all repos — grayfriday @ 8e90e8b6455f1608d6a0948d0eb07409bcf874d4

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	"fmt"
 24	"strings"
 25	"unicode/utf8"
 26)
 27
 28const VERSION = "1.4"
 29
 30type Extensions int
 31
 32// These are the supported markdown parsing extensions.
 33// OR these values together to select multiple extensions.
 34const (
 35	NoExtensions           Extensions = 0
 36	NoIntraEmphasis        Extensions = 1 << iota // Ignore emphasis markers inside words
 37	Tables                                        // Render tables
 38	FencedCode                                    // Render fenced code blocks
 39	Autolink                                      // Detect embedded URLs that are not explicitly marked
 40	Strikethrough                                 // Strikethrough text using ~~test~~
 41	LaxHTMLBlocks                                 // Loosen up HTML block parsing rules
 42	SpaceHeaders                                  // Be strict about prefix header rules
 43	HardLineBreak                                 // Translate newlines into line breaks
 44	TabSizeEight                                  // Expand tabs to eight spaces instead of four
 45	Footnotes                                     // Pandoc-style footnotes
 46	NoEmptyLineBeforeBlock                        // No need to insert an empty line to start a (code, quote, ordered list, unordered list) block
 47	HeaderIDs                                     // specify header IDs  with {#id}
 48	Titleblock                                    // Titleblock ala pandoc
 49	AutoHeaderIDs                                 // Create the header ID from the text
 50	BackslashLineBreak                            // Translate trailing backslashes into line breaks
 51	DefinitionLists                               // Render definition lists
 52
 53	commonHtmlFlags HtmlFlags = UseXHTML | UseSmartypants |
 54		SmartypantsFractions | SmartypantsDashes | SmartypantsLatexDashes
 55
 56	commonExtensions Extensions = NoIntraEmphasis | Tables | FencedCode |
 57		Autolink | Strikethrough | SpaceHeaders | HeaderIDs |
 58		BackslashLineBreak | DefinitionLists
 59)
 60
 61type LinkType int
 62
 63// These are the possible flag values for the link 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	LinkTypeNotAutolink LinkType = iota
 68	LinkTypeNormal
 69	LinkTypeEmail
 70)
 71
 72type ListType int
 73
 74// These are the possible flag values for the ListItem renderer.
 75// Multiple flag values may be ORed together.
 76// These are mostly of interest if you are writing a new output format.
 77const (
 78	ListTypeOrdered ListType = 1 << iota
 79	ListTypeDefinition
 80	ListTypeTerm
 81
 82	ListItemContainsBlock
 83	ListItemBeginningOfList
 84	ListItemEndOfList
 85)
 86
 87type TableFlags int
 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	TableAlignmentLeft = 1 << iota
 94	TableAlignmentRight
 95	TableAlignmentCenter = (TableAlignmentLeft | TableAlignmentRight)
 96)
 97
 98// The size of a tab stop.
 99const (
100	TabSizeDefault = 4
101	TabSizeDouble  = 8
102)
103
104// blockTags is a set of 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]struct{}{
107	"blockquote": struct{}{},
108	"del":        struct{}{},
109	"div":        struct{}{},
110	"dl":         struct{}{},
111	"fieldset":   struct{}{},
112	"form":       struct{}{},
113	"h1":         struct{}{},
114	"h2":         struct{}{},
115	"h3":         struct{}{},
116	"h4":         struct{}{},
117	"h5":         struct{}{},
118	"h6":         struct{}{},
119	"iframe":     struct{}{},
120	"ins":        struct{}{},
121	"math":       struct{}{},
122	"noscript":   struct{}{},
123	"ol":         struct{}{},
124	"pre":        struct{}{},
125	"p":          struct{}{},
126	"script":     struct{}{},
127	"style":      struct{}{},
128	"table":      struct{}{},
129	"ul":         struct{}{},
130
131	// HTML5
132	"address":    struct{}{},
133	"article":    struct{}{},
134	"aside":      struct{}{},
135	"canvas":     struct{}{},
136	"figcaption": struct{}{},
137	"figure":     struct{}{},
138	"footer":     struct{}{},
139	"header":     struct{}{},
140	"hgroup":     struct{}{},
141	"main":       struct{}{},
142	"nav":        struct{}{},
143	"output":     struct{}{},
144	"progress":   struct{}{},
145	"section":    struct{}{},
146	"video":      struct{}{},
147}
148
149// Renderer is the rendering interface.
