]> git.openstreetmap.org Git - osqa.git/blob - forum/skins/default/media/js/wmd/showdown.js
Initial commit
[osqa.git] / forum / skins / default / media / js / wmd / showdown.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 converter = new Attacklab.showdown.converter();
55 //   var html = converter.makeHtml(text);
56 //
57 //   alert(html);
58 //
59 // Note: move the sample code to the bottom of this
60 // file before uncommenting it.
61 //
62
63
64 //
65 // Attacklab namespace
66 //
67 var Attacklab = Attacklab || {}
68
69 //
70 // Showdown namespace
71 //
72 Attacklab.showdown = Attacklab.showdown || {}
73
74 //
75 // converter
76 //
77 // Wraps all "globals" so that the only thing
78 // exposed is makeHtml().
79 //
80 Attacklab.showdown.converter = function() {
81
82 //
83 // Globals:
84 //
85
86 // Global hashes, used by various utility routines
87 var g_urls;
88 var g_titles;
89 var g_html_blocks;
90
91 // Used to track when we're inside an ordered or unordered list
92 // (see _ProcessListItems() for details):
93 var g_list_level = 0;
94
95
96 this.makeHtml = function(text) {
97 //
98 // Main function. The order in which other subs are called here is
99 // essential. Link and image substitutions need to happen before
100 // _EscapeSpecialCharsWithinTagAttributes(), so that any *'s or _'s in the <a>
101 // and <img> tags get encoded.
102 //
103
104         // Clear the global hashes. If we don't clear these, you get conflicts
105         // from other articles when generating a page which contains more than
106         // one article (e.g. an index page that shows the N most recent
107         // articles):
108         g_urls = new Array();
109         g_titles = new Array();
110         g_html_blocks = new Array();
111
112         // attacklab: Replace ~ with ~T
113         // This lets us use tilde as an escape char to avoid md5 hashes
114         // The choice of character is arbitray; anything that isn't
115     // magic in Markdown will work.
116         text = text.replace(/~/g,"~T");
117
118         // attacklab: Replace $ with ~D
119         // RegExp interprets $ as a special character
120         // when it's in a replacement string
121         text = text.replace(/\$/g,"~D");
122
123         // Standardize line endings
124         text = text.replace(/\r\n/g,"\n"); // DOS to Unix
125         text = text.replace(/\r/g,"\n"); // Mac to Unix
126
127         // Make sure text begins and ends with a couple of newlines:
128         text = "\n\n" + text + "\n\n";
129
130         // Convert all tabs to spaces.
131         text = _Detab(text);
132
133         // Strip any lines consisting only of spaces and tabs.
134         // This makes subsequent regexen easier to write, because we can
135         // match consecutive blank lines with /\n+/ instead of something
136         // contorted like /[ \t]*\n+/ .
137         text = text.replace(/^[ \t]+$/mg,"");
138
139         // Turn block-level HTML blocks into hash entries
140         text = _HashHTMLBlocks(text);
141
142         // Strip link definitions, store in hashes.
143         text = _StripLinkDefinitions(text);
144
145         text = _RunBlockGamut(text);
146
147         text = _UnescapeSpecialChars(text);
148
149         // attacklab: Restore dollar signs
150         text = text.replace(/~D/g,"$$");
151
152         // attacklab: Restore tildes
153         text = text.replace(/~T/g,"~");
154
155         return text;
156 }
157
158 var _StripLinkDefinitions = function(text) {
159 //
160 // Strips link definitions from text, stores the URLs and titles in
161 // hash references.
162 //
163
164         // Link defs are in the form: ^[id]: url "optional title"
165
166         /*
167                 var text = text.replace(/
168                                 ^[ ]{0,3}\[(.+)\]:  // id = $1  attacklab: g_tab_width - 1
169                                   [ \t]*
170                                   \n?                           // maybe *one* newline
171                                   [ \t]*
172                                 <?(\S+?)>?                      // url = $2
173                                   [ \t]*
174                                   \n?                           // maybe one newline
175                                   [ \t]*
176                                 (?:
177                                   (\n*)                         // any lines skipped = $3 attacklab: lookbehind removed
178                                   ["(]
179                                   (.+?)                         // title = $4
180                                   [")]
181                                   [ \t]*
182                                 )?                                      // title is optional
183                                 (?:\n+|$)
184                           /gm,
185                           function(){...});
186         */
187         var text = text.replace(/^[ ]{0,3}\[(.+)\]:[ \t]*\n?[ \t]*<?(\S+?)>?[ \t]*\n?[ \t]*(?:(\n*)["(](.+?)[")][ \t]*)?(?:\n+|\Z)/gm,
188                 function (wholeMatch,m1,m2,m3,m4) {
189                         m1 = m1.toLowerCase();
190                         g_urls[m1] = _EncodeAmpsAndAngles(m2);  // Link IDs are case-insensitive
191                         if (m3) {
192                                 // Oops, found blank lines, so it's not a title.
193                                 // Put back the parenthetical statement we stole.
194                                 return m3+m4;
195                         } else if (m4) {
196                                 g_titles[m1] = m4.replace(/"/g,"&quot;");
197                         }
198                         
199                         // Completely remove the definition from the text
200                         return "";
201                 }
202         );
203
204         return text;
205 }
206
207 var _HashHTMLBlocks = function(text) {
208         // attacklab: Double up blank lines to reduce lookaround
209         text = text.replace(/\n/g,"\n\n");
210
211         // Hashify HTML blocks:
212         // We only want to do this for block-level HTML tags, such as headers,
213         // lists, and tables. That's because we still want to wrap <p>s around
214         // "paragraphs" that are wrapped in non-block-level tags, such as anchors,
215         // phrase emphasis, and spans. The list of tags we're looking for is
216         // hard-coded:
217         var block_tags_a = "p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math|ins|del"
218         var block_tags_b = "p|div|h[1-6]|blockquote|pre|table|dl|ol|ul|script|noscript|form|fieldset|iframe|math"
219
220         // First, look for nested blocks, e.g.:
221         //   <div>
222         //     <div>
223         //     tags for inner block must be indented.
