block.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 block-level elements.
12//
13
14package blackfriday
15
16import (
17 "bytes"
18 "html"
19 "regexp"
20 "strings"
21
22 "github.com/shurcooL/sanitized_anchor_name"
23)
24
25const (
26 charEntity = "&(?:#x[a-f0-9]{1,8}|#[0-9]{1,8}|[a-z][a-z0-9]{1,31});"
27 escapable = "[!\"#$%&'()*+,./:;<=>?@[\\\\\\]^_`{|}~-]"
28)
29
30var (
31 reBackslashOrAmp = regexp.MustCompile("[\\&]")
32 reEntityOrEscapedChar = regexp.MustCompile("(?i)\\\\" + escapable + "|" + charEntity)
33)
34
35// Parse block-level data.
36// Note: this function and many that it calls assume that
37// the input buffer ends with a newline.
38func (p *Markdown) block(data []byte) {
39 // this is called recursively: enforce a maximum depth
40 if p.nesting >= p.maxNesting {
41 return
42 }
43 p.nesting++
44
45 // parse out one block-level construct at a time
46 for len(data) > 0 {
47 // prefixed heading:
48 //
49 // # Heading 1
50 // ## Heading 2
51 // ...
52 // ###### Heading 6
53 if p.isPrefixHeading(data) {
54 data = data[p.prefixHeading(data):]
55 continue
56 }
57
58 // block of preformatted HTML:
59 //
60 // <div>
61 // ...
62 // </div>
63 if data[0] == '<' {
64 if i := p.html(data, true); i > 0 {
65 data = data[i:]
66 continue
67 }
68 }
69
70 // title block
71 //
72 // % stuff
73 // % more stuff
74 // % even more stuff
75 if p.extensions&Titleblock != 0 {
76 if data[0] == '%' {
77 if i := p.titleBlock(data, true); i > 0 {
78 data = data[i:]
79 continue
80 }
81 }
82 }
83
84 // blank lines. note: returns the # of bytes to skip
85 if i := p.isEmpty(data); i > 0 {
86 data = data[i:]
87 continue
88 }
89
90 // indented code block:
91 //
92 // func max(a, b int) int {
93 // if a > b {
94 // return a
95 // }
96 // return b
97 // }
98 if p.codePrefix(data) > 0 {
99 data = data[p.code(data):]
100 continue
101 }
102
103 // fenced code block:
104 //
105 // ``` go
106 // func fact(n int) int {
107 // if n <= 1 {
108 // return n
109 // }
110 // return n * fact(n-1)
111 // }
112 // ```
113 if p.extensions&FencedCode != 0 {
114 if i := p.fencedCodeBlock(data, true); i > 0 {
115 data = data[i:]
116 continue
117 }
118 }
119
120 // horizontal rule:
121 //
122 // ------
123 // or
124 // ******
125 // or
126 // ______
127 if p.isHRule(data) {
128 p.addBlock(HorizontalRule, nil)
129 var i int
130 for i = 0; i < len(data) && data[i] != '\n'; i++ {
131 }
132 data = data[i:]
133 continue
134 }
135
136 // block quote:
137 //
138 // > A big quote I found somewhere
139 // > on the web
140 if p.quotePrefix(data) > 0 {
141 data = data[p.quote(data):]
142 continue
143 }
144
145 // table:
146 //
147 // Name | Age | Phone
148 // ------|-----|---------
149 // Bob | 31 | 555-1234
150 // Alice | 27 | 555-4321
151 if p.extensions&Tables != 0 {
152 if i := p.table(data); i > 0 {
153 data = data[i:]
154 continue
155 }
156 }
157
158 // an itemized/unordered list:
159 //
160 // * Item 1
161 // * Item 2
162 //
163 // also works with + or -
164 if p.uliPrefix(data) > 0 {
165 data = data[p.list(data, 0):]
166 continue
167 }
168
169 // a numbered/ordered list:
170 //
171 // 1. Item 1
172 // 2. Item 2
173 if p.oliPrefix(data) > 0 {
174 data = data[p.list(data, ListTypeOrdered):]
175 continue
176 }
177
178 // definition lists:
179 //
180 // Term 1
181 // : Definition a
182 // : Definition b
183 //
184 // Term 2
185 // : Definition c
186 if p.extensions&DefinitionLists != 0 {
187 if p.dliPrefix(data) > 0 {
188 data = data[p.list(data, ListTypeDefinition):]
189 continue
190 }
191 }
192
193 // anything else must look like a normal paragraph
194 // note: this finds underlined headings, too
195 data = data[p.paragraph(data):]
196 }
197
198 p.nesting--
199}
200
201func (p *Markdown) addBlock(typ NodeType, content []byte) *Node {
202 p.closeUnmatchedBlocks()
203 container := p.addChild(typ, 0)
204 container.content = content
205 return container
206}
207
208func (p *Markdown) isPrefixHeading(data []byte) bool {
209 if data[0] != '#' {
210 return false
211 }
212
213 if p.extensions&SpaceHeadings != 0 {
214 level := 0
215 for level < 6 && level < len(data) && data[level] == '#' {
216 level++
217 }
218 if level == len(data) || data[level] != ' ' {
219 return false
220 }
221 }
222 return true
223}
224
225func (p *Markdown) prefixHeading(data []byte) int {
226 level := 0
227 for level < 6 && level < len(data) && data[level] == '#' {
228 level++
229 }
230 i := skipChar(data, level, ' ')
231 end := skipUntilChar(data, i, '\n')
232 skip := end
233 id := ""
234 if p.extensions&HeadingIDs != 0 {
235 j, k := 0, 0
236 // find start/end of heading id
237 for j = i; j < end-1 && (data[j] != '{' || data[j+1] != '#'); j++ {
238 }
239 for k = j + 1; k < end && data[k] != '}'; k++ {
