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