|
1 // |
|
2 // Blackfriday Markdown Processor |
|
3 // Available at http://github.com/russross/blackfriday |
|
4 // |
|
5 // Copyright © 2011 Russ Ross <russ@russross.com>. |
|
6 // Distributed under the Simplified BSD License. |
|
7 // See README.md for details. |
|
8 // |
|
9 |
|
10 // |
|
11 // Functions to parse inline elements. |
|
12 // |
|
13 |
|
14 package blackfriday |
|
15 |
|
16 import ( |
|
17 "bytes" |
|
18 "regexp" |
|
19 "strconv" |
|
20 ) |
|
21 |
|
22 var ( |
|
23 urlRe = `((https?|ftp):\/\/|\/)[-A-Za-z0-9+&@#\/%?=~_|!:,.;\(\)]+` |
|
24 anchorRe = regexp.MustCompile(`^(<a\shref="` + urlRe + `"(\stitle="[^"<>]+")?\s?>` + urlRe + `<\/a>)`) |
|
25 |
|
26 // https://www.w3.org/TR/html5/syntax.html#character-references |
|
27 // highest unicode code point in 17 planes (2^20): 1,114,112d = |
|
28 // 7 dec digits or 6 hex digits |
|
29 // named entity references can be 2-31 characters with stuff like < |
|
30 // at one end and ∳ at the other. There |
|
31 // are also sometimes numbers at the end, although this isn't inherent |
|
32 // in the specification; there are never numbers anywhere else in |
|
33 // current character references, though; see ¾ and ▒, etc. |
|
34 // https://www.w3.org/TR/html5/syntax.html#named-character-references |
|
35 // |
|
36 // entity := "&" (named group | number ref) ";" |
|
37 // named group := [a-zA-Z]{2,31}[0-9]{0,2} |
|
38 // number ref := "#" (dec ref | hex ref) |
|
39 // dec ref := [0-9]{1,7} |
|
40 // hex ref := ("x" | "X") [0-9a-fA-F]{1,6} |
|
41 htmlEntityRe = regexp.MustCompile(`&([a-zA-Z]{2,31}[0-9]{0,2}|#([0-9]{1,7}|[xX][0-9a-fA-F]{1,6}));`) |
|
42 ) |
|
43 |
|
44 // Functions to parse text within a block |
|
45 // Each function returns the number of chars taken care of |
|
46 // data is the complete block being rendered |
|
47 // offset is the number of valid chars before the current cursor |
|
48 |
|
49 func (p *Markdown) inline(currBlock *Node, data []byte) { |
|
50 // handlers might call us recursively: enforce a maximum depth |
|
51 if p.nesting >= p.maxNesting || len(data) == 0 { |
|
52 return |
|
53 } |
|
54 p.nesting++ |
|
55 beg, end := 0, 0 |
|
56 for end < len(data) { |
|
57 handler := p.inlineCallback[data[end]] |
|
58 if handler != nil { |
|
59 if consumed, node := handler(p, data, end); consumed == 0 { |
|
60 // No action from the callback. |
|
61 end++ |
|
62 } else { |
|
63 // Copy inactive chars into the output. |
|
64 currBlock.AppendChild(text(data[beg:end])) |
|
65 if node != nil { |
|
66 currBlock.AppendChild(node) |
|
67 } |
|
68 // Skip past whatever the callback used. |
|
69 beg = end + consumed |
|
70 end = beg |
|
71 } |
|
72 } else { |
|
73 end++ |
|
74 } |
|
75 } |
|
76 if beg < len(data) { |
|
77 if data[end-1] == '\n' { |
|
78 end-- |
|
79 } |
|
80 currBlock.AppendChild(text(data[beg:end])) |
|
81 } |
|
82 p.nesting-- |
|
83 } |
|
84 |
|
85 // single and double emphasis parsing |
|
86 func emphasis(p *Markdown, data []byte, offset int) (int, *Node) { |
|
87 data = data[offset:] |
|
88 c := data[0] |
|
89 |
|
90 if len(data) > 2 && data[1] != c { |
|
91 // whitespace cannot follow an opening emphasis; |
|
92 // strikethrough only takes two characters '~~' |
|
93 if c == '~' || isspace(data[1]) { |
|
94 return 0, nil |
|
95 } |
|
96 ret, node := helperEmphasis(p, data[1:], c) |
|
97 if ret == 0 { |
|
98 return 0, nil |
|
99 } |
|
100 |
|
101 return ret + 1, node |
|
102 } |
|
103 |
|
104 if len(data) > 3 && data[1] == c && data[2] != c { |
|
105 if isspace(data[2]) { |
|
106 return 0, nil |
|
107 } |
|
108 ret, node := helperDoubleEmphasis(p, data[2:], c) |
|
109 if ret == 0 { |
|
110 return 0, nil |
|
111 } |
|
112 |
|
113 return ret + 2, node |
|
114 } |
|
115 |
|
116 if len(data) > 4 && data[1] == c && data[2] == c && data[3] != c { |
|
117 if c == '~' || isspace(data[3]) { |
|
118 return 0, nil |
|
119 } |
|
120 ret, node := helperTripleEmphasis(p, data, 3, c) |
|
121 if ret == 0 { |
|
122 return 0, nil |
|
123 } |
|
124 |
|
125 return ret + 3, node |
|
126 } |
|
127 |
|
128 return 0, nil |
|
129 } |
|
130 |
|
131 func codeSpan(p *Markdown, data []byte, offset int) (int, *Node) { |
|
132 data = data[offset:] |
