTimeout(doWork, 250);\r
      } else if ('function' === typeof opt_whenDone) {\r
        opt_whenDone();\r
      }\r
    }\r
\r
    doWork();\r
  }\r
\r
  /**\r
   * Contains functions for creating and registering new language handlers.\r
   * @type {Object}\r
   */\r
  var PR = win['PR'] = {\r
        'createSimpleLexer': createSimpleLexer,\r
        'registerLangHandler': registerLangHandler,\r
        'sourceDecorator': sourceDecorator,\r
        'PR_ATTRIB_NAME': PR_ATTRIB_NAME,\r
        'PR_ATTRIB_VALUE': PR_ATTRIB_VALUE,\r
        'PR_COMMENT': PR_COMMENT,\r
        'PR_DECLARATION': PR_DECLARATION,\r
        'PR_KEYWORD': PR_KEYWORD,\r
        'PR_LITERAL': PR_LITERAL,\r
        'PR_NOCODE': PR_NOCODE,\r
        'PR_PLAIN': PR_PLAIN,\r
        'PR_PUNCTUATION': PR_PUNCTUATION,\r
        'PR_SOURCE': PR_SOURCE,\r
        'PR_STRING': PR_STRING,\r
        'PR_TAG': PR_TAG,\r
        'PR_TYPE': PR_TYPE,\r
        'prettyPrintOne':\r
           IN_GLOBAL_SCOPE\r
             ? (win['prettyPrintOne'] = $prettyPrintOne)\r
             : (prettyPrintOne = $prettyPrintOne),\r
        'prettyPrint': prettyPrint =\r
           IN_GLOBAL_SCOPE\r
             ? (win['prettyPrint'] = $prettyPrint)\r
             : (prettyPrint = $prettyPrint)\r
      };\r
\r
  // Make PR available via the Asynchronous Module Definition (AMD) API.\r
  // Per https://github.com/amdjs/amdjs-api/wiki/AMD:\r
  // The Asynchronous Module Definition (AMD) API specifies a\r
  // mechanism for defining modules such that the module and its\r
  // dependencies can be asynchronously loaded.\r
  // ...\r
  // To allow a clear indicator that a global define function (as\r
  // needed for script src browser loading) conforms to the AMD API,\r
  // any global define function SHOULD have a property called "amd"\r
  // whose value is an object. This helps avoid conflict with any\r
  // other existing JavaScript code that could have defined a define()\r
  // function that does not conform to the AMD API.\r
  if (typeof define === "function" && define['amd']) {\r
    define("google-code-prettify", [], function () {\r
      return PR; \r
    });\r
  }\r
})();\r