224         //     </div>
225         //   </div>
226         //
227         // The outermost tags must start at the left margin for this to match, and
228         // the inner nested divs must be indented.
229         // We need to do this before the next, more liberal match, because the next
230         // match will start at the first `<div>` and stop at the first `</div>`.
231
232         // attacklab: This regex can be expensive when it fails.
233         /*
234                 var text = text.replace(/
235                 (                                               // save in $1
236                         ^                                       // start of line  (with /m)
237                         <($block_tags_a)        // start tag = $2
238                         \b                                      // word break
239                                                                 // attacklab: hack around khtml/pcre bug...
240                         [^\r]*?\n                       // any number of lines, minimally matching
241                         </\2>                           // the matching end tag
242                         [ \t]*                          // trailing spaces/tabs
243                         (?=\n+)                         // followed by a newline
244                 )                                               // attacklab: there are sentinel newlines at end of document
245                 /gm,function(){...}};
246         */
247         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);
248
249         //
250         // Now match more liberally, simply from `\n<tag>` to `</tag>\n`
251         //
252
253         /*
254                 var text = text.replace(/
255                 (                                               // save in $1
256                         ^                                       // start of line  (with /m)
257                         <($block_tags_b)        // start tag = $2
258                         \b                                      // word break
259                                                                 // attacklab: hack around khtml/pcre bug...
260                         [^\r]*?                         // any number of lines, minimally matching
261                         .*</\2>                         // the matching end tag
262                         [ \t]*                          // trailing spaces/tabs
263                         (?=\n+)                         // followed by a newline
264                 )                                               // attacklab: there are sentinel newlines at end of document
265                 /gm,function(){...}};
266         */
267         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);
268
269         // Special case just for <hr />. It was easier to make a special case than
270         // to make the other regex more complicated.  
271
272         /*
273                 text = text.replace(/
274                 (                                               // save in $1
275                         \n\n                            // Starting after a blank line
276                         [ ]{0,3}
277                         (<(hr)                          // start tag = $2
278                         \b                                      // word break
279                         ([^<>])*?                       // 
280                         \/?>)                           // the matching end tag
281                         [ \t]*
282                         (?=\n{2,})                      // followed by a blank line
283                 )
284                 /g,hashElement);
285         */
286         text = text.replace(/(\n[ ]{0,3}(<(hr)\b([^<>])*?\/?>)[ \t]*(?=\n{2,}))/g,hashElement);
287
288         // Special case for standalone HTML comments:
289
290         /*
291                 text = text.replace(/
292                 (                                               // save in $1
293                         \n\n                            // Starting after a blank line
294                         [ ]{0,3}                        // attacklab: g_tab_width - 1
295                         <!
296                         (--[^\r]*?--\s*)+
297                         >
298                         [ \t]*
299                         (?=\n{2,})                      // followed by a blank line
300                 )
301                 /g,hashElement);
302         */
303         text = text.replace(/(\n\n[ ]{0,3}<!(--[^\r]*?--\s*)+>[ \t]*(?=\n{2,}))/g,hashElement);
304
305         // PHP and ASP-style processor instructions (<?...?> and <%...%>)
306
307         /*
308                 text = text.replace(/
309                 (?:
310                         \n\n                            // Starting after a blank line
311                 )
312                 (                                               // save in $1
313                         [ ]{0,3}                        // attacklab: g_tab_width - 1
314                         (?:
315                                 <([?%])                 // $2
316                                 [^\r]*?
317                                 \2>
318                         )
319                         [ \t]*
320                         (?=\n{2,})                      // followed by a blank line
321                 )
322                 /g,hashElement);
323         */
324         text = text.replace(/(?:\n\n)([ ]{0,3}(?:<([?%])[^\r]*?\2>)[ \t]*(?=\n{2,}))/g,hashElement);
325
326         // attacklab: Undo double lines (see comment at top of this function)
327         text = text.replace(/\n\n/g,"\n");
328         return text;
329 }
330
331 var hashElement = function(wholeMatch,m1) {
332         var blockText = m1;
333
334         // Undo double lines
335         blockText = blockText.replace(/\n\n/g,"\n");
336         blockText = blockText.replace(/^\n/,"");
337         
338         // strip trailing blank lines
339         blockText = blockText.replace(/\n+$/g,"");
340         
341         // Replace the element text with a marker ("~KxK" where x is its key)
342         blockText = "\n\n~K" + (g_html_blocks.push(blockText)-1) + "K\n\n";
343         
344         return blockText;
345 };
346
347 var _RunBlockGamut = function(text) {
348 //
349 // These are all the transformations that form block-level
350 // tags like paragraphs, headers, and list items.
351 //
352         text = _DoHeaders(text);
353
354         // Do Horizontal Rules:
355         var key = hashBlock("<hr />");
356         text = text.replace(/^[ ]{0,2}([ ]?\*[ ]?){3,}[ \t]*$/gm,key);
357         text = text.replace(/^[ ]{0,2}([ ]?-[ ]?){3,}[ \t]*$/gm,key);
358         text = text.replace(/^[ ]{0,2}([ ]?_[ ]?){3,}[ \t]*$/gm,key);
359
360         text = _DoLists(text);
361         text = _DoCodeBlocks(text);
362         text = _DoBlockQuotes(text);
363
364         // We already ran _HashHTMLBlocks() before, in Markdown(), but that
365         // was to escape raw HTML in the original Markdown source. This time,
366         // we're escaping the markup we've just created, so that we don't wrap
367         // <p> tags around block-level tags.