|
133 |
|
134 nb := 0 |
|
135 |
|
136 // count the number of backticks in the delimiter |
|
137 for nb < len(data) && data[nb] == '`' { |
|
138 nb++ |
|
139 } |
|
140 |
|
141 // find the next delimiter |
|
142 i, end := 0, 0 |
|
143 for end = nb; end < len(data) && i < nb; end++ { |
|
144 if data[end] == '`' { |
|
145 i++ |
|
146 } else { |
|
147 i = 0 |
|
148 } |
|
149 } |
|
150 |
|
151 // no matching delimiter? |
|
152 if i < nb && end >= len(data) { |
|
153 return 0, nil |
|
154 } |
|
155 |
|
156 // trim outside whitespace |
|
157 fBegin := nb |
|
158 for fBegin < end && data[fBegin] == ' ' { |
|
159 fBegin++ |
|
160 } |
|
161 |
|
162 fEnd := end - nb |
|
163 for fEnd > fBegin && data[fEnd-1] == ' ' { |
|
164 fEnd-- |
|
165 } |
|
166 |
|
167 // render the code span |
|
168 if fBegin != fEnd { |
|
169 code := NewNode(Code) |
|
170 code.Literal = data[fBegin:fEnd] |
|
171 return end, code |
|
172 } |
|
173 |
|
174 return end, nil |
|
175 } |
|
176 |
|
177 // newline preceded by two spaces becomes <br> |
|
178 func maybeLineBreak(p *Markdown, data []byte, offset int) (int, *Node) { |
|
179 origOffset := offset |
|
180 for offset < len(data) && data[offset] == ' ' { |
|
181 offset++ |
|
182 } |
|
183 |
|
184 if offset < len(data) && data[offset] == '\n' { |
|
185 if offset-origOffset >= 2 { |
|
186 return offset - origOffset + 1, NewNode(Hardbreak) |
|
187 } |
|
188 return offset - origOffset, nil |
|
189 } |
|
190 return 0, nil |
|
191 } |
|
192 |
|
193 // newline without two spaces works when HardLineBreak is enabled |
|
194 func lineBreak(p *Markdown, data []byte, offset int) (int, *Node) { |
|
195 if p.extensions&HardLineBreak != 0 { |
|
196 return 1, NewNode(Hardbreak) |
|
197 } |
|
198 return 0, nil |
|
199 } |
|
200 |
|
201 type linkType int |
|
202 |
|
203 const ( |
|
204 linkNormal linkType = iota |
|
205 linkImg |
|
206 linkDeferredFootnote |
|
207 linkInlineFootnote |
|
208 ) |
|
209 |
|
210 func isReferenceStyleLink(data []byte, pos int, t linkType) bool { |
|
211 if t == linkDeferredFootnote { |
|
212 return false |
|
213 } |
|
214 return pos < len(data)-1 && data[pos] == '[' && data[pos+1] != '^' |
|
215 } |
|
216 |
|
217 func maybeImage(p *Markdown, data []byte, offset int) (int, *Node) { |
|
218 if offset < len(data)-1 && data[offset+1] == '[' { |
|
219 return link(p, data, offset) |
|
220 } |
|
221 return 0, nil |
|
222 } |
|
223 |
|
224 func maybeInlineFootnote(p *Markdown, data []byte, offset int) (int, *Node) { |
|
225 if offset < len(data)-1 && data[offset+1] == '[' { |
|
226 return link(p, data, offset) |
|
227 } |
|
228 return 0, nil |
|
229 } |
|
230 |
|
231 // '[': parse a link or an image or a footnote |
|
232 func link(p *Markdown, data []byte, offset int) (int, *Node) { |
|
233 // no links allowed inside regular links, footnote, and deferred footnotes |
|
234 if p.insideLink && (offset > 0 && data[offset-1] == '[' || len(data)-1 > offset && data[offset+1] == '^') { |
|
235 return 0, nil |
|
236 } |
|
237 |
|
238 var t linkType |
|
239 switch { |
|
240 // special case: ![^text] == deferred footnote (that follows something with |
|
241 // an exclamation point) |
|
242 case p.extensions&Footnotes != 0 && len(data)-1 > offset && data[offset+1] == '^': |
|
243 t = linkDeferredFootnote |
|
244 // ![alt] == image |
|
245 case offset >= 0 && data[offset] == '!': |
|
246 t = linkImg |
|
247 offset++ |
|
248 // ^[text] == inline footnote |
|
249 // [^refId] == deferred footnote |
|
250 case p.extensions&Footnotes != 0: |
|
251 if offset >= 0 && data[offset] == '^' { |
|
252 t = linkInlineFootnote |
|
253 offset++ |
|
254 } else if len(data)-1 > offset && data[offset+1] == '^' { |
|
255 t = linkDeferredFootnote |
|
256 } |
|
257 // [text] == regular link |
|
258 default: |
|
259 t = linkNormal |
|
260 } |
|
261 |
|
262 data = data[offset:] |
|
263 |
|
264 var ( |
|
265 i = 1 |
|
266 noteID int |
|
267 title, link, altContent []byte |
|
268 textHasNl = false |
|
269 ) |
|
270 |
|
271 if t == linkDeferredFootnote { |
|
272 i++ |
|
273 } |
|
274 |
|
275 // look for the matching closing bracket |
|
276 for level := 1; level > 0 && i < len(data); i++ { |
|
277 switch { |
|
