diff --git a/src/IDF/Views/Source.php b/src/IDF/Views/Source.php index cd4fc0d..3c9775b 100644 --- a/src/IDF/Views/Source.php +++ b/src/IDF/Views/Source.php @@ -31,6 +31,12 @@ Pluf::loadFunction('Pluf_Shortcuts_GetFormForModel'); */ class IDF_Views_Source { + public static $supportedExtenstions = array('c', 'cc', 'cpp', 'cs', 'css', + 'cyc', 'java', 'bsh', 'csh', + 'sh', 'cv', 'py', 'perl', 'php', + 'pl', 'pm', 'rb', 'js', 'html', + 'html', 'xhtml', 'xml', 'xsl'); + public $changeLog_precond = array('IDF_Precondition::accessSource'); public function changeLog($request, $match) { @@ -378,16 +384,32 @@ class IDF_Views_Source public static function highLight($fileinfo, $content) { + $openPre = ''; + $closePre = ''; + if (IDF_Views_Source::isSupportedExtension($fileinfo[2])) { + $openPre = '
';
+            $closePre = '
'; + } $table = array(); $i = 1; foreach (preg_split("/\015\012|\015|\012/", $content) as $line) { $table[] = ''.$i.'' - .''.IDF_Diff::padLine(Pluf_esc($line)).''; + .''.$openPre.IDF_Diff::padLine(Pluf_esc($line)).$closePre.''; $i++; } return Pluf_Template::markSafe(implode("\n", $table)); } + /** + * @param string the extension to test + * + * @return + */ + public static function isSupportedExtension($extension) + { + return in_array($extension, IDF_Views_Source::$supportedExtenstions); + } + /** * Get the scm type for page title * diff --git a/src/IDF/templates/source/git/file.html b/src/IDF/templates/source/git/file.html index a147c59..1dfe773 100644 --- a/src/IDF/templates/source/git/file.html +++ b/src/IDF/templates/source/git/file.html @@ -1,4 +1,5 @@ {extends "source/base.html"} +{block extraheader}{/block} {block docclass}yui-t1{assign $inSourceTree=true}{/block} {block body}

{trans 'Root'}/{if $breadcrumb}{$breadcrumb|safe}{/if}

@@ -19,7 +20,6 @@ {aurl 'url', 'IDF_Views_Source::getFile', array($project.shortname, $commit, $fullpath)}

{trans 'Archive'} {trans 'Download this file'}

- {/block} {block context}

{trans 'Branches:'}
@@ -29,3 +29,8 @@ {/foreach}

{/block} + +{block javascript} + + +{/block} diff --git a/src/IDF/templates/source/svn/file.html b/src/IDF/templates/source/svn/file.html index fc78c4d..2e9de70 100644 --- a/src/IDF/templates/source/svn/file.html +++ b/src/IDF/templates/source/svn/file.html @@ -1,4 +1,5 @@ {extends "source/base.html"} +{block extraheader}{/block} {block docclass}yui-t1{assign $inSourceTree=true}{/block} {block body}

{trans 'Root'}/{if $breadcrumb}{$breadcrumb|safe}{/if}

@@ -30,7 +31,6 @@ {aurl 'url', 'IDF_Views_Source::getFile', array($project.shortname, $commit, $fullpath)}

{trans 'Archive'} {trans 'Download this file'}

- {/block} {block context}
@@ -42,3 +42,8 @@
{/block} + +{block javascript} + + +{/block} diff --git a/www/media/idf/css/prettify.css b/www/media/idf/css/prettify.css new file mode 100644 index 0000000..13e9c52 --- /dev/null +++ b/www/media/idf/css/prettify.css @@ -0,0 +1,27 @@ +/* Pretty printing styles. Used with prettify.js. */ + +.str { color: #080; } +.kwd { color: #008; } +.com { color: #800; } +.typ { color: #606; } +.lit { color: #066; } +.pun { color: #660; } +.pln { color: #000; } +.tag { color: #008; } +.atn { color: #606; } +.atv { color: #080; } +.dec { color: #606; } +pre.prettyprint { border:0px; padding:0; margin:0; } + +@media print { + .str { color: #060; } + .kwd { color: #006; font-weight: bold; } + .com { color: #600; font-style: italic; } + .typ { color: #404; font-weight: bold; } + .lit { color: #044; } + .pun { color: #440; } + .pln { color: #000; } + .tag { color: #006; font-weight: bold; } + .atn { color: #404; } + .atv { color: #060; } +} diff --git a/www/media/idf/js/prettify.js b/www/media/idf/js/prettify.js new file mode 100644 index 0000000..6a18f43 --- /dev/null +++ b/www/media/idf/js/prettify.js @@ -0,0 +1,1348 @@ +// Copyright (C) 2006 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + + +/** + * @fileoverview + * some functions for browser-side pretty printing of code contained in html. + * + * The lexer should work on a number of languages including C and friends, + * Java, Python, Bash, SQL, HTML, XML, CSS, Javascript, and Makefiles. + * It works passably on Ruby, PHP and Awk and a decent subset of Perl, but, + * because of commenting conventions, doesn't work on Smalltalk, Lisp-like, or + * CAML-like languages. + * + * If there's a language not mentioned here, then I don't know it, and don't + * know whether it works. If it has a C-like, Bash-like, or XML-like syntax + * then it should work passably. + * + * Usage: + * 1) include this source file in an html page via + * + * 2) define style rules. See the example page for examples. + * 3) mark the
 and  tags in your source with class=prettyprint.
+ *    You can also use the (html deprecated)  tag, but the pretty printer
+ *    needs to do more substantial DOM manipulations to support that, so some
+ *    css styles may not be preserved.
+ * That's it.  I wanted to keep the API as simple as possible, so there's no
+ * need to specify which language the code is in.
