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
665// check if the specified position is preceeded by an odd number of backslashes
666func isBackslashEscaped(data []byte, i int) bool {
667 backslashes := 0
668 for i-backslashes-1 >= 0 && data[i-backslashes-1] == '\\' {
669 backslashes++
670 }
671 return backslashes&1 == 1
672}
673
674func (p *parser) tableHeader(out *bytes.Buffer, data []byte) (size int, columns []int) {
675 i := 0
676 colCount := 1
677 for i = 0; data[i] != '\n'; i++ {
678 if data[i] == '|' && !isBackslashEscaped(data, i) {
679 colCount++
680 }
681 }
682
683 // doesn't look like a table header
684 if colCount == 1 {
685 return
686 }
687
688 // include the newline in the data sent to tableRow
689 header := data[:i+1]
690
691 // column count ignores pipes at beginning or end of line
692 if data[0] == '|' {
693 colCount--
694 }
695 if i > 2 && data[i-1] == '|' && !isBackslashEscaped(data, i-1) {
696 colCount--
697 }
698
699 columns = make([]int, colCount)
700
701 // move on to the header underline
702 i++
703 if i >= len(data) {
704 return
705 }
706
707 if data[i] == '|' && !isBackslashEscaped(data, i) {
708 i++
709 }
710 for data[i] == ' ' {
711 i++
712 }
713
714 // each column header is of form: / *:?-+:? *|/ with # dashes + # colons >= 3
715 // and trailing | optional on last column
716 col := 0
717 for data[i] != '\n' {
718 dashes := 0
719
720 if data[i] == ':' {
721 i++
722 columns[col] |= TABLE_ALIGNMENT_LEFT
723 dashes++
724 }
725 for data[i] == '-' {
726 i++
727 dashes++
728 }
729 if data[i] == ':' {
730 i++
731 columns[col] |= TABLE_ALIGNMENT_RIGHT
732 dashes++
733 }
734 for data[i] == ' ' {
735 i++
736 }
737
738 // end of column test is messy
739 switch {
740 case dashes < 3:
741 // not a valid column
742 return
743
744 case data[i] == '|' && !isBackslashEscaped(data, i):
745 // marker found, now skip past trailing whitespace
746 col++
747 i++
748 for data[i] == ' ' {
749 i++
750 }
751
752 // trailing junk found after last column
753 if col >= colCount && data[i] != '\n' {
754 return
755 }
756
757 case (data[i] != '|' || isBackslashEscaped(data, i)) && col+1 < colCount:
758 // something else found where marker was required
759 return
760
761 case data[i] == '\n':
762 // marker is optional for the last column
763 col++
764
765 default:
766 // trailing junk found after last column
767 return
768 }
769 }
770 if col != colCount {
771 return
772 }
773
774 p.tableRow(out, header, columns)
775 size = i + 1
776 return
777}
778
779func (p *parser) tableRow(out *bytes.Buffer, data []byte, columns []int) {
780 i, col := 0, 0
781 var rowWork bytes.Buffer
782
783 if data[i] == '|' && !isBackslashEscaped(data, i) {
784 i++
785 }
786
787 for col = 0; col < len(columns) && i < len(data); col++ {
788 for data[i] == ' ' {
789 i++
790 }
791
792 cellStart := i
793
794 for (data[i] != '|' || isBackslashEscaped(data, i)) && data[i] != '\n' {
795 i++
796 }
797
798 cellEnd := i
799
800 // skip the end-of-cell marker, possibly taking us past end of buffer
801 i++
802
803 for cellEnd > cellStart && data[cellEnd-1] == ' ' {
804 cellEnd--
805 }
806
807 var cellWork bytes.Buffer
808 p.inline(&cellWork, data[cellStart:cellEnd])
809 p.r.TableCell(&rowWork, cellWork.Bytes(), columns[col])
810 }
811
812 // pad it out with empty columns to get the right number
813 for ; col < len(columns); col++ {
814 p.r.TableCell(&rowWork, nil, columns[col])
815 }
816
817 // silently ignore rows with too many cells
818
819 p.r.TableRow(out, rowWork.Bytes())