150// This is mostly of interest if you are implementing a new rendering format.
151//
152// When a byte slice is provided, it contains the (rendered) contents of the
153// element.
154//
155// When a callback is provided instead, it will write the contents of the
156// respective element directly to the output buffer and return true on success.
157// If the callback returns false, the rendering function should reset the
158// output buffer as though it had never been called.
159//
160// Currently Html and Latex implementations are provided
161type Renderer interface {
162	// block-level callbacks
163	BlockCode(out *bytes.Buffer, text []byte, lang string)
164	BlockQuote(out *bytes.Buffer, text []byte)
165	BlockHtml(out *bytes.Buffer, text []byte)
166	BeginHeader(out *bytes.Buffer, level int, id string) int
167	EndHeader(out *bytes.Buffer, level int, id string, tocMarker int)
168	HRule(out *bytes.Buffer)
169	BeginList(out *bytes.Buffer, flags ListType)
170	EndList(out *bytes.Buffer, flags ListType)
171	ListItem(out *bytes.Buffer, text []byte, flags ListType)
172	BeginParagraph(out *bytes.Buffer)
173	EndParagraph(out *bytes.Buffer)
174	Table(out *bytes.Buffer, header []byte, body []byte, columnData []int)
175	TableRow(out *bytes.Buffer, text []byte)
176	TableHeaderCell(out *bytes.Buffer, text []byte, flags int)
177	TableCell(out *bytes.Buffer, text []byte, flags int)
178	BeginFootnotes(out *bytes.Buffer)
179	EndFootnotes(out *bytes.Buffer)
180	FootnoteItem(out *bytes.Buffer, name, text []byte, flags ListType)
181	TitleBlock(out *bytes.Buffer, text []byte)
182
183	// Span-level callbacks
184	AutoLink(out *bytes.Buffer, link []byte, kind LinkType)
185	CodeSpan(out *bytes.Buffer, text []byte)
186	DoubleEmphasis(out *bytes.Buffer, text []byte)
187	Emphasis(out *bytes.Buffer, text []byte)
188	Image(out *bytes.Buffer, link []byte, title []byte, alt []byte)
189	LineBreak(out *bytes.Buffer)
190	Link(out *bytes.Buffer, link []byte, title []byte, content []byte)
191	RawHtmlTag(out *bytes.Buffer, tag []byte)
192	TripleEmphasis(out *bytes.Buffer, text []byte)
193	StrikeThrough(out *bytes.Buffer, text []byte)
194	FootnoteRef(out *bytes.Buffer, ref []byte, id int)
195
196	// Low-level callbacks
197	Entity(out *bytes.Buffer, entity []byte)
198	NormalText(out *bytes.Buffer, text []byte)
199
200	// Header and footer
201	DocumentHeader(out *bytes.Buffer)
202	DocumentFooter(out *bytes.Buffer)
203
204	GetFlags() HtmlFlags
205}
206
207// Callback functions for inline parsing. One such function is defined
208// for each character that triggers a response when parsing inline data.
209type inlineParser func(p *parser, out *bytes.Buffer, data []byte, offset int) int
210
211// Parser holds runtime state used by the parser.
212// This is constructed by the Markdown function.