+ *
+ * Change log:
+ * cbeust, 2006/08/22
+ *   Java annotations (start with "@") are now captured as literals ("lit")
+ */
+
+var PR_keywords = new Object();
+/** initialize the keyword list for our target languages. */
+(function () {
+  var CPP_KEYWORDS = (
+    "bool break case catch char class const const_cast continue default " +
+    "delete deprecated dllexport dllimport do double dynamic_cast else enum " +
+    "explicit extern false float for friend goto if inline int long mutable " +
+    "naked namespace new noinline noreturn nothrow novtable operator private " +
+    "property protected public register reinterpret_cast return selectany " +
+    "short signed sizeof static static_cast struct switch template this " +
+    "thread throw true try typedef typeid typename union unsigned using " +
+    "declaration, using directive uuid virtual void volatile while typeof");
+  var JAVA_KEYWORDS = (
+    "abstract default goto package synchronized boolean do if private this " +
+    "break double implements protected throw byte else import public throws " +
+    "case enum instanceof return transient catch extends int short try char " +
+    "final interface static void class finally long strictfp volatile const " +
+    "float native super while continue for new switch");
+  var PYTHON_KEYWORDS = (
+    "and assert break class continue def del elif else except exec finally " +
+    "for from global if import in is lambda not or pass print raise return " +
+    "try while yield False True None");
+  var JSCRIPT_KEYWORDS = (
+    "abstract boolean break byte case catch char class const continue " +
+    "debugger default delete do double else enum export extends false final " +
+    "finally float for function goto if implements import in instanceof int " +
+    "interface long native new null package private protected public return " +
+    "short static super switch synchronized this throw throws transient " +
+    "true try typeof var void volatile while with NaN Infinity");
+  var PERL_KEYWORDS = (
+    "foreach require sub unless until use elsif BEGIN END");
+  var SH_KEYWORDS = (
+    "if then do else fi end");
+  var RUBY_KEYWORDS = (
+      "if then elsif else end begin do rescue ensure while for class module " +
+      "def yield raise until unless and or not when case super undef break " +
+      "next redo retry in return alias defined");
+  var KEYWORDS = [CPP_KEYWORDS, JAVA_KEYWORDS, JSCRIPT_KEYWORDS, PERL_KEYWORDS,
+                  PYTHON_KEYWORDS, RUBY_KEYWORDS, SH_KEYWORDS];
+  for (var k = 0; k < KEYWORDS.length; k++) {
+    var kw = KEYWORDS[k].split(' ');
+    for (var i = 0; i < kw.length; i++) {
+      if (kw[i]) { PR_keywords[kw[i]] = true; }
+    }
+  }
+}).call(this);
+
+// token style names.  correspond to css classes
+/** token style for a string literal */
+var PR_STRING = 'str';
+/** token style for a keyword */
+var PR_KEYWORD = 'kwd';
+/** token style for a comment */
+var PR_COMMENT = 'com';
+/** token style for a type */
+var PR_TYPE = 'typ';
+/** token style for a literal value.  e.g. 1, null, true. */
+var PR_LITERAL = 'lit';
+/** token style for a punctuation string. */
+var PR_PUNCTUATION = 'pun';
+/** token style for a punctuation string. */
+var PR_PLAIN = 'pln';
+
+/** token style for an sgml tag. */
+var PR_TAG = 'tag';
+/** token style for a markup declaration such as a DOCTYPE. */
+var PR_DECLARATION = 'dec';
+/** token style for embedded source. */
+var PR_SOURCE = 'src';
+/** token style for an sgml attribute name. */
+var PR_ATTRIB_NAME = 'atn';
+/** token style for an sgml attribute value. */
+var PR_ATTRIB_VALUE = 'atv';
+
+/** the position of the end of a token during.  A division of a string into
+  * n tokens can be represented as a series n - 1 token ends, as long as
+  * runs of whitespace warrant their own token.
+  * @private
+  */
+function PR_TokenEnd(end, style) {
+  if (undefined === style) { throw new Error('BAD'); }
+  if ('number' != typeof(end)) { throw new Error('BAD'); }
+  this.end = end;
+  this.style = style;
+}
+PR_TokenEnd.prototype.toString = function () {
+  return '[PR_TokenEnd ' + this.end +
+    (this.style ? ':' + this.style : '') + ']';
+};
+
+
+/** a chunk of text with a style.  These are used to represent both the output
+  * from the lexing functions as well as intermediate results.
+  * @constructor
+  * @param token the token text
+  * @param style one of the token styles defined in designdoc-template, or null
+  *   for a styleless token, such as an embedded html tag.
+  * @private
+  */
+function PR_Token(token, style) {
+  if (undefined === style) { throw new Error('BAD'); }
+  this.token = token;
+  this.style = style;
+}
+
+PR_Token.prototype.toString = function () {
+  return '[PR_Token ' + this.token + (this.style ? ':' + this.style : '') + ']';
+};
+
+
+/** a helper class that decodes common html entities used to escape source and
+  * markup punctuation characters in html.
+  * @constructor
+  * @private
+  */
+function PR_DecodeHelper() {
+  this.next = 0;
+  this.ch = '\0';
+}
+
+PR_DecodeHelper.prototype.decode = function (s, i) {
+  var next = i + 1;
+  var ch = s.charAt(i);
+  if ('&' == ch) {
+    var semi = s.indexOf(';', next);
+    if (semi >= 0 && semi < next + 4) {
+      var entityName = s.substring(next, semi).toLowerCase();
+      next = semi + 1;
+      if ('lt' == entityName) {
+        ch = '<';
+      } else if ('gt' == entityName) {
+        ch = '>';
+      } else if ('quot' == entityName) {
+        ch = '"';
+      } else if ('apos' == entityName) {
+        ch = '\'';
+      } else if ('amp' == entityName) {
+        ch = '&';
+      } else {
+        next = i + 1;
+      }
+    }
+  }
+  this.next = next;
+  this.ch = ch;
+  return this.ch;
+}
+
+
+// some string utilities
+function PR_isWordChar(ch) {
+  return (ch >= 'a' && ch <= 'z') || (ch >= 'A' && ch <= 'Z');
+}
+
+function PR_isIdentifierStart(ch) {
+  return PR_isWordChar(ch) || ch == '_' || ch == '$' || ch == '@';
+}
+
+function PR_isIdentifierPart(ch) {
+  return PR_isIdentifierStart(ch) || PR_isDigitChar(ch);
+}
+
+function PR_isSpaceChar(ch) {
+  return "\t \r\n".indexOf(ch) >= 0;
+}
+
+function PR_isDigitChar(ch) {
+  return ch >= '0' && ch <= '9';
+}
+
+function PR_trim(s) {
+  var i = 0, j = s.length - 1;
+  while (i <= j && PR_isSpaceChar(s.charAt(i))) { ++i; }
+  while (j > i && PR_isSpaceChar(s.charAt(j))) { --j; }
+  return s.substring(i, j + 1);
+}
+
+function PR_startsWith(s, prefix) {
+  return s.length >= prefix.length && prefix == s.substring(0, prefix.length);
+}
+
+function PR_endsWith(s, suffix) {
+  return s.length >= suffix.length &&
+         suffix == s.substring(s.length - suffix.length, s.length);
+}
+
+/** true iff prefix matches the first prefix characters in chars[0:len].