278 case data[i] == '\n': |
|
279 textHasNl = true |
|
280 |
|
281 case data[i-1] == '\\': |
|
282 continue |
|
283 |
|
284 case data[i] == '[': |
|
285 level++ |
|
286 |
|
287 case data[i] == ']': |
|
288 level-- |
|
289 if level <= 0 { |
|
290 i-- // compensate for extra i++ in for loop |
|
291 } |
|
292 } |
|
293 } |
|
294 |
|
295 if i >= len(data) { |
|
296 return 0, nil |
|
297 } |
|
298 |
|
299 txtE := i |
|
300 i++ |
|
301 var footnoteNode *Node |
|
302 |
|
303 // skip any amount of whitespace or newline |
|
304 // (this is much more lax than original markdown syntax) |
|
305 for i < len(data) && isspace(data[i]) { |
|
306 i++ |
|
307 } |
|
308 |
|
309 // inline style link |
|
310 switch { |
|
311 case i < len(data) && data[i] == '(': |
|
312 // skip initial whitespace |
|
313 i++ |
|
314 |
|
315 for i < len(data) && isspace(data[i]) { |
|
316 i++ |
|
317 } |
|
318 |
|
319 linkB := i |
|
320 |
|
321 // look for link end: ' " ) |
|
322 findlinkend: |
|
323 for i < len(data) { |
|
324 switch { |
|
325 case data[i] == '\\': |
|
326 i += 2 |
|
327 |
|
328 case data[i] == ')' || data[i] == '\'' || data[i] == '"': |
|
329 break findlinkend |
|
330 |
|
331 default: |
|
332 i++ |
|
333 } |
|
334 } |
|
335 |
|
336 if i >= len(data) { |
|
337 return 0, nil |
|
338 } |
|
339 linkE := i |
|
340 |
|
341 // look for title end if present |
|
342 titleB, titleE := 0, 0 |
|
343 if data[i] == '\'' || data[i] == '"' { |
|
344 i++ |
|
345 titleB = i |
|
346 |
|
347 findtitleend: |
|
348 for i < len(data) { |
|
349 switch { |
|
350 case data[i] == '\\': |
|
351 i += 2 |
|
352 |
|
353 case data[i] == ')': |
|
354 break findtitleend |
|
355 |
|
356 default: |
|
357 i++ |
|
358 } |
|
359 } |
|
360 |
|
361 if i >= len(data) { |
|
362 return 0, nil |
|
363 } |
|
364 |
|
365 // skip whitespace after title |
|
366 titleE = i - 1 |
|
367 for titleE > titleB && isspace(data[titleE]) { |
|
368 titleE-- |
|
369 } |
|
370 |
|
371 // check for closing quote presence |
|
372 if data[titleE] != '\'' && data[titleE] != '"' { |
|
373 titleB, titleE = 0, 0 |
|
374 linkE = i |
|
375 } |
|
376 } |
|
377 |
|
378 // remove whitespace at the end of the link |
|
379 for linkE > linkB && isspace(data[linkE-1]) { |
|
380 linkE-- |
|
381 } |
|
382 |
|
383 // remove optional angle brackets around the link |
|
384 if data[linkB] == '<' { |
|
385 linkB++ |
|
386 } |
|
387 if data[linkE-1] == '>' { |
|
388 linkE-- |
|
389 } |
|
390 |
|
391 // build escaped link and title |
|
392 if linkE > linkB { |
|
393 link = data[linkB:linkE] |
|
394 } |
|
395 |
|
396 if titleE > titleB { |
|
397 title = data[titleB:titleE] |
|
398 } |
|
399 |
|
400 i++ |
|
401 |
|
402 // reference style link |
|
403 case isReferenceStyleLink(data, i, t): |
|
404 var id []byte |
|
405 altContentConsidered := false |
|
406 |
|
407 // look for the id |
|
408 i++ |
|
409 linkB := i |
|
410 for i < len(data) && data[i] != ']' { |
|
411 i++ |
|
412 } |
|
413 if i >= len(data) { |
|
414 return 0, nil |
|
415 } |
|
416 linkE := i |
|
417 |
|
418 // find the reference |
|
419 if linkB == linkE { |
|
420 if textHasNl { |
|
421 var b bytes.Buffer |
|
422 |
|
423 for j := 1; j < txtE; j++ { |
|
424 switch { |
|
425 case data[j] != '\n': |
|
426 b.WriteByte(data[j]) |
|
427 case data[j-1] != ' ': |
|
428 b.WriteByte(' ') |
|
429 } |
|
430 } |
|
431 |
|
432 id = b.Bytes() |
|
433 } else { |
|
434 id = data[1:txtE] |
|
435 altContentConsidered = true |
|
436 } |
|
437 } else { |
|
438 id = data[linkB:linkE] |
|
439 } |
|
440 |
|
441 // find the reference with matching id |
|
442 lr, ok := p.getRef(string(id)) |
|
443 if !ok { |
|
444 return 0, nil |
|
445 } |
|
446 |
|
447 // keep link and title from reference |
|
448 link = lr.link |
|
449 title = lr.title |
|
450 if altContentConsidered { |
|
451 altContent = lr.text |
|
452 } |
|
453 i++ |
|
454 |
|
455 // shortcut reference style link or reference or inline footnote |
|
456 default: |
|
457 var id []byte |
|
458 |
|
459 // craft the id |
|
460 if textHasNl { |
|
461 var b bytes.Buffer |
|
462 |
|
463 for j := 1; j < txtE; j++ { |
|
464 switch { |
|
465 case data[j] != '\n': |
|
466 b.WriteByte(data[j]) |
|
467 case data[j-1] != ' ': |