213type parser struct {
214	r              Renderer
215	refOverride    ReferenceOverrideFunc
216	refs           map[string]*reference
217	inlineCallback [256]inlineParser
218	flags          Extensions
219	nesting        int
220	maxNesting     int
221	insideLink     bool
222
223	// Footnotes need to be ordered as well as available to quickly check for
224	// presence. If a ref is also a footnote, it's stored both in refs and here
225	// in notes. Slice is nil if footnotes not enabled.
226	notes []*reference
227}
228
229func (p *parser) getRef(refid string) (ref *reference, found bool) {
230	if p.refOverride != nil {
231		r, overridden := p.refOverride(refid)
232		if overridden {
233			if r == nil {
234				return nil, false
235			}
236			return &reference{
237				link:     []byte(r.Link),
238				title:    []byte(r.Title),
239				noteId:   0,
240				hasBlock: false,
241				text:     []byte(r.Text)}, true
242		}
243	}
244	// refs are case insensitive
245	ref, found = p.refs[strings.ToLower(refid)]
246	return ref, found
247}
248
249//
250//
251// Public interface
252//
253//
254
255// Reference represents the details of a link.
256// See the documentation in Options for more details on use-case.
257type Reference struct {
258	// Link is usually the URL the reference points to.
259	Link string
260	// Title is the alternate text describing the link in more detail.
261	Title string
262	// Text is the optional text to override the ref with if the syntax used was
263	// [refid][]
264	Text string
265}
266
267// ReferenceOverrideFunc is expected to be called with a reference string and
268// return either a valid Reference type that the reference string maps to or
269// nil. If overridden is false, the default reference logic will be executed.
270// See the documentation in Options for more details on use-case.
271type ReferenceOverrideFunc func(reference string) (ref *Reference, overridden bool)
272
273// Options represents configurable overrides and callbacks (in addition to the
274// extension flag set) for configuring a Markdown parse.
275type Options struct {
276	// Extensions is a flag set of bit-wise ORed extension bits. See the
277	// Extensions flags defined in this package.
278	Extensions Extensions
279
280	// ReferenceOverride is an optional function callback that is called every
281	// time a reference is resolved.
282	//
283	// In Markdown, the link reference syntax can be made to resolve a link to
284	// a reference instead of an inline URL, in one of the following ways:
285	//
286	//  * [link text][refid]
287	//  * [refid][]
288	//
289	// Usually, the refid is defined at the bottom of the Markdown document. If
290	// this override function is provided, the refid is passed to the override
291	// function first, before consulting the defined refids at the bottom. If
292	// the override function indicates an override did not occur, the refids at
293	// the bottom will be used to fill in the link details.
294	ReferenceOverride ReferenceOverrideFunc
295}
296
297// MarkdownBasic is a convenience function for simple rendering.
298// It processes markdown input with no extensions enabled.
299func MarkdownBasic(input []byte) []byte {
300	// set up the HTML renderer
301	htmlFlags := UseXHTML
302	renderer := HtmlRenderer(htmlFlags, "", "")
303
304	// set up the parser
305	return MarkdownOptions(input, renderer, Options{Extensions: 0})
306}
307
308// Call Markdown with most useful extensions enabled
309// MarkdownCommon is a convenience function for simple rendering.
310// It processes markdown input with common extensions enabled, including:
311//
312// * Smartypants processing with smart fractions and LaTeX dashes
313//
314// * Intra-word emphasis suppression
315//
316// * Tables
317//
318// * Fenced code blocks
319//
320// * Autolinking
321//
322// * Strikethrough support
323//
324// * Strict header parsing
325//
326// * Custom Header IDs
327func MarkdownCommon(input []byte) []byte {
328	// set up the HTML renderer
329	renderer := HtmlRenderer(commonHtmlFlags, "", "")
330	return MarkdownOptions(input, renderer, Options{
331		Extensions: commonExtensions})
332}
333
334// Markdown is the main rendering function.
335// It parses and renders a block of markdown-encoded text.
336// The supplied Renderer is used to format the output, and extensions dictates
337// which non-standard extensions are enabled.
338//
339// To use the supplied Html or LaTeX renderers, see HtmlRenderer and
340// LatexRenderer, respectively.