240 }
241 // extract heading id iff found
242 if j < end && k < end {
243 id = string(data[j+2 : k])
244 end = j
245 skip = k + 1
246 for end > 0 && data[end-1] == ' ' {
247 end--
248 }
249 }
250 }
251 for end > 0 && data[end-1] == '#' {
252 if isBackslashEscaped(data, end-1) {
253 break
254 }
255 end--
256 }
257 for end > 0 && data[end-1] == ' ' {
258 end--
259 }
260 if end > i {
261 if id == "" && p.extensions&AutoHeadingIDs != 0 {
262 id = sanitized_anchor_name.Create(string(data[i:end]))
263 }
264 block := p.addBlock(Heading, data[i:end])
265 block.HeadingID = id
266 block.Level = level
267 }
268 return skip
269}
270
271func (p *Markdown) isUnderlinedHeading(data []byte) int {
272 // test of level 1 heading
273 if data[0] == '=' {
274 i := skipChar(data, 1, '=')
275 i = skipChar(data, i, ' ')
276 if i < len(data) && data[i] == '\n' {
277 return 1
278 }
279 return 0
280 }
281
282 // test of level 2 heading
283 if data[0] == '-' {
284 i := skipChar(data, 1, '-')
285 i = skipChar(data, i, ' ')
286 if i < len(data) && data[i] == '\n' {
287 return 2
288 }
289 return 0
290 }
291
292 return 0
293}
294
295func (p *Markdown) titleBlock(data []byte, doRender bool) int {
296 if data[0] != '%' {
297 return 0
298 }
299 splitData := bytes.Split(data, []byte("\n"))
300 var i int
301 for idx, b := range splitData {
302 if !bytes.HasPrefix(b, []byte("%")) {
303 i = idx // - 1
304 break
305 }
306 }
307
308 data = bytes.Join(splitData[0:i], []byte("\n"))
309 consumed := len(data)
310 data = bytes.TrimPrefix(data, []byte("% "))
311 data = bytes.Replace(data, []byte("\n% "), []byte("\n"), -1)
312 block := p.addBlock(Heading, data)
313 block.Level = 1
314 block.IsTitleblock = true
315
316 return consumed
317}
318
319func (p *Markdown) html(data []byte, doRender bool) int {
320 var i, j int
321
322 // identify the opening tag
323 if data[0] != '<' {
324 return 0
325 }
326 curtag, tagfound := p.htmlFindTag(data[1:])
327
328 // handle special cases
329 if !tagfound {
330 // check for an HTML comment
331 if size := p.htmlComment(data, doRender); size > 0 {
332 return size
333 }
334
335 // check for an <hr> tag
336 if size := p.htmlHr(data, doRender); size > 0 {
337 return size
338 }
339
340 // no special case recognized
341 return 0
342 }
343
344 // look for an unindented matching closing tag
345 // followed by a blank line
346 found := false
347 /*
348 closetag := []byte("\n</" + curtag + ">")
349 j = len(curtag) + 1
350 for !found {
351 // scan for a closing tag at the beginning of a line
352 if skip := bytes.Index(data[j:], closetag); skip >= 0 {
353 j += skip + len(closetag)
354 } else {
355 break
356 }
357
358 // see if it is the only thing on the line
359 if skip := p.isEmpty(data[j:]); skip > 0 {
360 // see if it is followed by a blank line/eof
361 j += skip
362 if j >= len(data) {
363 found = true
364 i = j
365 } else {
366 if skip := p.isEmpty(data[j:]); skip > 0 {
367 j += skip
368 found = true
369 i = j
370 }
371 }
372 }
373 }
374 */
375
376 // if not found, try a second pass looking for indented match
377 // but not if tag is "ins" or "del" (following original Markdown.pl)
378 if !found && curtag != "ins" && curtag != "del" {
379 i = 1
380 for i < len(data) {
381 i++
382 for i < len(data) && !(data[i-1] == '<' && data[i] == '/') {
383 i++
384 }
385
386 if i+2+len(curtag) >= len(data) {
387 break
388 }
389
390 j = p.htmlFindEnd(curtag, data[i-1:])
391
392 if j > 0 {
393 i += j - 1
394 found = true
395 break
396 }
397 }
398 }
399
400 if !found {
401 return 0
402 }
403
404 // the end of the block has been found
405 if doRender {
406 // trim newlines
407 end := i
408 for end > 0 && data[end-1] == '\n' {
409 end--
410 }
411 finalizeHTMLBlock(p.addBlock(HTMLBlock, data[:end]))
412 }
413
414 return i
415}
416
417func finalizeHTMLBlock(block *Node) {
418 block.Literal = block.content
419 block.content = nil
420}
421
422// HTML comment, lax form
423func (p *Markdown) htmlComment(data []byte, doRender bool) int {
424 i := p.inlineHTMLComment(data)
425 // needs to end with a blank line
426 if j := p.isEmpty(data[i:]); j > 0 {
427 size := i + j
428 if doRender {
429 // trim trailing newlines
430 end := size
431 for end > 0 && data[end-1] == '\n' {
432 end--
433 }
434 block := p.addBlock(HTMLBlock, data[:end])
435 finalizeHTMLBlock(block)
436 }
437 return size
438 }
439 return 0
440}
441
442// HR, which is the only self-closing block tag considered
443func (p *Markdown) htmlHr(data []byte, doRender bool) int {
444 if len(data) < 4 {
445 return 0
446 }
447 if data[0] != '<' || (data[1] != 'h' && data[1] != 'H') || (data[2] != 'r' && data[2] != 'R') {
448 return 0
449 }
450 if data[3] != ' ' && data[3] != '/' && data[3] != '>' {
