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)
19
20// Parse block-level data.
21// Note: this function and many that it calls assume that
22// the input buffer ends with a newline.
23func (p *parser) block(out *bytes.Buffer, data []byte) {
24 if len(data) == 0 || data[len(data)-1] != '\n' {
25 panic("block input is missing terminating newline")
26 }
27
28 // this is called recursively: enforce a maximum depth
29 if p.nesting >= p.maxNesting {
30 return
31 }
32 p.nesting++
33
34 // parse out one block-level construct at a time
35 for len(data) > 0 {
36 // prefixed header:
37 //
38 // # Header 1
39 // ## Header 2
40 // ...
41 // ###### Header 6
42 if p.isPrefixHeader(data) {
43 data = data[p.prefixHeader(out, data):]
44 continue
45 }
46
47 // block of preformatted HTML:
48 //
49 // <div>
50 // ...
51 // </div>
52 if data[0] == '<' {
53 if i := p.html(out, data, true); i > 0 {
54 data = data[i:]
55 continue
56 }
57 }
58
59 // blank lines. note: returns the # of bytes to skip
60 if i := p.isEmpty(data); i > 0 {
61 data = data[i:]
62 continue
63 }
64
65 // indented code block:
66 //
67 // func max(a, b int) int {
68 // if a > b {
69 // return a
70 // }
71 // return b
72 // }
73 if p.codePrefix(data) > 0 {
74 data = data[p.code(out, data):]
75 continue
76 }
77
78 // fenced code block:
79 //
80 // ``` go
81 // func fact(n int) int {
82 // if n <= 1 {
83 // return n
84 // }
85 // return n * fact(n-1)
86 // }
87 // ```
88 if p.flags&EXTENSION_FENCED_CODE != 0 {
89 if i := p.fencedCode(out, data, true); i > 0 {
90 data = data[i:]
91 continue
92 }
93 }
94
95 // horizontal rule:
96 //
97 // ------
98 // or
99 // ******
100 // or
101 // ______
102 if p.isHRule(data) {
103 p.r.HRule(out)
104 var i int
105 for i = 0; data[i] != '\n'; i++ {
106 }
107 data = data[i:]
108 continue
109 }
110
111 // block quote:
112 //
113 // > A big quote I found somewhere
114 // > on the web
115 if p.quotePrefix(data) > 0 {
116 data = data[p.quote(out, data):]
117 continue
118 }
119
120 // table:
121 //
122 // Name | Age | Phone
123 // ------|-----|---------
124 // Bob | 31 | 555-1234
125 // Alice | 27 | 555-4321
126 if p.flags&EXTENSION_TABLES != 0 {
127 if i := p.table(out, data); i > 0 {
128 data = data[i:]
129 continue
130 }
131 }
132
133 // an itemized/unordered list:
134 //
135 // * Item 1
136 // * Item 2
137 //
138 // also works with + or -
139 if p.uliPrefix(data) > 0 {
140 data = data[p.list(out, data, 0):]
141 continue
142 }
143
144 // a numbered/ordered list:
145 //
146 // 1. Item 1
147 // 2. Item 2
148 if p.oliPrefix(data) > 0 {
149 data = data[p.list(out, data, LIST_TYPE_ORDERED):]
150 continue
151 }
152
153 // anything else must look like a normal paragraph
154 // note: this finds underlined headers, too
155 data = data[p.paragraph(out, data):]
156 }
157
158 p.nesting--
159}
160
161func (p *parser) isPrefixHeader(data []byte) bool {
162 if data[0] != '#' {
163 return false
164 }
165
166 if p.flags&EXTENSION_SPACE_HEADERS != 0 {
167 level := 0
168 for level < 6 && data[level] == '#' {
169 level++
170 }
171 if data[level] != ' ' {
172 return false
173 }
174 }
175 return true
176}
177
178func (p *parser) prefixHeader(out *bytes.Buffer, data []byte) int {
179 level := 0
180 for level < 6 && data[level] == '#' {
181 level++
182 }
183 i, end := 0, 0
184 for i = level; data[i] == ' '; i++ {
185 }
186 for end = i; data[end] != '\n'; end++ {
187 }
