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; i < len(data) && data[i] != '\n'; i++ {
462 if data[i] != ' ' && data[i] != '\t' {
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, false)
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, true)
775 size = i + 1
776 return
777}
778
779func (p *parser) tableRow(out *bytes.Buffer, data []byte, columns []int, header bool) {
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
810 if header {
811 p.r.TableHeaderCell(&rowWork, cellWork.Bytes(), columns[col])
812 } else {
813 p.r.TableCell(&rowWork, cellWork.Bytes(), columns[col])
814 }
815 }
816
817 // pad it out with empty columns to get the right number
818 for ; col < len(columns); col++ {
819 if header {
820 p.r.TableHeaderCell(&rowWork, nil, columns[col])
821 } else {
822 p.r.TableCell(&rowWork, nil, columns[col])
823 }
824 }
825
826 // silently ignore rows with too many cells
827
828 p.r.TableRow(out, rowWork.Bytes())
829}
830
831// returns blockquote prefix length
832func (p *parser) quotePrefix(data []byte) int {
833 i := 0
834 for i < 3 && data[i] == ' ' {
835 i++
836 }
837 if data[i] == '>' {
838 if data[i+1] == ' ' {
839 return i + 2
840 }
841 return i + 1
842 }
843 return 0
844}
845
846// parse a blockquote fragment
847func (p *parser) quote(out *bytes.Buffer, data []byte) int {
848 var raw bytes.Buffer
849 beg, end := 0, 0
850 for beg < len(data) {
851 end = beg
852 for data[end] != '\n' {
853 end++
854 }
855 end++
856
857 if pre := p.quotePrefix(data[beg:]); pre > 0 {
858 // skip the prefix
859 beg += pre
860 } else if p.isEmpty(data[beg:]) > 0 &&
861 (end >= len(data) ||
862 (p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0)) {
863 // blockquote ends with at least one blank line
864 // followed by something without a blockquote prefix
865 break
866 }
867
868 // this line is part of the blockquote
869 raw.Write(data[beg:end])
870 beg = end
871 }
872
873 var cooked bytes.Buffer
874 p.block(&cooked, raw.Bytes())
875 p.r.BlockQuote(out, cooked.Bytes())
876 return end
877}
878
879// returns prefix length for block code
880func (p *parser) codePrefix(data []byte) int {
881 if data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
882 return 4
883 }
884 return 0
885}
886
887func (p *parser) code(out *bytes.Buffer, data []byte) int {
888 var work bytes.Buffer
889
890 i := 0
891 for i < len(data) {
892 beg := i
893 for data[i] != '\n' {
894 i++
895 }
896 i++
897
898 blankline := p.isEmpty(data[beg:i]) > 0
899 if pre := p.codePrefix(data[beg:i]); pre > 0 {
900 beg += pre
901 } else if !blankline {
902 // non-empty, non-prefixed line breaks the pre
903 i = beg
904 break
905 }
906
907 // verbatim copy to the working buffeu
908 if blankline {
909 work.WriteByte('\n')
910 } else {
911 work.Write(data[beg:i])
912 }
913 }
914
915 // trim all the \n off the end of work
916 workbytes := work.Bytes()
917 eol := len(workbytes)
918 for eol > 0 && workbytes[eol-1] == '\n' {
919 eol--
920 }
921 if eol != len(workbytes) {
922 work.Truncate(eol)
923 }
924
925 work.WriteByte('\n')
926
927 p.r.BlockCode(out, work.Bytes(), "")
928
929 return i
930}
931
932// returns unordered list item prefix
933func (p *parser) uliPrefix(data []byte) int {
934 i := 0
935
936 // start with up to 3 spaces
937 for i < 3 && data[i] == ' ' {
938 i++
939 }
940
941 // need a *, +, or - followed by a space
942 if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
943 data[i+1] != ' ' {
944 return 0
945 }
946 return i + 2
947}
948
949// returns ordered list item prefix
950func (p *parser) oliPrefix(data []byte) int {
951 i := 0
952
953 // start with up to 3 spaces
954 for i < 3 && data[i] == ' ' {
955 i++
956 }
957
958 // count the digits
959 start := i
960 for data[i] >= '0' && data[i] <= '9' {
961 i++
962 }
963
964 // we need >= 1 digits followed by a dot and a space
965 if start == i || data[i] != '.' || data[i+1] != ' ' {
966 return 0
967 }
968 return i + 2
969}
970
971// parse ordered or unordered list block
972func (p *parser) list(out *bytes.Buffer, data []byte, flags int) int {
973 i := 0
974 flags |= LIST_ITEM_BEGINNING_OF_LIST
975 work := func() bool {
976 for i < len(data) {
977 skip := p.listItem(out, data[i:], &flags)
978 i += skip
979
980 if skip == 0 || flags&LIST_ITEM_END_OF_LIST != 0 {
981 break
982 }
983 flags &= ^LIST_ITEM_BEGINNING_OF_LIST
984 }
985 return true
986 }
987
988 p.r.List(out, work, flags)
989 return i
990}
991
992// Parse a single list item.