+  * @private
+  */
+function PR_prefixMatch(chars, len, prefix) {
+  if (len < prefix.length) { return false; }
+  for (var i = 0, n = prefix.length; i < n; ++i) {
+    if (prefix.charAt(i) != chars[i]) { return false; }
+  }
+  return true;
+}
+
+/** used to convert html special characters embedded in XMP tags into html. */
+function PR_textToHtml(str) {
+  return str.replace(/&/g, '&amp;').replace(/</g, '&lt;').replace(/>/g, '&gt;');
+}
+
+
+/** split markup into chunks of html tags (style null) and
+  * plain text (style {@link #PR_PLAIN}).
+  *
+  * @param s a String of html.
+  * @return an Array of PR_Tokens of style PR_PLAIN and null.
+  * @private
+  */
+function PR_chunkify(s) {
+  var chunks = new Array();
+  var state = 0;
+  var start = 0;
+  var pos = -1;
+  for (var i = 0, n = s.length; i < n; ++i) {
+    var ch = s.charAt(i);
+    switch (state) {
+      case 0:
+        if ('<' == ch) { state = 1; }
+        break;
+      case 1:
+        pos = i - 1;
+        if ('/' == ch) { state = 2; }
+        else if (PR_isWordChar(ch)) { state = 3; }
+        else if ('<' == ch) { state = 1; }
+        else { state = 0; }
+        break;
+      case 2:
+        if (PR_isWordChar(ch)) { state = 3; }
+        else if ('<' == ch) { state = 1; }
+        else { state = 0; }
+        break;
+      case 3:
+        if ('>' == ch) {
+          if (pos > start) {
+            chunks.push(new PR_Token(s.substring(start, pos), PR_PLAIN));
+          }
+          chunks.push(new PR_Token(s.substring(pos, i + 1), null));
+          start = i + 1;
+          pos = -1;
+          state = 0;
+        }
+        break;
+    }
+  }
+  if (s.length > start) {
+    chunks.push(new PR_Token(s.substring(start, s.length), PR_PLAIN));
+  }
+  return chunks;
+}
+
+/** splits chunks around entities.
+  * @private
+  */
+function PR_splitEntities(chunks) {
+  var chunksOut = new Array();
+  var state = 0;
+  for (var ci = 0, nc = chunks.length; ci < nc; ++ci) {
+    var chunk = chunks[ci];
+    if (PR_PLAIN != chunk.style) {
+      chunksOut.push(chunk);
+      continue;
+    }
+    var s = chunk.token;
+    var pos = 0;
+    var start;
+    for (var i = 0; i < s.length; ++i) {
+      var ch = s.charAt(i);
+      switch (state) {
+        case 0:
+          if ('&' == ch) { state = 1; }
+          break;
+        case 1:
+          if ('#' == ch || PR_isWordChar(ch)) {
+            start = i - 1;
+            state = 2;
+          } else {
+            state = 0;
+          }
+          break;
+        case 2:
+          if (';' == ch) {
+            if (start > pos) {
+              chunksOut.push(
+                  new PR_Token(s.substring(pos, start), chunk.style));
+            }
+            chunksOut.push(new PR_Token(s.substring(start, i + 1), null));
+            pos = i + 1;
+            state = 0;
+          }
+          break;
+      }
+    }
+    if (s.length > pos) {
+      chunksOut.push(pos ?
+                     new PR_Token(s.substring(pos, s.length), chunk.style) :
+                     chunk);
+    }
+  }
+  return chunksOut;
+}
+
+/** walk the tokenEnds list and the chunk list in parallel to generate a list
+  * of split tokens.
+  * @private
+  */
+function PR_splitChunks(chunks, tokenEnds) {
+  var tokens = new Array();  // the output
+
+  var ci = 0;  // index into chunks
+  // position of beginning of amount written so far in absolute space.
+  var posAbs = 0;
+  // position of amount written so far in chunk space
+  var posChunk = 0;
+
+  // current chunk
+  var chunk = new PR_Token('', null);
+
+  for (var ei = 0, ne = tokenEnds.length; ei < ne; ++ei) {
+    var tokenEnd = tokenEnds[ei];
+    var end = tokenEnd.end;
+
+    var tokLen = end - posAbs;
+    var remainingInChunk = chunk.token.length - posChunk;
+
+    // FIX Nicolas LASSALLE : remainingInChunk and tokLen can both be equals so that the while is infinite !
+    if (remainingInChunk != tokLen) {
+        while (remainingInChunk <= tokLen) {
+            if (remainingInChunk > 0) {
+                tokens.push(
+                            new PR_Token(chunk.token.substring(posChunk, chunk.token.length),
+                                         null == chunk.style ? null : tokenEnd.style));
+            }
+            posAbs += remainingInChunk;
+            posChunk = 0;
+            if (ci < chunks.length) { chunk = chunks[ci++]; }
+            
+            tokLen = end - posAbs;
+            remainingInChunk = chunk.token.length - posChunk;
+        }
+    }
+
+    if (tokLen) {
+      tokens.push(
+          new PR_Token(chunk.token.substring(posChunk, posChunk + tokLen),
+                       tokenEnd.style));
+      posAbs += tokLen;
+      posChunk += tokLen;
+    }
+  }
+
+  return tokens;
+}
+
+/** splits markup tokens into declarations, tags, and source chunks.
+  * @private
+  */
+function PR_splitMarkup(chunks) {
+  // A state machine to split out declarations, tags, etc.
+  // This state machine deals with absolute space in the text, indexed by k,
+  // and position in the current chunk, indexed by pos and tokenStart to
+  // generate a list of the ends of tokens.
+  // Absolute space is calculated by considering the chunks as appended into
+  // one big string, as they were before being split.
+
+  // Known failure cases
+  // Server side scripting sections such as <?...?> in attributes.
+  // i.e. <span class="<? foo ?>">
+  // Handling this would require a stack, and we don't use PHP.
