all repos — grayfriday @ 9c9c590f7ec7f639ca6d885ecb0659f9f61fa7fc

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	"strings"
 24	"unicode/utf8"
 25)
 26
 27const VERSION = "1.4"
 28
 29// These are the supported markdown parsing extensions.
 30// OR these values together to select multiple extensions.
 31const (
 32	EXTENSION_NO_INTRA_EMPHASIS          = 1 << iota // ignore emphasis markers inside words
 33	EXTENSION_TABLES                                 // render tables
 34	EXTENSION_FENCED_CODE                            // render fenced code blocks
 35	EXTENSION_AUTOLINK                               // detect embedded URLs that are not explicitly marked
 36	EXTENSION_STRIKETHROUGH                          // strikethrough text using ~~test~~
 37	EXTENSION_LAX_HTML_BLOCKS                        // loosen up HTML block parsing rules
 38	EXTENSION_SPACE_HEADERS                          // be strict about prefix header rules
 39	EXTENSION_HARD_LINE_BREAK                        // translate newlines into line breaks
 40	EXTENSION_TAB_SIZE_EIGHT                         // expand tabs to eight spaces instead of four
 41	EXTENSION_FOOTNOTES                              // Pandoc-style footnotes
 42	EXTENSION_NO_EMPTY_LINE_BEFORE_BLOCK             // No need to insert an empty line to start a (code, quote, ordered list, unordered list) block
 43	EXTENSION_HEADER_IDS                             // specify header IDs  with {#id}
 44	EXTENSION_TITLEBLOCK                             // Titleblock ala pandoc
 45	EXTENSION_AUTO_HEADER_IDS                        // Create the header ID from the text
 46	EXTENSION_BACKSLASH_LINE_BREAK                   // translate trailing backslashes into line breaks
 47	EXTENSION_DEFINITION_LISTS                       // render definition lists
 48
 49	commonHtmlFlags = 0 |
 50		HTML_USE_XHTML |
 51		HTML_USE_SMARTYPANTS |
 52		HTML_SMARTYPANTS_FRACTIONS |
 53		HTML_SMARTYPANTS_DASHES |
 54		HTML_SMARTYPANTS_LATEX_DASHES
 55
 56	commonExtensions = 0 |
 57		EXTENSION_NO_INTRA_EMPHASIS |
 58		EXTENSION_TABLES |
 59		EXTENSION_FENCED_CODE |
 60		EXTENSION_AUTOLINK |
 61		EXTENSION_STRIKETHROUGH |
 62		EXTENSION_SPACE_HEADERS |
 63		EXTENSION_HEADER_IDS |
 64		EXTENSION_BACKSLASH_LINE_BREAK |
 65		EXTENSION_DEFINITION_LISTS
 66)
 67
 68// These are the possible flag values for the link renderer.
 69// Only a single one of these values will be used; they are not ORed together.
 70// These are mostly of interest if you are writing a new output format.
 71const (
 72	LINK_TYPE_NOT_AUTOLINK = iota
 73	LINK_TYPE_NORMAL
 74	LINK_TYPE_EMAIL
 75)
 76
 77// These are the possible flag values for the ListItem renderer.
 78// Multiple flag values may be ORed together.
 79// These are mostly of interest if you are writing a new output format.
 80const (
 81	LIST_TYPE_ORDERED = 1 << iota
 82	LIST_TYPE_DEFINITION
 83	LIST_TYPE_TERM
 84	LIST_ITEM_CONTAINS_BLOCK
 85	LIST_ITEM_BEGINNING_OF_LIST
 86	LIST_ITEM_END_OF_LIST
 87)
 88
 89// These are the possible flag values for the table cell renderer.
 90// Only a single one of these values will be used; they are not ORed together.
 91// These are mostly of interest if you are writing a new output format.
 92const (
 93	TABLE_ALIGNMENT_LEFT = 1 << iota
 94	TABLE_ALIGNMENT_RIGHT
 95	TABLE_ALIGNMENT_CENTER = (TABLE_ALIGNMENT_LEFT | TABLE_ALIGNMENT_RIGHT)
 96)
 97
 98// The size of a tab stop.