188 skip := end
189 id := ""
190 if p.flags&EXTENSION_HEADER_IDS != 0 {
191 j, k := 0, 0
192 // find start/end of header id
193 for j = i; j < end - 1 && (data[j] != '{' || data[j+1] != '#'); j++ {
194 }
195 for k = j + 1; k < end && data[k] != '}'; k++ {
196 }
197 // extract header id iff found
198 if j < end && k < end {
199 id = string(data[j+2:k])
200 end = j
201 skip = k + 1
202 for end > 0 && data[end-1] == ' ' {
203 end--
204 }
205 }
206 }
207 for end > 0 && data[end-1] == '#' {
208 end--
209 }
210 for end > 0 && data[end-1] == ' ' {
211 end--
212 }
213 if end > i {
214 work := func() bool {
215 p.inline(out, data[i:end])
216 return true
217 }
218 p.r.Header(out, work, level, id)
219 }
220 return skip
221}
222
223func (p *parser) isUnderlinedHeader(data []byte) int {
224 // test of level 1 header
225 if data[0] == '=' {
226 i := 1
227 for data[i] == '=' {
228 i++
229 }
230 for data[i] == ' ' {
231 i++
232 }
233 if data[i] == '\n' {
234 return 1
235 } else {
236 return 0
237 }
238 }
239
240 // test of level 2 header
241 if data[0] == '-' {
242 i := 1
243 for data[i] == '-' {
244 i++
245 }
246 for data[i] == ' ' {
247 i++
248 }
249 if data[i] == '\n' {
250 return 2
251 } else {
252 return 0
253 }
254 }
255
256 return 0
257}
258
259func (p *parser) html(out *bytes.Buffer, data []byte, doRender bool) int {
260 var i, j int
261
262 // identify the opening tag
263 if data[0] != '<' {
264 return 0
265 }
266 curtag, tagfound := p.htmlFindTag(data[1:])
267
268 // handle special cases
269 if !tagfound {
270 // check for an HTML comment
271 if size := p.htmlComment(out, data, doRender); size > 0 {
272 return size
273 }
274
275 // check for an <hr> tag
276 if size := p.htmlHr(out, data, doRender); size > 0 {
277 return size
278 }
279
280 // no special case recognized
281 return 0
282 }
283
284 // look for an unindented matching closing tag
285 // followed by a blank line
286 found := false
287 /*
288 closetag := []byte("\n</" + curtag + ">")
289 j = len(curtag) + 1
290 for !found {
291 // scan for a closing tag at the beginning of a line
292 if skip := bytes.Index(data[j:], closetag); skip >= 0 {
293 j += skip + len(closetag)
294 } else {
295 break
296 }
297
298 // see if it is the only thing on the line
299 if skip := p.isEmpty(data[j:]); skip > 0 {
300 // see if it is followed by a blank line/eof
301 j += skip
302 if j >= len(data) {
303 found = true
304 i = j
305 } else {
306 if skip := p.isEmpty(data[j:]); skip > 0 {
307 j += skip
308 found = true
309 i = j
310 }
311 }
312 }
313 }
314 */
315
316 // if not found, try a second pass looking for indented match
317 // but not if tag is "ins" or "del" (following original Markdown.pl)
318 if !found && curtag != "ins" && curtag != "del" {
319 i = 1
320 for i < len(data) {
321 i++
322 for i < len(data) && !(data[i-1] == '<' && data[i] == '/') {
323 i++
324 }
325
326 if i+2+len(curtag) >= len(data) {
327 break
328 }
329
330 j = p.htmlFindEnd(curtag, data[i-1:])
331
332 if j > 0 {
333 i += j - 1
334 found = true
335 break
336 }
337 }
338 }
339
340 if !found {
341 return 0
342 }
343
344 // the end of the block has been found
345 if doRender {
346 // trim newlines
347 end := i
348 for end > 0 && data[end-1] == '\n' {
349 end--
350 }
351 p.r.BlockHtml(out, data[:end])
352 }
353
354 return i
355}
356
357// HTML comment, lax form
358func (p *parser) htmlComment(out *bytes.Buffer, data []byte, doRender bool) int {