368         text = _HashHTMLBlocks(text);
369         text = _FormParagraphs(text);
370
371         return text;
372 }
373
374
375 var _RunSpanGamut = function(text) {
376 //
377 // These are all the transformations that occur *within* block-level
378 // tags like paragraphs, headers, and list items.
379 //
380
381         text = _DoCodeSpans(text);
382         text = _EscapeSpecialCharsWithinTagAttributes(text);
383         text = _EncodeBackslashEscapes(text);
384
385         // Process anchor and image tags. Images must come first,
386         // because ![foo][f] looks like an anchor.
387         text = _DoImages(text);
388         text = _DoAnchors(text);
389
390         // Make links out of things like `<http://example.com/>`
391         // Must come after _DoAnchors(), because you can use < and >
392         // delimiters in inline links like [this](<url>).
393         text = _DoAutoLinks(text);
394         text = _EncodeAmpsAndAngles(text);
395         text = _DoItalicsAndBold(text);
396
397         // Do hard breaks:
398         text = text.replace(/  +\n/g," <br />\n");
399
400         return text;
401 }
402
403 var _EscapeSpecialCharsWithinTagAttributes = function(text) {
404 //
405 // Within tags -- meaning between < and > -- encode [\ ` * _] so they
406 // don't conflict with their use in Markdown for code, italics and strong.
407 //
408
409         // Build a regex to find HTML tags and comments.  See Friedl's 
410         // "Mastering Regular Expressions", 2nd Ed., pp. 200-201.
411         var regex = /(<[a-z\/!$]("[^"]*"|'[^']*'|[^'">])*>|<!(--.*?--\s*)+>)/gi;
412
413         text = text.replace(regex, function(wholeMatch) {
414                 var tag = wholeMatch.replace(/(.)<\/?code>(?=.)/g,"$1`");
415                 tag = escapeCharacters(tag,"\\`*_");
416                 return tag;
417         });
418
419         return text;
420 }
421
422 var _DoAnchors = function(text) {
423 //
424 // Turn Markdown link shortcuts into XHTML <a> tags.
425 //
426         //
427         // First, handle reference-style links: [link text] [id]
428         //
429
430         /*
431                 text = text.replace(/
432                 (                                                       // wrap whole match in $1
433                         \[
434                         (
435                                 (?:
436                                         \[[^\]]*\]              // allow brackets nested one level
437                                         |
438                                         [^\[]                   // or anything else
439                                 )*
440                         )
441                         \]
442
443                         [ ]?                                    // one optional space
444                         (?:\n[ ]*)?                             // one optional newline followed by spaces
445
446                         \[
447                         (.*?)                                   // id = $3
448                         \]
449                 )()()()()                                       // pad remaining backreferences
450                 /g,_DoAnchors_callback);
451         */
452         text = text.replace(/(\[((?:\[[^\]]*\]|[^\[\]])*)\][ ]?(?:\n[ ]*)?\[(.*?)\])()()()()/g,writeAnchorTag);
453
454         //
455         // Next, inline-style links: [link text](url "optional title")
456         //
457
458         /*
459                 text = text.replace(/
460                         (                                               // wrap whole match in $1
461                                 \[
462                                 (
463                                         (?:
464                                                 \[[^\]]*\]      // allow brackets nested one level
465                                         |
466                                         [^\[\]]                 // or anything else
467                                 )
468                         )
469                         \]
470                         \(                                              // literal paren
471                         [ \t]*
472                         ()                                              // no id, so leave $3 empty
473                         <?(.*?)>?                               // href = $4
474                         [ \t]*
475                         (                                               // $5
476                                 (['"])                          // quote char = $6
477                                 (.*?)                           // Title = $7
478                                 \6                                      // matching quote
479                                 [ \t]*                          // ignore any spaces/tabs between closing quote and )
480                         )?                                              // title is optional
481                         \)
482                 )
483                 /g,writeAnchorTag);
484         */
485         text = text.replace(/(\[((?:\[[^\]]*\]|[^\[\]])*)\]\([ \t]*()<?(.*?)>?[ \t]*((['"])(.*?)\6[ \t]*)?\))/g,writeAnchorTag);
486
487         //
488         // Last, handle reference-style shortcuts: [link text]
489         // These must come last in case you've also got [link test][1]
490         // or [link test](/foo)
491         //
492
493         /*
494                 text = text.replace(/
495                 (                                                       // wrap whole match in $1
496                         \[
497                         ([^\[\]]+)                              // link text = $2; can't contain '[' or ']'
498                         \]
499                 )()()()()()                                     // pad rest of backreferences
500                 /g, writeAnchorTag);
501         */
502         text = text.replace(/(\[([^\[\]]+)\])()()()()()/g, writeAnchorTag);
503
504         return text;
505 }
506
507 var writeAnchorTag = function(wholeMatch,m1,m2,m3,m4,m5,m6,m7) {
508         if (m7 == undefined) m7 = "";
509         var whole_match = m1;
510         var link_text   = m2;
511         var link_id      = m3.toLowerCase();
512         var url         = m4;
513         var title       = m7;
514         
515         if (url == "") {
516                 if (link_id == "") {
517                         // lower-case and turn embedded newlines into spaces
518                         link_id = link_text.toLowerCase().replace(/ ?\n/g," ");
519                 }
520                 url = "#"+link_id;
521                 
522                 if (g_urls[link_id] != undefined) {
523                         url = g_urls[link_id];
524                         if (g_titles[link_id] != undefined) {
525                                 title = g_titles[link_id];
526                         }
527                 }
528                 else {
529                         if (whole_match.search(/\(\s*\)$/m)>-1) {
530                                 // Special case for explicit empty url
531                                 url = "";
532                         } else {
533                                 return whole_match;
534                         }
535                 }
536         }       
537         
538         url = escapeCharacters(url,"*_");
539         var result = "<a href=\"" + url + "\"";
540         
541         if (title != "") {
542                 title = title.replace(/"/g,"&quot;");
543                 title = escapeCharacters(title,"*_");
544                 result +=  " title=\"" + title + "\"";
545         }
546         
547         result += ">" + link_text + "</a>";
548         
549         return result;
550 }
551
552
553 var _DoImages = function(text) {
554 //
555 // Turn Markdown image shortcuts into <img> tags.
