all repos — grayfriday @ 8e739c7f0d044677352d1530993879dc319b0cff

blackfriday fork with a few changes

markdown.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//
 12// Markdown parsing and processing
 13//
 14//
 15
 16// Blackfriday markdown processor.
 17//
 18// Translates plain text with simple formatting rules into HTML or LaTeX.
 19package blackfriday
 20
 21import (
 22	"bytes"
 23	"unicode/utf8"
 24)
 25
 26const VERSION = "1.1"
 27
 28// These are the supported markdown parsing extensions.
 29// OR these values together to select multiple extensions.
 30const (
 31	EXTENSION_NO_INTRA_EMPHASIS          = 1 << iota // ignore emphasis markers inside words
 32	EXTENSION_TABLES                                 // render tables
 33	EXTENSION_FENCED_CODE                            // render fenced code blocks
 34	EXTENSION_AUTOLINK                               // detect embedded URLs that are not explicitly marked
 35	EXTENSION_STRIKETHROUGH                          // strikethrough text using ~~test~~
 36	EXTENSION_LAX_HTML_BLOCKS                        // loosen up HTML block parsing rules
 37	EXTENSION_SPACE_HEADERS                          // be strict about prefix header rules
 38	EXTENSION_HARD_LINE_BREAK                        // translate newlines into line breaks
 39	EXTENSION_TAB_SIZE_EIGHT                         // expand tabs to eight spaces instead of four
 40	EXTENSION_FOOTNOTES                              // Pandoc-style footnotes
 41	EXTENSION_NO_EMPTY_LINE_BEFORE_BLOCK             // No need to insert an empty line to start a (code, quote, order list, unorder list)block
 42	EXTENSION_HEADER_IDS                             // specify header IDs  with {#id}
 43	EXTENSION_TITLEBLOCK                             // Titleblock ala pandoc
 44)
 45
 46// These are the possible flag values for the link renderer.
 47// Only a single one of these values will be used; they are not ORed together.
 48// These are mostly of interest if you are writing a new output format.
 49const (
 50	LINK_TYPE_NOT_AUTOLINK = iota
 51	LINK_TYPE_NORMAL
 52	LINK_TYPE_EMAIL
 53)
 54
 55// These are the possible flag values for the ListItem renderer.
 56// Multiple flag values may be ORed together.
 57// These are mostly of interest if you are writing a new output format.
 58const (
 59	LIST_TYPE_ORDERED = 1 << iota
 60	LIST_ITEM_CONTAINS_BLOCK
 61	LIST_ITEM_BEGINNING_OF_LIST
 62	LIST_ITEM_END_OF_LIST
 63)
 64
 65// These are the possible flag values for the table cell renderer.
 66// Only a single one of these values will be used; they are not ORed together.
 67// These are mostly of interest if you are writing a new output format.
 68const (
 69	TABLE_ALIGNMENT_LEFT = 1 << iota
 70	TABLE_ALIGNMENT_RIGHT
 71	TABLE_ALIGNMENT_CENTER = (TABLE_ALIGNMENT_LEFT | TABLE_ALIGNMENT_RIGHT)
 72)
 73
 74// The size of a tab stop.
 75const (
 76	TAB_SIZE_DEFAULT = 4
 77	TAB_SIZE_EIGHT   = 8
 78)
 79
 80// These are the tags that are recognized as HTML block tags.
 81// Any of these can be included in markdown text without special escaping.
 82var blockTags = map[string]bool{
 83	"p":          true,
 84	"dl":         true,
 85	"h1":         true,
 86	"h2":         true,
 87	"h3":         true,
 88	"h4":         true,
 89	"h5":         true,
 90	"h6":         true,
 91	"ol":         true,
 92	"ul":         true,
 93	"del":        true,
 94	"div":        true,
 95	"ins":        true,
 96	"pre":        true,
 97	"form":       true,
 98	"math":       true,
 99	"table":      true,
100	"iframe":     true,
101	"script":     true,
102	"fieldset":   true,
103	"noscript":   true,
104	"blockquote": true,
105
106	// HTML5
107	"video":      true,
108	"aside":      true,
109	"canvas":     true,
110	"figure":     true,
111	"footer":     true,
112	"header":     true,
113	"hgroup":     true,
114	"output":     true,
115	"article":    true,
116	"section":    true,
117	"progress":   true,
118	"figcaption": true,
119}
120
121// Renderer is the rendering interface.