 99const (
100	TAB_SIZE_DEFAULT = 4
101	TAB_SIZE_EIGHT   = 8
102)
103
104// These are the tags that are recognized as HTML block tags.
105// Any of these can be included in markdown text without special escaping.
106var blockTags = map[string]bool{
107	"p":          true,
108	"dl":         true,
109	"h1":         true,
110	"h2":         true,
111	"h3":         true,
112	"h4":         true,
113	"h5":         true,
114	"h6":         true,
115	"ol":         true,
116	"ul":         true,
117	"del":        true,
118	"div":        true,
119	"ins":        true,
120	"pre":        true,
121	"form":       true,
122	"math":       true,
123	"table":      true,
124	"iframe":     true,
125	"script":     true,
126	"fieldset":   true,
127	"noscript":   true,
128	"blockquote": true,
129
130	// HTML5
131	"video":      true,
132	"aside":      true,
133	"canvas":     true,
134	"figure":     true,
135	"footer":     true,
136	"header":     true,
137	"hgroup":     true,
138	"output":     true,
139	"article":    true,
140	"section":    true,
141	"progress":   true,
142	"figcaption": true,
143}
144
145// Renderer is the rendering interface.
146// This is mostly of interest if you are implementing a new rendering format.
147//
148// When a byte slice is provided, it contains the (rendered) contents of the
149// element.
150//
151// When a callback is provided instead, it will write the contents of the
152// respective element directly to the output buffer and return true on success.
153// If the callback returns false, the rendering function should reset the
154// output buffer as though it had never been called.
155//
156// Currently Html and Latex implementations are provided
157type Renderer interface {
158	// block-level callbacks
159	BlockCode(out *bytes.Buffer, text []byte, lang string)
160	BlockQuote(out *bytes.Buffer, text []byte)
161	BlockHtml(out *bytes.Buffer, text []byte)
162	Header(out *bytes.Buffer, text func() bool, level int, id string)
163	HRule(out *bytes.Buffer)
164	List(out *bytes.Buffer, text func() bool, flags int)
165	ListItem(out *bytes.Buffer, text []byte, flags int)
166	Paragraph(out *bytes.Buffer, text func() bool)
167	Table(out *bytes.Buffer, header []byte, body []byte, columnData []int)
168	TableRow(out *bytes.Buffer, text []byte)
169	TableHeaderCell(out *bytes.Buffer, text []byte, flags int)
170	TableCell(out *bytes.Buffer, text []byte, flags int)
171	Footnotes(out *bytes.Buffer, text func() bool)
172	FootnoteItem(out *bytes.Buffer, name, text []byte, flags int)
173	TitleBlock(out *bytes.Buffer, text []byte)
174
175	// Span-level callbacks
176	AutoLink(out *bytes.Buffer, link []byte, kind int)
177	CodeSpan(out *bytes.Buffer, text []byte)
178	DoubleEmphasis(out *bytes.Buffer, text []byte)
179	Emphasis(out *bytes.Buffer, text []byte)
180	Image(out *bytes.Buffer, link []byte, title []byte, alt []byte)
181	LineBreak(out *bytes.Buffer)
182	Link(out *bytes.Buffer, link []byte, title []byte, content []byte)
183	RawHtmlTag(out *bytes.Buffer, tag []byte)
184	TripleEmphasis(out *bytes.Buffer, text []byte)
185	StrikeThrough(out *bytes.Buffer, text []byte)
186	FootnoteRef(out *bytes.Buffer, ref []byte, id int)
187
188	// Low-level callbacks
189	Entity(out *bytes.Buffer, entity []byte)
190	NormalText(out *bytes.Buffer, text []byte)
191
192	// Header and footer
193	DocumentHeader(out *bytes.Buffer)
194	DocumentFooter(out *bytes.Buffer)
195
196	GetFlags() int
197}
198
199// Callback functions for inline parsing. One such function is defined
200// for each character that triggers a response when parsing inline data.