556 //
557
558         //
559         // First, handle reference-style labeled images: ![alt text][id]
560         //
561
562         /*
563                 text = text.replace(/
564                 (                                               // wrap whole match in $1
565                         !\[
566                         (.*?)                           // alt text = $2
567                         \]
568
569                         [ ]?                            // one optional space
570                         (?:\n[ ]*)?                     // one optional newline followed by spaces
571
572                         \[
573                         (.*?)                           // id = $3
574                         \]
575                 )()()()()                               // pad rest of backreferences
576                 /g,writeImageTag);
577         */
578         text = text.replace(/(!\[(.*?)\][ ]?(?:\n[ ]*)?\[(.*?)\])()()()()/g,writeImageTag);
579
580         //
581         // Next, handle inline images:  ![alt text](url "optional title")
582         // Don't forget: encode * and _
583
584         /*
585                 text = text.replace(/
586                 (                                               // wrap whole match in $1
587                         !\[
588                         (.*?)                           // alt text = $2
589                         \]
590                         \s?                                     // One optional whitespace character
591                         \(                                      // literal paren
592                         [ \t]*
593                         ()                                      // no id, so leave $3 empty
594                         <?(\S+?)>?                      // src url = $4
595                         [ \t]*
596                         (                                       // $5
597                                 (['"])                  // quote char = $6
598                                 (.*?)                   // title = $7
599                                 \6                              // matching quote
600                                 [ \t]*
601                         )?                                      // title is optional
602                 \)
603                 )
604                 /g,writeImageTag);
605         */
606         text = text.replace(/(!\[(.*?)\]\s?\([ \t]*()<?(\S+?)>?[ \t]*((['"])(.*?)\6[ \t]*)?\))/g,writeImageTag);
607
608         return text;
609 }
610
611 var writeImageTag = function(wholeMatch,m1,m2,m3,m4,m5,m6,m7) {
612         var whole_match = m1;
613         var alt_text   = m2;
614         var link_id      = m3.toLowerCase();
615         var url         = m4;
616         var title       = m7;
617
618         if (!title) title = "";
619         
620         if (url == "") {
621                 if (link_id == "") {
622                         // lower-case and turn embedded newlines into spaces
623                         link_id = alt_text.toLowerCase().replace(/ ?\n/g," ");
624                 }
625                 url = "#"+link_id;
626                 
627                 if (g_urls[link_id] != undefined) {
628                         url = g_urls[link_id];
629                         if (g_titles[link_id] != undefined) {
630                                 title = g_titles[link_id];
631                         }
632                 }
633                 else {
634                         return whole_match;
635                 }
636         }       
637         
638         alt_text = alt_text.replace(/"/g,"&quot;");
639         url = escapeCharacters(url,"*_");
640         var result = "<img src=\"" + url + "\" alt=\"" + alt_text + "\"";
641
642         // attacklab: Markdown.pl adds empty title attributes to images.
643         // Replicate this bug.
644
645         //if (title != "") {
646                 title = title.replace(/"/g,"&quot;");
647                 title = escapeCharacters(title,"*_");
648                 result +=  " title=\"" + title + "\"";
649         //}
650         
651         result += " />";
652         
653         return result;
654 }
655
656
657 var _DoHeaders = function(text) {
658
659         // Setext-style headers:
660         //      Header 1
661         //      ========
662         //  
663         //      Header 2
664         //      --------
665         //
666         text = text.replace(/^(.+)[ \t]*\n=+[ \t]*\n+/gm,
667                 function(wholeMatch,m1){return hashBlock("<h1>" + _RunSpanGamut(m1) + "</h1>");});
668
669         text = text.replace(/^(.+)[ \t]*\n-+[ \t]*\n+/gm,
670                 function(matchFound,m1){return hashBlock("<h2>" + _RunSpanGamut(m1) + "</h2>");});
671
672         // atx-style headers:
673         //  # Header 1
674         //  ## Header 2
675         //  ## Header 2 with closing hashes ##
676         //  ...
677         //  ###### Header 6
678         //
679
680         /*
681                 text = text.replace(/
682                         ^(\#{1,6})                              // $1 = string of #'s
683                         [ \t]*
684                         (.+?)                                   // $2 = Header text
685                         [ \t]*
686                         \#*                                             // optional closing #'s (not counted)
687                         \n+
688                 /gm, function() {...});
689         */
690
691         text = text.replace(/^(\#{1,6})[ \t]*(.+?)[ \t]*\#*\n+/gm,
692                 function(wholeMatch,m1,m2) {
693                         var h_level = m1.length;
694                         return hashBlock("<h" + h_level + ">" + _RunSpanGamut(m2) + "</h" + h_level + ">");
695                 });
696
697         return text;
698 }
699
700 // This declaration keeps Dojo compressor from outputting garbage:
701 var _ProcessListItems;
702
703 var _DoLists = function(text) {
704 //
705 // Form HTML ordered (numbered) and unordered (bulleted) lists.