451 // not an <hr> tag after all; at least not a valid one
452 return 0
453 }
454 i := 3
455 for i < len(data) && data[i] != '>' && data[i] != '\n' {
456 i++
457 }
458 if i < len(data) && data[i] == '>' {
459 i++
460 if j := p.isEmpty(data[i:]); j > 0 {
461 size := i + j
462 if doRender {
463 // trim newlines
464 end := size
465 for end > 0 && data[end-1] == '\n' {
466 end--
467 }
468 finalizeHTMLBlock(p.addBlock(HTMLBlock, data[:end]))
469 }
470 return size
471 }
472 }
473 return 0
474}
475
476func (p *Markdown) htmlFindTag(data []byte) (string, bool) {
477 i := 0
478 for i < len(data) && isalnum(data[i]) {
479 i++
480 }
481 key := string(data[:i])
482 if _, ok := blockTags[key]; ok {
483 return key, true
484 }
485 return "", false
486}
487
488func (p *Markdown) htmlFindEnd(tag string, data []byte) int {
489 // assume data[0] == '<' && data[1] == '/' already tested
490 if tag == "hr" {
491 return 2
492 }
493 // check if tag is a match
494 closetag := []byte("</" + tag + ">")
495 if !bytes.HasPrefix(data, closetag) {
496 return 0
497 }
498 i := len(closetag)
499
500 // check that the rest of the line is blank
501 skip := 0
502 if skip = p.isEmpty(data[i:]); skip == 0 {
503 return 0
504 }
505 i += skip
506 skip = 0
507
508 if i >= len(data) {
509 return i
510 }
511
512 if p.extensions&LaxHTMLBlocks != 0 {
513 return i
514 }
515 if skip = p.isEmpty(data[i:]); skip == 0 {
516 // following line must be blank
517 return 0
518 }
519
520 return i + skip
521}
522
523func (*Markdown) isEmpty(data []byte) int {
524 // it is okay to call isEmpty on an empty buffer
525 if len(data) == 0 {
526 return 0
527 }
528
529 var i int
530 for i = 0; i < len(data) && data[i] != '\n'; i++ {
531 if data[i] != ' ' && data[i] != '\t' {
532 return 0
533 }
534 }
535 if i < len(data) && data[i] == '\n' {
536 i++
537 }
538 return i
539}
540
541func (*Markdown) isHRule(data []byte) bool {
542 i := 0
543
544 // skip up to three spaces
545 for i < 3 && data[i] == ' ' {
546 i++
547 }
548
549 // look at the hrule char
550 if data[i] != '*' && data[i] != '-' && data[i] != '_' {
551 return false
552 }
553 c := data[i]
554
555 // the whole line must be the char or whitespace
556 n := 0
557 for i < len(data) && data[i] != '\n' {
558 switch {
559 case data[i] == c:
560 n++
561 case data[i] != ' ':
562 return false
563 }
564 i++
565 }
566
567 return n >= 3
568}
569
570// isFenceLine checks if there's a fence line (e.g., ``` or ``` go) at the beginning of data,
571// and returns the end index if so, or 0 otherwise. It also returns the marker found.
572// If info is not nil, it gets set to the syntax specified in the fence line.
573func isFenceLine(data []byte, info *string, oldmarker string) (end int, marker string) {
574 i, size := 0, 0
575
576 // skip up to three spaces
577 for i < len(data) && i < 3 && data[i] == ' ' {
578 i++
579 }
580
581 // check for the marker characters: ~ or `
582 if i >= len(data) {
583 return 0, ""
584 }
585 if data[i] != '~' && data[i] != '`' {
586 return 0, ""
587 }
588
589 c := data[i]
590
591 // the whole line must be the same char or whitespace
592 for i < len(data) && data[i] == c {
593 size++
594 i++
595 }
596
597 // the marker char must occur at least 3 times
598 if size < 3 {
599 return 0, ""
600 }
601 marker = string(data[i-size : i])
602
603 // if this is the end marker, it must match the beginning marker
604 if oldmarker != "" && marker != oldmarker {
605 return 0, ""
606 }
607
608 // TODO(shurcooL): It's probably a good idea to simplify the 2 code paths here
609 // into one, always get the info string, and discard it if the caller doesn't care.
610 if info != nil {
611 infoLength := 0
612 i = skipChar(data, i, ' ')
613
614 if i >= len(data) {
615 if i == len(data) {
616 return i, marker
617 }
618 return 0, ""
619 }
620
621 infoStart := i
622
623 if data[i] == '{' {
624 i++
625 infoStart++
626
627 for i < len(data) && data[i] != '}' && data[i] != '\n' {
628 infoLength++
629 i++
630 }
631
632 if i >= len(data) || data[i] != '}' {
633 return 0, ""
634 }
635
636 // strip all whitespace at the beginning and the end
637 // of the {} block
638 for infoLength > 0 && isspace(data[infoStart]) {
639 infoStart++
640 infoLength--
641 }
642
643 for infoLength > 0 && isspace(data[infoStart+infoLength-1]) {
644 infoLength--
645 }
646 i++
647 i = skipChar(data, i, ' ')
648 } else {
649 for i < len(data) && !isverticalspace(data[i]) {
650 infoLength++
651 i++
652 }
653 }
654
655 *info = strings.TrimSpace(string(data[infoStart : infoStart+infoLength]))
656 }
657
658 if i == len(data) {
659 return i, marker
660 }
661 if i > len(data) || data[i] != '\n' {
662 return 0, ""
663 }
664 return i + 1, marker // Take newline into account.