+
+  // The output: a list of pairs of PR_TokenEnd instances
+  var tokenEnds = new Array();
+
+  var state = 0;  // FSM state variable
+  var k = 0;  // position in absolute space of the start of the current chunk
+  var tokenStart = -1;  // the start of the current token
+
+  // Try to find a closing tag for any open <style> or <script> tags
+  // We can't do this at a later stage because then the following case
+  // would fail:
+  // <script>document.writeln('<!--');</script>
+
+  // We use tokenChars[:tokenCharsI] to accumulate the tag name so that we
+  // can check whether to enter into a no scripting section when the tag ends.
+  var tokenChars = new Array(12);
+  var tokenCharsI = 0;
+  // if non null, the tag prefix that we need to see to break out.
+  var endScriptTag = null;
+  var decodeHelper = new PR_DecodeHelper();
+
+  for (var ci = 0, nc = chunks.length; ci < nc; ++ci) {
+    var chunk = chunks[ci];
+    if (PR_PLAIN != chunk.style) {
+      k += chunk.token.length;
+      continue;
+    }
+
+    var s = chunk.token;
+    var pos = 0;  // the position past the last character processed so far in s
+
+    for (var i = 0, n = s.length; i < n; /* i = next at bottom */) {
+      decodeHelper.decode(s, i);
+      var ch = decodeHelper.ch;
+      var next = decodeHelper.next;
+
+      var tokenStyle = null;
+      switch (state) {
+        case 0:
+          if ('<' == ch) { state = 1; }
+          break;
+        case 1:
+          tokenCharsI = 0;
+          if ('/' == ch) {  // only consider close tags if we're in script/style
+            state = 7;
+          } else if (null == endScriptTag) {
+            if ('!' == ch) {
+              state = 2;
+            } else if (PR_isWordChar(ch)) {
+              state = 8;
+            } else if ('?' == ch) {
+              state = 9;
+            } else if ('%' == ch) {
+              state = 11;
+            } else if ('<' != ch) {
+              state = 0;
+            }
+          } else if ('<' != ch) {
+            state = 0;
+          }
+          break;
+        case 2:
+          if ('-' == ch) {
+            state = 4;
+          } else if (PR_isWordChar(ch)) {
+            state = 3;
+          } else if ('<' == ch) {
+            state = 1;
+          } else {
+            state = 0;
+          }
+          break;
+        case 3:
+          if ('>' == ch) {
+            state = 0;
+            tokenStyle = PR_DECLARATION;
+          }
+          break;
+        case 4:
+          if ('-' == ch) { state = 5; }
+          break;
+        case 5:
+          if ('-' == ch) { state = 6; }
+          break;
+        case 6:
+          if ('>' == ch) {
+            state = 0;
+            tokenStyle = PR_COMMENT;
+          } else if ('-' == ch) {
+            state = 6;
+          } else {
+            state = 4;
+          }
+          break;
+        case 7:
+          if (PR_isWordChar(ch)) {
+            state = 8;
+          } else if ('<' == ch) {
+            state = 1;
+          } else {
+            state = 0;
+          }
+          break;
+        case 8:
+          if ('>' == ch) {
+            state = 0;
+            tokenStyle = PR_TAG;
+          }
+          break;
+        case 9:
+          if ('?' == ch) { state = 10; }
+          break;
+        case 10:
+          if ('>' == ch) {
+            state = 0;
+            tokenStyle = PR_SOURCE;
+          } else if ('?' != ch) {
+            state = 9;
+          }
+          break;
+        case 11:
+          if ('%' == ch) { state = 12; }
+          break;
+        case 12:
+          if ('>' == ch) {
+            state = 0;
+            tokenStyle = PR_SOURCE;
+          } else if ('%' != ch) {
+            state = 11;
+          }
+          break;
+      }
+
+      if (tokenCharsI < tokenChars.length) {
+        tokenChars[tokenCharsI++] = ch.toLowerCase();
+      }
+      if (1 == state) { tokenStart = k + i; }
+      i = next;
+      if (tokenStyle != null) {
+        if (null != tokenStyle) {
+          if (endScriptTag) {
+            if (PR_prefixMatch(tokenChars, tokenCharsI, endScriptTag)) {
+              endScriptTag = null;
+            }
+          } else {
+            if (PR_prefixMatch(tokenChars, tokenCharsI, 'script')) {
+              endScriptTag = '/script';
+            } else if (PR_prefixMatch(tokenChars, tokenCharsI, 'style')) {
+              endScriptTag = '/style';
+            } else if (PR_prefixMatch(tokenChars, tokenCharsI, 'xmp')) {
+              endScriptTag = '/xmp';
+            }
+          }
+          // disallow the tag if endScriptTag is set and this was not an open
+          // tag.
+          if (endScriptTag && tokenCharsI && '/' == tokenChars[0]) {
+            tokenStyle = null;
+          }
+        }
+        if (null != tokenStyle) {
+          tokenEnds.push(new PR_TokenEnd(tokenStart, PR_PLAIN));
+          tokenEnds.push(new PR_TokenEnd(k + next, tokenStyle));
+        }
+      }
+    }
+    k += chunk.token.length;
+  }
+  tokenEnds.push(new PR_TokenEnd(k, PR_PLAIN));
+
+  return tokenEnds;
+}
+
+/** splits the given string into comment, string, and "other" tokens.
+  * @return an array of PR_Tokens with style in
+  *   (PR_STRING, PR_COMMENT, PR_PLAIN, null)
+  *   The result array may contain spurious zero length tokens.  Ignore them.