359 if data[0] != '<' || data[1] != '!' || data[2] != '-' || data[3] != '-' {
360 return 0
361 }
362
363 i := 5
364
365 // scan for an end-of-comment marker, across lines if necessary
366 for i < len(data) && !(data[i-2] == '-' && data[i-1] == '-' && data[i] == '>') {
367 i++
368 }
369 i++
370
371 // no end-of-comment marker
372 if i >= len(data) {
373 return 0
374 }
375
376 // needs to end with a blank line
377 if j := p.isEmpty(data[i:]); j > 0 {
378 size := i + j
379 if doRender {
380 // trim trailing newlines
381 end := size
382 for end > 0 && data[end-1] == '\n' {
383 end--
384 }
385 p.r.BlockHtml(out, data[:end])
386 }
387 return size
388 }
389
390 return 0
391}
392
393// HR, which is the only self-closing block tag considered
394func (p *parser) htmlHr(out *bytes.Buffer, data []byte, doRender bool) int {
395 if data[0] != '<' || (data[1] != 'h' && data[1] != 'H') || (data[2] != 'r' && data[2] != 'R') {
396 return 0
397 }
398 if data[3] != ' ' && data[3] != '/' && data[3] != '>' {
399 // not an <hr> tag after all; at least not a valid one
400 return 0
401 }
402
403 i := 3
404 for data[i] != '>' && data[i] != '\n' {
405 i++
406 }
407
408 if data[i] == '>' {
409 i++
410 if j := p.isEmpty(data[i:]); j > 0 {
411 size := i + j
412 if doRender {
413 // trim newlines
414 end := size
415 for end > 0 && data[end-1] == '\n' {
416 end--
417 }
418 p.r.BlockHtml(out, data[:end])
419 }
420 return size
421 }
422 }
423
424 return 0
425}
426
427func (p *parser) htmlFindTag(data []byte) (string, bool) {
428 i := 0
429 for isalnum(data[i]) {
430 i++
431 }
432 key := string(data[:i])
433 if blockTags[key] {
434 return key, true
435 }
436 return "", false
437}
438
439func (p *parser) htmlFindEnd(tag string, data []byte) int {
440 // assume data[0] == '<' && data[1] == '/' already tested
441
442 // check if tag is a match
443 closetag := []byte("</" + tag + ">")
444 if !bytes.HasPrefix(data, closetag) {
445 return 0
446 }
447 i := len(closetag)
448
449 // check that the rest of the line is blank
450 skip := 0
451 if skip = p.isEmpty(data[i:]); skip == 0 {
452 return 0
453 }
454 i += skip
455 skip = 0
456
457 if i >= len(data) {
458 return i
459 }
460
461 if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
462 return i
463 }
464 if skip = p.isEmpty(data[i:]); skip == 0 {
465 // following line must be blank
466 return 0
467 }
468
469 return i + skip
470}
471
472func (p *parser) isEmpty(data []byte) int {
473 // it is okay to call isEmpty on an empty buffer
474 if len(data) == 0 {
475 return 0
476 }
477
478 var i int
479 for i = 0; i < len(data) && data[i] != '\n'; i++ {
480 if data[i] != ' ' && data[i] != '\t' {
481 return 0
482 }
483 }
484 return i + 1
485}
486
487func (p *parser) isHRule(data []byte) bool {
488 i := 0
489
490 // skip up to three spaces
491 for i < 3 && data[i] == ' ' {
492 i++
493 }
494
495 // look at the hrule char
496 if data[i] != '*' && data[i] != '-' && data[i] != '_' {
497 return false
498 }
499 c := data[i]
500
501 // the whole line must be the char or whitespace
502 n := 0
503 for data[i] != '\n' {
504 switch {
505 case data[i] == c:
506 n++
507 case data[i] != ' ':
508 return false
509 }
510 i++
511 }
512
513 return n >= 3
514}
515
516func (p *parser) isFencedCode(data []byte, syntax **string, oldmarker string) (skip int, marker string) {
517 i, size := 0, 0
518 skip = 0
519
520 // skip up to three spaces
