web/query/vendor/couchapp/lib/markdown.js

   1 //
   2 // showdown.js -- A javascript port of Markdown.
   3 //
   4 // Copyright (c) 2007 John Fraser.
   5 //
   6 // Original Markdown Copyright (c) 2004-2005 John Gruber
   7 //   <http://daringfireball.net/projects/markdown/>
   8 //
   9 // Redistributable under a BSD-style open source license.
  10 // See license.txt for more information.
  11 //
  12 // The full source distribution is at:
  13 //
  14 //                              A A L
  15 //                              T C A
  16 //                              T K B
  17 //
  18 //   <http://www.attacklab.net/>
  19 //
  20
  21 //
  22 // Wherever possible, Showdown is a straight, line-by-line port
  23 // of the Perl version of Markdown.
  24 //
  25 // This is not a normal parser design; it's basically just a
  26 // series of string substitutions.  It's hard to read and
  27 // maintain this way,  but keeping Showdown close to the original
  28 // design makes it easier to port new features.
  29 //
  30 // More importantly, Showdown behaves like markdown.pl in most
  31 // edge cases.  So web applications can do client-side preview
  32 // in Javascript, and then build identical HTML on the server.
  33 //
  34 // This port needs the new RegExp functionality of ECMA 262,
  35 // 3rd Edition (i.e. Javascript 1.5).  Most modern web browsers
  36 // should do fine.  Even with the new regular expression features,
  37 // We do a lot of work to emulate Perl's regex functionality.
  38 // The tricky changes in this file mostly have the "attacklab:"
  39 // label.  Major or self-explanatory changes don't.
  40 //
  41 // Smart diff tools like Araxis Merge will be able to match up
  42 // this file with markdown.pl in a useful way.  A little tweaking
  43 // helps: in a copy of markdown.pl, replace "#" with "//" and
  44 // replace "$text" with "text".  Be sure to ignore whitespace
  45 // and line endings.
  46 //
  47
  48
  49 //
  50 // Showdown usage:
  51 //
  52 //   var text = "Markdown *rocks*.";
  53 //
  54 //   var markdown = require("markdown");
  55 //   var html = markdown.encode(text);
  56 //
  57 //   print(html);
  58 //
  59 // Note: move the sample code to the bottom of this
  60 // file before uncommenting it.
  61 //
  62
  63
  64 //
  65 // Globals:
  66 //
  67
  68 // Global hashes, used by various utility routines
  69 var g_urls;
  70 var g_titles;
  71 var g_html_blocks;
  72
  73 // Used to track when we're inside an ordered or unordered list
  74 // (see _ProcessListItems() for details):
  75 var g_list_level = 0;
  76
  77
  78 exports.makeHtml = function(text) {
  79 //
  80 // Main function. The order in which other subs are called here is
  81 // essential. Link and image substitutions need to happen before
  82 // _EscapeSpecialCharsWithinTagAttributes(), so that any *'s or _'s in the <a>
  83 // and <img> tags get encoded.
  84 //
  85
  86         // Clear the global hashes. If we don't clear these, you get conflicts
  87         // from other articles when generating a page which contains more than
  88         // one article (e.g. an index page that shows the N most recent
  89         // articles):
  90         g_urls = new Array();
  91         g_titles = new Array();
  92         g_html_blocks = new Array();
  93
  94         // attacklab: Replace ~ with ~T
  95         // This lets us use tilde as an escape char to avoid md5 hashes
  96         // The choice of character is arbitray; anything that isn't
  97     // magic in Markdown will work.
  98         text = text.replace(/~/g,"~T");
  99
 100         // attacklab: Replace $ with ~D
 101         // RegExp interprets $ as a special character
 102         // when it's in a replacement string
 103         text = text.replace(/\$/g,"~D");
 104
 105         // Standardize line endings
 106         text = text.replace(/\r\n/g,"\n"); // DOS to Unix
 107         text = text.replace(/\r/g,"\n"); // Mac to Unix
 108
 109         // Make sure text begins and ends with a couple of newlines:
 110         text = "\n\n" + text + "\n\n";
 111
 112         // Convert all tabs to spaces.
 113         text = _Detab(text);
 114
 115         // Strip any lines consisting only of spaces and tabs.
 116         // This makes subsequent regexen easier to write, because we can
 117         // match consecutive blank lines with /\n+/ instead of something
 118         // contorted like /[ \t]*\n+/ .
 119         text = text.replace(/^[ \t]+$/mg,"");
 120
 121         // Turn block-level HTML blocks into hash entries
 122         text = _HashHTMLBlocks(text);
 123
 124         // Strip link definitions, store in hashes.
 125         text = _StripLinkDefinitions(text);
 126
 127         text = _RunBlockGamut(text);
 128
 129         text = _UnescapeSpecialChars(text);
 130
 131         // attacklab: Restore dollar signs
 132         text = text.replace(/~D/g,"$$");
 133
 134         // attacklab: Restore tildes
 135         text = text.replace(/~T/g,"~");
 136         return text;
 137 }
 138
 139
 140 var _StripLinkDefinitions = function(text) {
 141 //
 142 // Strips link definitions from text, stores the URLs and titles in
 143 // hash references.
 144 //
 145
 146         // Link defs are in the form: ^[id]: url "optional title"
 147
 148         /*
 149                 var text = text.replace(/
 150                                 ^[ ]{0,3}\[(.+)\]:  // id = $1  attacklab: g_tab_width - 1
 151                                   [ \t]*
 152                                   \n?                           // maybe *one* newline
 153                                   [ \t]*
 154                                 <?(\S+?)>?                      // url = $2
 155                                   [ \t]*
 156                                   \n?                           // maybe one newline
 157                                   [ \t]*
 158                                 (?:
 159                                   (\n*)                         // any lines skipped = $3 attacklab: lookbehind removed
 160                                   ["(]
 161                                   (.+?)                         // title = $4
 162                                   [")]
 163                                   [ \t]*
 164                                 )?                                      // title is optional
 165                                 (?:\n+|$)
 166                           /gm,
 167                           function(){...});
 168         */
 169         var text = text.replace(/^[ ]{0,3}\[(.+)\]:[ \t]*\n?[ \t]*<?(\S+?)>?[ \t]*\n?[ \t]*(?:(\n*)["(](.+?)[")][ \t]*)?(?:\n+|\Z)/gm,
 170                 function (wholeMatch,m1,m2,m3,m4) {
 171                         m1 = m1.toLowerCase();
 172                         g_urls[m1] = _EncodeAmpsAndAngles(m2);  // Link IDs are case-insensitive
 173                         if (m3) {
 174                                 // Oops, found blank lines, so it's not a title.