|
468 b.WriteByte(' ') |
|
469 } |
|
470 } |
|
471 |
|
472 id = b.Bytes() |
|
473 } else { |
|
474 if t == linkDeferredFootnote { |
|
475 id = data[2:txtE] // get rid of the ^ |
|
476 } else { |
|
477 id = data[1:txtE] |
|
478 } |
|
479 } |
|
480 |
|
481 footnoteNode = NewNode(Item) |
|
482 if t == linkInlineFootnote { |
|
483 // create a new reference |
|
484 noteID = len(p.notes) + 1 |
|
485 |
|
486 var fragment []byte |
|
487 if len(id) > 0 { |
|
488 if len(id) < 16 { |
|
489 fragment = make([]byte, len(id)) |
|
490 } else { |
|
491 fragment = make([]byte, 16) |
|
492 } |
|
493 copy(fragment, slugify(id)) |
|
494 } else { |
|
495 fragment = append([]byte("footnote-"), []byte(strconv.Itoa(noteID))...) |
|
496 } |
|
497 |
|
498 ref := &reference{ |
|
499 noteID: noteID, |
|
500 hasBlock: false, |
|
501 link: fragment, |
|
502 title: id, |
|
503 footnote: footnoteNode, |
|
504 } |
|
505 |
|
506 p.notes = append(p.notes, ref) |
|
507 |
|
508 link = ref.link |
|
509 title = ref.title |
|
510 } else { |
|
511 // find the reference with matching id |
|
512 lr, ok := p.getRef(string(id)) |
|
513 if !ok { |
|
514 return 0, nil |
|
515 } |
|
516 |
|
517 if t == linkDeferredFootnote { |
|
518 lr.noteID = len(p.notes) + 1 |
|
519 lr.footnote = footnoteNode |
|
520 p.notes = append(p.notes, lr) |
|
521 } |
|
522 |
|
523 // keep link and title from reference |
|
524 link = lr.link |
|
525 // if inline footnote, title == footnote contents |
|
526 title = lr.title |
|
527 noteID = lr.noteID |
|
528 } |
|
529 |
|
530 // rewind the whitespace |
|
531 i = txtE + 1 |
|
532 } |
|
533 |
|
534 var uLink []byte |
|
535 if t == linkNormal || t == linkImg { |
|
536 if len(link) > 0 { |
|
537 var uLinkBuf bytes.Buffer |
|
538 unescapeText(&uLinkBuf, link) |
|
539 uLink = uLinkBuf.Bytes() |
|
540 } |
|
541 |
|
542 // links need something to click on and somewhere to go |
|
543 if len(uLink) == 0 || (t == linkNormal && txtE <= 1) { |
|
544 return 0, nil |
|
545 } |
|
546 } |
|
547 |
|
548 // call the relevant rendering function |
|
549 var linkNode *Node |
|
550 switch t { |
|
551 case linkNormal: |
|
552 linkNode = NewNode(Link) |
|
553 linkNode.Destination = normalizeURI(uLink) |
|
554 linkNode.Title = title |
|
555 if len(altContent) > 0 { |
|
556 linkNode.AppendChild(text(altContent)) |
|
557 } else { |
|
558 // links cannot contain other links, so turn off link parsing |
|
559 // temporarily and recurse |
|
560 insideLink := p.insideLink |
|
561 p.insideLink = true |
|
562 p.inline(linkNode, data[1:txtE]) |
|
563 p.insideLink = insideLink |
|
564 } |
|
565 |
|
566 case linkImg: |
|
567 linkNode = NewNode(Image) |
|
568 linkNode.Destination = uLink |
|
569 linkNode.Title = title |
|
570 linkNode.AppendChild(text(data[1:txtE])) |
|
571 i++ |
|
572 |
|
573 case linkInlineFootnote, linkDeferredFootnote: |
|
574 linkNode = NewNode(Link) |
|
575 linkNode.Destination = link |
|
576 linkNode.Title = title |
|
577 linkNode.NoteID = noteID |
|
578 linkNode.Footnote = footnoteNode |
|
579 if t == linkInlineFootnote { |
|
580 i++ |
|
581 } |
|
582 |
|
583 default: |
|
584 return 0, nil |
|
585 } |
|
586 |
|
587 return i, linkNode |
|
588 } |
|
589 |
|
590 func (p *Markdown) inlineHTMLComment(data []byte) int { |
|
591 if len(data) < 5 { |
|
592 return 0 |
|
593 } |
|
594 if data[0] != '<' || data[1] != '!' || data[2] != '-' || data[3] != '-' { |
|
595 return 0 |
|
596 } |
|
597 i := 5 |
|
598 // scan for an end-of-comment marker, across lines if necessary |
|
599 for i < len(data) && !(data[i-2] == '-' && data[i-1] == '-' && data[i] == '>') { |
|
600 i++ |
|
601 } |
|
602 // no end-of-comment marker |
|
603 if i >= len(data) { |
|
604 return 0 |
|
605 } |
|
606 return i + 1 |
|
607 } |
|
608 |
|
609 func stripMailto(link []byte) []byte { |
|
610 if bytes.HasPrefix(link, []byte("mailto://")) { |
|
611 return link[9:] |
|
612 } else if bytes.HasPrefix(link, []byte("mailto:")) { |
|
613 return link[7:] |
|
614 } else { |
|
615 return link |
|
616 } |
|
617 } |
|
618 |
|
619 // autolinkType specifies a kind of autolink that gets detected. |
|
620 type autolinkType int |
|
621 |
|
622 // These are the possible flag values for the autolink renderer. |