201type inlineParser func(p *parser, out *bytes.Buffer, data []byte, offset int) int
202
203// Parser holds runtime state used by the parser.
204// This is constructed by the Markdown function.
205type parser struct {
206	r              Renderer
207	refOverride    ReferenceOverrideFunc
208	refs           map[string]*reference
209	inlineCallback [256]inlineParser
210	flags          int
211	nesting        int
212	maxNesting     int
213	insideLink     bool
214
215	// Footnotes need to be ordered as well as available to quickly check for
216	// presence. If a ref is also a footnote, it's stored both in refs and here
217	// in notes. Slice is nil if footnotes not enabled.
218	notes []*reference
219}
220
221func (p *parser) getRef(refid string) (ref *reference, found bool) {
222	if p.refOverride != nil {
223		r, overridden := p.refOverride(refid)
224		if overridden {
225			if r == nil {
226				return nil, false
227			}
228			return &reference{
229				link:     []byte(r.Link),
230				title:    []byte(r.Title),
231				noteId:   0,
232				hasBlock: false,
233				text:     []byte(r.Text)}, true
234		}
235	}
236	// refs are case insensitive
237	ref, found = p.refs[strings.ToLower(refid)]
238	return ref, found
239}
240
241//
242//
243// Public interface
244//
245//
246
247// Reference represents the details of a link.
248// See the documentation in Options for more details on use-case.
249type Reference struct {
250	// Link is usually the URL the reference points to.
251	Link string
252	// Title is the alternate text describing the link in more detail.
253	Title string
254	// Text is the optional text to override the ref with if the syntax used was
255	// [refid][]
256	Text string
257}
258
259// ReferenceOverrideFunc is expected to be called with a reference string and
260// return either a valid Reference type that the reference string maps to or
261// nil. If overridden is false, the default reference logic will be executed.
262// See the documentation in Options for more details on use-case.
263type ReferenceOverrideFunc func(reference string) (ref *Reference, overridden bool)
264
265// Options represents configurable overrides and callbacks (in addition to the
266// extension flag set) for configuring a Markdown parse.
267type Options struct {
268	// Extensions is a flag set of bit-wise ORed extension bits. See the
269	// EXTENSION_* flags defined in this package.
270	Extensions int
271
272	// ReferenceOverride is an optional function callback that is called every
273	// time a reference is resolved.
274	//
275	// In Markdown, the link reference syntax can be made to resolve a link to
276	// a reference instead of an inline URL, in one of the following ways:
277	//
278	//  * [link text][refid]
279	//  * [refid][]
280	//
281	// Usually, the refid is defined at the bottom of the Markdown document. If
282	// this override function is provided, the refid is passed to the override
283	// function first, before consulting the defined refids at the bottom. If
284	// the override function indicates an override did not occur, the refids at
285	// the bottom will be used to fill in the link details.
286	ReferenceOverride ReferenceOverrideFunc
287}
288
289// MarkdownBasic is a convenience function for simple rendering.
290// It processes markdown input with no extensions enabled.
291func MarkdownBasic(input []byte) []byte {
292	// set up the HTML renderer
293	htmlFlags := HTML_USE_XHTML
294	renderer := HtmlRenderer(htmlFlags, "", "")
295
296	// set up the parser
297	return MarkdownOptions(input, renderer, Options{Extensions: 0})
298}
299
300// Call Markdown with most useful extensions enabled
301// MarkdownCommon is a convenience function for simple rendering.
302// It processes markdown input with common extensions enabled, including:
303//
304// * Smartypants processing with smart fractions and LaTeX dashes
305//
306// * Intra-word emphasis suppression
307//
308// * Tables
309//
310// * Fenced code blocks
311//
312// * Autolinking
313//
314// * Strikethrough support
315//
316// * Strict header parsing
317//
318// * Custom Header IDs
319func MarkdownCommon(input []byte) []byte {
320	// set up the HTML renderer
321	renderer := HtmlRenderer(commonHtmlFlags, "", "")
322	return MarkdownOptions(input, renderer, Options{
323		Extensions: commonExtensions})
324}
325
326// Markdown is the main rendering function.