 175                                 // Put back the parenthetical statement we stole.
 176                                 return m3+m4;
 177                         } else if (m4) {
 178                                 g_titles[m1] = m4.replace(/"/g,"&quot;");
 179                         }
 180
 181                         // Completely remove the definition from the text
 182                         return "";
 183                 }
 184         );
 185
 186         return text;
 187 }
 188
 189
 190 var _HashHTMLBlocks = function(text) {
 191         // attacklab: Double up blank lines to reduce lookaround
 192         text = text.replace(/\n/g,"\n\n");
 193
 194         // Hashify HTML blocks:
 195         // We only want to do this for block-level HTML tags, such as headers,
 196         // lists, and tables. That's because we still want to wrap <p>s around
 197         // "paragraphs" that are wrapped in non-block-level tags, such as anchors,
 198         // phrase emphasis, and spans. The list of tags we're looking for is
 199         // hard-coded:
 200         var block_tags_a = "p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math|ins|del"
 201         var block_tags_b = "p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math"
 202
 203         // First, look for nested blocks, e.g.:
 204         //   <div>
 205         //     <div>
 206         //     tags for inner block must be indented.
 207         //     </div>
 208         //   </div>
 209         //
 210         // The outermost tags must start at the left margin for this to match, and
 211         // the inner nested divs must be indented.
 212         // We need to do this before the next, more liberal match, because the next
 213         // match will start at the first `<div>` and stop at the first `</div>`.
 214
 215         // attacklab: This regex can be expensive when it fails.
 216         /*
 217                 var text = text.replace(/
 218                 (                                               // save in $1
 219                         ^                                       // start of line  (with /m)
 220                         <($block_tags_a)        // start tag = $2
 221                         \b                                      // word break
 222                                                                 // attacklab: hack around khtml/pcre bug...
 223                         [^\r]*?\n                       // any number of lines, minimally matching
 224                         </\2>                           // the matching end tag
 225                         [ \t]*                          // trailing spaces/tabs
 226                         (?=\n+)                         // followed by a newline
 227                 )                                               // attacklab: there are sentinel newlines at end of document
 228                 /gm,function(){...}};
 229         */
 230         text = text.replace(/^(<(p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math|ins|del)\b[^\r]*?\n<\/\2>[ \t]*(?=\n+))/gm,hashElement);
 231
 232         //
 233         // Now match more liberally, simply from `\n<tag>` to `</tag>\n`
 234         //
 235
 236         /*
 237                 var text = text.replace(/
 238                 (                                               // save in $1
 239                         ^                                       // start of line  (with /m)
 240                         <($block_tags_b)        // start tag = $2
 241                         \b                                      // word break
 242                                                                 // attacklab: hack around khtml/pcre bug...
 243                         [^\r]*?                         // any number of lines, minimally matching
 244                         .*</\2>                         // the matching end tag
 245                         [ \t]*                          // trailing spaces/tabs
 246                         (?=\n+)                         // followed by a newline
 247                 )                                               // attacklab: there are sentinel newlines at end of document
 248                 /gm,function(){...}};
 249         */
 250         text = text.replace(/^(<(p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math)\b[^\r]*?.*<\/\2>[ \t]*(?=\n+)\n)/gm,hashElement);
 251
 252         // Special case just for <hr />. It was easier to make a special case than
 253         // to make the other regex more complicated.
 254
 255         /*
 256                 text = text.replace(/
 257                 (                                               // save in $1
 258                         \n\n                            // Starting after a blank line
 259                         [ ]{0,3}
 260                         (<(hr)                          // start tag = $2
 261                         \b                                      // word break
 262                         ([^<>])*?                       //
 263                         \/?>)                           // the matching end tag
 264                         [ \t]*
 265                         (?=\n{2,})                      // followed by a blank line
 266                 )
 267                 /g,hashElement);
 268         */
 269         text = text.replace(/(\n[ ]{0,3}(<(hr)\b([^<>])*?\/?>)[ \t]*(?=\n{2,}))/g,hashElement);
 270
 271         // Special case for standalone HTML comments:
 272
 273         /*
 274                 text = text.replace(/
 275                 (                                               // save in $1
 276                         \n\n                            // Starting after a blank line
 277                         [ ]{0,3}                        // attacklab: g_tab_width - 1
 278                         <!
 279                         (--[^\r]*?--\s*)+
 280                         >
 281                         [ \t]*
 282                         (?=\n{2,})                      // followed by a blank line
 283                 )
 284                 /g,hashElement);
 285         */
 286         text = text.replace(/(\n\n[ ]{0,3}<!(--[^\r]*?--\s*)+>[ \t]*(?=\n{2,}))/g,hashElement);
 287
 288         // PHP and ASP-style processor instructions (<?...?> and <%...%>)
 289
 290         /*
 291                 text = text.replace(/
 292                 (?:
 293                         \n\n                            // Starting after a blank line
 294                 )
 295                 (                                               // save in $1
 296                         [ ]{0,3}                        // attacklab: g_tab_width - 1
 297                         (?:
 298                                 <([?%])                 // $2
 299                                 [^\r]*?