521 for i < 3 && data[i] == ' ' {
522 i++
523 }
524
525 // check for the marker characters: ~ or `
526 if data[i] != '~' && data[i] != '`' {
527 return
528 }
529
530 c := data[i]
531
532 // the whole line must be the same char or whitespace
533 for data[i] == c {
534 size++
535 i++
536 }
537
538 // the marker char must occur at least 3 times
539 if size < 3 {
540 return
541 }
542 marker = string(data[i-size : i])
543
544 // if this is the end marker, it must match the beginning marker
545 if oldmarker != "" && marker != oldmarker {
546 return
547 }
548
549 if syntax != nil {
550 syn := 0
551
552 for data[i] == ' ' {
553 i++
554 }
555
556 syntaxStart := i
557
558 if data[i] == '{' {
559 i++
560 syntaxStart++
561
562 for data[i] != '}' && data[i] != '\n' {
563 syn++
564 i++
565 }
566
567 if data[i] != '}' {
568 return
569 }
570
571 // strip all whitespace at the beginning and the end
572 // of the {} block
573 for syn > 0 && isspace(data[syntaxStart]) {
574 syntaxStart++
575 syn--
576 }
577
578 for syn > 0 && isspace(data[syntaxStart+syn-1]) {
579 syn--
580 }
581
582 i++
583 } else {
584 for !isspace(data[i]) {
585 syn++
586 i++
587 }
588 }
589
590 language := string(data[syntaxStart : syntaxStart+syn])
591 *syntax = &language
592 }
593
594 for data[i] == ' ' {
595 i++
596 }
597 if data[i] != '\n' {
598 return
599 }
600
601 skip = i + 1
602 return
603}
604
605func (p *parser) fencedCode(out *bytes.Buffer, data []byte, doRender bool) int {
606 var lang *string
607 beg, marker := p.isFencedCode(data, &lang, "")
608 if beg == 0 || beg >= len(data) {
609 return 0
610 }
611
612 var work bytes.Buffer
613
614 for {
615 // safe to assume beg < len(data)
616
617 // check for the end of the code block
618 fenceEnd, _ := p.isFencedCode(data[beg:], nil, marker)
619 if fenceEnd != 0 {
620 beg += fenceEnd
621 break
622 }
623
624 // copy the current line
625 end := beg
626 for data[end] != '\n' {
627 end++
628 }
629 end++
630
631 // did we reach the end of the buffer without a closing marker?
632 if end >= len(data) {
633 return 0
634 }
635
636 // verbatim copy to the working buffer
637 if doRender {
638 work.Write(data[beg:end])
639 }
640 beg = end
641 }
642
643 syntax := ""
644 if lang != nil {
645 syntax = *lang
646 }
647
648 if doRender {
649 p.r.BlockCode(out, work.Bytes(), syntax)
650 }
651
652 return beg
653}
654
655func (p *parser) table(out *bytes.Buffer, data []byte) int {
656 var header bytes.Buffer
657 i, columns := p.tableHeader(&header, data)
658 if i == 0 {
659 return 0
660 }
661
662 var body bytes.Buffer
663
664 for i < len(data) {
665 pipes, rowStart := 0, i
666 for ; data[i] != '\n'; i++ {
667 if data[i] == '|' {
668 pipes++
669 }
670 }
671
672 if pipes == 0 {
673 i = rowStart
674 break
675 }
676
677 // include the newline in data sent to tableRow
678 i++
679 p.tableRow(&body, data[rowStart:i], columns, false)
680 }
681
682 p.r.Table(out, header.Bytes(), body.Bytes(), columns)
683
684 return i
685}
686
687// check if the specified position is preceeded by an odd number of backslashes
688func isBackslashEscaped(data []byte, i int) bool {
689 backslashes := 0
690 for i-backslashes-1 >= 0 && data[i-backslashes-1] == '\\' {
691 backslashes++
692 }
693 return backslashes&1 == 1
694}
695
696func (p *parser) tableHeader(out *bytes.Buffer, data []byte) (size int, columns []int) {
697 i := 0
698 colCount := 1
699 for i = 0; data[i] != '\n'; i++ {