706 //
707
708         // attacklab: add sentinel to hack around khtml/safari bug:
709         // http://bugs.webkit.org/show_bug.cgi?id=11231
710         text += "~0";
711
712         // Re-usable pattern to match any entirel ul or ol list:
713
714         /*
715                 var whole_list = /
716                 (                                                                       // $1 = whole list
717                         (                                                               // $2
718                                 [ ]{0,3}                                        // attacklab: g_tab_width - 1
719                                 ([*+-]|\d+[.])                          // $3 = first list item marker
720                                 [ \t]+
721                         )
722                         [^\r]+?
723                         (                                                               // $4
724                                 ~0                                                      // sentinel for workaround; should be $
725                         |
726                                 \n{2,}
727                                 (?=\S)
728                                 (?!                                                     // Negative lookahead for another list item marker
729                                         [ \t]*
730                                         (?:[*+-]|\d+[.])[ \t]+
731                                 )
732                         )
733                 )/g
734         */
735         var whole_list = /^(([ ]{0,3}([*+-]|\d+[.])[ \t]+)[^\r]+?(~0|\n{2,}(?=\S)(?![ \t]*(?:[*+-]|\d+[.])[ \t]+)))/gm;
736
737         if (g_list_level) {
738                 text = text.replace(whole_list,function(wholeMatch,m1,m2) {
739                         var list = m1;
740                         var list_type = (m2.search(/[*+-]/g)>-1) ? "ul" : "ol";
741
742                         // Turn double returns into triple returns, so that we can make a
743                         // paragraph for the last item in a list, if necessary:
744                         list = list.replace(/\n{2,}/g,"\n\n\n");;
745                         var result = _ProcessListItems(list);
746         
747                         // Trim any trailing whitespace, to put the closing `</$list_type>`
748                         // up on the preceding line, to get it past the current stupid
749                         // HTML block parser. This is a hack to work around the terrible
750                         // hack that is the HTML block parser.
751                         result = result.replace(/\s+$/,"");
752                         result = "<"+list_type+">" + result + "</"+list_type+">\n";
753                         return result;
754                 });
755         } else {
756                 whole_list = /(\n\n|^\n?)(([ ]{0,3}([*+-]|\d+[.])[ \t]+)[^\r]+?(~0|\n{2,}(?=\S)(?![ \t]*(?:[*+-]|\d+[.])[ \t]+)))/g;
757                 text = text.replace(whole_list,function(wholeMatch,m1,m2,m3) {
758                         var runup = m1;
759                         var list = m2;
760
761                         var list_type = (m3.search(/[*+-]/g)>-1) ? "ul" : "ol";
762                         // Turn double returns into triple returns, so that we can make a
763                         // paragraph for the last item in a list, if necessary:
764                         var list = list.replace(/\n{2,}/g,"\n\n\n");;
765                         var result = _ProcessListItems(list);
766                         result = runup + "<"+list_type+">\n" + result + "</"+list_type+">\n";   
767                         return result;
768                 });
769         }
770
771         // attacklab: strip sentinel
772         text = text.replace(/~0/,"");
773
774         return text;
775 }
776
777 _ProcessListItems = function(list_str) {
778 //
779 //  Process the contents of a single ordered or unordered list, splitting it
780 //  into individual list items.
781 //
782         // The $g_list_level global keeps track of when we're inside a list.
783         // Each time we enter a list, we increment it; when we leave a list,
784         // we decrement. If it's zero, we're not in a list anymore.
785         //
786         // We do this because when we're not inside a list, we want to treat
787         // something like this:
788         //
789         //    I recommend upgrading to version
790         //    8. Oops, now this line is treated
791         //    as a sub-list.
792         //
793         // As a single paragraph, despite the fact that the second line starts
794         // with a digit-period-space sequence.
795         //
796         // Whereas when we're inside a list (or sub-list), that line will be
797         // treated as the start of a sub-list. What a kludge, huh? This is
798         // an aspect of Markdown's syntax that's hard to parse perfectly
799         // without resorting to mind-reading. Perhaps the solution is to
800         // change the syntax rules such that sub-lists must start with a
801         // starting cardinal number; e.g. "1." or "a.".
802
803         g_list_level++;
804
805         // trim trailing blank lines:
806         list_str = list_str.replace(/\n{2,}$/,"\n");
807
808         // attacklab: add sentinel to emulate \z
809         list_str += "~0";
810
811         /*
812                 list_str = list_str.replace(/
813                         (\n)?                                                   // leading line = $1
814                         (^[ \t]*)                                               // leading whitespace = $2
815                         ([*+-]|\d+[.]) [ \t]+                   // list marker = $3
816                         ([^\r]+?                                                // list item text   = $4
817                         (\n{1,2}))
818                         (?= \n* (~0 | \2 ([*+-]|\d+[.]) [ \t]+))
819                 /gm, function(){...});
820         */
821         list_str = list_str.replace(/(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+([^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm,
822                 function(wholeMatch,m1,m2,m3,m4){
823                         var item = m4;
824                         var leading_line = m1;
825                         var leading_space = m2;
826
827                         if (leading_line || (item.search(/\n{2,}/)>-1)) {
828                                 item = _RunBlockGamut(_Outdent(item));
829                         }
830                         else {
831                                 // Recursion for sub-lists:
832                                 item = _DoLists(_Outdent(item));
833                                 item = item.replace(/\n$/,""); // chomp(item)
834                                 item = _RunSpanGamut(item);
835                         }
836
837                         return  "<li>" + item + "</li>\n";
838                 }
839         );
840
841         // attacklab: strip sentinel
842         list_str = list_str.replace(/~0/g,"");
843
844         g_list_level--;
845         return list_str;
846 }
847
848
849 var _DoCodeBlocks = function(text) {
850 //
851 //  Process Markdown `<pre><code>` blocks.