 300                                 \2>
 301                         )
 302                         [ \t]*
 303                         (?=\n{2,})                      // followed by a blank line
 304                 )
 305                 /g,hashElement);
 306         */
 307         text = text.replace(/(?:\n\n)([ ]{0,3}(?:<([?%])[^\r]*?\2>)[ \t]*(?=\n{2,}))/g,hashElement);
 308
 309         // attacklab: Undo double lines (see comment at top of this function)
 310         text = text.replace(/\n\n/g,"\n");
 311         return text;
 312 }
 313
 314 var hashElement = function(wholeMatch,m1) {
 315         var blockText = m1;
 316
 317         // Undo double lines
 318         blockText = blockText.replace(/\n\n/g,"\n");
 319         blockText = blockText.replace(/^\n/,"");
 320
 321         // strip trailing blank lines
 322         blockText = blockText.replace(/\n+$/g,"");
 323
 324         // Replace the element text with a marker ("~KxK" where x is its key)
 325         blockText = "\n\n~K" + (g_html_blocks.push(blockText)-1) + "K\n\n";
 326
 327         return blockText;
 328 };
 329
 330 var _RunBlockGamut = function(text) {
 331 //
 332 // These are all the transformations that form block-level
 333 // tags like paragraphs, headers, and list items.
 334 //
 335         text = _DoHeaders(text);
 336
 337         // Do Horizontal Rules:
 338         var key = hashBlock("<hr />");
 339         text = text.replace(/^[ ]{0,2}([ ]?\*[ ]?){3,}[ \t]*$/gm,key);
 340         text = text.replace(/^[ ]{0,2}([ ]?\-[ ]?){3,}[ \t]*$/gm,key);
 341         text = text.replace(/^[ ]{0,2}([ ]?\_[ ]?){3,}[ \t]*$/gm,key);
 342
 343         text = _DoLists(text);
 344         text = _DoCodeBlocks(text);
 345         text = _DoBlockQuotes(text);
 346
 347         // We already ran _HashHTMLBlocks() before, in Markdown(), but that
 348         // was to escape raw HTML in the original Markdown source. This time,
 349         // we're escaping the markup we've just created, so that we don't wrap
 350         // <p> tags around block-level tags.
 351         text = _HashHTMLBlocks(text);
 352         text = _FormParagraphs(text);
 353
 354         return text;
 355 }
 356
 357
 358 var _RunSpanGamut = function(text) {
 359 //
 360 // These are all the transformations that occur *within* block-level
 361 // tags like paragraphs, headers, and list items.
 362 //
 363
 364         text = _DoCodeSpans(text);
 365         text = _EscapeSpecialCharsWithinTagAttributes(text);
 366         text = _EncodeBackslashEscapes(text);
 367
 368         // Process anchor and image tags. Images must come first,
 369         // because ![foo][f] looks like an anchor.
 370         text = _DoImages(text);
 371         text = _DoAnchors(text);
 372
 373         // Make links out of things like `<http://example.com/>`
 374         // Must come after _DoAnchors(), because you can use < and >
 375         // delimiters in inline links like [this](<url>).
 376         text = _DoAutoLinks(text);
 377         text = _EncodeAmpsAndAngles(text);
 378         text = _DoItalicsAndBold(text);
 379
 380         // Do hard breaks:
 381         text = text.replace(/  +\n/g," <br />\n");
 382
 383         return text;
 384 }
 385
 386 var _EscapeSpecialCharsWithinTagAttributes = function(text) {
 387 //
 388 // Within tags -- meaning between < and > -- encode [\ ` * _] so they
 389 // don't conflict with their use in Markdown for code, italics and strong.
 390 //
 391
 392         // Build a regex to find HTML tags and comments.  See Friedl's
 393         // "Mastering Regular Expressions", 2nd Ed., pp. 200-201.
 394         var regex = /(<[a-z\/!$]("[^"]*"|'[^']*'|[^'">])*>|<!(--.*?--\s*)+>)/gi;
 395
 396         text = text.replace(regex, function(wholeMatch) {
 397                 var tag = wholeMatch.replace(/(.)<\/?code>(?=.)/g,"$1`");
 398                 tag = escapeCharacters(tag,"\\`*_");
 399                 return tag;
 400         });
 401
 402         return text;
 403 }
 404
 405 var _DoAnchors = function(text) {
 406 //
 407 // Turn Markdown link shortcuts into XHTML <a> tags.
 408 //
 409         //
 410         // First, handle reference-style links: [link text] [id]
 411         //
 412
 413         /*
 414                 text = text.replace(/
 415                 (                                                       // wrap whole match in $1
 416                         \[
 417                         (
 418                                 (?:
 419                                         \[[^\]]*\]              // allow brackets nested one level
 420                                         |
 421                                         [^\[]                   // or anything else
 422                                 )*
 423                         )
 424                         \]
 425
 426                         [ ]?                                    // one optional space
 427                         (?:\n[ ]*)?                             // one optional newline followed by spaces
 428
 429                         \[
 430                         (.*?)                                   // id = $3
 431                         \]
 432                 )()()()()                                       // pad remaining backreferences
 433                 /g,_DoAnchors_callback);
 434         */
 435         text = text.replace(/(\[((?:\[[^\]]*\]|[^\[\]])*)\][ ]?(?:\n[ ]*)?\[(.*?)\])()()()()/g,writeAnchorTag);
 436
 437         //
 438         // Next, inline-style links: [link text](url "optional title")
 439         //
 440
 441         /*
 442                 text = text.replace(/
 443                         (                                               // wrap whole match in $1
 444                                 \[
 445                                 (
 446                                         (?:
 447                                                 \[[^\]]*\]      // allow brackets nested one level
 448                                         |
 449                                         [^\[\]]                 // or anything else
 450                                 )
 451                         )
 452                         \]
 453                         \(                                              // literal paren
 454                         [ \t]*
 455                         ()                                              // no id, so leave $3 empty
 456                         <?(.*?)>?                               // href = $4
 457                         [ \t]*
 458                         (                                               // $5
 459                                 (['"])                          // quote char = $6
 460                                 (.*?)                           // Title = $7
 461                                 \6                                      // matching quote
 462                                 [ \t]*                          // ignore any spaces/tabs between closing quote and )
 463                         )?                                              // title is optional
 464                         \)
 465                 )
 466                 /g,writeAnchorTag);
 467         */
 468         text = text.replace(/(\[((?:\[[^\]]*\]|[^\[\]])*)\]\([ \t]*()<?(.*?)>?[ \t]*((['"])(.*?)\6[ \t]*)?\))/g,writeAnchorTag);
 469
 470         //
 471         // Last, handle reference-style shortcuts: [link text]
 472         // These must come last in case you've also got [link test][1]