700 if data[i] == '|' && !isBackslashEscaped(data, i) {
701 colCount++
702 }
703 }
704
705 // doesn't look like a table header
706 if colCount == 1 {
707 return
708 }
709
710 // include the newline in the data sent to tableRow
711 header := data[:i+1]
712
713 // column count ignores pipes at beginning or end of line
714 if data[0] == '|' {
715 colCount--
716 }
717 if i > 2 && data[i-1] == '|' && !isBackslashEscaped(data, i-1) {
718 colCount--
719 }
720
721 columns = make([]int, colCount)
722
723 // move on to the header underline
724 i++
725 if i >= len(data) {
726 return
727 }
728
729 if data[i] == '|' && !isBackslashEscaped(data, i) {
730 i++
731 }
732 for data[i] == ' ' {
733 i++
734 }
735
736 // each column header is of form: / *:?-+:? *|/ with # dashes + # colons >= 3
737 // and trailing | optional on last column
738 col := 0
739 for data[i] != '\n' {
740 dashes := 0
741
742 if data[i] == ':' {
743 i++
744 columns[col] |= TABLE_ALIGNMENT_LEFT
745 dashes++
746 }
747 for data[i] == '-' {
748 i++
749 dashes++
750 }
751 if data[i] == ':' {
752 i++
753 columns[col] |= TABLE_ALIGNMENT_RIGHT
754 dashes++
755 }
756 for data[i] == ' ' {
757 i++
758 }
759
760 // end of column test is messy
761 switch {
762 case dashes < 3:
763 // not a valid column
764 return
765
766 case data[i] == '|' && !isBackslashEscaped(data, i):
767 // marker found, now skip past trailing whitespace
768 col++
769 i++
770 for data[i] == ' ' {
771 i++
772 }
773
774 // trailing junk found after last column
775 if col >= colCount && data[i] != '\n' {
776 return
777 }
778
779 case (data[i] != '|' || isBackslashEscaped(data, i)) && col+1 < colCount:
780 // something else found where marker was required
781 return
782
783 case data[i] == '\n':
784 // marker is optional for the last column
785 col++
786
787 default:
788 // trailing junk found after last column
789 return
790 }
791 }
792 if col != colCount {
793 return
794 }
795
796 p.tableRow(out, header, columns, true)
797 size = i + 1
798 return
799}
800
801func (p *parser) tableRow(out *bytes.Buffer, data []byte, columns []int, header bool) {
802 i, col := 0, 0
803 var rowWork bytes.Buffer
804
805 if data[i] == '|' && !isBackslashEscaped(data, i) {
806 i++
807 }
808
809 for col = 0; col < len(columns) && i < len(data); col++ {
810 for data[i] == ' ' {
811 i++
812 }
813
814 cellStart := i
815
816 for (data[i] != '|' || isBackslashEscaped(data, i)) && data[i] != '\n' {
817 i++
818 }
819
820 cellEnd := i
821
822 // skip the end-of-cell marker, possibly taking us past end of buffer
823 i++
824
825 for cellEnd > cellStart && data[cellEnd-1] == ' ' {
826 cellEnd--
827 }
828
829 var cellWork bytes.Buffer
830 p.inline(&cellWork, data[cellStart:cellEnd])
831
832 if header {
833 p.r.TableHeaderCell(&rowWork, cellWork.Bytes(), columns[col])
834 } else {
835 p.r.TableCell(&rowWork, cellWork.Bytes(), columns[col])
836 }
837 }
838
839 // pad it out with empty columns to get the right number
840 for ; col < len(columns); col++ {
841 if header {
842 p.r.TableHeaderCell(&rowWork, nil, columns[col])
843 } else {
844 p.r.TableCell(&rowWork, nil, columns[col])
845 }
846 }
847
848 // silently ignore rows with too many cells
849
850 p.r.TableRow(out, rowWork.Bytes())
851}
852
853// returns blockquote prefix length
854func (p *parser) quotePrefix(data []byte) int {
855 i := 0
856 for i < 3 && data[i] == ' ' {
857 i++
858 }
859 if data[i] == '>' {