122// This is mostly of interest if you are implementing a new rendering format.
123//
124// When a byte slice is provided, it contains the (rendered) contents of the
125// element.
126//
127// When a callback is provided instead, it will write the contents of the
128// respective element directly to the output buffer and return true on success.
129// If the callback returns false, the rendering function should reset the
130// output buffer as though it had never been called.
131//
132// Currently Html and Latex implementations are provided
133type Renderer interface {
134	// block-level callbacks
135	BlockCode(out *bytes.Buffer, text []byte, lang string)
136	BlockQuote(out *bytes.Buffer, text []byte)
137	BlockHtml(out *bytes.Buffer, text []byte)
138	Header(out *bytes.Buffer, text func() bool, level int, id string)
139	HRule(out *bytes.Buffer)
140	List(out *bytes.Buffer, text func() bool, flags int)
141	ListItem(out *bytes.Buffer, text []byte, flags int)
142	Paragraph(out *bytes.Buffer, text func() bool)
143	Table(out *bytes.Buffer, header []byte, body []byte, columnData []int)
144	TableRow(out *bytes.Buffer, text []byte)
145	TableHeaderCell(out *bytes.Buffer, text []byte, flags int)
146	TableCell(out *bytes.Buffer, text []byte, flags int)
147	Footnotes(out *bytes.Buffer, text func() bool)
148	FootnoteItem(out *bytes.Buffer, name, text []byte, flags int)
149	TitleBlock(out *bytes.Buffer, text []byte)
150
151	// Span-level callbacks
152	AutoLink(out *bytes.Buffer, link []byte, kind int)
153	CodeSpan(out *bytes.Buffer, text []byte)
154	DoubleEmphasis(out *bytes.Buffer, text []byte)
155	Emphasis(out *bytes.Buffer, text []byte)
156	Image(out *bytes.Buffer, link []byte, title []byte, alt []byte)
157	LineBreak(out *bytes.Buffer)
158	Link(out *bytes.Buffer, link []byte, title []byte, content []byte)
159	RawHtmlTag(out *bytes.Buffer, tag []byte)
160	TripleEmphasis(out *bytes.Buffer, text []byte)
161	StrikeThrough(out *bytes.Buffer, text []byte)
162	FootnoteRef(out *bytes.Buffer, ref []byte, id int)
163
164	// Low-level callbacks
165	Entity(out *bytes.Buffer, entity []byte)
166	NormalText(out *bytes.Buffer, text []byte)
167
168	// Header and footer
169	DocumentHeader(out *bytes.Buffer)
170	DocumentFooter(out *bytes.Buffer)
171
172	GetFlags() int
173}
174
175// Callback functions for inline parsing. One such function is defined
176// for each character that triggers a response when parsing inline data.
177type inlineParser func(p *parser, out *bytes.Buffer, data []byte, offset int) int
178
179// Parser holds runtime state used by the parser.
180// This is constructed by the Markdown function.
181type parser struct {
182	r              Renderer
183	refs           map[string]*reference
184	inlineCallback [256]inlineParser
185	flags          int
186	nesting        int
187	maxNesting     int
188	insideLink     bool
189
190	// Footnotes need to be ordered as well as available to quickly check for
191	// presence. If a ref is also a footnote, it's stored both in refs and here
192	// in notes. Slice is nil if footnotes not enabled.
193	notes []*reference
194}
195
196//
197//
198// Public interface
199//
200//
201
202// MarkdownBasic is a convenience function for simple rendering.
203// It processes markdown input with no extensions enabled.
204func MarkdownBasic(input []byte) []byte {
205	// set up the HTML renderer
206	htmlFlags := HTML_USE_XHTML
207	renderer := HtmlRenderer(htmlFlags, "", "")
208
209	// set up the parser
210	extensions := 0
211
212	return Markdown(input, renderer, extensions)
213}
214
215// Call Markdown with most useful extensions enabled
216// MarkdownCommon is a convenience function for simple rendering.