 473         // or [link test](/foo)
 474         //
 475
 476         /*
 477                 text = text.replace(/
 478                 (                                                       // wrap whole match in $1
 479                         \[
 480                         ([^\[\]]+)                              // link text = $2; can't contain '[' or ']'
 481                         \]
 482                 )()()()()()                                     // pad rest of backreferences
 483                 /g, writeAnchorTag);
 484         */
 485         text = text.replace(/(\[([^\[\]]+)\])()()()()()/g, writeAnchorTag);
 486
 487         return text;
 488 }
 489
 490 var writeAnchorTag = function(wholeMatch,m1,m2,m3,m4,m5,m6,m7) {
 491         if (m7 == undefined) m7 = "";
 492         var whole_match = m1;
 493         var link_text   = m2;
 494         var link_id      = m3.toLowerCase();
 495         var url         = m4;
 496         var title       = m7;
 497
 498         if (url == "") {
 499                 if (link_id == "") {
 500                         // lower-case and turn embedded newlines into spaces
 501                         link_id = link_text.toLowerCase().replace(/ ?\n/g," ");
 502                 }
 503                 url = "#"+link_id;
 504
 505                 if (g_urls[link_id] != undefined) {
 506                         url = g_urls[link_id];
 507                         if (g_titles[link_id] != undefined) {
 508                                 title = g_titles[link_id];
 509                         }
 510                 }
 511                 else {
 512                         if (whole_match.search(/\(\s*\)$/m)>-1) {
 513                                 // Special case for explicit empty url
 514                                 url = "";
 515                         } else {
 516                                 return whole_match;
 517                         }
 518                 }
 519         }
 520
 521         url = escapeCharacters(url,"*_");
 522         var result = "<a href=\"" + url + "\"";
 523
 524         if (title != "") {
 525                 title = title.replace(/"/g,"&quot;");
 526                 title = escapeCharacters(title,"*_");
 527                 result +=  " title=\"" + title + "\"";
 528         }
 529
 530         result += ">" + link_text + "</a>";
 531
 532         return result;
 533 }
 534
 535
 536 var _DoImages = function(text) {
 537 //
 538 // Turn Markdown image shortcuts into <img> tags.
 539 //
 540
 541         //
 542         // First, handle reference-style labeled images: ![alt text][id]
 543         //
 544
 545         /*
 546                 text = text.replace(/
 547                 (                                               // wrap whole match in $1
 548                         !\[
 549                         (.*?)                           // alt text = $2
 550                         \]
 551
 552                         [ ]?                            // one optional space
 553                         (?:\n[ ]*)?                     // one optional newline followed by spaces
 554
 555                         \[
 556                         (.*?)                           // id = $3
 557                         \]
 558                 )()()()()                               // pad rest of backreferences
 559                 /g,writeImageTag);
 560         */
 561         text = text.replace(/(!\[(.*?)\][ ]?(?:\n[ ]*)?\[(.*?)\])()()()()/g,writeImageTag);
 562
 563         //
 564         // Next, handle inline images:  ![alt text](url "optional title")
 565         // Don't forget: encode * and _
 566
 567         /*
 568                 text = text.replace(/
 569                 (                                               // wrap whole match in $1
 570                         !\[
 571                         (.*?)                           // alt text = $2
 572                         \]
 573                         \s?                                     // One optional whitespace character
 574                         \(                                      // literal paren
 575                         [ \t]*
 576                         ()                                      // no id, so leave $3 empty
 577                         <?(\S+?)>?                      // src url = $4
 578                         [ \t]*
 579                         (                                       // $5
 580                                 (['"])                  // quote char = $6
 581                                 (.*?)                   // title = $7
 582                                 \6                              // matching quote
 583                                 [ \t]*
 584                         )?                                      // title is optional
 585                 \)
 586                 )
 587                 /g,writeImageTag);
 588         */
 589         text = text.replace(/(!\[(.*?)\]\s?\([ \t]*()<?(\S+?)>?[ \t]*((['"])(.*?)\6[ \t]*)?\))/g,writeImageTag);
 590
 591         return text;
 592 }
 593
 594 var writeImageTag = function(wholeMatch,m1,m2,m3,m4,m5,m6,m7) {
 595         var whole_match = m1;
 596         var alt_text   = m2;
 597         var link_id      = m3.toLowerCase();
 598         var url         = m4;
 599         var title       = m7;
 600
 601         if (!title) title = "";
 602
 603         if (url == "") {
 604                 if (link_id == "") {
 605                         // lower-case and turn embedded newlines into spaces
 606                         link_id = alt_text.toLowerCase().replace(/ ?\n/g," ");
 607                 }
 608                 url = "#"+link_id;
 609
 610                 if (g_urls[link_id] != undefined) {
 611                         url = g_urls[link_id];
 612                         if (g_titles[link_id] != undefined) {
 613                                 title = g_titles[link_id];
 614                         }
 615                 }
 616                 else {
 617                         return whole_match;
 618                 }
 619         }
 620
 621         alt_text = alt_text.replace(/"/g,"&quot;");
 622         url = escapeCharacters(url,"*_");
 623         var result = "<img src=\"" + url + "\" alt=\"" + alt_text + "\"";
 624
 625         // attacklab: Markdown.pl adds empty title attributes to images.
 626         // Replicate this bug.
 627
 628         //if (title != "") {
 629                 title = title.replace(/"/g,"&quot;");
 630                 title = escapeCharacters(title,"*_");
 631                 result +=  " title=\"" + title + "\"";
 632         //}
 633
 634         result += " />";
 635
 636         return result;
 637 }
 638
 639
 640 var _DoHeaders = function(text) {
 641
 642         // Setext-style headers:
 643         //      Header 1
 644         //      ========
 645         //
 646         //      Header 2
 647         //      --------
 648         //
 649         text = text.replace(/^(.+)[ \t]*\n=+[ \t]*\n+/gm,
 650                 function(wholeMatch,m1){return hashBlock("<h1>" + _RunSpanGamut(m1) + "</h1>");});
 651
 652         text = text.replace(/^(.+)[ \t]*\n-+[ \t]*\n+/gm,
 653                 function(matchFound,m1){return hashBlock("<h2>" + _RunSpanGamut(m1) + "</h2>");});
 654
 655         // atx-style headers:
 656         //  # Header 1
 657         //  ## Header 2
 658         //  ## Header 2 with closing hashes ##
 659         //  ...