665}
666
667// fencedCodeBlock returns the end index if data contains a fenced code block at the beginning,
668// or 0 otherwise. It writes to out if doRender is true, otherwise it has no side effects.
669// If doRender is true, a final newline is mandatory to recognize the fenced code block.
670func (p *Markdown) fencedCodeBlock(data []byte, doRender bool) int {
671 var info string
672 beg, marker := isFenceLine(data, &info, "")
673 if beg == 0 || beg >= len(data) {
674 return 0
675 }
676
677 var work bytes.Buffer
678 work.Write([]byte(info))
679 work.WriteByte('\n')
680
681 for {
682 // safe to assume beg < len(data)
683
684 // check for the end of the code block
685 fenceEnd, _ := isFenceLine(data[beg:], nil, marker)
686 if fenceEnd != 0 {
687 beg += fenceEnd
688 break
689 }
690
691 // copy the current line
692 end := skipUntilChar(data, beg, '\n') + 1
693
694 // did we reach the end of the buffer without a closing marker?
695 if end >= len(data) {
696 return 0
697 }
698
699 // verbatim copy to the working buffer
700 if doRender {
701 work.Write(data[beg:end])
702 }
703 beg = end
704 }
705
706 if doRender {
707 block := p.addBlock(CodeBlock, work.Bytes()) // TODO: get rid of temp buffer
708 block.IsFenced = true
709 finalizeCodeBlock(block)
710 }
711
712 return beg
713}
714
715func unescapeChar(str []byte) []byte {
716 if str[0] == '\\' {
717 return []byte{str[1]}
718 }
719 return []byte(html.UnescapeString(string(str)))
720}
721
722func unescapeString(str []byte) []byte {
723 if reBackslashOrAmp.Match(str) {
724 return reEntityOrEscapedChar.ReplaceAllFunc(str, unescapeChar)
725 }
726 return str
727}
728
729func finalizeCodeBlock(block *Node) {
730 if block.IsFenced {
731 newlinePos := bytes.IndexByte(block.content, '\n')
732 firstLine := block.content[:newlinePos]
733 rest := block.content[newlinePos+1:]
734 block.Info = unescapeString(bytes.Trim(firstLine, "\n"))
735 block.Literal = rest
736 } else {
737 block.Literal = block.content
738 }
739 block.content = nil
740}
741
742func (p *Markdown) table(data []byte) int {
743 table := p.addBlock(Table, nil)
744 i, columns := p.tableHeader(data)
745 if i == 0 {
746 p.tip = table.Parent
747 table.Unlink()
748 return 0
749 }
750
751 p.addBlock(TableBody, nil)
752
753 for i < len(data) {
754 pipes, rowStart := 0, i
755 for ; i < len(data) && data[i] != '\n'; i++ {
756 if data[i] == '|' {
757 pipes++
758 }
759 }
760
761 if pipes == 0 {
762 i = rowStart
763 break
764 }
765
766 // include the newline in data sent to tableRow
767 if i < len(data) && data[i] == '\n' {
768 i++
769 }
770 p.tableRow(data[rowStart:i], columns, false)
771 }
772
773 return i
774}
775
776// check if the specified position is preceded by an odd number of backslashes
777func isBackslashEscaped(data []byte, i int) bool {
778 backslashes := 0
779 for i-backslashes-1 >= 0 && data[i-backslashes-1] == '\\' {
780 backslashes++
781 }
782 return backslashes&1 == 1
783}
784
785func (p *Markdown) tableHeader(data []byte) (size int, columns []CellAlignFlags) {
786 i := 0
787 colCount := 1
788 for i = 0; i < len(data) && data[i] != '\n'; i++ {
789 if data[i] == '|' && !isBackslashEscaped(data, i) {
790 colCount++
791 }
792 }
793
794 // doesn't look like a table header
795 if colCount == 1 {
796 return
797 }
798
799 // include the newline in the data sent to tableRow
800 j := i
801 if j < len(data) && data[j] == '\n' {
802 j++
803 }
804 header := data[:j]
805
806 // column count ignores pipes at beginning or end of line
807 if data[0] == '|' {
808 colCount--
809 }
810 if i > 2 && data[i-1] == '|' && !isBackslashEscaped(data, i-1) {
811 colCount--
812 }
813
814 columns = make([]CellAlignFlags, colCount)
815
816 // move on to the header underline
817 i++
818 if i >= len(data) {
819 return
820 }
821
822 if data[i] == '|' && !isBackslashEscaped(data, i) {
823 i++
824 }
825 i = skipChar(data, i, ' ')
826
827 // each column header is of form: / *:?-+:? *|/ with # dashes + # colons >= 3
828 // and trailing | optional on last column
829 col := 0
830 for i < len(data) && data[i] != '\n' {
831 dashes := 0
832
833 if data[i] == ':' {
834 i++
835 columns[col] |= TableAlignmentLeft
836 dashes++
837 }
838 for i < len(data) && data[i] == '-' {
839 i++
840 dashes++
841 }
842 if i < len(data) && data[i] == ':' {
843 i++
844 columns[col] |= TableAlignmentRight
845 dashes++
846 }
847 for i < len(data) && data[i] == ' ' {
848 i++
849 }
850 if i == len(data) {
851 return