217// It processes markdown input with common extensions enabled, including:
218//
219// * Smartypants processing with smart fractions and LaTeX dashes
220//
221// * Intra-word emphasis suppression
222//
223// * Tables
224//
225// * Fenced code blocks
226//
227// * Autolinking
228//
229// * Strikethrough support
230//
231// * Strict header parsing
232//
233// * Custom Header IDs
234func MarkdownCommon(input []byte) []byte {
235	// set up the HTML renderer
236	htmlFlags := 0
237	htmlFlags |= HTML_USE_XHTML
238	htmlFlags |= HTML_USE_SMARTYPANTS
239	htmlFlags |= HTML_SMARTYPANTS_FRACTIONS
240	htmlFlags |= HTML_SMARTYPANTS_LATEX_DASHES
241	renderer := HtmlRenderer(htmlFlags, "", "")
242
243	// set up the parser
244	extensions := 0
245	extensions |= EXTENSION_NO_INTRA_EMPHASIS
246	extensions |= EXTENSION_TABLES
247	extensions |= EXTENSION_FENCED_CODE
248	extensions |= EXTENSION_AUTOLINK
249	extensions |= EXTENSION_STRIKETHROUGH
250	extensions |= EXTENSION_SPACE_HEADERS
251	extensions |= EXTENSION_HEADER_IDS
252
253	return Markdown(input, renderer, extensions)
254}
255
256// Markdown is the main rendering function.
257// It parses and renders a block of markdown-encoded text.
258// The supplied Renderer is used to format the output, and extensions dictates
259// which non-standard extensions are enabled.
260//
261// To use the supplied Html or LaTeX renderers, see HtmlRenderer and
262// LatexRenderer, respectively.
263func Markdown(input []byte, renderer Renderer, extensions int) []byte {
264	// no point in parsing if we can't render
265	if renderer == nil {
266		return nil
267	}
268
269	// fill in the render structure
270	p := new(parser)
271	p.r = renderer
272	p.flags = extensions
273	p.refs = make(map[string]*reference)
274	p.maxNesting = 16
275	p.insideLink = false
276
277	// register inline parsers
278	p.inlineCallback['*'] = emphasis
279	p.inlineCallback['_'] = emphasis
280	if extensions&EXTENSION_STRIKETHROUGH != 0 {
281		p.inlineCallback['~'] = emphasis
282	}
283	p.inlineCallback['`'] = codeSpan
284	p.inlineCallback['\n'] = lineBreak
285	p.inlineCallback['['] = link
286	p.inlineCallback['<'] = leftAngle
287	p.inlineCallback['\\'] = escape
288	p.inlineCallback['&'] = entity
289
290	if extensions&EXTENSION_AUTOLINK != 0 {
291		p.inlineCallback[':'] = autoLink
292	}
293
294	if extensions&EXTENSION_FOOTNOTES != 0 {
295		p.notes = make([]*reference, 0)
296	}
297
298	first := firstPass(p, input)
299	second := secondPass(p, first)
300	return second
301}
302
303// first pass:
304// - extract references
305// - expand tabs
306// - normalize newlines
307// - copy everything else
308// - add missing newlines before fenced code blocks
309func firstPass(p *parser, input []byte) []byte {
310	var out bytes.Buffer
311	tabSize := TAB_SIZE_DEFAULT
312	if p.flags&EXTENSION_TAB_SIZE_EIGHT != 0 {
313		tabSize = TAB_SIZE_EIGHT
314	}
315	beg, end := 0, 0
316	lastLineWasBlank := false
317	lastFencedCodeBlockEnd := 0
318	for beg < len(input) { // iterate over lines
319		if end = isReference(p, input[beg:], tabSize); end > 0 {
320			beg += end
321		} else { // skip to the next line
322			end = beg
323			for end < len(input) && input[end] != '\n' && input[end] != '\r' {
324				end++
325			}
326
327			if p.flags&EXTENSION_FENCED_CODE != 0 {
328				// when last line was none blank and a fenced code block comes after
329				if beg >= lastFencedCodeBlockEnd {
330					// tmp var so we don't modify beyond bounds of `input`
331					var tmp = make([]byte, len(input[beg:]), len(input[beg:])+1)
332					copy(tmp, input[beg:])
333					if i := p.fencedCode(&out, append(tmp, '\n'), false); i > 0 {
334						if !lastLineWasBlank {
335							out.WriteByte('\n') // need to inject additional linebreak
336						}
337						lastFencedCodeBlockEnd = beg + i
338					}
339				}
340				lastLineWasBlank = end == beg
341			}
342
343			// add the line body if present
344			if end > beg {
345				if end < lastFencedCodeBlockEnd { // Do not expand tabs while inside fenced code blocks.