327// It parses and renders a block of markdown-encoded text.
328// The supplied Renderer is used to format the output, and extensions dictates
329// which non-standard extensions are enabled.
330//
331// To use the supplied Html or LaTeX renderers, see HtmlRenderer and
332// LatexRenderer, respectively.
333func Markdown(input []byte, renderer Renderer, extensions int) []byte {
334	return MarkdownOptions(input, renderer, Options{
335		Extensions: extensions})
336}
337
338// MarkdownOptions is just like Markdown but takes additional options through
339// the Options struct.
340func MarkdownOptions(input []byte, renderer Renderer, opts Options) []byte {
341	// no point in parsing if we can't render
342	if renderer == nil {
343		return nil
344	}
345
346	extensions := opts.Extensions
347
348	// fill in the render structure
349	p := new(parser)
350	p.r = renderer
351	p.flags = extensions
352	p.refOverride = opts.ReferenceOverride
353	p.refs = make(map[string]*reference)
354	p.maxNesting = 16
355	p.insideLink = false
356
357	// register inline parsers
358	p.inlineCallback['*'] = emphasis
359	p.inlineCallback['_'] = emphasis
360	if extensions&EXTENSION_STRIKETHROUGH != 0 {
361		p.inlineCallback['~'] = emphasis
362	}
363	p.inlineCallback['`'] = codeSpan
364	p.inlineCallback['\n'] = lineBreak
365	p.inlineCallback['['] = link
366	p.inlineCallback['<'] = leftAngle
367	p.inlineCallback['\\'] = escape
368	p.inlineCallback['&'] = entity
369
370	if extensions&EXTENSION_AUTOLINK != 0 {
371		p.inlineCallback[':'] = autoLink
372	}
373
374	if extensions&EXTENSION_FOOTNOTES != 0 {
375		p.notes = make([]*reference, 0)
376	}
377
378	first := firstPass(p, input)
379	second := secondPass(p, first)
380	return second
381}
382
383// first pass:
384// - extract references
385// - expand tabs
386// - normalize newlines
387// - copy everything else
388func firstPass(p *parser, input []byte) []byte {
389	var out bytes.Buffer
390	tabSize := TAB_SIZE_DEFAULT
391	if p.flags&EXTENSION_TAB_SIZE_EIGHT != 0 {
392		tabSize = TAB_SIZE_EIGHT
393	}
394	beg, end := 0, 0
395	lastFencedCodeBlockEnd := 0
396	for beg < len(input) { // iterate over lines
397		if end = isReference(p, input[beg:], tabSize); end > 0 {
398			beg += end
399		} else { // skip to the next line
400			end = beg
401			for end < len(input) && input[end] != '\n' && input[end] != '\r' {
402				end++
403			}
404
405			if p.flags&EXTENSION_FENCED_CODE != 0 {
406				// track fenced code block boundaries to suppress tab expansion
407				// inside them:
408				if beg >= lastFencedCodeBlockEnd {
409					if i := p.fencedCode(&out, input[beg:], false); i > 0 {
410						lastFencedCodeBlockEnd = beg + i
411					}
412				}
413			}
414
415			// add the line body if present
416			if end > beg {
417				if end < lastFencedCodeBlockEnd { // Do not expand tabs while inside fenced code blocks.
418					out.Write(input[beg:end])
419				} else {
420					expandTabs(&out, input[beg:end], tabSize)
421				}
422			}
423			out.WriteByte('\n')
424
425			if end < len(input) && input[end] == '\r' {
426				end++
427			}
428			if end < len(input) && input[end] == '\n' {
429				end++
430			}
431
432			beg = end
433		}
434	}
435
436	// empty input?