852 }
853 // end of column test is messy
854 switch {
855 case dashes < 3:
856 // not a valid column
857 return
858
859 case data[i] == '|' && !isBackslashEscaped(data, i):
860 // marker found, now skip past trailing whitespace
861 col++
862 i++
863 for i < len(data) && data[i] == ' ' {
864 i++
865 }
866
867 // trailing junk found after last column
868 if col >= colCount && i < len(data) && data[i] != '\n' {
869 return
870 }
871
872 case (data[i] != '|' || isBackslashEscaped(data, i)) && col+1 < colCount:
873 // something else found where marker was required
874 return
875
876 case data[i] == '\n':
877 // marker is optional for the last column
878 col++
879
880 default:
881 // trailing junk found after last column
882 return
883 }
884 }
885 if col != colCount {
886 return
887 }
888
889 p.addBlock(TableHead, nil)
890 p.tableRow(header, columns, true)
891 size = i
892 if size < len(data) && data[size] == '\n' {
893 size++
894 }
895 return
896}
897
898func (p *Markdown) tableRow(data []byte, columns []CellAlignFlags, header bool) {
899 p.addBlock(TableRow, nil)
900 i, col := 0, 0
901
902 if data[i] == '|' && !isBackslashEscaped(data, i) {
903 i++
904 }
905
906 for col = 0; col < len(columns) && i < len(data); col++ {
907 for i < len(data) && data[i] == ' ' {
908 i++
909 }
910
911 cellStart := i
912
913 for i < len(data) && (data[i] != '|' || isBackslashEscaped(data, i)) && data[i] != '\n' {
914 i++
915 }
916
917 cellEnd := i
918
919 // skip the end-of-cell marker, possibly taking us past end of buffer
920 i++
921
922 for cellEnd > cellStart && cellEnd-1 < len(data) && data[cellEnd-1] == ' ' {
923 cellEnd--
924 }
925
926 cell := p.addBlock(TableCell, data[cellStart:cellEnd])
927 cell.IsHeader = header
928 cell.Align = columns[col]
929 }
930
931 // pad it out with empty columns to get the right number
932 for ; col < len(columns); col++ {
933 cell := p.addBlock(TableCell, nil)
934 cell.IsHeader = header
935 cell.Align = columns[col]
936 }
937
938 // silently ignore rows with too many cells
939}
940
941// returns blockquote prefix length
942func (p *Markdown) quotePrefix(data []byte) int {
943 i := 0
944 for i < 3 && i < len(data) && data[i] == ' ' {
945 i++
946 }
947 if i < len(data) && data[i] == '>' {
948 if i+1 < len(data) && data[i+1] == ' ' {
949 return i + 2
950 }
951 return i + 1
952 }
953 return 0
954}
955
956// blockquote ends with at least one blank line
957// followed by something without a blockquote prefix
958func (p *Markdown) terminateBlockquote(data []byte, beg, end int) bool {
959 if p.isEmpty(data[beg:]) <= 0 {
960 return false
961 }
962 if end >= len(data) {
963 return true
964 }
965 return p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0
966}
967
968// parse a blockquote fragment
969func (p *Markdown) quote(data []byte) int {
970 block := p.addBlock(BlockQuote, nil)
971 var raw bytes.Buffer
972 beg, end := 0, 0
973 for beg < len(data) {
974 end = beg
975 // Step over whole lines, collecting them. While doing that, check for
976 // fenced code and if one's found, incorporate it altogether,
977 // irregardless of any contents inside it
978 for end < len(data) && data[end] != '\n' {
979 if p.extensions&FencedCode != 0 {
980 if i := p.fencedCodeBlock(data[end:], false); i > 0 {
981 // -1 to compensate for the extra end++ after the loop:
982 end += i - 1
983 break
984 }
985 }
986 end++
987 }
988 if end < len(data) && data[end] == '\n' {
989 end++
990 }
991 if pre := p.quotePrefix(data[beg:]); pre > 0 {
992 // skip the prefix
993 beg += pre
994 } else if p.terminateBlockquote(data, beg, end) {
995 break
996 }
997 // this line is part of the blockquote
998 raw.Write(data[beg:end])
999 beg = end
1000 }
1001 p.block(raw.Bytes())
1002 p.finalize(block)
1003 return end
1004}
1005
1006// returns prefix length for block code
1007func (p *Markdown) codePrefix(data []byte) int {
1008 if len(data) >= 1 && data[0] == '\t' {
1009 return 1
1010 }
1011 if len(data) >= 4 && data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
1012 return 4
1013 }
1014 return 0
1015}
1016
1017func (p *Markdown) code(data []byte) int {
1018 var work bytes.Buffer
1019
1020 i := 0
1021 for i < len(data) {
1022 beg := i
1023 for i < len(data) && data[i] != '\n' {
1024 i++
1025 }
1026 if i < len(data) && data[i] == '\n' {
1027 i++
1028 }
1029
1030 blankline := p.isEmpty(data[beg:i]) > 0
1031 if pre := p.codePrefix(data[beg:i]); pre > 0 {
1032 beg += pre
1033 } else if !blankline {