852 //  
853
854         /*
855                 text = text.replace(text,
856                         /(?:\n\n|^)
857                         (                                                               // $1 = the code block -- one or more lines, starting with a space/tab
858                                 (?:
859                                         (?:[ ]{4}|\t)                   // Lines must start with a tab or a tab-width of spaces - attacklab: g_tab_width
860                                         .*\n+
861                                 )+
862                         )
863                         (\n*[ ]{0,3}[^ \t\n]|(?=~0))    // attacklab: g_tab_width
864                 /g,function(){...});
865         */
866
867         // attacklab: sentinel workarounds for lack of \A and \Z, safari\khtml bug
868         text += "~0";
869         
870         text = text.replace(/(?:\n\n|^)((?:(?:[ ]{4}|\t).*\n+)+)(\n*[ ]{0,3}[^ \t\n]|(?=~0))/g,
871                 function(wholeMatch,m1,m2) {
872                         var codeblock = m1;
873                         var nextChar = m2;
874                 
875                         codeblock = _EncodeCode( _Outdent(codeblock));
876                         codeblock = _Detab(codeblock);
877                         codeblock = codeblock.replace(/^\n+/g,""); // trim leading newlines
878                         codeblock = codeblock.replace(/\n+$/g,""); // trim trailing whitespace
879
880                         codeblock = "<pre><code>" + codeblock + "\n</code></pre>";
881
882                         return hashBlock(codeblock) + nextChar;
883                 }
884         );
885
886         // attacklab: strip sentinel
887         text = text.replace(/~0/,"");
888
889         return text;
890 }
891
892 var hashBlock = function(text) {
893         text = text.replace(/(^\n+|\n+$)/g,"");
894         return "\n\n~K" + (g_html_blocks.push(text)-1) + "K\n\n";
895 }
896
897
898 var _DoCodeSpans = function(text) {
899 //
900 //   *  Backtick quotes are used for <code></code> spans.
901 // 
902 //   *  You can use multiple backticks as the delimiters if you want to
903 //       include literal backticks in the code span. So, this input:
904 //       
905 //               Just type ``foo `bar` baz`` at the prompt.
906 //       
907 //         Will translate to:
908 //       
909 //               <p>Just type <code>foo `bar` baz</code> at the prompt.</p>
910 //       
911 //      There's no arbitrary limit to the number of backticks you
912 //      can use as delimters. If you need three consecutive backticks
913 //      in your code, use four for delimiters, etc.
914 //
915 //  *  You can use spaces to get literal backticks at the edges:
916 //       
917 //               ... type `` `bar` `` ...
918 //       
919 //         Turns to:
920 //       
921 //               ... type <code>`bar`</code> ...
922 //
923
924         /*
925                 text = text.replace(/
926                         (^|[^\\])                                       // Character before opening ` can't be a backslash
927                         (`+)                                            // $2 = Opening run of `
928                         (                                                       // $3 = The code block
929                                 [^\r]*?
930                                 [^`]                                    // attacklab: work around lack of lookbehind
931                         )
932                         \2                                                      // Matching closer
933                         (?!`)
934                 /gm, function(){...});
935         */
936
937         text = text.replace(/(^|[^\\])(`+)([^\r]*?[^`])\2(?!`)/gm,
938                 function(wholeMatch,m1,m2,m3,m4) {
939                         var c = m3;
940                         c = c.replace(/^([ \t]*)/g,""); // leading whitespace
941                         c = c.replace(/[ \t]*$/g,"");   // trailing whitespace
942                         c = _EncodeCode(c);
943                         return m1+"<code>"+c+"</code>";
944                 });
945
946         return text;
947 }
948
949
950 var _EncodeCode = function(text) {
951 //
952 // Encode/escape certain characters inside Markdown code runs.
953 // The point is that in code, these characters are literals,
954 // and lose their special Markdown meanings.
955 //
956         // Encode all ampersands; HTML entities are not
957         // entities within a Markdown code span.