346					out.Write(input[beg:end])
347				} else {
348					expandTabs(&out, input[beg:end], tabSize)
349				}
350			}
351			out.WriteByte('\n')
352
353			if end < len(input) && input[end] == '\r' {
354				end++
355			}
356			if end < len(input) && input[end] == '\n' {
357				end++
358			}
359
360			beg = end
361		}
362	}
363
364	// empty input?
365	if out.Len() == 0 {
366		out.WriteByte('\n')
367	}
368
369	return out.Bytes()
370}
371
372// second pass: actual rendering
373func secondPass(p *parser, input []byte) []byte {
374	var output bytes.Buffer
375
376	p.r.DocumentHeader(&output)
377	p.block(&output, input)
378
379	if p.flags&EXTENSION_FOOTNOTES != 0 && len(p.notes) > 0 {
380		p.r.Footnotes(&output, func() bool {
381			flags := LIST_ITEM_BEGINNING_OF_LIST
382			for _, ref := range p.notes {
383				var buf bytes.Buffer
384				if ref.hasBlock {
385					flags |= LIST_ITEM_CONTAINS_BLOCK
386					p.block(&buf, ref.title)
387				} else {
388					p.inline(&buf, ref.title)
389				}
390				p.r.FootnoteItem(&output, ref.link, buf.Bytes(), flags)
391				flags &^= LIST_ITEM_BEGINNING_OF_LIST | LIST_ITEM_CONTAINS_BLOCK
392			}
393
394			return true
395		})
396	}
397
398	p.r.DocumentFooter(&output)
399
400	if p.nesting != 0 {
401		panic("Nesting level did not end at zero")
402	}
403
404	return output.Bytes()
405}
406
407//
408// Link references
409//
410// This section implements support for references that (usually) appear
411// as footnotes in a document, and can be referenced anywhere in the document.
412// The basic format is:
413//
414//    [1]: http://www.google.com/ "Google"
415//    [2]: http://www.github.com/ "Github"
416//
417// Anywhere in the document, the reference can be linked by referring to its
418// label, i.e., 1 and 2 in this example, as in:
419//
420//    This library is hosted on [Github][2], a git hosting site.
421//
422// Actual footnotes as specified in Pandoc and supported by some other Markdown
423// libraries such as php-markdown are also taken care of. They look like this:
424//
425//    This sentence needs a bit of further explanation.[^note]
426//
427//    [^note]: This is the explanation.
428//
429// Footnotes should be placed at the end of the document in an ordered list.
430// Inline footnotes such as:
431//
432//    Inline footnotes^[Not supported.] also exist.
433//
434// are not yet supported.
435
436// References are parsed and stored in this struct.
437type reference struct {
438	link     []byte
439	title    []byte
440	noteId   int // 0 if not a footnote ref
441	hasBlock bool
442}
443
444// Check whether or not data starts with a reference link.
445// If so, it is parsed and stored in the list of references
446// (in the render struct).
447// Returns the number of bytes to skip to move past it,
448// or zero if the first line is not a reference.