+  *
+  * @private
+  */
+function PR_splitStringAndCommentTokens(chunks) {
+  // a state machine to split out comments, strings, and other stuff
+  var tokenEnds = new Array();  // positions of ends of tokens in absolute space
+  var state = 0;  // FSM state variable
+  var delim = -1;  // string delimiter
+  var k = 0;  // absolute position of beginning of current chunk
+  for (var ci = 0, nc = chunks.length; ci < nc; ++ci) {
+    var chunk = chunks[ci];
+    var s = chunk.token;
+    if (PR_PLAIN == chunk.style) {
+      for (var i = 0, n = s.length; i < n; ++i) {
+        var ch = s.charAt(i);
+        if (0 == state) {
+          if (ch == '"' || ch == '\'' || ch == '`') {
+            tokenEnds.push(new PR_TokenEnd(k + i, PR_PLAIN));
+            state = 1;
+            delim = ch;
+          } else if (ch == '/') {
+            state = 3;
+          } else if (ch == '#') {
+            tokenEnds.push(new PR_TokenEnd(k + i, PR_PLAIN));
+            state = 4;
+          }
+        } else if (1 == state) {
+          if (ch == delim) {
+            state = 0;
+            tokenEnds.push(new PR_TokenEnd(k + i + 1, PR_STRING));
+          } else if (ch == '\\') {
+            state = 2;
+          }
+        } else if (2 == state) {
+          state = 1;
+        } else if (3 == state) {
+          if (ch == '/') {
+            state = 4;
+            tokenEnds.push(new PR_TokenEnd(k + i - 1, PR_PLAIN));
+          } else if (ch == '*') {
+            state = 5;
+            tokenEnds.push(new PR_TokenEnd(k + i - 1, PR_PLAIN));
+          } else {
+            state = 0;
+            // next loop will reenter state 0 without same value of i, so
+            // ch will be reconsidered as start of new token.
+            --i;
+          }
+        } else if (4 == state) {
+          if (ch == '\r' || ch == '\n') {
+            state = 0;
+            tokenEnds.push(new PR_TokenEnd(k + i, PR_COMMENT));
+          }
+        } else if (5 == state) {
+          if (ch == '*') {
+            state = 6;
+          }
+        } else if (6 == state) {
+          if (ch == '/') {
+            state = 0;
+            tokenEnds.push(new PR_TokenEnd(k + i + 1, PR_COMMENT));
+          } else if (ch != '*') {
+            state = 5;
+          }
+        }
+      }
+    }
+    k += s.length;
+  }
+  tokenEnds.push(new PR_TokenEnd(k, PR_PLAIN));  // a token ends at the end
+
+  return PR_splitChunks(chunks, tokenEnds);
+}
+
+/** used by lexSource to split a non string, non comment token.
+  * @private
+  */
+function PR_splitNonStringNonCommentToken(s, outlist) {
+  var pos = 0;
+  var state = 0;
+  for (var i = 0; i <= s.length; i++) {
+    var ch = s.charAt(i);
+    // the next state.
+    // if set to -1 then it will cause a reentry to state 0 without consuming
+    // another character.
+    var nstate = state;
+
+    if (i == s.length) {
+      // nstate will not be equal to state, so it will append the token
+      nstate = -2;
+    } else {
+      switch (state) {
+      case 0:  // whitespace state
+        if (PR_isIdentifierStart(ch)) {
+          nstate = 1;
+        } else if (PR_isDigitChar(ch)) {
+          nstate = 2;
+        } else if (!PR_isSpaceChar(ch)) {
+          nstate = 3;
+        }
+        if (nstate && pos < i) {
+          var t = s.substring(pos, i);
+          outlist.push(new PR_Token(t, PR_PLAIN));
+          pos = i;
+        }
+        break;
+      case 1:  // identifier state
+        if (!PR_isIdentifierPart(ch)) {
+          nstate = -1;
+        }
+        break;
+      case 2:  // number literal state
+        // handle numeric literals like
+        // 0x7f 300UL 100_000
+
+        // this does not treat floating point values as a single literal
+        //   0.1 and 3e-6
+        // are each split into multiple tokens
+        if (!(PR_isDigitChar(ch) || PR_isWordChar(ch) || ch == '_')) {
+          nstate = -1;
+        }
+        break;
+      case 3:  // punctuation state
+        if (PR_isIdentifierStart(ch) || PR_isDigitChar(ch) ||
+            PR_isSpaceChar(ch)) {
+          nstate = -1;
+        }
+        break;
+      }
+    }
+
+    if (nstate != state) {
+      if (nstate < 0) {
+        if (i > pos) {
+          var t = s.substring(pos, i);
+          var ch0 = t.charAt(0);
+          var style;
+          if (PR_isIdentifierStart(ch0)) {
+            if (PR_keywords[t]) {
+              style = PR_KEYWORD;
+            } else if (ch0 === '@') {
+              style = PR_LITERAL;
+            } else {
+              // Treat any word that starts with an uppercase character and
+              // contains at least one lowercase character as a type, or
+              // ends with _t.
+              // This works perfectly for Java, pretty well for C++, and
+              // passably for Python.  The _t catches C structs.
+              var isType = false;
+              if (ch0 >= 'A' && ch0 <= 'Z') {
+                for (var j = 1; j < t.length; j++) {
+                  var ch1 = t.charAt(j);
+                  if (ch1 >= 'a' && ch1 <= 'z') {
+                    isType = true;
+                    break;
+                  }
+                }
+                if (!isType && t.length >= 2 &&
+                    t.substring(t.length - 2) == '_t') {
+                  isType = true;
+                }
+              }
+              style = isType ? PR_TYPE : PR_PLAIN;
+            }
+          } else if (PR_isDigitChar(ch0)) {
+            style = PR_LITERAL;
+          } else if (!PR_isSpaceChar(ch0)) {
+            style = PR_PUNCTUATION;
+          } else {
+            style = PR_PLAIN;
+          }
+          pos = i;
+          outlist.push(new PR_Token(t, style));
+        }
+
+        state = 0;
+        if (nstate == -1) {
+          // don't increment.  This allows us to use state 0 to redispatch based
+          // on the current character.
+          i--;
+          continue;
+        }
+      }
+      state = nstate;
+    }
+  }
+}
+
+/** split a group of chunks of markup.
+  * @private
+  */
+function PR_tokenizeMarkup(chunks) {
+  if (!(chunks && chunks.length)) { return chunks; }
+
+  var tokenEnds = PR_splitMarkup(chunks);
+  return PR_splitChunks(chunks, tokenEnds);
+}
+
+/** split tags attributes and their values out from the tag name, and
+  * recursively lex source chunks.
+  * @private
+  */
+function PR_splitTagAttributes(tokens) {
+  var tokensOut = new Array();
+  var state = 0;
+  var stateStyle = PR_TAG;
+  var delim = null;  // attribute delimiter for quoted value state.