993// Assumes initial prefix is already removed if this is a sublist.
994func (p *parser) listItem(out *bytes.Buffer, data []byte, flags *int) int {
995 // keep track of the indentation of the first line
996 itemIndent := 0
997 for itemIndent < 3 && data[itemIndent] == ' ' {
998 itemIndent++
999 }
1000
1001 i := p.uliPrefix(data)
1002 if i == 0 {
1003 i = p.oliPrefix(data)
1004 }
1005 if i == 0 {
1006 return 0
1007 }
1008
1009 // skip leading whitespace on first line
1010 for data[i] == ' ' {
1011 i++
1012 }
1013
1014 // find the end of the line
1015 line := i
1016 for data[i-1] != '\n' {
1017 i++
1018 }
1019
1020 // get working buffer
1021 var raw bytes.Buffer
1022
1023 // put the first line into the working buffer
1024 raw.Write(data[line:i])
1025 line = i
1026
1027 // process the following lines
1028 containsBlankLine := false
1029 sublist := 0
1030
1031gatherlines:
1032 for line < len(data) {
1033 i++
1034
1035 // find the end of this line
1036 for data[i-1] != '\n' {
1037 i++
1038 }
1039
1040 // if it is an empty line, guess that it is part of this item
1041 // and move on to the next line
1042 if p.isEmpty(data[line:i]) > 0 {
1043 containsBlankLine = true
1044 line = i
1045 continue
1046 }
1047
1048 // calculate the indentation
1049 indent := 0
1050 for indent < 4 && line+indent < i && data[line+indent] == ' ' {
1051 indent++
1052 }
1053
1054 chunk := data[line+indent : i]
1055
1056 // evaluate how this line fits in
1057 switch {
1058 // is this a nested list item?
1059 case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) ||
1060 p.oliPrefix(chunk) > 0:
1061
1062 if containsBlankLine {
1063 *flags |= LIST_ITEM_CONTAINS_BLOCK
1064 }
1065
1066 // to be a nested list, it must be indented more
1067 // if not, it is the next item in the same list
1068 if indent <= itemIndent {
1069 break gatherlines
1070 }
1071
1072 // is this the first item in the the nested list?
1073 if sublist == 0 {
1074 sublist = raw.Len()
1075 }
1076
1077 // is this a nested prefix header?
1078 case p.isPrefixHeader(chunk):
1079 // if the header is not indented, it is not nested in the list
1080 // and thus ends the list
1081 if containsBlankLine && indent < 4 {
1082 *flags |= LIST_ITEM_END_OF_LIST
1083 break gatherlines
1084 }
1085 *flags |= LIST_ITEM_CONTAINS_BLOCK
1086
1087 // anything following an empty line is only part
1088 // of this item if it is indented 4 spaces
1089 // (regardless of the indentation of the beginning of the item)
1090 case containsBlankLine && indent < 4:
1091 *flags |= LIST_ITEM_END_OF_LIST
1092 break gatherlines
1093
1094 // a blank line means this should be parsed as a block
1095 case containsBlankLine:
1096 raw.WriteByte('\n')
1097 *flags |= LIST_ITEM_CONTAINS_BLOCK
1098 }
1099
1100 // if this line was preceeded by one or more blanks,
1101 // re-introduce the blank into the buffer
1102 if containsBlankLine {
1103 containsBlankLine = false
1104 raw.WriteByte('\n')
1105 }
1106
1107 // add the line into the working buffer without prefix
1108 raw.Write(data[line+indent : i])
1109
1110 line = i
1111 }
1112
1113 rawBytes := raw.Bytes()
1114
1115 // render the contents of the list item
1116 var cooked bytes.Buffer
1117 if *flags&LIST_ITEM_CONTAINS_BLOCK != 0 {
1118 // intermediate render of block li
1119 if sublist > 0 {
1120 p.block(&cooked, rawBytes[:sublist])
1121 p.block(&cooked, rawBytes[sublist:])
1122 } else {
1123 p.block(&cooked, rawBytes)
1124 }
1125 } else {
1126 // intermediate render of inline li
1127 if sublist > 0 {
1128 p.inline(&cooked, rawBytes[:sublist])