449func isReference(p *parser, data []byte, tabSize int) int {
450	// up to 3 optional leading spaces
451	if len(data) < 4 {
452		return 0
453	}
454	i := 0
455	for i < 3 && data[i] == ' ' {
456		i++
457	}
458
459	noteId := 0
460
461	// id part: anything but a newline between brackets
462	if data[i] != '[' {
463		return 0
464	}
465	i++
466	if p.flags&EXTENSION_FOOTNOTES != 0 {
467		if data[i] == '^' {
468			// we can set it to anything here because the proper noteIds will
469			// be assigned later during the second pass. It just has to be != 0
470			noteId = 1
471			i++
472		}
473	}
474	idOffset := i
475	for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
476		i++
477	}
478	if i >= len(data) || data[i] != ']' {
479		return 0
480	}
481	idEnd := i
482
483	// spacer: colon (space | tab)* newline? (space | tab)*
484	i++
485	if i >= len(data) || data[i] != ':' {
486		return 0
487	}
488	i++
489	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
490		i++
491	}
492	if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
493		i++
494		if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
495			i++
496		}
497	}
498	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
499		i++
500	}
501	if i >= len(data) {
502		return 0
503	}
504
505	var (
506		linkOffset, linkEnd   int
507		titleOffset, titleEnd int
508		lineEnd               int
509		raw                   []byte
510		hasBlock              bool
511	)
512
513	if p.flags&EXTENSION_FOOTNOTES != 0 && noteId != 0 {
514		linkOffset, linkEnd, raw, hasBlock = scanFootnote(p, data, i, tabSize)
515		lineEnd = linkEnd
516	} else {
517		linkOffset, linkEnd, titleOffset, titleEnd, lineEnd = scanLinkRef(p, data, i)
518	}
519	if lineEnd == 0 {
520		return 0
521	}
522
523	// a valid ref has been found
524
525	ref := &reference{
526		noteId:   noteId,
527		hasBlock: hasBlock,
528	}
529
530	if noteId > 0 {
531		// reusing the link field for the id since footnotes don't have links
532		ref.link = data[idOffset:idEnd]
533		// if footnote, it's not really a title, it's the contained text
534		ref.title = raw
535	} else {
536		ref.link = data[linkOffset:linkEnd]
537		ref.title = data[titleOffset:titleEnd]
538	}
539
540	// id matches are case-insensitive
541	id := string(bytes.ToLower(data[idOffset:idEnd]))
542
543	p.refs[id] = ref
544
545	return lineEnd
546}
547
548func scanLinkRef(p *parser, data []byte, i int) (linkOffset, linkEnd, titleOffset, titleEnd, lineEnd int) {
549	// link: whitespace-free sequence, optionally between angle brackets
550	if data[i] == '<' {
551		i++
552	}
553	linkOffset = i
554	for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
555		i++
556	}
557	linkEnd = i
558	if data[linkOffset] == '<' && data[linkEnd-1] == '>' {
559		linkOffset++
560		linkEnd--
561	}
562
563	// optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
564	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
565		i++
566	}
567	if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
568		return
569	}
570
571	// compute end-of-line
572	if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
573		lineEnd = i
574	}
575	if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
576		lineEnd++
577	}
578
579	// optional (space|tab)* spacer after a newline
580	if lineEnd > 0 {
581		i = lineEnd + 1
582		for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
583			i++
584		}
585	}
586
587	// optional title: any non-newline sequence enclosed in '"() alone on its line
588	if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
589		i++
590		titleOffset = i
591
592		// look for EOL
593		for i < len(data) && data[i] != '\n' && data[i] != '\r' {
594			i++
595		}
596		if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
597			titleEnd = i + 1
598		} else {
599			titleEnd = i
600		}
601
602		// step back
603		i--
604		for i > titleOffset && (data[i] == ' ' || data[i] == '\t') {
605			i--
606		}
607		if i > titleOffset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
608			lineEnd = titleEnd
609			titleEnd = i
610		}
611	}
612
613	return
614}
615
616// The first bit of this logic is the same as (*parser).listItem, but the rest
617// is much simpler. This function simply finds the entire block and shifts it
618// over by one tab if it is indeed a block (just returns the line if it's not).
619// blockEnd is the end of the section in the input buffer, and contents is the
620// extracted text that was shifted over one tab. It will need to be rendered at
621// the end of the document.
622func scanFootnote(p *parser, data []byte, i, indentSize int) (blockStart, blockEnd int, contents []byte, hasBlock bool) {
623	if i == 0 || len(data) == 0 {
624		return
625	}
626
627	// skip leading whitespace on first line
628	for i < len(data) && data[i] == ' ' {
629		i++
630	}
631
632	blockStart = i
633
634	// find the end of the line
635	blockEnd = i
636	for i < len(data) && data[i-1] != '\n' {
637		i++
638	}
639
640	// get working buffer
641	var raw bytes.Buffer
642
643	// put the first line into the working buffer
644	raw.Write(data[blockEnd:i])
645	blockEnd = i
646
647	// process the following lines
648	containsBlankLine := false
649
650gatherLines:
651	for blockEnd < len(data) {
652		i++
653
654		// find the end of this line
655		for i < len(data) && data[i-1] != '\n' {
656			i++
657		}
658
659		// if it is an empty line, guess that it is part of this item
660		// and move on to the next line
661		if p.isEmpty(data[blockEnd:i]) > 0 {
662			containsBlankLine = true
663			blockEnd = i
664			continue
665		}
666
667		n := 0
668		if n = isIndented(data[blockEnd:i], indentSize); n == 0 {
669			// this is the end of the block.