820}
821
822// returns blockquote prefix length
823func (p *parser) quotePrefix(data []byte) int {
824 i := 0
825 for i < 3 && data[i] == ' ' {
826 i++
827 }
828 if data[i] == '>' {
829 if data[i+1] == ' ' {
830 return i + 2
831 }
832 return i + 1
833 }
834 return 0
835}
836
837// parse a blockquote fragment
838func (p *parser) quote(out *bytes.Buffer, data []byte) int {
839 var raw bytes.Buffer
840 beg, end := 0, 0
841 for beg < len(data) {
842 end = beg
843 for data[end] != '\n' {
844 end++
845 }
846 end++
847
848 if pre := p.quotePrefix(data[beg:]); pre > 0 {
849 // skip the prefix
850 beg += pre
851 } else if p.isEmpty(data[beg:]) > 0 &&
852 (end >= len(data) ||
853 (p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0)) {
854 // blockquote ends with at least one blank line
855 // followed by something without a blockquote prefix
856 break
857 }
858
859 // this line is part of the blockquote
860 raw.Write(data[beg:end])
861 beg = end
862 }
863
864 var cooked bytes.Buffer
865 p.block(&cooked, raw.Bytes())
866 p.r.BlockQuote(out, cooked.Bytes())
867 return end
868}
869
870// returns prefix length for block code
871func (p *parser) codePrefix(data []byte) int {
872 if data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
873 return 4
874 }
875 return 0
876}
877
878func (p *parser) code(out *bytes.Buffer, data []byte) int {
879 var work bytes.Buffer
880
881 i := 0
882 for i < len(data) {
883 beg := i
884 for data[i] != '\n' {
885 i++
886 }
887 i++
888
889 blankline := p.isEmpty(data[beg:i]) > 0
890 if pre := p.codePrefix(data[beg:i]); pre > 0 {
891 beg += pre
892 } else if !blankline {
893 // non-empty, non-prefixed line breaks the pre
894 i = beg
895 break
896 }
897
898 // verbatim copy to the working buffeu
899 if blankline {
900 work.WriteByte('\n')
901 } else {
902 work.Write(data[beg:i])
903 }
904 }
905
906 // trim all the \n off the end of work
907 workbytes := work.Bytes()
908 eol := len(workbytes)
909 for eol > 0 && workbytes[eol-1] == '\n' {
910 eol--
911 }
912 if eol != len(workbytes) {
913 work.Truncate(eol)
914 }
915
916 work.WriteByte('\n')
917
918 p.r.BlockCode(out, work.Bytes(), "")
919
920 return i
921}
922
923// returns unordered list item prefix
924func (p *parser) uliPrefix(data []byte) int {
925 i := 0
926
927 // start with up to 3 spaces
928 for i < 3 && data[i] == ' ' {
929 i++
930 }
931
932 // need a *, +, or - followed by a space
933 if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
934 data[i+1] != ' ' {
935 return 0
936 }
937 return i + 2
938}
939
940// returns ordered list item prefix
941func (p *parser) oliPrefix(data []byte) int {
942 i := 0
943
944 // start with up to 3 spaces
945 for i < 3 && data[i] == ' ' {
946 i++
947 }
948
949 // count the digits
950 start := i
951 for data[i] >= '0' && data[i] <= '9' {
952 i++
953 }
954
955 // we need >= 1 digits followed by a dot and a space
956 if start == i || data[i] != '.' || data[i+1] != ' ' {
957 return 0
958 }
959 return i + 2
960}
961
962// parse ordered or unordered list block
963func (p *parser) list(out *bytes.Buffer, data []byte, flags int) int {
964 i := 0
965 flags |= LIST_ITEM_BEGINNING_OF_LIST
966 work := func() bool {
967 for i < len(data) {
968 skip := p.listItem(out, data[i:], &flags)
969 i += skip
970
971 if skip == 0 || flags&LIST_ITEM_END_OF_LIST != 0 {
972 break
973 }
974 flags &= ^LIST_ITEM_BEGINNING_OF_LIST
975 }
976 return true
977 }
978
979 p.r.List(out, work, flags)
980 return i
981}
982
983// Parse a single list item.