+  var decodeHelper = new PR_DecodeHelper();
+  for (var ci = 0; ci < tokens.length; ++ci) {
+    var tok = tokens[ci];
+    if (PR_TAG == tok.style) {
+      var s = tok.token;
+      var start = 0;
+      for (var i = 0; i < s.length; /* i = next at bottom */) {
+        decodeHelper.decode(s, i);
+        var ch = decodeHelper.ch;
+        var next = decodeHelper.next;
+
+        var emitEnd = null;  // null or position of end of chunk to emit.
+        var nextStyle = null;  // null or next value of stateStyle
+        if (ch == '>') {
+          if (PR_TAG != stateStyle) {
+            emitEnd = i;
+            nextStyle = PR_TAG;
+          }
+        } else {
+          switch (state) {
+            case 0:
+              if ('<' == ch) { state = 1; }
+              break;
+            case 1:
+              if (PR_isSpaceChar(ch)) { state = 2; }
+              break;
+            case 2:
+              if (!PR_isSpaceChar(ch)) {
+                nextStyle = PR_ATTRIB_NAME;
+                emitEnd = i;
+                state = 3;
+              }
+              break;
+            case 3:
+              if ('=' == ch) {
+                emitEnd = i;
+                nextStyle = PR_TAG;
+                state = 5;
+              } else if (PR_isSpaceChar(ch)) {
+                emitEnd = i;
+                nextStyle = PR_TAG;
+                state = 4;
+              }
+              break;
+            case 4:
+              if ('=' == ch) {
+                state = 5;
+              } else if (!PR_isSpaceChar(ch)) {
+                emitEnd = i;
+                nextStyle = PR_ATTRIB_NAME;
+                state = 3;
+              }
+              break;
+            case 5:
+              if ('"' == ch || '\'' == ch) {
+                emitEnd = i;
+                nextStyle = PR_ATTRIB_VALUE;
+                state = 6;
+                delim = ch;
+              } else if (!PR_isSpaceChar(ch)) {
+                emitEnd = i;
+                nextStyle = PR_ATTRIB_VALUE;
+                state = 7;
+              }
+              break;
+            case 6:
+              if (ch == delim) {
+                emitEnd = next;
+                nextStyle = PR_TAG;
+                state = 2;
+              }
+              break;
+            case 7:
+              if (PR_isSpaceChar(ch)) {
+                emitEnd = i;
+                nextStyle = PR_TAG;
+                state = 2;
+              }
+              break;
+          }
+        }
+        if (emitEnd) {
+          if (emitEnd > start) {
+            tokensOut.push(
+                new PR_Token(s.substring(start, emitEnd), stateStyle));
+            start = emitEnd;
+          }
+          stateStyle = nextStyle;
+        }
+        i = next;
+      }
+      if (s.length > start) {
+        tokensOut.push(new PR_Token(s.substring(start, s.length), stateStyle));
+      }
+    } else {
+      if (tok.style) {
+        state = 0;
+        stateStyle = PR_TAG;
+      }
+      tokensOut.push(tok);
+    }
+  }
+  return tokensOut;
+}
+
+/** identify regions of markup that are really source code, and recursivley
+  * lex them.
+  * @private
+  */
+function PR_splitSourceNodes(tokens) {
+  var tokensOut = new Array();
+  // when we see a <script> tag, store '/' here so that we know to end the
+  // source processing
+  var endScriptTag = null;
+  var decodeHelper = new PR_DecodeHelper();
+
+  var sourceChunks = null;
+
+  for (var ci = 0, nc = tokens.length; ci < nc; ++ci) {
+    var tok = tokens[ci];
+    if (null == tok.style) {
+      tokens.push(tok);
+      continue;
+    }
+
+    var s = tok.token;
+
+    if (null == endScriptTag) {
+      if (PR_SOURCE == tok.style) {
+        // split off any starting and trailing <?, <%
+        if ('<' == decodeHelper.decode(s, 0)) {
+          decodeHelper.decode(s, decodeHelper.next);
+          if ('%' == decodeHelper.ch || '?' == decodeHelper.ch) {
+            endScriptTag = decodeHelper.ch;
+            tokensOut.push(new PR_Token(s.substring(0, decodeHelper.next),
+                                        PR_TAG));
+            s = s.substring(decodeHelper.next, s.length);
+          }
+        }
+      } else if (PR_TAG == tok.style) {
+        if ('<' == decodeHelper.decode(s, 0) &&
+            '/' != s.charAt(decodeHelper.next)) {
+          var tagContent = s.substring(decodeHelper.next).toLowerCase();
+          // FIXME(msamuel): this does not mirror exactly the code in
+          // in PR_splitMarkup that defers splitting tags inside script and
+          // style blocks.
+          if (PR_startsWith(tagContent, 'script') ||
+              PR_startsWith(tagContent, 'style') ||
+              PR_startsWith(tagContent, 'xmp')) {
+            endScriptTag = '/';
+          }
+        }
+      }
+    }
+
+    if (null != endScriptTag) {
+      var endTok = null;
+      if (PR_SOURCE == tok.style) {
+        if (endScriptTag == '%' || endScriptTag == '?') {
+          var pos = s.lastIndexOf(endScriptTag);
+          if (pos >= 0 && '>' == decodeHelper.decode(s, pos + 1) &&
+              s.length == decodeHelper.next) {
+            endTok = new PR_Token(s.substring(pos, s.length), PR_TAG);
+            s = s.substring(0, pos);
+          }
+        }
+        if (null == sourceChunks) { sourceChunks = new Array(); }
+        sourceChunks.push(new PR_Token(s, PR_PLAIN));
+      } else if (PR_PLAIN == tok.style) {
+        if (null == sourceChunks) { sourceChunks = new Array(); }
+        sourceChunks.push(tok);
+      } else if (PR_TAG == tok.style) {
+        // if it starts with </ then it must be the end tag.
+        if ('<' == decodeHelper.decode(tok.token, 0) &&
+            tok.token.length > decodeHelper.next &&
+            '/' == decodeHelper.decode(tok.token, decodeHelper.next)) {
+          endTok = tok;
+        } else {
+          tokensOut.push(tok);
+        }
+      } else {
+        if (sourceChunks) {
+          sourceChunks.push(tok);
+        } else {
+          // push remaining tag and attribute tokens from the opening tag
+          tokensOut.push(tok);
+        }
+      }
+      if (endTok) {
+        if (sourceChunks) {
+          var sourceTokens = PR_lexSource(sourceChunks);
+          tokensOut.push(new PR_Token('<span class=embsrc>', null));
+          for (var si = 0, ns = sourceTokens.length; si < ns; ++si) {
+            tokensOut.push(sourceTokens[si]);
+          }
+          tokensOut.push(new PR_Token('</span>', null));
+          sourceChunks = null;
+        }
+        tokensOut.push(endTok);
+        endScriptTag = null;
+      }
+    } else {
+      tokensOut.push(tok);
+    }
+  }
+  return tokensOut;
+}
+
+/** splits the quotes from an attribute value.