341func Markdown(input []byte, renderer Renderer, extensions Extensions) []byte {
342	return MarkdownOptions(input, renderer, Options{
343		Extensions: extensions})
344}
345
346// MarkdownOptions is just like Markdown but takes additional options through
347// the Options struct.
348func MarkdownOptions(input []byte, renderer Renderer, opts Options) []byte {
349	// no point in parsing if we can't render
350	if renderer == nil {
351		return nil
352	}
353
354	extensions := opts.Extensions
355
356	// fill in the render structure
357	p := new(parser)
358	p.r = renderer
359	p.flags = extensions
360	p.refOverride = opts.ReferenceOverride
361	p.refs = make(map[string]*reference)
362	p.maxNesting = 16
363	p.insideLink = false
364
365	// register inline parsers
366	p.inlineCallback['*'] = emphasis
367	p.inlineCallback['_'] = emphasis
368	if extensions&Strikethrough != 0 {
369		p.inlineCallback['~'] = emphasis
370	}
371	p.inlineCallback['`'] = codeSpan
372	p.inlineCallback['\n'] = lineBreak
373	p.inlineCallback['['] = link
374	p.inlineCallback['<'] = leftAngle
375	p.inlineCallback['\\'] = escape
376	p.inlineCallback['&'] = entity
377	p.inlineCallback['!'] = maybeImage
378	p.inlineCallback['^'] = maybeInlineFootnote
379	p.inlineCallback[' '] = maybeLineBreak
380
381	if extensions&Autolink != 0 {
382		p.inlineCallback[':'] = autoLink
383	}
384
385	if extensions&Footnotes != 0 {
386		p.notes = make([]*reference, 0)
387	}
388
389	first := firstPass(p, input)
390	second := secondPass(p, first)
391	return second
392}
393
394// first pass:
395// - extract references
396// - expand tabs
397// - normalize newlines
398// - copy everything else
399func firstPass(p *parser, input []byte) []byte {
400	var out bytes.Buffer
401	tabSize := TabSizeDefault
402	if p.flags&TabSizeEight != 0 {
403		tabSize = TabSizeDouble
404	}
405	beg, end := 0, 0
406	lastFencedCodeBlockEnd := 0
407	for beg < len(input) { // iterate over lines
408		if end = isReference(p, input[beg:], tabSize); end > 0 {
409			beg += end
410		} else { // skip to the next line
411			end = beg
412			for end < len(input) && input[end] != '\n' && input[end] != '\r' {
413				end++
414			}
415
416			if p.flags&FencedCode != 0 {
417				// track fenced code block boundaries to suppress tab expansion
418				// inside them:
419				if beg >= lastFencedCodeBlockEnd {
420					if i := p.fencedCode(&out, input[beg:], false); i > 0 {
421						lastFencedCodeBlockEnd = beg + i
422					}
423				}
424			}
425
426			// add the line body if present
427			if end > beg {
428				if end < lastFencedCodeBlockEnd { // Do not expand tabs while inside fenced code blocks.
429					out.Write(input[beg:end])
430				} else {
431					expandTabs(&out, input[beg:end], tabSize)
432				}
433			}
434			out.WriteByte('\n')
435
436			if end < len(input) && input[end] == '\r' {
437				end++
438			}
439			if end < len(input) && input[end] == '\n' {
440				end++
441			}
442
443			beg = end
444		}
445	}
446
447	// empty input?