984// Assumes initial prefix is already removed if this is a sublist.
985func (p *parser) listItem(out *bytes.Buffer, data []byte, flags *int) int {
986 // keep track of the indentation of the first line
987 itemIndent := 0
988 for itemIndent < 3 && data[itemIndent] == ' ' {
989 itemIndent++
990 }
991
992 i := p.uliPrefix(data)
993 if i == 0 {
994 i = p.oliPrefix(data)
995 }
996 if i == 0 {
997 return 0
998 }
999
1000 // skip leading whitespace on first line
1001 for data[i] == ' ' {
1002 i++
1003 }
1004
1005 // find the end of the line
1006 line := i
1007 for data[i-1] != '\n' {
1008 i++
1009 }
1010
1011 // get working buffer
1012 var raw bytes.Buffer
1013
1014 // put the first line into the working buffer
1015 raw.Write(data[line:i])
1016 line = i
1017
1018 // process the following lines
1019 containsBlankLine := false
1020 sublist := 0
1021
1022gatherlines:
1023 for line < len(data) {
1024 i++
1025
1026 // find the end of this line
1027 for data[i-1] != '\n' {
1028 i++
1029 }
1030
1031 // if it is an empty line, guess that it is part of this item
1032 // and move on to the next line
1033 if p.isEmpty(data[line:i]) > 0 {
1034 containsBlankLine = true
1035 line = i
1036 continue
1037 }
1038
1039 // calculate the indentation
1040 indent := 0
1041 for indent < 4 && line+indent < i && data[line+indent] == ' ' {
1042 indent++
1043 }
1044
1045 chunk := data[line+indent : i]
1046
1047 // evaluate how this line fits in
1048 switch {
1049 // is this a nested list item?
1050 case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) ||
1051 p.oliPrefix(chunk) > 0:
1052
1053 if containsBlankLine {
1054 *flags |= LIST_ITEM_CONTAINS_BLOCK
1055 }
1056
1057 // to be a nested list, it must be indented more
1058 // if not, it is the next item in the same list
1059 if indent <= itemIndent {
1060 break gatherlines
1061 }
1062
1063 // is this the first item in the the nested list?
1064 if sublist == 0 {
1065 sublist = raw.Len()
1066 }
1067
1068 // is this a nested prefix header?
1069 case p.isPrefixHeader(chunk):
1070 // if the header is not indented, it is not nested in the list
1071 // and thus ends the list
1072 if containsBlankLine && indent < 4 {
1073 *flags |= LIST_ITEM_END_OF_LIST
1074 break gatherlines
1075 }
1076 *flags |= LIST_ITEM_CONTAINS_BLOCK
1077
1078 // anything following an empty line is only part
1079 // of this item if it is indented 4 spaces
1080 // (regardless of the indentation of the beginning of the item)
1081 case containsBlankLine && indent < 4:
1082 *flags |= LIST_ITEM_END_OF_LIST
1083 break gatherlines
1084
1085 // a blank line means this should be parsed as a block
1086 case containsBlankLine:
1087 raw.WriteByte('\n')
1088 *flags |= LIST_ITEM_CONTAINS_BLOCK
1089 }
1090
1091 // if this line was preceeded by one or more blanks,
1092 // re-introduce the blank into the buffer
1093 if containsBlankLine {
1094 containsBlankLine = false
1095 raw.WriteByte('\n')
1096 }
1097
1098 // add the line into the working buffer without prefix
1099 raw.Write(data[line+indent : i])
1100
1101 line = i
1102 }
1103
1104 // render the contents of the list item
1105 rawBytes := raw.Bytes()
1106 var cooked bytes.Buffer
1107 if *flags&LIST_ITEM_CONTAINS_BLOCK != 0 {
1108 // intermediate render of block li
1109 if sublist > 0 {
1110 p.block(&cooked, rawBytes[:sublist])
1111 p.block(&cooked, rawBytes[sublist:])