+  * ['"foo"'] -> ['"', 'foo', '"']
+  * @private
+  */
+function PR_splitAttributeQuotes(tokens) {
+  var firstPlain = null, lastPlain = null;
+  for (var i = 0; i < tokens.length; ++i) {
+    if (PR_PLAIN == tokens[i].style) {
+      firstPlain = i;
+      break;
+    }
+  }
+  for (var i = tokens.length; --i >= 0;) {
+    if (PR_PLAIN == tokens[i].style) {
+      lastPlain = i;
+      break;
+    }
+  }
+  if (null == firstPlain) { return tokens; }
+
+  var decodeHelper = new PR_DecodeHelper();
+  var fs = tokens[firstPlain].token;
+  var fc = decodeHelper.decode(fs, 0);
+  if ('"' != fc && '\'' != fc) {
+    return tokens;
+  }
+  var fpos = decodeHelper.next;
+
+  var ls = tokens[lastPlain].token;
+  var lpos = ls.lastIndexOf('&');
+  if (lpos < 0) { lpos = ls.length - 1; }
+  var lc = decodeHelper.decode(ls, lpos);
+  if (lc != fc || decodeHelper.next != ls.length) {
+    lc = null;
+    lpos = ls.length;
+  }
+
+  var tokensOut = new Array();
+  for (var i = 0; i < firstPlain; ++i) {
+    tokensOut.push(tokens[i]);
+  }
+  tokensOut.push(new PR_Token(fs.substring(0, fpos), PR_ATTRIB_VALUE));
+  if (lastPlain == firstPlain) {
+    tokensOut.push(new PR_Token(fs.substring(fpos, lpos), PR_PLAIN));
+  } else {
+    tokensOut.push(new PR_Token(fs.substring(fpos, fs.length), PR_PLAIN));
+    for (var i = firstPlain + 1; i < lastPlain; ++i) {
+      tokensOut.push(tokens[i]);
+    }
+    if (lc) {
+      tokens.push(new PR_Token(ls.substring(0, lpos), PR_PLAIN));
+    } else {
+      tokens.push(tokens[lastPlain]);
+    }
+  }
+  if (lc) {
+    tokensOut.push(new PR_Token(ls.substring(lpos, ls.length), PR_PLAIN));
+  }
+  for (var i = lastPlain + 1; i < tokens.length; ++i) {
+    tokensOut.push(tokens[i]);
+  }
+  return tokensOut;
+}
+
+/** identify attribute values that really contain source code and recursively
+  * lex them.
+  * @private
+  */
+function PR_splitSourceAttributes(tokens) {
+  var tokensOut = new Array();
+
+  var sourceChunks = null;
+  var inSource = false;
+  var name = '';
+
+  for (var ci = 0, nc = tokens.length; ci < nc; ++ci) {
+    var tok = tokens[ci];
+    var outList = tokensOut;
+    if (PR_TAG == tok.style) {
+      if (inSource) {
+        inSource = false;
+        name = '';
+        if (sourceChunks) {
+          tokensOut.push(new PR_Token('<span class=embsrc>', null));
+          var sourceTokens =
+            PR_lexSource(PR_splitAttributeQuotes(sourceChunks));
+          for (var si = 0, ns = sourceTokens.length; si < ns; ++si) {
+            tokensOut.push(sourceTokens[si]);
+          }
+          tokensOut.push(new PR_Token('</span>', null));
+          sourceChunks = null;
+        }
+      } else if (name && tok.token.indexOf('=') >= 0) {
+        var nameLower = name.toLowerCase();
+        if (PR_startsWith(nameLower, 'on') || 'style' == nameLower) {
+          inSource = true;
+        }
+      } else {
+        name = '';
+      }
+    } else if (PR_ATTRIB_NAME == tok.style) {
+      name += tok.token;
+    } else if (PR_ATTRIB_VALUE == tok.style) {
+      if (inSource) {
+        if (null == sourceChunks) { sourceChunks = new Array(); }
+        outList = sourceChunks;
+        tok = new PR_Token(tok.token, PR_PLAIN);
+      }
+    } else {
+      if (sourceChunks) {
+        outList = sourceChunks;
+      }
+    }
+    outList.push(tok);
+  }
+  return tokensOut;
+}
+
+/** returns a list of PR_Token objects given chunks of source code.
+  *
+  * This code assumes that < tokens are html escaped, but " are not.
+  * It will do a resonable job with <, but will not recognize an &quot;
+  * as starting a string.
+  *
+  * This code treats ", ', and ` as string delimiters, and \ as a string escape.
+  * It does not recognize double delimiter escapes, or perl's qq() style
+  * strings.
+  *
+  * It recognizes C, C++, and shell style comments.
+  *
+  * @param chunks PR_Tokens with style in (null, PR_PLAIN)
+  */
+function PR_lexSource(chunks) {
+  // positions of ends of tokens in order
+  var tokensIn = PR_splitStringAndCommentTokens(chunks);
+
+  // split entities out of so that we know to treat them as single units.
+  tokensIn = PR_splitEntities(tokensIn);
+
+  // split non comment|string tokens on whitespace and word boundaries
+  var tokensOut = new Array();
+  for (var i = 0; i < tokensIn.length; ++i) {
+    var tok = tokensIn[i];
+    var t = tok.token;
+    var s = tok.style;
+
+    if (PR_PLAIN == s) {
+      PR_splitNonStringNonCommentToken(t, tokensOut);
+      continue;
+    }
+    tokensOut.push(tok);
+  }
+
+  return tokensOut;
+}
+
+/** returns a list of PR_Token objects given a string of markup.
+  *
+  * This code assumes that < tokens are html escaped, but " are not.
+  * It will do a resonable job with <, but will not recognize an &quot;
+  * as starting a string.
+  *
+  * This code recognizes a number of constructs.
