all repos — grayfriday @ ee98bc0bf4ddd4670dfad41751d5162470c18dca

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