958         text = text.replace(/&/g,"&amp;");
959
960         // Do the angle bracket song and dance:
961         text = text.replace(/</g,"&lt;");
962         text = text.replace(/>/g,"&gt;");
963
964         // Now, escape characters that are magic in Markdown:
965         text = escapeCharacters(text,"\*_{}[]\\",false);
966
967 // jj the line above breaks this:
968 //---
969
970 //* Item
971
972 //   1. Subitem
973
974 //            special char: *
975 //---
976
977         return text;
978 }
979
980
981 var _DoItalicsAndBold = function(text) {
982
983         // <strong> must go first:
984         text = text.replace(/(\*\*|__)(?=\S)([^\r]*?\S[\*_]*)\1/g,
985                 "<strong>$2</strong>");
986
987         text = text.replace(/(\*|_)(?=\S)([^\r]*?\S)\1/g,
988                 "<em>$2</em>");
989
990         return text;
991 }
992
993
994 var _DoBlockQuotes = function(text) {
995
996         /*
997                 text = text.replace(/
998                 (                                                               // Wrap whole match in $1
999                         (
1000                                 ^[ \t]*>[ \t]?                  // '>' at the start of a line
1001                                 .+\n                                    // rest of the first line
1002                                 (.+\n)*                                 // subsequent consecutive lines
1003                                 \n*                                             // blanks
1004                         )+
1005                 )
1006                 /gm, function(){...});
1007         */
1008
1009         text = text.replace(/((^[ \t]*>[ \t]?.+\n(.+\n)*\n*)+)/gm,
1010                 function(wholeMatch,m1) {
1011                         var bq = m1;
1012
1013                         // attacklab: hack around Konqueror 3.5.4 bug:
1014                         // "----------bug".replace(/^-/g,"") == "bug"
1015
1016                         bq = bq.replace(/^[ \t]*>[ \t]?/gm,"~0");       // trim one level of quoting
1017
1018                         // attacklab: clean up hack
1019                         bq = bq.replace(/~0/g,"");
1020
1021                         bq = bq.replace(/^[ \t]+$/gm,"");               // trim whitespace-only lines
1022                         bq = _RunBlockGamut(bq);                                // recurse
1023                         
1024                         bq = bq.replace(/(^|\n)/g,"$1  ");
1025                         // These leading spaces screw with <pre> content, so we need to fix that:
1026                         bq = bq.replace(
1027                                         /(\s*<pre>[^\r]+?<\/pre>)/gm,
1028                                 function(wholeMatch,m1) {
1029                                         var pre = m1;
1030                                         // attacklab: hack around Konqueror 3.5.4 bug:
1031                                         pre = pre.replace(/^  /mg,"~0");
1032                                         pre = pre.replace(/~0/g,"");
1033                                         return pre;
1034                                 });
1035                         
1036                         return hashBlock("<blockquote>\n" + bq + "\n</blockquote>");
1037                 });
1038         return text;
1039 }
1040
1041
1042 var _FormParagraphs = function(text) {
1043 //
1044 //  Params:
1045 //    $text - string to process with html <p> tags
1046 //
1047
1048         // Strip leading and trailing lines:
1049         text = text.replace(/^\n+/g,"");
1050         text = text.replace(/\n+$/g,"");
1051
1052         var grafs = text.split(/\n{2,}/g);
1053         var grafsOut = new Array();
1054
1055         //
1056         // Wrap <p> tags.
1057         //
1058         var end = grafs.length;
1059         for (var i=0; i<end; i++) {
1060                 var str = grafs[i];
1061
1062                 // if this is an HTML marker, copy it
1063                 if (str.search(/~K(\d+)K/g) >= 0) {
1064                         grafsOut.push(str);
1065                 }
1066                 else if (str.search(/\S/) >= 0) {
1067                         str = _RunSpanGamut(str);
1068                         str = str.replace(/^([ \t]*)/g,"<p>");
1069                         str += "</p>"
1070                         grafsOut.push(str);
1071                 }
1072
1073         }
1074
1075         //
1076         // Unhashify HTML blocks
1077         //
1078         end = grafsOut.length;
1079         for (var i=0; i<end; i++) {
1080                 // if this is a marker for an html block...
1081                 while (grafsOut[i].search(/~K(\d+)K/) >= 0) {
1082                         var blockText = g_html_blocks[RegExp.$1];
1083                         blockText = blockText.replace(/\$/g,"$$$$"); // Escape any dollar signs
1084                         grafsOut[i] = grafsOut[i].replace(/~K\d+K/,blockText);
1085                 }
1086         }
1087
1088         return grafsOut.join("\n\n");
1089 }
1090
1091
1092 var _EncodeAmpsAndAngles = function(text) {
1093 // Smart processing for ampersands and angle brackets that need to be encoded.
1094         
1095         // Ampersand-encoding based entirely on Nat Irons's Amputator MT plugin:
1096         //   http://bumppo.net/projects/amputator/
1097         text = text.replace(/&(?!#?[xX]?(?:[0-9a-fA-F]+|\w+);)/g,"&amp;");
1098         
1099         // Encode naked <'s
1100         text = text.replace(/<(?![a-z\/?\$!])/gi,"&lt;");
1101         
1102         return text;
1103 }
1104
1105
1106 var _EncodeBackslashEscapes = function(text) {
1107 //
1108 //   Parameter:  String.
1109 //   Returns:   The string, with after processing the following backslash
1110 //                         escape sequences.
1111 //
1112
1113         // attacklab: The polite way to do this is with the new
1114         // escapeCharacters() function:
1115         //
1116         //      text = escapeCharacters(text,"\\",true);
1117         //      text = escapeCharacters(text,"`*_{}[]()>#+-.!",true);
1118         //
1119         // ...but we're sidestepping its use of the (slow) RegExp constructor
1120         // as an optimization for Firefox.  This function gets called a LOT.
1121
1122         text = text.replace(/\\(\\)/g,escapeCharacters_callback);
1123         text = text.replace(/\\([`*_{}\[\]()>#+-.!])/g,escapeCharacters_callback);
1124         return text;
1125 }
1126
1127
1128 var _DoAutoLinks = function(text) {
1129
1130         text = text.replace(/<((https?|ftp|dict):[^'">\s]+)>/gi,"<a href=\"$1\">$1</a>");
1131
1132         // Email addresses: <address@domain.foo>
1133
1134         /*
1135                 text = text.replace(/
1136                         <
1137                         (?:mailto:)?
1138                         (
1139                                 [-.\w]+
1140                                 \@
1141                                 [-a-z0-9]+(\.[-a-z0-9]+)*\.[a-z]+
1142                         )
1143                         >
1144                 /gi, _DoAutoLinks_callback());
1145         */
1146         text = text.replace(/<(?:mailto:)?([-.\w]+\@[-a-z0-9]+(\.[-a-z0-9]+)*\.[a-z]+)>/gi,
1147                 function(wholeMatch,m1) {
1148                         return _EncodeEmailAddress( _UnescapeSpecialChars(m1) );
1149                 }
1150         );
1151
1152         return text;
1153 }
1154
1155
1156 var _EncodeEmailAddress = function(addr) {
1157 //
1158 //  Input: an email address, e.g. "foo@example.com"
1159 //
1160 //  Output: the email address as a mailto link, with each character
1161 //      of the address encoded as either a decimal or hex entity, in
1162 //      the hopes of foiling most address harvesting spam bots. E.g.:
1163 //
1164 //      <a href="&#x6D;&#97;&#105;&#108;&#x74;&#111;:&#102;&#111;&#111;&#64;&#101;
1165 //         x&#x61;&#109;&#x70;&#108;&#x65;&#x2E;&#99;&#111;&#109;">&#102;&#111;&#111;
1166 //         &#64;&#101;x&#x61;&#109;&#x70;&#108;&#x65;&#x2E;&#99;&#111;&#109;</a>
1167 //
1168 //  Based on a filter by Matthew Wickline, posted to the BBEdit-Talk
1169 //  mailing list: <http://tinyurl.com/yu7ue>
1170 //
1171
1172         // attacklab: why can't javascript speak hex?