+  * <!-- ... --> comment
+  * <!\w ... >   declaration
+  * <\w ... >    tag
+  * </\w ... >   tag
+  * <?...?>      embedded source
+  * &[#\w]...;   entity
+  *
+  * It does not recognizes %foo; entities.
+  *
+  * It will recurse into any <style>, <script>, and on* attributes using
+  * PR_lexSource.
+  */
+function PR_lexMarkup(chunks) {
+  // This function works as follows:
+  // 1) Start by splitting the markup into text and tag chunks
+  //    Input:  String s
+  //    Output: List<PR_Token> where style in (PR_PLAIN, null)
+  // 2) Then split the text chunks further into comments, declarations,
+  //    tags, etc.
+  //    After each split, consider whether the token is the start of an
+  //    embedded source section, i.e. is an open <script> tag.  If it is,
+  //    find the corresponding close token, and don't bother to lex in between.
+  //    Input:  List<String>
+  //    Output: List<PR_Token> with style in (PR_TAG, PR_PLAIN, PR_SOURCE, null)
+  // 3) Finally go over each tag token and split out attribute names and values.
+  //    Input:  List<PR_Token>
+  //    Output: List<PR_Token> where style in
+  //            (PR_TAG, PR_PLAIN, PR_SOURCE, NAME, VALUE, null)
+  var tokensOut = PR_tokenizeMarkup(chunks);
+  tokensOut = PR_splitTagAttributes(tokensOut);
+  tokensOut = PR_splitSourceNodes(tokensOut);
+  tokensOut = PR_splitSourceAttributes(tokensOut);
+  return tokensOut;
+}
+
+/** classify the string as either source or markup and lex appropriately. */
+function PR_lexOne(s) {
+  var chunks = PR_chunkify(s);
+  // treat it as markup if the first non whitespace character is a < and the
+  // last non-whitespace character is a >
+  var isMarkup = false;
+  for (var i = 0; i < chunks.length; ++i) {
+    if (PR_PLAIN == chunks[i].style) {
+      if (PR_startsWith(PR_trim(chunks[i].token), '&lt;')) {
+        for (var j = chunks.length; --j >= 0;) {
+          if (PR_PLAIN == chunks[j].style) {
+            isMarkup = PR_endsWith(PR_trim(chunks[j].token), '&gt;');
+            break;
+          }
+        }
+      }
+      break;
+    }
+  }
+  return isMarkup ? PR_lexMarkup(chunks) : PR_lexSource(chunks);
+}
+
+/** pretty print a chunk of code.
+  *
+  * @param s code as html
+  * @return code as html, but prettier
+  */
+function prettyPrintOne(s) {
+  try {
+    var tokens = PR_lexOne(s);
+    var out = '';
+    var lastStyle = null;
+    for (var i = 0; i < tokens.length; i++) {
+      var t = tokens[i];
+      if (t.style != lastStyle) {
+        if (lastStyle != null) {
+          out += '</span>';
+        }
+        if (t.style != null) {
+          out += '<span class=' + t.style + '>';
+        }
+        lastStyle = t.style;
+      }
+      var html = t.token;
+      if (null != t.style) {
+        // This interacts badly with the wiki which introduces paragraph tags
+        // int pre blocks for some strange reason.
+        // It's necessary for IE though which seems to lose the preformattedness
+        // of <pre> tags when their innerHTML is assigned.
+        html = html.replace(/(?:\r\n?)|\n/g, '<br>').replace(/  /g, '&nbsp; ');
+      }
+      out += html;
+    }
+    if (lastStyle != null) {
+      out += '</span>';
+    }
+    return out;
+  } catch (e) {
+    //alert(e.stack);  // DISABLE in production
+    return s;
+  }
+}
+
+/** find all the < pre > and < code > tags in the DOM with class=prettyprint and
+  * prettify them.
+  */
+function prettyPrint() {
+  // fetch a list of nodes to rewrite
+  var codeSegments = [
+      document.getElementsByTagName('pre'),
+      document.getElementsByTagName('code'),
+      document.getElementsByTagName('xmp') ];
+  var elements = [];
+  for (var i = 0; i < codeSegments.length; ++i) {
+    for (var j = 0; j < codeSegments[i].length; ++j) {
+      elements.push(codeSegments[i][j]);
+    }
+  }
+  codeSegments = null;
+
+  // the loop is broken into a series of continuations to make sure that we
+  // don't make the browser unresponsive when rewriting a large page.
+  var k = 0;
+
+  function doWork() {
+    var endTime = new Date().getTime() + 250;
+    for (; k < elements.length && new Date().getTime() < endTime; k++) {
+      var cs = elements[k];
+      if (cs.className && cs.className.indexOf('prettyprint') >= 0) {
+
+        // make sure this is not nested in an already prettified element
+        var nested = false;
+        for (var p = cs.parentNode; p != null; p = p.parentNode) {
+          if ((p.tagName == 'pre' || p.tagName == 'code' ||
+               p.tagName == 'xmp') &&
+              p.className && p.className.indexOf('prettyprint') >= 0) {
+            nested = true;
+            break;
+          }
+        }
+        if (!nested) {
+          // XMP tags contain unescaped entities so require special handling.
+          var isRawContent = 'XMP' == cs.tagName;
+
+          // fetch the content as a snippet of properly escaped HTML
+          var content = cs.innerHTML;
+          if (isRawContent) {
+            content = PR_textToHtml(content);
+          }
+
+          // do the pretty printing
+          var newContent = prettyPrintOne(content);
+
+          // push the prettified html back into the tag.
+          if (!isRawContent) {
+            // just replace the old html with the new
+            cs.innerHTML = newContent;
+          } else {
+            // we need to change the tag to a <pre> since <xmp>s do not allow
+            // embedded tags such as the span tags used to attach styles to
+            // sections of source code.
+            var pre = document.createElement('PRE');
+            for (var i = 0; i < cs.attributes.length; ++i) {
+              var a = cs.attributes[i];
+              if (a.specified) {
+                pre.setAttribute(a.name, a.value);
+              }
+            }
+            pre.innerHTML = newContent;
+            // remove the old
+            cs.parentNode.replaceChild(pre, cs);
+          }
+        }
+      }
+    }
+    if (k < elements.length) {
+      // finish up in a continuation
+      setTimeout(doWork, 250);
+    }
+  }
+
+  doWork();
+}
+