670			// we don't want to include this last line in the index.
671			break gatherLines
672		}
673
674		// if there were blank lines before this one, insert a new one now
675		if containsBlankLine {
676			raw.WriteByte('\n')
677			containsBlankLine = false
678		}
679
680		// get rid of that first tab, write to buffer
681		raw.Write(data[blockEnd+n : i])
682		hasBlock = true
683
684		blockEnd = i
685	}
686
687	if data[blockEnd-1] != '\n' {
688		raw.WriteByte('\n')
689	}
690
691	contents = raw.Bytes()
692
693	return
694}
695
696//
697//
698// Miscellaneous helper functions
699//
700//
701
702// Test if a character is a punctuation symbol.
703// Taken from a private function in regexp in the stdlib.
704func ispunct(c byte) bool {
705	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
706		if c == r {
707			return true
708		}
709	}
710	return false
711}
712
713// Test if a character is a whitespace character.
714func isspace(c byte) bool {
715	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
716}
717
718// Test if a character is letter.
719func isletter(c byte) bool {
720	return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
721}
722
723// Test if a character is a letter or a digit.
724// TODO: check when this is looking for ASCII alnum and when it should use unicode
725func isalnum(c byte) bool {
726	return (c >= '0' && c <= '9') || isletter(c)
727}
728
729// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
730// always ends output with a newline
731func expandTabs(out *bytes.Buffer, line []byte, tabSize int) {
732	// first, check for common cases: no tabs, or only tabs at beginning of line
733	i, prefix := 0, 0
734	slowcase := false
735	for i = 0; i < len(line); i++ {
736		if line[i] == '\t' {
737			if prefix == i {
738				prefix++
739			} else {
740				slowcase = true
741				break
742			}
743		}
744	}
745
746	// no need to decode runes if all tabs are at the beginning of the line
747	if !slowcase {
748		for i = 0; i < prefix*tabSize; i++ {
749			out.WriteByte(' ')
750		}
751		out.Write(line[prefix:])
752		return
753	}
754
755	// the slow case: we need to count runes to figure out how
756	// many spaces to insert for each tab
757	column := 0
758	i = 0
759	for i < len(line) {
760		start := i
761		for i < len(line) && line[i] != '\t' {
762			_, size := utf8.DecodeRune(line[i:])
763			i += size
764			column++
765		}
766
767		if i > start {
768			out.Write(line[start:i])
769		}
770
771		if i >= len(line) {
772			break
773		}
774
775		for {
776			out.WriteByte(' ')
777			column++
778			if column%tabSize == 0 {
779				break
780			}
781		}
782
783		i++
784	}
785}
786
787// Find if a line counts as indented or not.
788// Returns number of characters the indent is (0 = not indented).
789func isIndented(data []byte, indentSize int) int {
790	if len(data) == 0 {
791		return 0
792	}
793	if data[0] == '\t' {
794		return 1
795	}
796	if len(data) < indentSize {
797		return 0
798	}
799	for i := 0; i < indentSize; i++ {
800		if data[i] != ' ' {
801			return 0
802		}
803	}
804	return indentSize
805}
806
807// Create a url-safe slug for fragments
808func slugify(in []byte) []byte {
809	if len(in) == 0 {
810		return in
811	}
812	out := make([]byte, 0, len(in))
813	sym := false
814
815	for _, ch := range in {
816		if isalnum(ch) {
817			sym = false
818			out = append(out, ch)
819		} else if sym {
820			continue
821		} else {
822			out = append(out, '-')
823			sym = true
824		}
825	}
826	var a, b int
827	var ch byte
828	for a, ch = range out {
829		if ch != '-' {
830			break
831		}
832	}
833	for b = len(out) - 1; b > 0; b-- {
834		if out[b] != '-' {
835			break
836		}
837	}
838	return out[a : b+1]
839}