437	if out.Len() == 0 {
438		out.WriteByte('\n')
439	}
440
441	return out.Bytes()
442}
443
444// second pass: actual rendering
445func secondPass(p *parser, input []byte) []byte {
446	var output bytes.Buffer
447
448	p.r.DocumentHeader(&output)
449	p.block(&output, input)
450
451	if p.flags&EXTENSION_FOOTNOTES != 0 && len(p.notes) > 0 {
452		p.r.Footnotes(&output, func() bool {
453			flags := LIST_ITEM_BEGINNING_OF_LIST
454			for _, ref := range p.notes {
455				var buf bytes.Buffer
456				if ref.hasBlock {
457					flags |= LIST_ITEM_CONTAINS_BLOCK
458					p.block(&buf, ref.title)
459				} else {
460					p.inline(&buf, ref.title)
461				}
462				p.r.FootnoteItem(&output, ref.link, buf.Bytes(), flags)
463				flags &^= LIST_ITEM_BEGINNING_OF_LIST | LIST_ITEM_CONTAINS_BLOCK
464			}
465
466			return true
467		})
468	}
469
470	p.r.DocumentFooter(&output)
471
472	if p.nesting != 0 {
473		panic("Nesting level did not end at zero")
474	}
475
476	return output.Bytes()
477}
478
479//
480// Link references
481//
482// This section implements support for references that (usually) appear
483// as footnotes in a document, and can be referenced anywhere in the document.
484// The basic format is:
485//
486//    [1]: http://www.google.com/ "Google"
487//    [2]: http://www.github.com/ "Github"
488//
489// Anywhere in the document, the reference can be linked by referring to its
490// label, i.e., 1 and 2 in this example, as in:
491//
492//    This library is hosted on [Github][2], a git hosting site.
493//
494// Actual footnotes as specified in Pandoc and supported by some other Markdown
495// libraries such as php-markdown are also taken care of. They look like this:
496//
497//    This sentence needs a bit of further explanation.[^note]
498//
499//    [^note]: This is the explanation.
500//
501// Footnotes should be placed at the end of the document in an ordered list.
502// Inline footnotes such as:
503//
504//    Inline footnotes^[Not supported.] also exist.
505//
506// are not yet supported.
507
508// References are parsed and stored in this struct.
509type reference struct {
510	link     []byte
511	title    []byte
512	noteId   int // 0 if not a footnote ref
513	hasBlock bool
514	text     []byte
515}
516
517// Check whether or not data starts with a reference link.
518// If so, it is parsed and stored in the list of references
519// (in the render struct).
520// Returns the number of bytes to skip to move past it,
521// or zero if the first line is not a reference.
522func isReference(p *parser, data []byte, tabSize int) int {
523	// up to 3 optional leading spaces
524	if len(data) < 4 {
525		return 0
526	}
527	i := 0
528	for i < 3 && data[i] == ' ' {
529		i++
530	}
531
532	noteId := 0
533
534	// id part: anything but a newline between brackets
535	if data[i] != '[' {
536		return 0
537	}
538	i++
539	if p.flags&EXTENSION_FOOTNOTES != 0 {
540		if i < len(data) && data[i] == '^' {
541			// we can set it to anything here because the proper noteIds will
542			// be assigned later during the second pass. It just has to be != 0
543			noteId = 1
544			i++
545		}
546	}
547	idOffset := i
548	for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
549		i++
550	}
551	if i >= len(data) || data[i] != ']' {
552		return 0
553	}
554	idEnd := i
555
556	// spacer: colon (space | tab)* newline? (space | tab)*
557	i++
558	if i >= len(data) || data[i] != ':' {
559		return 0
560	}
561	i++
562	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
563		i++
564	}
565	if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
566		i++
567		if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
568			i++
569		}
570	}
571	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
572		i++
573	}
574	if i >= len(data) {
575		return 0
576	}
577
578	var (
579		linkOffset, linkEnd   int
580		titleOffset, titleEnd int
581		lineEnd               int
582		raw                   []byte
583		hasBlock              bool
584	)
585
586	if p.flags&EXTENSION_FOOTNOTES != 0 && noteId != 0 {
587		linkOffset, linkEnd, raw, hasBlock = scanFootnote(p, data, i, tabSize)
588		lineEnd = linkEnd