1034 // non-empty, non-prefixed line breaks the pre
1035 i = beg
1036 break
1037 }
1038
1039 // verbatim copy to the working buffer
1040 if blankline {
1041 work.WriteByte('\n')
1042 } else {
1043 work.Write(data[beg:i])
1044 }
1045 }
1046
1047 // trim all the \n off the end of work
1048 workbytes := work.Bytes()
1049 eol := len(workbytes)
1050 for eol > 0 && workbytes[eol-1] == '\n' {
1051 eol--
1052 }
1053 if eol != len(workbytes) {
1054 work.Truncate(eol)
1055 }
1056
1057 work.WriteByte('\n')
1058
1059 block := p.addBlock(CodeBlock, work.Bytes()) // TODO: get rid of temp buffer
1060 block.IsFenced = false
1061 finalizeCodeBlock(block)
1062
1063 return i
1064}
1065
1066// returns unordered list item prefix
1067func (p *Markdown) uliPrefix(data []byte) int {
1068 i := 0
1069 // start with up to 3 spaces
1070 for i < len(data) && i < 3 && data[i] == ' ' {
1071 i++
1072 }
1073 if i >= len(data)-1 {
1074 return 0
1075 }
1076 // need one of {'*', '+', '-'} followed by a space or a tab
1077 if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
1078 (data[i+1] != ' ' && data[i+1] != '\t') {
1079 return 0
1080 }
1081 return i + 2
1082}
1083
1084// returns ordered list item prefix
1085func (p *Markdown) oliPrefix(data []byte) int {
1086 i := 0
1087
1088 // start with up to 3 spaces
1089 for i < 3 && i < len(data) && data[i] == ' ' {
1090 i++
1091 }
1092
1093 // count the digits
1094 start := i
1095 for i < len(data) && data[i] >= '0' && data[i] <= '9' {
1096 i++
1097 }
1098 if start == i || i >= len(data)-1 {
1099 return 0
1100 }
1101
1102 // we need >= 1 digits followed by a dot and a space or a tab
1103 if data[i] != '.' || !(data[i+1] == ' ' || data[i+1] == '\t') {
1104 return 0
1105 }
1106 return i + 2
1107}
1108
1109// returns definition list item prefix
1110func (p *Markdown) dliPrefix(data []byte) int {
1111 if len(data) < 2 {
1112 return 0
1113 }
1114 i := 0
1115 // need a ':' followed by a space or a tab
1116 if data[i] != ':' || !(data[i+1] == ' ' || data[i+1] == '\t') {
1117 return 0
1118 }
1119 for i < len(data) && data[i] == ' ' {
1120 i++
1121 }
1122 return i + 2
1123}
1124
1125// parse ordered or unordered list block
1126func (p *Markdown) list(data []byte, flags ListType) int {
1127 i := 0
1128 flags |= ListItemBeginningOfList
1129 block := p.addBlock(List, nil)
1130 block.ListFlags = flags
1131 block.Tight = true
1132
1133 for i < len(data) {
1134 skip := p.listItem(data[i:], &flags)
1135 if flags&ListItemContainsBlock != 0 {
1136 block.ListData.Tight = false
1137 }
1138 i += skip
1139 if skip == 0 || flags&ListItemEndOfList != 0 {
1140 break
1141 }
1142 flags &= ^ListItemBeginningOfList
1143 }
1144
1145 above := block.Parent
1146 finalizeList(block)
1147 p.tip = above
1148 return i
1149}
1150
1151// Returns true if the list item is not the same type as its parent list
1152func (p *Markdown) listTypeChanged(data []byte, flags *ListType) bool {
1153 if p.dliPrefix(data) > 0 && *flags&ListTypeDefinition == 0 {
1154 return true
1155 } else if p.oliPrefix(data) > 0 && *flags&ListTypeOrdered == 0 {
1156 return true
1157 } else if p.uliPrefix(data) > 0 && (*flags&ListTypeOrdered != 0 || *flags&ListTypeDefinition != 0) {
1158 return true
1159 }
1160 return false
1161}
1162
1163// Returns true if block ends with a blank line, descending if needed
1164// into lists and sublists.
1165func endsWithBlankLine(block *Node) bool {
1166 // TODO: figure this out. Always false now.
1167 for block != nil {
1168 //if block.lastLineBlank {
1169 //return true
1170 //}
1171 t := block.Type
1172 if t == List || t == Item {
1173 block = block.LastChild
1174 } else {
1175 break
1176 }
1177 }
1178 return false
1179}
1180
1181func finalizeList(block *Node) {
1182 block.open = false
1183 item := block.FirstChild
1184 for item != nil {
1185 // check for non-final list item ending with blank line:
1186 if endsWithBlankLine(item) && item.Next != nil {
1187 block.ListData.Tight = false
1188 break
1189 }
1190 // recurse into children of list item, to see if there are spaces
1191 // between any of them:
1192 subItem := item.FirstChild
1193 for subItem != nil {
1194 if endsWithBlankLine(subItem) && (item.Next != nil || subItem.Next != nil) {
1195 block.ListData.Tight = false
1196 break
1197 }
1198 subItem = subItem.Next
1199 }
1200 item = item.Next
1201 }
1202}
1203
1204// Parse a single list item.
1205// Assumes initial prefix is already removed if this is a sublist.