 660         //  ###### Header 6
 661         //
 662
 663         /*
 664                 text = text.replace(/
 665                         ^(\#{1,6})                              // $1 = string of #'s
 666                         [ \t]*
 667                         (.+?)                                   // $2 = Header text
 668                         [ \t]*
 669                         \#*                                             // optional closing #'s (not counted)
 670                         \n+
 671                 /gm, function() {...});
 672         */
 673
 674         text = text.replace(/^(\#{1,6})[ \t]*(.+?)[ \t]*\#*\n+/gm,
 675                 function(wholeMatch,m1,m2) {
 676                         var h_level = m1.length;
 677                         return hashBlock("<h" + h_level + ">" + _RunSpanGamut(m2) + "</h" + h_level + ">");
 678                 });
 679
 680         return text;
 681 }
 682
 683 // This declaration keeps Dojo compressor from outputting garbage:
 684 var _ProcessListItems;
 685
 686 var _DoLists = function(text) {
 687 //
 688 // Form HTML ordered (numbered) and unordered (bulleted) lists.
 689 //
 690
 691         // attacklab: add sentinel to hack around khtml/safari bug:
 692         // http://bugs.webkit.org/show_bug.cgi?id=11231
 693         text += "~0";
 694
 695         // Re-usable pattern to match any entirel ul or ol list:
 696
 697         /*
 698                 var whole_list = /
 699                 (                                                                       // $1 = whole list
 700                         (                                                               // $2
 701                                 [ ]{0,3}                                        // attacklab: g_tab_width - 1
 702                                 ([*+-]|\d+[.])                          // $3 = first list item marker
 703                                 [ \t]+
 704                         )
 705                         [^\r]+?
 706                         (                                                               // $4
 707                                 ~0                                                      // sentinel for workaround; should be $
 708                         |
 709                                 \n{2,}
 710                                 (?=\S)
 711                                 (?!                                                     // Negative lookahead for another list item marker
 712                                         [ \t]*
 713                                         (?:[*+-]|\d+[.])[ \t]+
 714                                 )
 715                         )
 716                 )/g
 717         */
 718         var whole_list = /^(([ ]{0,3}([*+-]|\d+[.])[ \t]+)[^\r]+?(~0|\n{2,}(?=\S)(?![ \t]*(?:[*+-]|\d+[.])[ \t]+)))/gm;
 719
 720         if (g_list_level) {
 721                 text = text.replace(whole_list,function(wholeMatch,m1,m2) {
 722                         var list = m1;
 723                         var list_type = (m2.search(/[*+-]/g)>-1) ? "ul" : "ol";
 724
 725                         // Turn double returns into triple returns, so that we can make a
 726                         // paragraph for the last item in a list, if necessary:
 727                         list = list.replace(/\n{2,}/g,"\n\n\n");;
 728                         var result = _ProcessListItems(list);
 729
 730                         // Trim any trailing whitespace, to put the closing `</$list_type>`
 731                         // up on the preceding line, to get it past the current stupid
 732                         // HTML block parser. This is a hack to work around the terrible
 733                         // hack that is the HTML block parser.
 734                         result = result.replace(/\s+$/,"");
 735                         result = "<"+list_type+">" + result + "</"+list_type+">\n";
 736                         return result;
 737                 });
 738         } else {
 739                 whole_list = /(\n\n|^\n?)(([ ]{0,3}([*+-]|\d+[.])[ \t]+)[^\r]+?(~0|\n{2,}(?=\S)(?![ \t]*(?:[*+-]|\d+[.])[ \t]+)))/g;
 740                 text = text.replace(whole_list,function(wholeMatch,m1,m2,m3) {
 741                         var runup = m1;
 742                         var list = m2;
 743
 744                         var list_type = (m3.search(/[*+-]/g)>-1) ? "ul" : "ol";
 745                         // Turn double returns into triple returns, so that we can make a
 746                         // paragraph for the last item in a list, if necessary:
 747                         var list = list.replace(/\n{2,}/g,"\n\n\n");;
 748                         var result = _ProcessListItems(list);
 749                         result = runup + "<"+list_type+">\n" + result + "</"+list_type+">\n";
 750                         return result;
 751                 });
 752         }
 753
 754         // attacklab: strip sentinel
 755         text = text.replace(/~0/,"");
 756
 757         return text;
 758 }
 759
 760 _ProcessListItems = function(list_str) {
 761 //
 762 //  Process the contents of a single ordered or unordered list, splitting it
 763 //  into individual list items.
 764 //
 765         // The $g_list_level global keeps track of when we're inside a list.
 766         // Each time we enter a list, we increment it; when we leave a list,
 767         // we decrement. If it's zero, we're not in a list anymore.
 768         //
 769         // We do this because when we're not inside a list, we want to treat
 770         // something like this:
 771         //
 772         //    I recommend upgrading to version
 773         //    8. Oops, now this line is treated
 774         //    as a sub-list.
 775         //
 776         // As a single paragraph, despite the fact that the second line starts
 777         // with a digit-period-space sequence.
 778         //
 779         // Whereas when we're inside a list (or sub-list), that line will be
 780         // treated as the start of a sub-list. What a kludge, huh? This is
 781         // an aspect of Markdown's syntax that's hard to parse perfectly
 782         // without resorting to mind-reading. Perhaps the solution is to
 783         // change the syntax rules such that sub-lists must start with a
 784         // starting cardinal number; e.g. "1." or "a.".