860 if data[i+1] == ' ' {
861 return i + 2
862 }
863 return i + 1
864 }
865 return 0
866}
867
868// parse a blockquote fragment
869func (p *parser) quote(out *bytes.Buffer, data []byte) int {
870 var raw bytes.Buffer
871 beg, end := 0, 0
872 for beg < len(data) {
873 end = beg
874 for data[end] != '\n' {
875 end++
876 }
877 end++
878
879 if pre := p.quotePrefix(data[beg:]); pre > 0 {
880 // skip the prefix
881 beg += pre
882 } else if p.isEmpty(data[beg:]) > 0 &&
883 (end >= len(data) ||
884 (p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0)) {
885 // blockquote ends with at least one blank line
886 // followed by something without a blockquote prefix
887 break
888 }
889
890 // this line is part of the blockquote
891 raw.Write(data[beg:end])
892 beg = end
893 }
894
895 var cooked bytes.Buffer
896 p.block(&cooked, raw.Bytes())
897 p.r.BlockQuote(out, cooked.Bytes())
898 return end
899}
900
901// returns prefix length for block code
902func (p *parser) codePrefix(data []byte) int {
903 if data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
904 return 4
905 }
906 return 0
907}
908
909func (p *parser) code(out *bytes.Buffer, data []byte) int {
910 var work bytes.Buffer
911
912 i := 0
913 for i < len(data) {
914 beg := i
915 for data[i] != '\n' {
916 i++
917 }
918 i++
919
920 blankline := p.isEmpty(data[beg:i]) > 0
921 if pre := p.codePrefix(data[beg:i]); pre > 0 {
922 beg += pre
923 } else if !blankline {
924 // non-empty, non-prefixed line breaks the pre
925 i = beg
926 break
927 }
928
929 // verbatim copy to the working buffeu
930 if blankline {
931 work.WriteByte('\n')
932 } else {
933 work.Write(data[beg:i])
934 }
935 }
936
937 // trim all the \n off the end of work
938 workbytes := work.Bytes()
939 eol := len(workbytes)
940 for eol > 0 && workbytes[eol-1] == '\n' {
941 eol--
942 }
943 if eol != len(workbytes) {
944 work.Truncate(eol)
945 }
946
947 work.WriteByte('\n')
948
949 p.r.BlockCode(out, work.Bytes(), "")
950
951 return i
952}
953
954// returns unordered list item prefix
955func (p *parser) uliPrefix(data []byte) int {
956 i := 0
957
958 // start with up to 3 spaces
959 for i < 3 && data[i] == ' ' {
960 i++
961 }
962
963 // need a *, +, or - followed by a space
964 if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
965 data[i+1] != ' ' {
966 return 0
967 }
968 return i + 2
969}
970
971// returns ordered list item prefix
972func (p *parser) oliPrefix(data []byte) int {
973 i := 0
974
975 // start with up to 3 spaces
976 for i < 3 && data[i] == ' ' {
977 i++
978 }
979
980 // count the digits
981 start := i
982 for data[i] >= '0' && data[i] <= '9' {
983 i++
984 }
985
986 // we need >= 1 digits followed by a dot and a space
987 if start == i || data[i] != '.' || data[i+1] != ' ' {
988 return 0
989 }
990 return i + 2
991}
992
993// parse ordered or unordered list block
994func (p *parser) list(out *bytes.Buffer, data []byte, flags int) int {
995 i := 0
996 flags |= LIST_ITEM_BEGINNING_OF_LIST
997 work := func() bool {
998 for i < len(data) {
999 skip := p.listItem(out, data[i:], &flags)
1000 i += skip
1001
1002 if skip == 0 || flags&LIST_ITEM_END_OF_LIST != 0 {
1003 break
1004 }
1005 flags &= ^LIST_ITEM_BEGINNING_OF_LIST
1006 }
1007 return true
1008 }
1009
1010 p.r.List(out, work, flags)
1011 return i
1012}
1013
1014// Parse a single list item.
1015// Assumes initial prefix is already removed if this is a sublist.