448	if out.Len() == 0 {
449		out.WriteByte('\n')
450	}
451
452	return out.Bytes()
453}
454
455// second pass: actual rendering
456func secondPass(p *parser, input []byte) []byte {
457	var output bytes.Buffer
458
459	p.r.DocumentHeader(&output)
460	p.block(&output, input)
461
462	if p.flags&Footnotes != 0 && len(p.notes) > 0 {
463		p.r.BeginFootnotes(&output)
464		flags := ListItemBeginningOfList
465		for i := 0; i < len(p.notes); i += 1 {
466			ref := p.notes[i]
467			var buf bytes.Buffer
468			if ref.hasBlock {
469				flags |= ListItemContainsBlock
470				p.block(&buf, ref.title)
471			} else {
472				p.inline(&buf, ref.title)
473			}
474			p.r.FootnoteItem(&output, ref.link, buf.Bytes(), flags)
475			flags &^= ListItemBeginningOfList | ListItemContainsBlock
476		}
477		p.r.EndFootnotes(&output)
478	}
479
480	p.r.DocumentFooter(&output)
481
482	if p.nesting != 0 {
483		panic("Nesting level did not end at zero")
484	}
485
486	return output.Bytes()
487}
488
489//
490// Link references
491//
492// This section implements support for references that (usually) appear
493// as footnotes in a document, and can be referenced anywhere in the document.
494// The basic format is:
495//
496//    [1]: http://www.google.com/ "Google"
497//    [2]: http://www.github.com/ "Github"
498//
499// Anywhere in the document, the reference can be linked by referring to its
500// label, i.e., 1 and 2 in this example, as in:
501//
502//    This library is hosted on [Github][2], a git hosting site.
503//
504// Actual footnotes as specified in Pandoc and supported by some other Markdown
505// libraries such as php-markdown are also taken care of. They look like this:
506//
507//    This sentence needs a bit of further explanation.[^note]
508//
509//    [^note]: This is the explanation.
510//
511// Footnotes should be placed at the end of the document in an ordered list.
512// Inline footnotes such as:
513//
514//    Inline footnotes^[Not supported.] also exist.
515//
516// are not yet supported.
517
518// References are parsed and stored in this struct.
519type reference struct {
520	link     []byte
521	title    []byte
522	noteId   int // 0 if not a footnote ref
523	hasBlock bool
524	text     []byte
525}
526
527func (r *reference) String() string {
528	return fmt.Sprintf("{link: %q, title: %q, text: %q, noteId: %d, hasBlock: %v}",
529		r.link, r.title, r.text, r.noteId, r.hasBlock)
530}
531
532// Check whether or not data starts with a reference link.
533// If so, it is parsed and stored in the list of references
534// (in the render struct).
535// Returns the number of bytes to skip to move past it,
536// or zero if the first line is not a reference.
537func isReference(p *parser, data []byte, tabSize int) int {
538	// up to 3 optional leading spaces
539	if len(data) < 4 {
540		return 0
541	}
542	i := 0
543	for i < 3 && data[i] == ' ' {
544		i++
545	}
546
547	noteId := 0
548
549	// id part: anything but a newline between brackets
550	if data[i] != '[' {
551		return 0
552	}
553	i++
554	if p.flags&Footnotes != 0 {
555		if i < len(data) && data[i] == '^' {
556			// we can set it to anything here because the proper noteIds will
557			// be assigned later during the second pass. It just has to be != 0
558			noteId = 1
559			i++
560		}
561	}
562	idOffset := i
563	for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
564		i++
565	}
566	if i >= len(data) || data[i] != ']' {
567		return 0
568	}
569	idEnd := i
570
571	// spacer: colon (space | tab)* newline? (space | tab)*
572	i++
573	if i >= len(data) || data[i] != ':' {
574		return 0
575	}
576	i++
577	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
578		i++
579	}
580	if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
581		i++
582		if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
583			i++
584		}
585	}
586	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
587		i++
588	}
589	if i >= len(data) {
590		return 0
591	}
592
593	var (
594		linkOffset, linkEnd   int
595		titleOffset, titleEnd int
596		lineEnd               int
597		raw                   []byte
598		hasBlock              bool
599	)
600
601	if p.flags&Footnotes != 0 && noteId != 0 {