 785
 786         g_list_level++;
 787
 788         // trim trailing blank lines:
 789         list_str = list_str.replace(/\n{2,}$/,"\n");
 790
 791         // attacklab: add sentinel to emulate \z
 792         list_str += "~0";
 793
 794         /*
 795                 list_str = list_str.replace(/
 796                         (\n)?                                                   // leading line = $1
 797                         (^[ \t]*)                                               // leading whitespace = $2
 798                         ([*+-]|\d+[.]) [ \t]+                   // list marker = $3
 799                         ([^\r]+?                                                // list item text   = $4
 800                         (\n{1,2}))
 801                         (?= \n* (~0 | \2 ([*+-]|\d+[.]) [ \t]+))
 802                 /gm, function(){...});
 803         */
 804         list_str = list_str.replace(/(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+([^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm,
 805                 function(wholeMatch,m1,m2,m3,m4){
 806                         var item = m4;
 807                         var leading_line = m1;
 808                         var leading_space = m2;
 809
 810                         if (leading_line || (item.search(/\n{2,}/)>-1)) {
 811                                 item = _RunBlockGamut(_Outdent(item));
 812                         }
 813                         else {
 814                                 // Recursion for sub-lists:
 815                                 item = _DoLists(_Outdent(item));
 816                                 item = item.replace(/\n$/,""); // chomp(item)
 817                                 item = _RunSpanGamut(item);
 818                         }
 819
 820                         return  "<li>" + item + "</li>\n";
 821                 }
 822         );
 823
 824         // attacklab: strip sentinel
 825         list_str = list_str.replace(/~0/g,"");
 826
 827         g_list_level--;
 828         return list_str;
 829 }
 830
 831
 832 var _DoCodeBlocks = function(text) {
 833 //
 834 //  Process Markdown `<pre><code>` blocks.
 835 //
 836
 837         /*
 838                 text = text.replace(text,
 839                         /(?:\n\n|^)
 840                         (                                                               // $1 = the code block -- one or more lines, starting with a space/tab
 841                                 (?:
 842                                         (?:[ ]{4}|\t)                   // Lines must start with a tab or a tab-width of spaces - attacklab: g_tab_width
 843                                         .*\n+
 844                                 )+
 845                         )
 846                         (\n*[ ]{0,3}[^ \t\n]|(?=~0))    // attacklab: g_tab_width
 847                 /g,function(){...});
 848         */
 849
 850         // attacklab: sentinel workarounds for lack of \A and \Z, safari\khtml bug
 851         text += "~0";
 852
 853         text = text.replace(/(?:\n\n|^)((?:(?:[ ]{4}|\t).*\n+)+)(\n*[ ]{0,3}[^ \t\n]|(?=~0))/g,
 854                 function(wholeMatch,m1,m2) {
 855                         var codeblock = m1;
 856                         var nextChar = m2;
 857
 858                         codeblock = _EncodeCode( _Outdent(codeblock));
 859                         codeblock = _Detab(codeblock);
 860                         codeblock = codeblock.replace(/^\n+/g,""); // trim leading newlines
 861                         codeblock = codeblock.replace(/\n+$/g,""); // trim trailing whitespace
 862
 863                         codeblock = "<pre><code>" + codeblock + "\n</code></pre>";
 864
 865                         return hashBlock(codeblock) + nextChar;
 866                 }
 867         );
 868
 869         // attacklab: strip sentinel
 870         text = text.replace(/~0/,"");
 871
 872         return text;
 873 }
 874
 875 var hashBlock = function(text) {
 876         text = text.replace(/(^\n+|\n+$)/g,"");
 877         return "\n\n~K" + (g_html_blocks.push(text)-1) + "K\n\n";
 878 }
 879
 880
 881 var _DoCodeSpans = function(text) {
 882 //
 883 //   *  Backtick quotes are used for <code></code> spans.
 884 //
 885 //   *  You can use multiple backticks as the delimiters if you want to
 886 //       include literal backticks in the code span. So, this input:
 887 //
 888 //               Just type ``foo `bar` baz`` at the prompt.
 889 //
 890 //         Will translate to:
 891 //
 892 //               <p>Just type <code>foo `bar` baz</code> at the prompt.</p>
 893 //
 894 //      There's no arbitrary limit to the number of backticks you
 895 //      can use as delimters. If you need three consecutive backticks
 896 //      in your code, use four for delimiters, etc.
 897 //
 898 //  *  You can use spaces to get literal backticks at the edges:
 899 //
 900 //               ... type `` `bar` `` ...
 901 //
 902 //         Turns to:
 903 //
 904 //               ... type <code>`bar`</code> ...
 905 //
 906
 907         /*
 908                 text = text.replace(/
 909                         (^|[^\\])                                       // Character before opening ` can't be a backslash
 910                         (`+)                                            // $2 = Opening run of `
 911                         (                                                       // $3 = The code block
 912                                 [^\r]*?
 913                                 [^`]                                    // attacklab: work around lack of lookbehind
 914                         )
 915                         \2                                                      // Matching closer
 916                         (?!`)
 917                 /gm, function(){...});
 918         */
 919
 920         text = text.replace(/(^|[^\\])(`+)([^\r]*?[^`])\2(?!`)/gm,
 921                 function(wholeMatch,m1,m2,m3,m4) {
 922                         var c = m3;
 923                         c = c.replace(/^([ \t]*)/g,""); // leading whitespace
 924                         c = c.replace(/[ \t]*$/g,"");   // trailing whitespace
 925                         c = _EncodeCode(c);
 926                         return m1+"<code>"+c+"</code>";
 927                 });
 928
 929         return text;
 930 }
 931
 932
 933 var _EncodeCode = function(text) {
 934 //
 935 // Encode/escape certain characters inside Markdown code runs.
 936 // The point is that in code, these characters are literals,
 937 // and lose their special Markdown meanings.
 938 //
 939         // Encode all ampersands; HTML entities are not
 940         // entities within a Markdown code span.