|
623 const ( |
|
624 notAutolink autolinkType = iota |
|
625 normalAutolink |
|
626 emailAutolink |
|
627 ) |
|
628 |
|
629 // '<' when tags or autolinks are allowed |
|
630 func leftAngle(p *Markdown, data []byte, offset int) (int, *Node) { |
|
631 data = data[offset:] |
|
632 altype, end := tagLength(data) |
|
633 if size := p.inlineHTMLComment(data); size > 0 { |
|
634 end = size |
|
635 } |
|
636 if end > 2 { |
|
637 if altype != notAutolink { |
|
638 var uLink bytes.Buffer |
|
639 unescapeText(&uLink, data[1:end+1-2]) |
|
640 if uLink.Len() > 0 { |
|
641 link := uLink.Bytes() |
|
642 node := NewNode(Link) |
|
643 node.Destination = link |
|
644 if altype == emailAutolink { |
|
645 node.Destination = append([]byte("mailto:"), link...) |
|
646 } |
|
647 node.AppendChild(text(stripMailto(link))) |
|
648 return end, node |
|
649 } |
|
650 } else { |
|
651 htmlTag := NewNode(HTMLSpan) |
|
652 htmlTag.Literal = data[:end] |
|
653 return end, htmlTag |
|
654 } |
|
655 } |
|
656 |
|
657 return end, nil |
|
658 } |
|
659 |
|
660 // '\\' backslash escape |
|
661 var escapeChars = []byte("\\`*_{}[]()#+-.!:|&<>~") |
|
662 |
|
663 func escape(p *Markdown, data []byte, offset int) (int, *Node) { |
|
664 data = data[offset:] |
|
665 |
|
666 if len(data) > 1 { |
|
667 if p.extensions&BackslashLineBreak != 0 && data[1] == '\n' { |
|
668 return 2, NewNode(Hardbreak) |
|
669 } |
|
670 if bytes.IndexByte(escapeChars, data[1]) < 0 { |
|
671 return 0, nil |
|
672 } |
|
673 |
|
674 return 2, text(data[1:2]) |
|
675 } |
|
676 |
|
677 return 2, nil |
|
678 } |
|
679 |
|
680 func unescapeText(ob *bytes.Buffer, src []byte) { |
|
681 i := 0 |
|
682 for i < len(src) { |
|
683 org := i |
|
684 for i < len(src) && src[i] != '\\' { |
|
685 i++ |
|
686 } |
|
687 |
|
688 if i > org { |
|
689 ob.Write(src[org:i]) |
|
690 } |
|
691 |
|
692 if i+1 >= len(src) { |
|
693 break |
|
694 } |
|
695 |
|
696 ob.WriteByte(src[i+1]) |
|
697 i += 2 |
|
698 } |
|
699 } |
|
700 |
|
701 // '&' escaped when it doesn't belong to an entity |
|
702 // valid entities are assumed to be anything matching &#?[A-Za-z0-9]+; |
|
703 func entity(p *Markdown, data []byte, offset int) (int, *Node) { |
|
704 data = data[offset:] |
|
705 |
|
706 end := 1 |
|
707 |
|
708 if end < len(data) && data[end] == '#' { |
|
709 end++ |
|
710 } |
|
711 |
|
712 for end < len(data) && isalnum(data[end]) { |
|
713 end++ |
|
714 } |
|
715 |
|
716 if end < len(data) && data[end] == ';' { |
|
717 end++ // real entity |
|
718 } else { |
|
719 return 0, nil // lone '&' |
|
720 } |
|
721 |
|
722 ent := data[:end] |
|
723 // undo & escaping or it will be converted to &amp; by another |
|
724 // escaper in the renderer |
|
725 if bytes.Equal(ent, []byte("&")) { |
|
726 ent = []byte{'&'} |
|
727 } |
|
728 |
|
729 return end, text(ent) |
|
730 } |
|
731 |
|
732 func linkEndsWithEntity(data []byte, linkEnd int) bool { |
|
733 entityRanges := htmlEntityRe.FindAllIndex(data[:linkEnd], -1) |
|
734 return entityRanges != nil && entityRanges[len(entityRanges)-1][1] == linkEnd |
|
735 } |
|
736 |
|
737 // hasPrefixCaseInsensitive is a custom implementation of |
|
738 // strings.HasPrefix(strings.ToLower(s), prefix) |
|
739 // we rolled our own because ToLower pulls in a huge machinery of lowercasing |
|
740 // anything from Unicode and that's very slow. Since this func will only be |
|
741 // used on ASCII protocol prefixes, we can take shortcuts. |
|
742 func hasPrefixCaseInsensitive(s, prefix []byte) bool { |
|
743 if len(s) < len(prefix) { |
|
744 return false |
|
745 } |
|
746 delta := byte('a' - 'A') |
|
747 for i, b := range prefix { |
|
748 if b != s[i] && b != s[i]+delta { |
|
749 return false |
|
750 } |
|
751 } |
|
752 return true |
|
753 } |
|
754 |
|
755 var protocolPrefixes = [][]byte{ |
|
756 []byte("http://"), |
|
757 []byte("https://"), |
|
758 []byte("ftp://"), |
|
759 []byte("file://"), |
|
760 []byte("mailto:"), |
|
761 } |
|
762 |
|
763 const shortestPrefix = 6 // len("ftp://"), the shortest of the above |
|
764 |
|
765 func maybeAutoLink(p *Markdown, data []byte, offset int) (int, *Node) { |
|
766 // quick check to rule out most false hits |