1112 } else {
1113 p.block(&cooked, rawBytes)
1114 }
1115 } else {
1116 // intermediate render of inline li
1117 if sublist > 0 {
1118 p.inline(&cooked, rawBytes[:sublist])
1119 p.block(&cooked, rawBytes[sublist:])
1120 } else {
1121 p.inline(&cooked, rawBytes)
1122 }
1123 }
1124
1125 // render the actual list item
1126 cookedBytes := cooked.Bytes()
1127 parsedEnd := len(cookedBytes)
1128
1129 // strip trailing newlines
1130 for parsedEnd > 0 && cookedBytes[parsedEnd-1] == '\n' {
1131 parsedEnd--
1132 }
1133 p.r.ListItem(out, cookedBytes[:parsedEnd], *flags)
1134
1135 return line
1136}
1137
1138// render a single paragraph that has already been parsed out
1139func (p *parser) renderParagraph(out *bytes.Buffer, data []byte) {
1140 if len(data) == 0 {
1141 return
1142 }
1143
1144 // trim leading spaces
1145 beg := 0
1146 for data[beg] == ' ' {
1147 beg++
1148 }
1149
1150 // trim trailing newline
1151 end := len(data) - 1
1152
1153 // trim trailing spaces
1154 for end > beg && data[end-1] == ' ' {
1155 end--
1156 }
1157
1158 work := func() bool {
1159 p.inline(out, data[beg:end])
1160 return true
1161 }
1162 p.r.Paragraph(out, work)
1163}
1164
1165func (p *parser) paragraph(out *bytes.Buffer, data []byte) int {
1166 // prev: index of 1st char of previous line
1167 // line: index of 1st char of current line
1168 // i: index of cursor/end of current line
1169 var prev, line, i int
1170
1171 // keep going until we find something to mark the end of the paragraph
1172 for i < len(data) {
1173 // mark the beginning of the current line
1174 prev = line
1175 current := data[i:]
1176 line = i
1177
1178 // did we find a blank line marking the end of the paragraph?
1179 if n := p.isEmpty(current); n > 0 {
1180 p.renderParagraph(out, data[:i])
1181 return i + n
1182 }
1183
1184 // an underline under some text marks a header, so our paragraph ended on prev line
1185 if i > 0 {
1186 if level := p.isUnderlinedHeader(current); level > 0 {
1187 // render the paragraph
1188 p.renderParagraph(out, data[:prev])
1189
1190 // ignore leading and trailing whitespace
1191 eol := i - 1
1192 for prev < eol && data[prev] == ' ' {
1193 prev++
1194 }
1195 for eol > prev && data[eol-1] == ' ' {
1196 eol--
1197 }
1198
1199 // render the header
1200 // this ugly double closure avoids forcing variables onto the heap
1201 work := func(o *bytes.Buffer, pp *parser, d []byte) func() bool {
1202 return func() bool {
1203 pp.inline(o, d)
1204 return true
1205 }
1206 }(out, p, data[prev:eol])
1207 p.r.Header(out, work, level)
1208
1209 // find the end of the underline
1210 for data[i] != '\n' {
1211 i++
1212 }
1213 return i
1214 }
1215 }
1216
1217 // if the next line starts a block of HTML, then the paragraph ends here
1218 if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
1219 if data[i] == '<' && p.html(out, current, false) > 0 {
1220 // rewind to before the HTML block
1221 p.renderParagraph(out, data[:i])
1222 return i
1223 }
1224 }
1225
1226 // if there's a prefixed header or a horizontal rule after this, paragraph is over
1227 if p.isPrefixHeader(current) || p.isHRule(current) {
1228 p.renderParagraph(out, data[:i])
1229 return i
1230 }
1231
1232 // otherwise, scan to the beginning of the next line
1233 for data[i] != '\n' {
1234 i++
1235 }
1236 i++
1237 }
1238
1239 p.renderParagraph(out, data[:i])
1240 return i
1241}