602		linkOffset, linkEnd, raw, hasBlock = scanFootnote(p, data, i, tabSize)
603		lineEnd = linkEnd
604	} else {
605		linkOffset, linkEnd, titleOffset, titleEnd, lineEnd = scanLinkRef(p, data, i)
606	}
607	if lineEnd == 0 {
608		return 0
609	}
610
611	// a valid ref has been found
612
613	ref := &reference{
614		noteId:   noteId,
615		hasBlock: hasBlock,
616	}
617
618	if noteId > 0 {
619		// reusing the link field for the id since footnotes don't have links
620		ref.link = data[idOffset:idEnd]
621		// if footnote, it's not really a title, it's the contained text
622		ref.title = raw
623	} else {
624		ref.link = data[linkOffset:linkEnd]
625		ref.title = data[titleOffset:titleEnd]
626	}
627
628	// id matches are case-insensitive
629	id := string(bytes.ToLower(data[idOffset:idEnd]))
630
631	p.refs[id] = ref
632
633	return lineEnd
634}
635
636func scanLinkRef(p *parser, data []byte, i int) (linkOffset, linkEnd, titleOffset, titleEnd, lineEnd int) {
637	// link: whitespace-free sequence, optionally between angle brackets
638	if data[i] == '<' {
639		i++
640	}
641	linkOffset = i
642	for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
643		i++
644	}
645	if i == len(data) {
646		return
647	}
648	linkEnd = i
649	if data[linkOffset] == '<' && data[linkEnd-1] == '>' {
650		linkOffset++
651		linkEnd--
652	}
653
654	// optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
655	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
656		i++
657	}
658	if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
659		return
660	}
661
662	// compute end-of-line
663	if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
664		lineEnd = i
665	}
666	if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
667		lineEnd++
668	}
669
670	// optional (space|tab)* spacer after a newline
671	if lineEnd > 0 {
672		i = lineEnd + 1
673		for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
674			i++
675		}
676	}
677
678	// optional title: any non-newline sequence enclosed in '"() alone on its line
679	if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
680		i++
681		titleOffset = i
682
683		// look for EOL
684		for i < len(data) && data[i] != '\n' && data[i] != '\r' {
685			i++
686		}
687		if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
688			titleEnd = i + 1
689		} else {
690			titleEnd = i
691		}
692
693		// step back
694		i--
695		for i > titleOffset && (data[i] == ' ' || data[i] == '\t') {
696			i--
697		}
698		if i > titleOffset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
699			lineEnd = titleEnd
700			titleEnd = i
701		}
702	}
703
704	return
705}
706
707// The first bit of this logic is the same as (*parser).listItem, but the rest
708// is much simpler. This function simply finds the entire block and shifts it
709// over by one tab if it is indeed a block (just returns the line if it's not).
710// blockEnd is the end of the section in the input buffer, and contents is the
711// extracted text that was shifted over one tab. It will need to be rendered at
712// the end of the document.
713func scanFootnote(p *parser, data []byte, i, indentSize int) (blockStart, blockEnd int, contents []byte, hasBlock bool) {
714	if i == 0 || len(data) == 0 {
715		return
716	}
717
718	// skip leading whitespace on first line
719	for i < len(data) && data[i] == ' ' {
720		i++
721	}
722
723	blockStart = i
724
725	// find the end of the line
726	blockEnd = i
727	for i < len(data) && data[i-1] != '\n' {
728		i++
729	}
730
731	// get working buffer
732	var raw bytes.Buffer
733
734	// put the first line into the working buffer
735	raw.Write(data[blockEnd:i])
736	blockEnd = i
737
738	// process the following lines
739	containsBlankLine := false
740
741gatherLines:
742	for blockEnd < len(data) {
743		i++
744
745		// find the end of this line
746		for i < len(data) && data[i-1] != '\n' {
747			i++
748		}
749
750		// if it is an empty line, guess that it is part of this item
751		// and move on to the next line
752		if p.isEmpty(data[blockEnd:i]) > 0 {
753			containsBlankLine = true
754			blockEnd = i
755			continue
756		}
757
758		n := 0
759		if n = isIndented(data[blockEnd:i], indentSize); n == 0 {
760			// this is the end of the block.