|
767 if p.insideLink || len(data) < offset+shortestPrefix { |
|
768 return 0, nil |
|
769 } |
|
770 for _, prefix := range protocolPrefixes { |
|
771 endOfHead := offset + 8 // 8 is the len() of the longest prefix |
|
772 if endOfHead > len(data) { |
|
773 endOfHead = len(data) |
|
774 } |
|
775 if hasPrefixCaseInsensitive(data[offset:endOfHead], prefix) { |
|
776 return autoLink(p, data, offset) |
|
777 } |
|
778 } |
|
779 return 0, nil |
|
780 } |
|
781 |
|
782 func autoLink(p *Markdown, data []byte, offset int) (int, *Node) { |
|
783 // Now a more expensive check to see if we're not inside an anchor element |
|
784 anchorStart := offset |
|
785 offsetFromAnchor := 0 |
|
786 for anchorStart > 0 && data[anchorStart] != '<' { |
|
787 anchorStart-- |
|
788 offsetFromAnchor++ |
|
789 } |
|
790 |
|
791 anchorStr := anchorRe.Find(data[anchorStart:]) |
|
792 if anchorStr != nil { |
|
793 anchorClose := NewNode(HTMLSpan) |
|
794 anchorClose.Literal = anchorStr[offsetFromAnchor:] |
|
795 return len(anchorStr) - offsetFromAnchor, anchorClose |
|
796 } |
|
797 |
|
798 // scan backward for a word boundary |
|
799 rewind := 0 |
|
800 for offset-rewind > 0 && rewind <= 7 && isletter(data[offset-rewind-1]) { |
|
801 rewind++ |
|
802 } |
|
803 if rewind > 6 { // longest supported protocol is "mailto" which has 6 letters |
|
804 return 0, nil |
|
805 } |
|
806 |
|
807 origData := data |
|
808 data = data[offset-rewind:] |
|
809 |
|
810 if !isSafeLink(data) { |
|
811 return 0, nil |
|
812 } |
|
813 |
|
814 linkEnd := 0 |
|
815 for linkEnd < len(data) && !isEndOfLink(data[linkEnd]) { |
|
816 linkEnd++ |
|
817 } |
|
818 |
|
819 // Skip punctuation at the end of the link |
|
820 if (data[linkEnd-1] == '.' || data[linkEnd-1] == ',') && data[linkEnd-2] != '\\' { |
|
821 linkEnd-- |
|
822 } |
|
823 |
|
824 // But don't skip semicolon if it's a part of escaped entity: |
|
825 if data[linkEnd-1] == ';' && data[linkEnd-2] != '\\' && !linkEndsWithEntity(data, linkEnd) { |
|
826 linkEnd-- |
|
827 } |
|
828 |
|
829 // See if the link finishes with a punctuation sign that can be closed. |
|
830 var copen byte |
|
831 switch data[linkEnd-1] { |
|
832 case '"': |
|
833 copen = '"' |
|
834 case '\'': |
|
835 copen = '\'' |
|
836 case ')': |
|
837 copen = '(' |
|
838 case ']': |
|
839 copen = '[' |
|
840 case '}': |
|
841 copen = '{' |
|
842 default: |
|
843 copen = 0 |
|
844 } |
|
845 |
|
846 if copen != 0 { |
|
847 bufEnd := offset - rewind + linkEnd - 2 |
|
848 |
|
849 openDelim := 1 |
|
850 |
|
851 /* Try to close the final punctuation sign in this same line; |
|
852 * if we managed to close it outside of the URL, that means that it's |
|
853 * not part of the URL. If it closes inside the URL, that means it |
|
854 * is part of the URL. |
|
855 * |
|
856 * Examples: |
|
857 * |
|
858 * foo http://www.pokemon.com/Pikachu_(Electric) bar |
|
859 * => http://www.pokemon.com/Pikachu_(Electric) |
|
860 * |
|
861 * foo (http://www.pokemon.com/Pikachu_(Electric)) bar |
|
862 * => http://www.pokemon.com/Pikachu_(Electric) |
|
863 * |
|
864 * foo http://www.pokemon.com/Pikachu_(Electric)) bar |
|
865 * => http://www.pokemon.com/Pikachu_(Electric)) |
|
866 * |
|
867 * (foo http://www.pokemon.com/Pikachu_(Electric)) bar |
|
868 * => foo http://www.pokemon.com/Pikachu_(Electric) |
|
869 */ |
|
870 |
|
871 for bufEnd >= 0 && origData[bufEnd] != '\n' && openDelim != 0 { |
|
872 if origData[bufEnd] == data[linkEnd-1] { |
|
873 openDelim++ |
|
874 } |
|
875 |
|
876 if origData[bufEnd] == copen { |
|
877 openDelim-- |
|
878 } |
|
879 |
|
880 bufEnd-- |
|
881 } |
|
882 |
|
883 if openDelim == 0 { |
|
884 linkEnd-- |
|
885 } |
|
886 } |
|
887 |
|
888 var uLink bytes.Buffer |
|
889 unescapeText(&uLink, data[:linkEnd]) |
|
890 |
|
891 if uLink.Len() > 0 { |
|
892 node := NewNode(Link) |
|
893 node.Destination = uLink.Bytes() |
|
894 node.AppendChild(text(uLink.Bytes())) |
|
895 return linkEnd, node |
|
896 } |
|
897 |
|
898 return linkEnd, nil |
|
899 } |
|
900 |
|
901 func isEndOfLink(char byte) bool { |
|
902 return isspace(char) || char == '<' |
|
903 } |
|
904 |
|
905 var validUris = [][]byte{[]byte("http://"), []byte("https://"), []byte("ftp://"), []byte("mailto://")} |