1016func (p *parser) listItem(out *bytes.Buffer, data []byte, flags *int) int {
1017 // keep track of the indentation of the first line
1018 itemIndent := 0
1019 for itemIndent < 3 && data[itemIndent] == ' ' {
1020 itemIndent++
1021 }
1022
1023 i := p.uliPrefix(data)
1024 if i == 0 {
1025 i = p.oliPrefix(data)
1026 }
1027 if i == 0 {
1028 return 0
1029 }
1030
1031 // skip leading whitespace on first line
1032 for data[i] == ' ' {
1033 i++
1034 }
1035
1036 // find the end of the line
1037 line := i
1038 for data[i-1] != '\n' {
1039 i++
1040 }
1041
1042 // get working buffer
1043 var raw bytes.Buffer
1044
1045 // put the first line into the working buffer
1046 raw.Write(data[line:i])
1047 line = i
1048
1049 // process the following lines
1050 containsBlankLine := false
1051 sublist := 0
1052
1053gatherlines:
1054 for line < len(data) {
1055 i++
1056
1057 // find the end of this line
1058 for data[i-1] != '\n' {
1059 i++
1060 }
1061
1062 // if it is an empty line, guess that it is part of this item
1063 // and move on to the next line
1064 if p.isEmpty(data[line:i]) > 0 {
1065 containsBlankLine = true
1066 line = i
1067 continue
1068 }
1069
1070 // calculate the indentation
1071 indent := 0
1072 for indent < 4 && line+indent < i && data[line+indent] == ' ' {
1073 indent++
1074 }
1075
1076 chunk := data[line+indent : i]
1077
1078 // evaluate how this line fits in
1079 switch {
1080 // is this a nested list item?
1081 case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) ||
1082 p.oliPrefix(chunk) > 0:
1083
1084 if containsBlankLine {
1085 *flags |= LIST_ITEM_CONTAINS_BLOCK
1086 }
1087
1088 // to be a nested list, it must be indented more
1089 // if not, it is the next item in the same list
1090 if indent <= itemIndent {
1091 break gatherlines
1092 }
1093
1094 // is this the first item in the the nested list?
1095 if sublist == 0 {
1096 sublist = raw.Len()
1097 }
1098
1099 // is this a nested prefix header?
1100 case p.isPrefixHeader(chunk):
1101 // if the header is not indented, it is not nested in the list
1102 // and thus ends the list
1103 if containsBlankLine && indent < 4 {
1104 *flags |= LIST_ITEM_END_OF_LIST
1105 break gatherlines
1106 }
1107 *flags |= LIST_ITEM_CONTAINS_BLOCK
1108
1109 // anything following an empty line is only part
1110 // of this item if it is indented 4 spaces
1111 // (regardless of the indentation of the beginning of the item)
1112 case containsBlankLine && indent < 4:
1113 *flags |= LIST_ITEM_END_OF_LIST
1114 break gatherlines
1115
1116 // a blank line means this should be parsed as a block
1117 case containsBlankLine:
1118 raw.WriteByte('\n')
1119 *flags |= LIST_ITEM_CONTAINS_BLOCK
1120 }
1121
1122 // if this line was preceeded by one or more blanks,
1123 // re-introduce the blank into the buffer
1124 if containsBlankLine {
1125 containsBlankLine = false
1126 raw.WriteByte('\n')
1127 }
1128
1129 // add the line into the working buffer without prefix
1130 raw.Write(data[line+indent : i])
1131
1132 line = i
1133 }
1134
1135 rawBytes := raw.Bytes()
1136
1137 // render the contents of the list item
1138 var cooked bytes.Buffer
1139 if *flags&LIST_ITEM_CONTAINS_BLOCK != 0 {
1140 // intermediate render of block li
1141 if sublist > 0 {
1142 p.block(&cooked, rawBytes[:sublist])
1143 p.block(&cooked, rawBytes[sublist:])
1144 } else {
1145 p.block(&cooked, rawBytes)
1146 }
1147 } else {
1148 // intermediate render of inline li
1149 if sublist > 0 {
1150 p.inline(&cooked, rawBytes[:sublist])