589	} else {
590		linkOffset, linkEnd, titleOffset, titleEnd, lineEnd = scanLinkRef(p, data, i)
591	}
592	if lineEnd == 0 {
593		return 0
594	}
595
596	// a valid ref has been found
597
598	ref := &reference{
599		noteId:   noteId,
600		hasBlock: hasBlock,
601	}
602
603	if noteId > 0 {
604		// reusing the link field for the id since footnotes don't have links
605		ref.link = data[idOffset:idEnd]
606		// if footnote, it's not really a title, it's the contained text
607		ref.title = raw
608	} else {
609		ref.link = data[linkOffset:linkEnd]
610		ref.title = data[titleOffset:titleEnd]
611	}
612
613	// id matches are case-insensitive
614	id := string(bytes.ToLower(data[idOffset:idEnd]))
615
616	p.refs[id] = ref
617
618	return lineEnd
619}
620
621func scanLinkRef(p *parser, data []byte, i int) (linkOffset, linkEnd, titleOffset, titleEnd, lineEnd int) {
622	// link: whitespace-free sequence, optionally between angle brackets
623	if data[i] == '<' {
624		i++
625	}
626	linkOffset = i
627	for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
628		i++
629	}
630	if i == len(data) {
631		return
632	}
633	linkEnd = i
634	if data[linkOffset] == '<' && data[linkEnd-1] == '>' {
635		linkOffset++
636		linkEnd--
637	}
638
639	// optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
640	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
641		i++
642	}
643	if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
644		return
645	}
646
647	// compute end-of-line
648	if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
649		lineEnd = i
650	}
651	if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
652		lineEnd++
653	}
654
655	// optional (space|tab)* spacer after a newline
656	if lineEnd > 0 {
657		i = lineEnd + 1
658		for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
659			i++
660		}
661	}
662
663	// optional title: any non-newline sequence enclosed in '"() alone on its line
664	if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
665		i++
666		titleOffset = i
667
668		// look for EOL
669		for i < len(data) && data[i] != '\n' && data[i] != '\r' {
670			i++
671		}
672		if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
673			titleEnd = i + 1
674		} else {
675			titleEnd = i
676		}
677
678		// step back
679		i--
680		for i > titleOffset && (data[i] == ' ' || data[i] == '\t') {
681			i--
682		}
683		if i > titleOffset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
684			lineEnd = titleEnd
685			titleEnd = i
686		}
687	}
688
689	return
690}
691
692// The first bit of this logic is the same as (*parser).listItem, but the rest
693// is much simpler. This function simply finds the entire block and shifts it
694// over by one tab if it is indeed a block (just returns the line if it's not).
695// blockEnd is the end of the section in the input buffer, and contents is the
696// extracted text that was shifted over one tab. It will need to be rendered at
697// the end of the document.
698func scanFootnote(p *parser, data []byte, i, indentSize int) (blockStart, blockEnd int, contents []byte, hasBlock bool) {
699	if i == 0 || len(data) == 0 {
700		return
701	}
702
703	// skip leading whitespace on first line
704	for i < len(data) && data[i] == ' ' {
705		i++
706	}
707
708	blockStart = i
709
710	// find the end of the line
711	blockEnd = i
712	for i < len(data) && data[i-1] != '\n' {
713		i++
714	}
715
716	// get working buffer
717	var raw bytes.Buffer
718
719	// put the first line into the working buffer
720	raw.Write(data[blockEnd:i])
721	blockEnd = i
722
723	// process the following lines
724	containsBlankLine := false
725
726gatherLines:
727	for blockEnd < len(data) {
728		i++
729
730		// find the end of this line
731		for i < len(data) && data[i-1] != '\n' {
732			i++
733		}
734
735		// if it is an empty line, guess that it is part of this item
736		// and move on to the next line
737		if p.isEmpty(data[blockEnd:i]) > 0 {
738			containsBlankLine = true
739			blockEnd = i
740			continue
741		}
742
743		n := 0
744		if n = isIndented(data[blockEnd:i], indentSize); n == 0 {
745			// this is the end of the block.