 941         text = text.replace(/&/g,"&amp;");
 942
 943         // Do the angle bracket song and dance:
 944         text = text.replace(/</g,"&lt;");
 945         text = text.replace(/>/g,"&gt;");
 946
 947         // Now, escape characters that are magic in Markdown:
 948         text = escapeCharacters(text,"\*_{}[]\\",false);
 949
 950 // jj the line above breaks this:
 951 //---
 952
 953 //* Item
 954
 955 //   1. Subitem
 956
 957 //            special char: *
 958 //---
 959
 960         return text;
 961 }
 962
 963
 964 var _DoItalicsAndBold = function(text) {
 965
 966         // <strong> must go first:
 967         text = text.replace(/(\*\*|__)(?=\S)([^\r]*?\S[*_]*)\1/g,
 968                 "<strong>$2</strong>");
 969
 970         text = text.replace(/(\*|_)(?=\S)([^\r]*?\S)\1/g,
 971                 "<em>$2</em>");
 972
 973         return text;
 974 }
 975
 976
 977 var _DoBlockQuotes = function(text) {
 978
 979         /*
 980                 text = text.replace(/
 981                 (                                                               // Wrap whole match in $1
 982                         (
 983                                 ^[ \t]*>[ \t]?                  // '>' at the start of a line
 984                                 .+\n                                    // rest of the first line
 985                                 (.+\n)*                                 // subsequent consecutive lines
 986                                 \n*                                             // blanks
 987                         )+
 988                 )
 989                 /gm, function(){...});
 990         */
 991
 992         text = text.replace(/((^[ \t]*>[ \t]?.+\n(.+\n)*\n*)+)/gm,
 993                 function(wholeMatch,m1) {
 994                         var bq = m1;
 995
 996                         // attacklab: hack around Konqueror 3.5.4 bug:
 997                         // "----------bug".replace(/^-/g,"") == "bug"
 998
 999                         bq = bq.replace(/^[ \t]*>[ \t]?/gm,"~0");       // trim one level of quoting
1000
1001                         // attacklab: clean up hack
1002                         bq = bq.replace(/~0/g,"");
1003
1004                         bq = bq.replace(/^[ \t]+$/gm,"");               // trim whitespace-only lines
1005                         bq = _RunBlockGamut(bq);                                // recurse
1006
1007                         bq = bq.replace(/(^|\n)/g,"$1  ");
1008                         // These leading spaces screw with <pre> content, so we need to fix that:
1009                         bq = bq.replace(
1010                                         /(\s*<pre>[^\r]+?<\/pre>)/gm,
1011                                 function(wholeMatch,m1) {
1012                                         var pre = m1;
1013                                         // attacklab: hack around Konqueror 3.5.4 bug:
1014                                         pre = pre.replace(/^  /mg,"~0");
1015                                         pre = pre.replace(/~0/g,"");
1016                                         return pre;
1017                                 });
1018
1019                         return hashBlock("<blockquote>\n" + bq + "\n</blockquote>");
1020                 });
1021         return text;
1022 }
1023
1024
1025 var _FormParagraphs = function(text) {
1026 //
1027 //  Params:
1028 //    $text - string to process with html <p> tags
1029 //
1030
1031         // Strip leading and trailing lines:
1032         text = text.replace(/^\n+/g,"");
1033         text = text.replace(/\n+$/g,"");
1034
1035         var grafs = text.split(/\n{2,}/g);
1036         var grafsOut = new Array();
1037
1038         //
1039         // Wrap <p> tags.
1040         //
1041         var end = grafs.length;
1042         for (var i=0; i<end; i++) {
1043                 var str = grafs[i];
1044
1045                 // if this is an HTML marker, copy it
1046                 if (str.search(/~K(\d+)K/g) >= 0) {
1047                         grafsOut.push(str);
1048                 }
1049                 else if (str.search(/\S/) >= 0) {
1050                         str = _RunSpanGamut(str);
1051                         str = str.replace(/^([ \t]*)/g,"<p>");
1052                         str += "</p>"
1053                         grafsOut.push(str);
1054                 }
1055
1056         }
1057
1058         //
1059         // Unhashify HTML blocks
1060         //
1061         end = grafsOut.length;
1062         for (var i=0; i<end; i++) {
1063                 // if this is a marker for an html block...
1064                 while (grafsOut[i].search(/~K(\d+)K/) >= 0) {
1065                         var blockText = g_html_blocks[RegExp.$1];
1066                         blockText = blockText.replace(/\$/g,"$$$$"); // Escape any dollar signs
1067                         grafsOut[i] = grafsOut[i].replace(/~K\d+K/,blockText);
1068                 }
1069         }
1070
1071         return grafsOut.join("\n\n");
1072 }
1073
1074
1075 var _EncodeAmpsAndAngles = function(text) {
1076 // Smart processing for ampersands and angle brackets that need to be encoded.
1077
1078         // Ampersand-encoding based entirely on Nat Irons's Amputator MT plugin:
1079         //   http://bumppo.net/projects/amputator/
1080         text = text.replace(/&(?!#?[xX]?(?:[0-9a-fA-F]+|\w+);)/g,"&amp;");
1081
1082         // Encode naked <'s
1083         text = text.replace(/<(?![a-z\/?\$!])/gi,"&lt;");
1084
1085         return text;
1086 }
1087
1088
1089 var _EncodeBackslashEscapes = function(text) {
1090 //
1091 //   Parameter:  String.
1092 //   Returns:   The string, with after processing the following backslash
1093 //                         escape sequences.
1094 //
1095
1096         // attacklab: The polite way to do this is with the new
1097         // escapeCharacters() function:
1098         //
1099         //      text = escapeCharacters(text,"\\",true);
1100         //      text = escapeCharacters(text,"`*_{}[]()>#+-.!",true);
1101         //
1102         // ...but we're sidestepping its use of the (slow) RegExp constructor
1103         // as an optimization for Firefox.  This function gets called a LOT.
1104
1105         text = text.replace(/\\(\\)/g,escapeCharacters_callback);
1106         text = text.replace(/\\([`*_{}\[\]()>#+-.!])/g,escapeCharacters_callback);
1107         return text;
1108 }
1109
1110
1111 var _DoAutoLinks = function(text) {
1112
1113         text = text.replace(/<((https?|ftp|dict):[^'">\s]+)>/gi,"<a href=\"$1\">$1</a>");
1114
1115         // Email addresses: <address@domain.foo>
1116
1117         /*
1118                 text = text.replace(/
1119                         <
1120                         (?:mailto:)?