1151 p.block(&cooked, rawBytes[sublist:])
1152 } else {
1153 p.inline(&cooked, rawBytes)
1154 }
1155 }
1156
1157 // render the actual list item
1158 cookedBytes := cooked.Bytes()
1159 parsedEnd := len(cookedBytes)
1160
1161 // strip trailing newlines
1162 for parsedEnd > 0 && cookedBytes[parsedEnd-1] == '\n' {
1163 parsedEnd--
1164 }
1165 p.r.ListItem(out, cookedBytes[:parsedEnd], *flags)
1166
1167 return line
1168}
1169
1170// render a single paragraph that has already been parsed out
1171func (p *parser) renderParagraph(out *bytes.Buffer, data []byte) {
1172 if len(data) == 0 {
1173 return
1174 }
1175
1176 // trim leading spaces
1177 beg := 0
1178 for data[beg] == ' ' {
1179 beg++
1180 }
1181
1182 // trim trailing newline
1183 end := len(data) - 1
1184
1185 // trim trailing spaces
1186 for end > beg && data[end-1] == ' ' {
1187 end--
1188 }
1189
1190 work := func() bool {
1191 p.inline(out, data[beg:end])
1192 return true
1193 }
1194 p.r.Paragraph(out, work)
1195}
1196
1197func (p *parser) paragraph(out *bytes.Buffer, data []byte) int {
1198 // prev: index of 1st char of previous line
1199 // line: index of 1st char of current line
1200 // i: index of cursor/end of current line
1201 var prev, line, i int
1202
1203 // keep going until we find something to mark the end of the paragraph
1204 for i < len(data) {
1205 // mark the beginning of the current line
1206 prev = line
1207 current := data[i:]
1208 line = i
1209
1210 // did we find a blank line marking the end of the paragraph?
1211 if n := p.isEmpty(current); n > 0 {
1212 p.renderParagraph(out, data[:i])
1213 return i + n
1214 }
1215
1216 // an underline under some text marks a header, so our paragraph ended on prev line
1217 if i > 0 {
1218 if level := p.isUnderlinedHeader(current); level > 0 {
1219 // render the paragraph
1220 p.renderParagraph(out, data[:prev])
1221
1222 // ignore leading and trailing whitespace
1223 eol := i - 1
1224 for prev < eol && data[prev] == ' ' {
1225 prev++
1226 }
1227 for eol > prev && data[eol-1] == ' ' {
1228 eol--
1229 }
1230
1231 // render the header
1232 // this ugly double closure avoids forcing variables onto the heap
1233 work := func(o *bytes.Buffer, pp *parser, d []byte) func() bool {
1234 return func() bool {
1235 pp.inline(o, d)
1236 return true
1237 }
1238 }(out, p, data[prev:eol])
1239 p.r.Header(out, work, level, "")
1240
1241 // find the end of the underline
1242 for data[i] != '\n' {
1243 i++
1244 }
1245 return i
1246 }
1247 }
1248
1249 // if the next line starts a block of HTML, then the paragraph ends here
1250 if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
1251 if data[i] == '<' && p.html(out, current, false) > 0 {
1252 // rewind to before the HTML block
1253 p.renderParagraph(out, data[:i])
1254 return i
1255 }
1256 }
1257
1258 // if there's a prefixed header or a horizontal rule after this, paragraph is over
1259 if p.isPrefixHeader(current) || p.isHRule(current) {
1260 p.renderParagraph(out, data[:i])
1261 return i
1262 }
1263
1264 // if there's a list after this, paragraph is over
1265 if p.flags&EXTENSION_NO_EMPTY_LINE_BEFORE_BLOCK != 0 {
1266 if p.uliPrefix(current) != 0 ||
1267 p.oliPrefix(current) != 0 ||
1268 p.quotePrefix(current) != 0 ||
1269 p.codePrefix(current) != 0 {
1270 p.renderParagraph(out, data[:i])
1271 return i
1272 }
1273 }
1274
1275 // otherwise, scan to the beginning of the next line
1276 for data[i] != '\n' {
1277 i++
1278 }
1279 i++
1280 }
1281
1282 p.renderParagraph(out, data[:i])
1283 return i
1284}