|
906 var validPaths = [][]byte{[]byte("/"), []byte("./"), []byte("../")} |
|
907 |
|
908 func isSafeLink(link []byte) bool { |
|
909 for _, path := range validPaths { |
|
910 if len(link) >= len(path) && bytes.Equal(link[:len(path)], path) { |
|
911 if len(link) == len(path) { |
|
912 return true |
|
913 } else if isalnum(link[len(path)]) { |
|
914 return true |
|
915 } |
|
916 } |
|
917 } |
|
918 |
|
919 for _, prefix := range validUris { |
|
920 // TODO: handle unicode here |
|
921 // case-insensitive prefix test |
|
922 if len(link) > len(prefix) && bytes.Equal(bytes.ToLower(link[:len(prefix)]), prefix) && isalnum(link[len(prefix)]) { |
|
923 return true |
|
924 } |
|
925 } |
|
926 |
|
927 return false |
|
928 } |
|
929 |
|
930 // return the length of the given tag, or 0 is it's not valid |
|
931 func tagLength(data []byte) (autolink autolinkType, end int) { |
|
932 var i, j int |
|
933 |
|
934 // a valid tag can't be shorter than 3 chars |
|
935 if len(data) < 3 { |
|
936 return notAutolink, 0 |
|
937 } |
|
938 |
|
939 // begins with a '<' optionally followed by '/', followed by letter or number |
|
940 if data[0] != '<' { |
|
941 return notAutolink, 0 |
|
942 } |
|
943 if data[1] == '/' { |
|
944 i = 2 |
|
945 } else { |
|
946 i = 1 |
|
947 } |
|
948 |
|
949 if !isalnum(data[i]) { |
|
950 return notAutolink, 0 |
|
951 } |
|
952 |
|
953 // scheme test |
|
954 autolink = notAutolink |
|
955 |
|
956 // try to find the beginning of an URI |
|
957 for i < len(data) && (isalnum(data[i]) || data[i] == '.' || data[i] == '+' || data[i] == '-') { |
|
958 i++ |
|
959 } |
|
960 |
|
961 if i > 1 && i < len(data) && data[i] == '@' { |
|
962 if j = isMailtoAutoLink(data[i:]); j != 0 { |
|
963 return emailAutolink, i + j |
|
964 } |
|
965 } |
|
966 |
|
967 if i > 2 && i < len(data) && data[i] == ':' { |
|
968 autolink = normalAutolink |
|
969 i++ |
|
970 } |
|
971 |
|
972 // complete autolink test: no whitespace or ' or " |
|
973 switch { |
|
974 case i >= len(data): |
|
975 autolink = notAutolink |
|
976 case autolink != notAutolink: |
|
977 j = i |
|
978 |
|
979 for i < len(data) { |
|
980 if data[i] == '\\' { |
|
981 i += 2 |
|
982 } else if data[i] == '>' || data[i] == '\'' || data[i] == '"' || isspace(data[i]) { |
|
983 break |
|
984 } else { |
|
985 i++ |
|
986 } |
|
987 |
|
988 } |
|
989 |
|
990 if i >= len(data) { |
|
991 return autolink, 0 |
|
992 } |
|
993 if i > j && data[i] == '>' { |
|
994 return autolink, i + 1 |
|
995 } |
|
996 |
|
997 // one of the forbidden chars has been found |
|
998 autolink = notAutolink |
|
999 } |
|
1000 i += bytes.IndexByte(data[i:], '>') |
|
1001 if i < 0 { |
|
1002 return autolink, 0 |
|
1003 } |
|
1004 return autolink, i + 1 |
|
1005 } |
|
1006 |
|
1007 // look for the address part of a mail autolink and '>' |
|
1008 // this is less strict than the original markdown e-mail address matching |
|
1009 func isMailtoAutoLink(data []byte) int { |
|
1010 nb := 0 |
|
1011 |
|
1012 // address is assumed to be: [-@._a-zA-Z0-9]+ with exactly one '@' |
|
1013 for i := 0; i < len(data); i++ { |
|
1014 if isalnum(data[i]) { |
|
1015 continue |
|
1016 } |
|
1017 |
|
1018 switch data[i] { |
|
1019 case '@': |
|
1020 nb++ |
|
1021 |
|
1022 case '-', '.', '_': |
|
1023 break |
|
1024 |
|
1025 case '>': |
|
1026 if nb == 1 { |
|
1027 return i + 1 |
|
1028 } |
|
1029 return 0 |
|
1030 default: |
|
1031 return 0 |
|
1032 } |
|
1033 } |
|
1034 |
|
1035 return 0 |
|
1036 } |
|
1037 |
|
1038 // look for the next emph char, skipping other constructs |
|
1039 func helperFindEmphChar(data []byte, c byte) int { |
|
1040 i := 0 |
|
1041 |
|
1042 for i < len(data) { |
|
1043 for i < len(data) && data[i] != c && data[i] != '`' && data[i] != '[' { |
|
1044 i++ |
|
1045 } |
|
1046 if i >= len(data) { |
|
1047 return 0 |
|
1048 } |
|
1049 // do not count escaped chars |
|
1050 if i != 0 && data[i-1] == '\\' { |
|
1051 i++ |
|
1052 continue |
|
1053 } |
|
1054 if data[i] == c { |
|
1055 return i |
|
1056 } |
|
1057 |
|
1058 if data[i] == '`' { |
|
1059 // skip a code span |
|
1060 tmpI := 0 |
|
1061 i++ |
|
1062 for i < len(data) && data[i] != '`' { |
|
