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