1129 p.block(&cooked, rawBytes[sublist:])
1130 } else {
1131 p.inline(&cooked, rawBytes)
1132 }
1133 }
1134
1135 // render the actual list item
1136 cookedBytes := cooked.Bytes()
1137 parsedEnd := len(cookedBytes)
1138
1139 // strip trailing newlines
1140 for parsedEnd > 0 && cookedBytes[parsedEnd-1] == '\n' {
1141 parsedEnd--
1142 }
1143 p.r.ListItem(out, cookedBytes[:parsedEnd], *flags)
1144
1145 return line
1146}
1147
1148// render a single paragraph that has already been parsed out
1149func (p *parser) renderParagraph(out *bytes.Buffer, data []byte) {
1150 if len(data) == 0 {
1151 return
1152 }
1153
1154 // trim leading spaces
1155 beg := 0
1156 for data[beg] == ' ' {
1157 beg++
1158 }
1159
1160 // trim trailing newline
1161 end := len(data) - 1
1162
1163 // trim trailing spaces
1164 for end > beg && data[end-1] == ' ' {
1165 end--
1166 }
1167
1168 work := func() bool {
1169 p.inline(out, data[beg:end])
1170 return true
1171 }
1172 p.r.Paragraph(out, work)
1173}
1174
1175func (p *parser) paragraph(out *bytes.Buffer, data []byte) int {
1176 // prev: index of 1st char of previous line
1177 // line: index of 1st char of current line
1178 // i: index of cursor/end of current line
1179 var prev, line, i int
1180
1181 // keep going until we find something to mark the end of the paragraph
1182 for i < len(data) {
1183 // mark the beginning of the current line
1184 prev = line
1185 current := data[i:]
1186 line = i
1187
1188 // did we find a blank line marking the end of the paragraph?
1189 if n := p.isEmpty(current); n > 0 {
1190 p.renderParagraph(out, data[:i])
1191 return i + n
1192 }
1193
1194 // an underline under some text marks a header, so our paragraph ended on prev line
1195 if i > 0 {
1196 if level := p.isUnderlinedHeader(current); level > 0 {
1197 // render the paragraph
1198 p.renderParagraph(out, data[:prev])
1199
1200 // ignore leading and trailing whitespace
1201 eol := i - 1
1202 for prev < eol && data[prev] == ' ' {
1203 prev++
1204 }
1205 for eol > prev && data[eol-1] == ' ' {
1206 eol--
1207 }
1208
1209 // render the header
1210 // this ugly double closure avoids forcing variables onto the heap
1211 work := func(o *bytes.Buffer, pp *parser, d []byte) func() bool {
1212 return func() bool {
1213 pp.inline(o, d)
1214 return true
1215 }
1216 }(out, p, data[prev:eol])
1217 p.r.Header(out, work, level)
1218
1219 // find the end of the underline
1220 for data[i] != '\n' {
1221 i++
1222 }
1223 return i
1224 }
1225 }
1226
1227 // if the next line starts a block of HTML, then the paragraph ends here
1228 if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
1229 if data[i] == '<' && p.html(out, current, false) > 0 {
1230 // rewind to before the HTML block
1231 p.renderParagraph(out, data[:i])
1232 return i
1233 }
1234 }
1235
1236 // if there's a prefixed header or a horizontal rule after this, paragraph is over
1237 if p.isPrefixHeader(current) || p.isHRule(current) {
1238 p.renderParagraph(out, data[:i])
1239 return i
1240 }
1241
1242 // if there's a list after this, paragraph is over
1243 if p.flags&EXTENSION_NO_EMPTY_LINE_BEFORE_BLOCK != 0 {
1244 if p.uliPrefix(current) != 0 ||
1245 p.oliPrefix(current) != 0 ||
1246 p.quotePrefix(current) != 0 ||
1247 p.codePrefix(current) != 0 {
1248 p.renderParagraph(out, data[:i])
1249 return i
1250 }
1251 }
1252
1253 // otherwise, scan to the beginning of the next line
1254 for data[i] != '\n' {
1255 i++
1256 }
1257 i++
1258 }
1259
1260 p.renderParagraph(out, data[:i])
1261 return i
1262}