1063 if tmpI == 0 && data[i] == c { |
|
1064 tmpI = i |
|
1065 } |
|
1066 i++ |
|
1067 } |
|
1068 if i >= len(data) { |
|
1069 return tmpI |
|
1070 } |
|
1071 i++ |
|
1072 } else if data[i] == '[' { |
|
1073 // skip a link |
|
1074 tmpI := 0 |
|
1075 i++ |
|
1076 for i < len(data) && data[i] != ']' { |
|
1077 if tmpI == 0 && data[i] == c { |
|
1078 tmpI = i |
|
1079 } |
|
1080 i++ |
|
1081 } |
|
1082 i++ |
|
1083 for i < len(data) && (data[i] == ' ' || data[i] == '\n') { |
|
1084 i++ |
|
1085 } |
|
1086 if i >= len(data) { |
|
1087 return tmpI |
|
1088 } |
|
1089 if data[i] != '[' && data[i] != '(' { // not a link |
|
1090 if tmpI > 0 { |
|
1091 return tmpI |
|
1092 } |
|
1093 continue |
|
1094 } |
|
1095 cc := data[i] |
|
1096 i++ |
|
1097 for i < len(data) && data[i] != cc { |
|
1098 if tmpI == 0 && data[i] == c { |
|
1099 return i |
|
1100 } |
|
1101 i++ |
|
1102 } |
|
1103 if i >= len(data) { |
|
1104 return tmpI |
|
1105 } |
|
1106 i++ |
|
1107 } |
|
1108 } |
|
1109 return 0 |
|
1110 } |
|
1111 |
|
1112 func helperEmphasis(p *Markdown, data []byte, c byte) (int, *Node) { |
|
1113 i := 0 |
|
1114 |
|
1115 // skip one symbol if coming from emph3 |
|
1116 if len(data) > 1 && data[0] == c && data[1] == c { |
|
1117 i = 1 |
|
1118 } |
|
1119 |
|
1120 for i < len(data) { |
|
1121 length := helperFindEmphChar(data[i:], c) |
|
1122 if length == 0 { |
|
1123 return 0, nil |
|
1124 } |
|
1125 i += length |
|
1126 if i >= len(data) { |
|
1127 return 0, nil |
|
1128 } |
|
1129 |
|
1130 if i+1 < len(data) && data[i+1] == c { |
|
1131 i++ |
|
1132 continue |
|
1133 } |
|
1134 |
|
1135 if data[i] == c && !isspace(data[i-1]) { |
|
1136 |
|
1137 if p.extensions&NoIntraEmphasis != 0 { |
|
1138 if !(i+1 == len(data) || isspace(data[i+1]) || ispunct(data[i+1])) { |
|
1139 continue |
|
1140 } |
|
1141 } |
|
1142 |
|
1143 emph := NewNode(Emph) |
|
1144 p.inline(emph, data[:i]) |
|
1145 return i + 1, emph |
|
1146 } |
|
1147 } |
|
1148 |
|
1149 return 0, nil |
|
1150 } |
|
1151 |
|
1152 func helperDoubleEmphasis(p *Markdown, data []byte, c byte) (int, *Node) { |
|
1153 i := 0 |
|
1154 |
|
1155 for i < len(data) { |
|
1156 length := helperFindEmphChar(data[i:], c) |
|
1157 if length == 0 { |
|
1158 return 0, nil |
|
1159 } |
|
1160 i += length |
|
1161 |
|
1162 if i+1 < len(data) && data[i] == c && data[i+1] == c && i > 0 && !isspace(data[i-1]) { |
|
1163 nodeType := Strong |
|
1164 if c == '~' { |
|
1165 nodeType = Del |
|
1166 } |
|
1167 node := NewNode(nodeType) |
|
1168 p.inline(node, data[:i]) |
|
1169 return i + 2, node |
|
1170 } |
|
1171 i++ |
|
1172 } |
|
1173 return 0, nil |
|
1174 } |
|
1175 |
|
1176 func helperTripleEmphasis(p *Markdown, data []byte, offset int, c byte) (int, *Node) { |
|
1177 i := 0 |
|
1178 origData := data |
|
1179 data = data[offset:] |
|
1180 |
|
1181 for i < len(data) { |
|
1182 length := helperFindEmphChar(data[i:], c) |
|
1183 if length == 0 { |
|
1184 return 0, nil |
|
1185 } |
|
1186 i += length |
|
1187 |
|
1188 // skip whitespace preceded symbols |
|
1189 if data[i] != c || isspace(data[i-1]) { |
|
1190 continue |
|
1191 } |
|
1192 |
|
1193 switch { |
|
1194 case i+2 < len(data) && data[i+1] == c && data[i+2] == c: |
|
1195 // triple symbol found |
|
1196 strong := NewNode(Strong) |
|
1197 em := NewNode(Emph) |
|
1198 strong.AppendChild(em) |
|
1199 p.inline(em, data[:i]) |
|
1200 return i + 3, strong |
|
1201 case (i+1 < len(data) && data[i+1] == c): |
|
1202 // double symbol found, hand over to emph1 |
|
1203 length, node := helperEmphasis(p, origData[offset-2:], c) |
|
1204 if length == 0 { |
|
1205 return 0, nil |
|
1206 } |
|
1207 return length - 2, node |
|
1208 default: |
|
1209 // single symbol found, hand over to emph2 |
|
1210 length, node := helperDoubleEmphasis(p, origData[offset-1:], c) |
|
1211 if length == 0 { |
|
1212 return 0, nil |
|
1213 } |
|
1214 return length - 1, node |
|
1215 } |
|
1216 } |
|
1217 return 0, nil |
|
1218 } |
|
1219 |
|
1220 func text(s []byte) *Node { |
|
1221 node := NewNode(Text) |
|
1222 node.Literal = s |
|
1223 return node |
|
1224 } |
|
1225 |
|
1226 func normalizeURI(s []byte) []byte { |
|
1227 return s // TODO: implement |
|
1228 } |