all repos — grayfriday @ ec90dd0fc486d1ab6d263d794eb10b86dc1ba0b4

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