1121                         (
1122                                 [-.\w]+
1123                                 \@
1124                                 [-a-z0-9]+(\.[-a-z0-9]+)*\.[a-z]+
1125                         )
1126                         >
1127                 /gi, _DoAutoLinks_callback());
1128         */
1129         text = text.replace(/<(?:mailto:)?([-.\w]+\@[-a-z0-9]+(\.[-a-z0-9]+)*\.[a-z]+)>/gi,
1130                 function(wholeMatch,m1) {
1131                         return _EncodeEmailAddress( _UnescapeSpecialChars(m1) );
1132                 }
1133         );
1134
1135         return text;
1136 }
1137
1138
1139 var _EncodeEmailAddress = function(addr) {
1140 //
1141 //  Input: an email address, e.g. "foo@example.com"
1142 //
1143 //  Output: the email address as a mailto link, with each character
1144 //      of the address encoded as either a decimal or hex entity, in
1145 //      the hopes of foiling most address harvesting spam bots. E.g.:
1146 //
1147 //      <a href="&#x6D;&#97;&#105;&#108;&#x74;&#111;:&#102;&#111;&#111;&#64;&#101;
1148 //         x&#x61;&#109;&#x70;&#108;&#x65;&#x2E;&#99;&#111;&#109;">&#102;&#111;&#111;
1149 //         &#64;&#101;x&#x61;&#109;&#x70;&#108;&#x65;&#x2E;&#99;&#111;&#109;</a>
1150 //
1151 //  Based on a filter by Matthew Wickline, posted to the BBEdit-Talk
1152 //  mailing list: <http://tinyurl.com/yu7ue>
1153 //
1154
1155         // attacklab: why can't javascript speak hex?
1156         function char2hex(ch) {
1157                 var hexDigits = '0123456789ABCDEF';
1158                 var dec = ch.charCodeAt(0);
1159                 return(hexDigits.charAt(dec>>4) + hexDigits.charAt(dec&15));
1160         }
1161
1162         var encode = [
1163                 function(ch){return "&#"+ch.charCodeAt(0)+";";},
1164                 function(ch){return "&#x"+char2hex(ch)+";";},
1165                 function(ch){return ch;}
1166         ];
1167
1168         addr = "mailto:" + addr;
1169
1170         addr = addr.replace(/./g, function(ch) {
1171                 if (ch == "@") {
1172                         // this *must* be encoded. I insist.
1173                         ch = encode[Math.floor(Math.random()*2)](ch);
1174                 } else if (ch !=":") {
1175                         // leave ':' alone (to spot mailto: later)
1176                         var r = Math.random();
1177                         // roughly 10% raw, 45% hex, 45% dec
1178                         ch =  (
1179                                         r > .9  ?       encode[2](ch)   :
1180                                         r > .45 ?       encode[1](ch)   :
1181                                                                 encode[0](ch)
1182                                 );
1183                 }
1184                 return ch;
1185         });
1186
1187         addr = "<a href=\"" + addr + "\">" + addr + "</a>";
1188         addr = addr.replace(/">.+:/g,"\">"); // strip the mailto: from the visible part
1189
1190         return addr;
1191 }
1192
1193
1194 var _UnescapeSpecialChars = function(text) {
1195 //
1196 // Swap back in all the special characters we've hidden.
1197 //
1198         text = text.replace(/~E(\d+)E/g,
1199                 function(wholeMatch,m1) {
1200                         var charCodeToReplace = parseInt(m1);
1201                         return String.fromCharCode(charCodeToReplace);
1202                 }
1203         );
1204         return text;
1205 }
1206
1207
1208 var _Outdent = function(text) {
1209 //
1210 // Remove one level of line-leading tabs or spaces
1211 //
1212
1213         // attacklab: hack around Konqueror 3.5.4 bug:
1214         // "----------bug".replace(/^-/g,"") == "bug"
1215
1216         text = text.replace(/^(\t|[ ]{1,4})/gm,"~0"); // attacklab: g_tab_width
1217
1218         // attacklab: clean up hack
1219         text = text.replace(/~0/g,"")
1220
1221         return text;
1222 }
1223
1224 var _Detab = function(text) {
1225 // attacklab: Detab's completely rewritten for speed.
1226 // In perl we could fix it by anchoring the regexp with \G.
1227 // In javascript we're less fortunate.
1228
1229         // expand first n-1 tabs
1230         text = text.replace(/\t(?=\t)/g,"    "); // attacklab: g_tab_width
1231
1232         // replace the nth with two sentinels
1233         text = text.replace(/\t/g,"~A~B");
1234
1235         // use the sentinel to anchor our regex so it doesn't explode
1236         text = text.replace(/~B(.+?)~A/g,
1237                 function(wholeMatch,m1,m2) {
1238                         var leadingText = m1;
1239                         var numSpaces = 4 - leadingText.length % 4;  // attacklab: g_tab_width
1240
1241                         // there *must* be a better way to do this:
1242                         for (var i=0; i<numSpaces; i++) leadingText+=" ";
1243
1244                         return leadingText;
1245                 }
1246         );
1247
1248         // clean up sentinels
1249         text = text.replace(/~A/g,"    ");  // attacklab: g_tab_width
1250         text = text.replace(/~B/g,"");
1251
1252         return text;
1253 }
1254
1255
1256 //
1257 //  attacklab: Utility functions
1258 //
1259
1260
1261 var escapeCharacters = function(text, charsToEscape, afterBackslash) {
1262         // First we have to escape the escape characters so that
1263         // we can build a character class out of them
1264         var regexString = "([" + charsToEscape.replace(/([\[\]\\])/g,"\\$1") + "])";
1265
1266         if (afterBackslash) {
1267                 regexString = "\\\\" + regexString;
1268         }
1269
1270         var regex = new RegExp(regexString,"g");
1271         text = text.replace(regex,escapeCharacters_callback);
1272
1273         return text;
1274 }
1275
1276
1277 var escapeCharacters_callback = function(wholeMatch,m1) {
1278         var charCodeToEscape = m1.charCodeAt(0);
1279         return "~E"+charCodeToEscape+"E";
1280 }
1281
1282 exports.encode = exports.markdown = function (src) {
1283    return exports.makeHtml(src);
1284 };
1285
1286 exports.main = function (system) {
1287     var command = system.args.shift();
1288     if (!system.args.length) {
1289         system.stdout.write(exports.markdown(system.stdin.read())).flush();
1290     } else {
1291         var arg;
1292         while (arg = system.args.shift()) {
1293             var out = system.fs.basename(arg, '.md') + '.html';
1294             print(out);
1295             system.fs.write(out, exports.markdown(system.fs.read(arg)));
1296         }
1297     }
1298 };