746			// we don't want to include this last line in the index.
747			break gatherLines
748		}
749
750		// if there were blank lines before this one, insert a new one now
751		if containsBlankLine {
752			raw.WriteByte('\n')
753			containsBlankLine = false
754		}
755
756		// get rid of that first tab, write to buffer
757		raw.Write(data[blockEnd+n : i])
758		hasBlock = true
759
760		blockEnd = i
761	}
762
763	if data[blockEnd-1] != '\n' {
764		raw.WriteByte('\n')
765	}
766
767	contents = raw.Bytes()
768
769	return
770}
771
772//
773//
774// Miscellaneous helper functions
775//
776//
777
778// Test if a character is a punctuation symbol.
779// Taken from a private function in regexp in the stdlib.
780func ispunct(c byte) bool {
781	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
782		if c == r {
783			return true
784		}
785	}
786	return false
787}
788
789// Test if a character is a whitespace character.
790func isspace(c byte) bool {
791	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
792}
793
794// Test if a character is letter.
795func isletter(c byte) bool {
796	return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
797}
798
799// Test if a character is a letter or a digit.
800// TODO: check when this is looking for ASCII alnum and when it should use unicode
801func isalnum(c byte) bool {
802	return (c >= '0' && c <= '9') || isletter(c)
803}
804
805// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
806// always ends output with a newline
807func expandTabs(out *bytes.Buffer, line []byte, tabSize int) {
808	// first, check for common cases: no tabs, or only tabs at beginning of line
809	i, prefix := 0, 0
810	slowcase := false
811	for i = 0; i < len(line); i++ {
812		if line[i] == '\t' {
813			if prefix == i {
814				prefix++
815			} else {
816				slowcase = true
817				break
818			}
819		}
820	}
821
822	// no need to decode runes if all tabs are at the beginning of the line
823	if !slowcase {
824		for i = 0; i < prefix*tabSize; i++ {
825			out.WriteByte(' ')
826		}
827		out.Write(line[prefix:])
828		return
829	}
830
831	// the slow case: we need to count runes to figure out how
832	// many spaces to insert for each tab
833	column := 0
834	i = 0
835	for i < len(line) {
836		start := i
837		for i < len(line) && line[i] != '\t' {
838			_, size := utf8.DecodeRune(line[i:])
839			i += size
840			column++
841		}
842
843		if i > start {
844			out.Write(line[start:i])
845		}
846
847		if i >= len(line) {
848			break
849		}
850
851		for {
852			out.WriteByte(' ')
853			column++
854			if column%tabSize == 0 {
855				break
856			}
857		}
858
859		i++
860	}
861}
862
863// Find if a line counts as indented or not.
864// Returns number of characters the indent is (0 = not indented).
865func isIndented(data []byte, indentSize int) int {
866	if len(data) == 0 {
867		return 0
868	}
869	if data[0] == '\t' {
870		return 1
871	}
872	if len(data) < indentSize {
873		return 0
874	}
875	for i := 0; i < indentSize; i++ {
876		if data[i] != ' ' {
877			return 0
878		}
879	}
880	return indentSize
881}
882
883// Create a url-safe slug for fragments
884func slugify(in []byte) []byte {
885	if len(in) == 0 {
886		return in
887	}
888	out := make([]byte, 0, len(in))
889	sym := false
890
891	for _, ch := range in {
892		if isalnum(ch) {
893			sym = false
894			out = append(out, ch)
895		} else if sym {
896			continue
897		} else {
898			out = append(out, '-')
899			sym = true
900		}
901	}
902	var a, b int
903	var ch byte
904	for a, ch = range out {
905		if ch != '-' {
906			break
907		}
908	}
909	for b = len(out) - 1; b > 0; b-- {
910		if out[b] != '-' {
911			break
912		}
913	}
914	return out[a : b+1]
915}