1206func (p *Markdown) listItem(data []byte, flags *ListType) int {
1207 // keep track of the indentation of the first line
1208 itemIndent := 0
1209 if data[0] == '\t' {
1210 itemIndent += 4
1211 } else {
1212 for itemIndent < 3 && data[itemIndent] == ' ' {
1213 itemIndent++
1214 }
1215 }
1216
1217 var bulletChar byte = '*'
1218 i := p.uliPrefix(data)
1219 if i == 0 {
1220 i = p.oliPrefix(data)
1221 } else {
1222 bulletChar = data[i-2]
1223 }
1224 if i == 0 {
1225 i = p.dliPrefix(data)
1226 // reset definition term flag
1227 if i > 0 {
1228 *flags &= ^ListTypeTerm
1229 }
1230 }
1231 if i == 0 {
1232 // if in definition list, set term flag and continue
1233 if *flags&ListTypeDefinition != 0 {
1234 *flags |= ListTypeTerm
1235 } else {
1236 return 0
1237 }
1238 }
1239
1240 // skip leading whitespace on first line
1241 for i < len(data) && data[i] == ' ' {
1242 i++
1243 }
1244
1245 // find the end of the line
1246 line := i
1247 for i > 0 && i < len(data) && data[i-1] != '\n' {
1248 i++
1249 }
1250
1251 // get working buffer
1252 var raw bytes.Buffer
1253
1254 // put the first line into the working buffer
1255 raw.Write(data[line:i])
1256 line = i
1257
1258 // process the following lines
1259 containsBlankLine := false
1260 sublist := 0
1261 codeBlockMarker := ""
1262
1263gatherlines:
1264 for line < len(data) {
1265 i++
1266
1267 // find the end of this line
1268 for i < len(data) && data[i-1] != '\n' {
1269 i++
1270 }
1271
1272 // if it is an empty line, guess that it is part of this item
1273 // and move on to the next line
1274 if p.isEmpty(data[line:i]) > 0 {
1275 containsBlankLine = true
1276 line = i
1277 continue
1278 }
1279
1280 // calculate the indentation
1281 indent := 0
1282 indentIndex := 0
1283 if data[line] == '\t' {
1284 indentIndex++
1285 indent += 4
1286 } else {
1287 for indent < 4 && line+indent < i && data[line+indent] == ' ' {
1288 indent++
1289 indentIndex++
1290 }
1291 }
1292
1293 chunk := data[line+indentIndex : i]
1294
1295 if p.extensions&FencedCode != 0 {
1296 // determine if in or out of codeblock
1297 // if in codeblock, ignore normal list processing
1298 _, marker := isFenceLine(chunk, nil, codeBlockMarker)
1299 if marker != "" {
1300 if codeBlockMarker == "" {
1301 // start of codeblock
1302 codeBlockMarker = marker
1303 } else {
1304 // end of codeblock.
1305 codeBlockMarker = ""
1306 }
1307 }
1308 // we are in a codeblock, write line, and continue
1309 if codeBlockMarker != "" || marker != "" {
1310 raw.Write(data[line+indentIndex : i])
1311 line = i
1312 continue gatherlines
1313 }
1314 }
1315
1316 // evaluate how this line fits in
1317 switch {
1318 // is this a nested list item?
1319 case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) ||
1320 p.oliPrefix(chunk) > 0 ||
1321 p.dliPrefix(chunk) > 0:
1322
1323 // to be a nested list, it must be indented more
1324 // if not, it is either a different kind of list
1325 // or the next item in the same list
1326 if indent <= itemIndent {
1327 if p.listTypeChanged(chunk, flags) {
1328 *flags |= ListItemEndOfList
1329 } else if containsBlankLine {
1330 *flags |= ListItemContainsBlock
1331 }
1332
1333 break gatherlines
1334 }
1335
1336 if containsBlankLine {
1337 *flags |= ListItemContainsBlock
1338 }
1339
1340 // is this the first item in the nested list?
1341 if sublist == 0 {
1342 sublist = raw.Len()
1343 }
1344
1345 // is this a nested prefix heading?
1346 case p.isPrefixHeading(chunk):
1347 // if the heading is not indented, it is not nested in the list
1348 // and thus ends the list
1349 if containsBlankLine && indent < 4 {
1350 *flags |= ListItemEndOfList
1351 break gatherlines
1352 }
1353 *flags |= ListItemContainsBlock
1354
1355 // anything following an empty line is only part
1356 // of this item if it is indented 4 spaces
1357 // (regardless of the indentation of the beginning of the item)
1358 case containsBlankLine && indent < 4:
1359 if *flags&ListTypeDefinition != 0 && i < len(data)-1 {
1360 // is the next item still a part of this list?
1361 next := i
1362 for next < len(data) && data[next] != '\n' {
1363 next++
1364 }
1365 for next < len(data)-1 && data[next] == '\n' {
1366 next++
1367 }
1368 if i < len(data)-1 && data[i] != ':' && data[next] != ':' {
1369 *flags |= ListItemEndOfList
1370 }
1371 } else {
1372 *flags |= ListItemEndOfList
1373 }
1374 break gatherlines
1375
1376 // a blank line means this should be parsed as a block
1377 case containsBlankLine:
1378 raw.WriteByte('\n')
1379 *flags |= ListItemContainsBlock
1380 }
1381
1382 // if this line was preceded by one or more blanks,
1383 // re-introduce the blank into the buffer
1384 if containsBlankLine {
1385 containsBlankLine = false
1386 raw.WriteByte('\n')
1387 }
1388
1389 // add the line into the working buffer without prefix
1390 raw.Write(data[line+indentIndex : i])
1391
1392 line = i
1393 }
1394
1395 rawBytes := raw.Bytes()
1396
1397 block := p.addBlock(Item, nil)
1398 block.ListFlags = *flags
1399 block.Tight = false
1400 block.BulletChar = bulletChar
1401 block.Delimiter = '.' // Only '.' is possible in Markdown, but ')' will also be possible in CommonMark