1173         function char2hex(ch) {
1174                 var hexDigits = '0123456789ABCDEF';
1175                 var dec = ch.charCodeAt(0);
1176                 return(hexDigits.charAt(dec>>4) + hexDigits.charAt(dec&15));
1177         }
1178
1179         var encode = [
1180                 function(ch){return "&#"+ch.charCodeAt(0)+";";},
1181                 function(ch){return "&#x"+char2hex(ch)+";";},
1182                 function(ch){return ch;}
1183         ];
1184
1185         addr = "mailto:" + addr;
1186
1187         addr = addr.replace(/./g, function(ch) {
1188                 if (ch == "@") {
1189                         // this *must* be encoded. I insist.
1190                         ch = encode[Math.floor(Math.random()*2)](ch);
1191                 } else if (ch !=":") {
1192                         // leave ':' alone (to spot mailto: later)
1193                         var r = Math.random();
1194                         // roughly 10% raw, 45% hex, 45% dec
1195                         ch =  (
1196                                         r > .9  ?       encode[2](ch)   :
1197                                         r > .45 ?       encode[1](ch)   :
1198                                                                 encode[0](ch)
1199                                 );
1200                 }
1201                 return ch;
1202         });
1203
1204         addr = "<a href=\"" + addr + "\">" + addr + "</a>";
1205         addr = addr.replace(/">.+:/g,"\">"); // strip the mailto: from the visible part
1206
1207         return addr;
1208 }
1209
1210
1211 var _UnescapeSpecialChars = function(text) {
1212 //
1213 // Swap back in all the special characters we've hidden.
1214 //
1215         text = text.replace(/~E(\d+)E/g,
1216                 function(wholeMatch,m1) {
1217                         var charCodeToReplace = parseInt(m1);
1218                         return String.fromCharCode(charCodeToReplace);
1219                 }
1220         );
1221         return text;
1222 }
1223
1224
1225 var _Outdent = function(text) {
1226 //
1227 // Remove one level of line-leading tabs or spaces
1228 //
1229
1230         // attacklab: hack around Konqueror 3.5.4 bug:
1231         // "----------bug".replace(/^-/g,"") == "bug"
1232
1233         text = text.replace(/^(\t|[ ]{1,4})/gm,"~0"); // attacklab: g_tab_width
1234
1235         // attacklab: clean up hack
1236         text = text.replace(/~0/g,"")
1237
1238         return text;
1239 }
1240
1241 var _Detab = function(text) {
1242 // attacklab: Detab's completely rewritten for speed.
1243 // In perl we could fix it by anchoring the regexp with \G.
1244 // In javascript we're less fortunate.
1245
1246         // expand first n-1 tabs
1247         text = text.replace(/\t(?=\t)/g,"    "); // attacklab: g_tab_width
1248
1249         // replace the nth with two sentinels
1250         text = text.replace(/\t/g,"~A~B");
1251
1252         // use the sentinel to anchor our regex so it doesn't explode
1253         text = text.replace(/~B(.+?)~A/g,
1254                 function(wholeMatch,m1,m2) {
1255                         var leadingText = m1;
1256                         var numSpaces = 4 - leadingText.length % 4;  // attacklab: g_tab_width
1257
1258                         // there *must* be a better way to do this:
1259                         for (var i=0; i<numSpaces; i++) leadingText+=" ";
1260
1261                         return leadingText;
1262                 }
1263         );
1264
1265         // clean up sentinels
1266         text = text.replace(/~A/g,"    ");  // attacklab: g_tab_width
1267         text = text.replace(/~B/g,"");
1268
1269         return text;
1270 }
1271
1272
1273 //
1274 //  attacklab: Utility functions
1275 //
1276
1277
1278 var escapeCharacters = function(text, charsToEscape, afterBackslash) {
1279         // First we have to escape the escape characters so that
1280         // we can build a character class out of them
1281         var regexString = "([" + charsToEscape.replace(/([\[\]\\])/g,"\\$1") + "])";
1282
1283         if (afterBackslash) {
1284                 regexString = "\\\\" + regexString;
1285         }
1286
1287         var regex = new RegExp(regexString,"g");
1288         text = text.replace(regex,escapeCharacters_callback);
1289
1290         return text;
1291 }
1292
1293
1294 var escapeCharacters_callback = function(wholeMatch,m1) {
1295         var charCodeToEscape = m1.charCodeAt(0);
1296         return "~E"+charCodeToEscape+"E";
1297 }
1298
1299 } // end of Attacklab.showdown.converter
1300
1301
1302 // Version 0.9 used the Showdown namespace instead of Attacklab.showdown
1303 // The old namespace is deprecated, but we'll support it for now:
1304 var Showdown = Attacklab.showdown;
1305
1306 // If anyone's interested, tell the world that this file's been loaded
1307 if (Attacklab.fileLoaded) {
1308         Attacklab.fileLoaded("showdown.js");
1309 }