761			// we don't want to include this last line in the index.
762			break gatherLines
763		}
764
765		// if there were blank lines before this one, insert a new one now
766		if containsBlankLine {
767			raw.WriteByte('\n')
768			containsBlankLine = false
769		}
770
771		// get rid of that first tab, write to buffer
772		raw.Write(data[blockEnd+n : i])
773		hasBlock = true
774
775		blockEnd = i
776	}
777
778	if data[blockEnd-1] != '\n' {
779		raw.WriteByte('\n')
780	}
781
782	contents = raw.Bytes()
783
784	return
785}
786
787//
788//
789// Miscellaneous helper functions
790//
791//
792
793// Test if a character is a punctuation symbol.
794// Taken from a private function in regexp in the stdlib.
795func ispunct(c byte) bool {
796	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
797		if c == r {
798			return true
799		}
800	}
801	return false
802}
803
804// Test if a character is a whitespace character.
805func isspace(c byte) bool {
806	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
807}
808
809// Test if a character is letter.
810func isletter(c byte) bool {
811	return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
812}
813
814// Test if a character is a letter or a digit.
815// TODO: check when this is looking for ASCII alnum and when it should use unicode
816func isalnum(c byte) bool {
817	return (c >= '0' && c <= '9') || isletter(c)
818}
819
820// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
821// always ends output with a newline
822func expandTabs(out *bytes.Buffer, line []byte, tabSize int) {
823	// first, check for common cases: no tabs, or only tabs at beginning of line
824	i, prefix := 0, 0
825	slowcase := false
826	for i = 0; i < len(line); i++ {
827		if line[i] == '\t' {
828			if prefix == i {
829				prefix++
830			} else {
831				slowcase = true
832				break
833			}
834		}
835	}
836
837	// no need to decode runes if all tabs are at the beginning of the line
838	if !slowcase {
839		for i = 0; i < prefix*tabSize; i++ {
840			out.WriteByte(' ')
841		}
842		out.Write(line[prefix:])
843		return
844	}
845
846	// the slow case: we need to count runes to figure out how
847	// many spaces to insert for each tab
848	column := 0
849	i = 0
850	for i < len(line) {
851		start := i
852		for i < len(line) && line[i] != '\t' {
853			_, size := utf8.DecodeRune(line[i:])
854			i += size
855			column++
856		}
857
858		if i > start {
859			out.Write(line[start:i])
860		}
861
862		if i >= len(line) {
863			break
864		}
865
866		for {
867			out.WriteByte(' ')
868			column++
869			if column%tabSize == 0 {
870				break
871			}
872		}
873
874		i++
875	}
876}
877
878// Find if a line counts as indented or not.
879// Returns number of characters the indent is (0 = not indented).
880func isIndented(data []byte, indentSize int) int {
881	if len(data) == 0 {
882		return 0
883	}
884	if data[0] == '\t' {
885		return 1
886	}
887	if len(data) < indentSize {
888		return 0
889	}
890	for i := 0; i < indentSize; i++ {
891		if data[i] != ' ' {
892			return 0
893		}
894	}
895	return indentSize
896}
897
898// Create a url-safe slug for fragments
899func slugify(in []byte) []byte {
900	if len(in) == 0 {
901		return in
902	}
903	out := make([]byte, 0, len(in))
904	sym := false
905
906	for _, ch := range in {
907		if isalnum(ch) {
908			sym = false
909			out = append(out, ch)
910		} else if sym {
911			continue
912		} else {
913			out = append(out, '-')
914			sym = true
915		}
916	}
917	var a, b int
918	var ch byte
919	for a, ch = range out {
920		if ch != '-' {
921			break
922		}
923	}
924	for b = len(out) - 1; b > 0; b-- {
925		if out[b] != '-' {
926			break
927		}
928	}
929	return out[a : b+1]
930}