1402
1403 // render the contents of the list item
1404 if *flags&ListItemContainsBlock != 0 && *flags&ListTypeTerm == 0 {
1405 // intermediate render of block item, except for definition term
1406 if sublist > 0 {
1407 p.block(rawBytes[:sublist])
1408 p.block(rawBytes[sublist:])
1409 } else {
1410 p.block(rawBytes)
1411 }
1412 } else {
1413 // intermediate render of inline item
1414 if sublist > 0 {
1415 child := p.addChild(Paragraph, 0)
1416 child.content = rawBytes[:sublist]
1417 p.block(rawBytes[sublist:])
1418 } else {
1419 child := p.addChild(Paragraph, 0)
1420 child.content = rawBytes
1421 }
1422 }
1423 return line
1424}
1425
1426// render a single paragraph that has already been parsed out
1427func (p *Markdown) renderParagraph(data []byte) {
1428 if len(data) == 0 {
1429 return
1430 }
1431
1432 // trim leading spaces
1433 beg := 0
1434 for data[beg] == ' ' {
1435 beg++
1436 }
1437
1438 end := len(data)
1439 // trim trailing newline
1440 if data[len(data)-1] == '\n' {
1441 end--
1442 }
1443
1444 // trim trailing spaces
1445 for end > beg && data[end-1] == ' ' {
1446 end--
1447 }
1448
1449 p.addBlock(Paragraph, data[beg:end])
1450}
1451
1452func (p *Markdown) paragraph(data []byte) int {
1453 // prev: index of 1st char of previous line
1454 // line: index of 1st char of current line
1455 // i: index of cursor/end of current line
1456 var prev, line, i int
1457 tabSize := TabSizeDefault
1458 if p.extensions&TabSizeEight != 0 {
1459 tabSize = TabSizeDouble
1460 }
1461 // keep going until we find something to mark the end of the paragraph
1462 for i < len(data) {
1463 // mark the beginning of the current line
1464 prev = line
1465 current := data[i:]
1466 line = i
1467
1468 // did we find a reference or a footnote? If so, end a paragraph
1469 // preceding it and report that we have consumed up to the end of that
1470 // reference:
1471 if refEnd := isReference(p, current, tabSize); refEnd > 0 {
1472 p.renderParagraph(data[:i])
1473 return i + refEnd
1474 }
1475
1476 // did we find a blank line marking the end of the paragraph?
1477 if n := p.isEmpty(current); n > 0 {
1478 // did this blank line followed by a definition list item?
1479 if p.extensions&DefinitionLists != 0 {
1480 if i < len(data)-1 && data[i+1] == ':' {
1481 return p.list(data[prev:], ListTypeDefinition)
1482 }
1483 }
1484
1485 p.renderParagraph(data[:i])
1486 return i + n
1487 }
1488
1489 // an underline under some text marks a heading, so our paragraph ended on prev line
1490 if i > 0 {
1491 if level := p.isUnderlinedHeading(current); level > 0 {
1492 // render the paragraph
1493 p.renderParagraph(data[:prev])
1494
1495 // ignore leading and trailing whitespace
1496 eol := i - 1
1497 for prev < eol && data[prev] == ' ' {
1498 prev++
1499 }
1500 for eol > prev && data[eol-1] == ' ' {
1501 eol--
1502 }
1503
1504 id := ""
1505 if p.extensions&AutoHeadingIDs != 0 {
1506 id = sanitized_anchor_name.Create(string(data[prev:eol]))
1507 }
1508
1509 block := p.addBlock(Heading, data[prev:eol])
1510 block.Level = level
1511 block.HeadingID = id
1512
1513 // find the end of the underline
1514 for i < len(data) && data[i] != '\n' {
1515 i++
1516 }
1517 return i
1518 }
1519 }
1520
1521 // if the next line starts a block of HTML, then the paragraph ends here
1522 if p.extensions&LaxHTMLBlocks != 0 {
1523 if data[i] == '<' && p.html(current, false) > 0 {
1524 // rewind to before the HTML block
1525 p.renderParagraph(data[:i])
1526 return i
1527 }
1528 }
1529
1530 // if there's a prefixed heading or a horizontal rule after this, paragraph is over
1531 if p.isPrefixHeading(current) || p.isHRule(current) {
1532 p.renderParagraph(data[:i])
1533 return i
1534 }
1535
1536 // if there's a fenced code block, paragraph is over
1537 if p.extensions&FencedCode != 0 {
1538 if p.fencedCodeBlock(current, false) > 0 {
1539 p.renderParagraph(data[:i])
1540 return i
1541 }
1542 }
1543
1544 // if there's a definition list item, prev line is a definition term
1545 if p.extensions&DefinitionLists != 0 {
1546 if p.dliPrefix(current) != 0 {
1547 ret := p.list(data[prev:], ListTypeDefinition)
1548 return ret
1549 }
1550 }
1551
1552 // if there's a list after this, paragraph is over
1553 if p.extensions&NoEmptyLineBeforeBlock != 0 {
1554 if p.uliPrefix(current) != 0 ||
1555 p.oliPrefix(current) != 0 ||
1556 p.quotePrefix(current) != 0 ||
1557 p.codePrefix(current) != 0 {
1558 p.renderParagraph(data[:i])
1559 return i
1560 }
1561 }
1562
1563 // otherwise, scan to the beginning of the next line
1564 nl := bytes.IndexByte(data[i:], '\n')
1565 if nl >= 0 {
1566 i += nl + 1
1567 } else {
1568 i += len(data[i:])
1569 }
1570 }
1571
1572 p.renderParagraph(data[:i])
1573 return i
1574}
1575
1576func skipChar(data []byte, start int, char byte) int {
1577 i := start
1578 for i < len(data) && data[i] == char {
1579 i++
1580 }
1581 return i
1582}
1583
1584func skipUntilChar(text []byte, start int, char byte) int {
1585 i := start
1586 for i < len(text) && text[i] != char {
1587 i++
1588 }
1589 return i
1590}