all repos — grayfriday @ 583b3c5e1dcb22f284b040675c11024081a4a66c

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	"utf8"
 24)
 25
 26const VERSION = "1.0"
 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)
 41
 42// These are the possible flag values for the link renderer.
 43// Only a single one of these values will be used; they are not ORed together.
 44// These are mostly of interest if you are writing a new output format.
 45const (
 46	LINK_TYPE_NOT_AUTOLINK = iota
 47	LINK_TYPE_NORMAL
 48	LINK_TYPE_EMAIL
 49)
 50
 51// These are the possible flag values for the ListItem renderer.
 52// Multiple flag values may be ORed together.
 53// These are mostly of interest if you are writing a new output format.
 54const (
 55	LIST_TYPE_ORDERED = 1 << iota
 56	LIST_ITEM_CONTAINS_BLOCK
 57	LIST_ITEM_BEGINNING_OF_LIST
 58	LIST_ITEM_END_OF_LIST
 59)
 60
 61// These are the possible flag values for the table cell renderer.
 62// Only a single one of these values will be used; they are not ORed together.
 63// These are mostly of interest if you are writing a new output format.
 64const (
 65	TABLE_ALIGNMENT_LEFT = 1 << iota
 66	TABLE_ALIGNMENT_RIGHT
 67	TABLE_ALIGNMENT_CENTER = (TABLE_ALIGNMENT_LEFT | TABLE_ALIGNMENT_RIGHT)
 68)
 69
 70// The size of a tab stop.
 71const (
 72	TAB_SIZE_DEFAULT = 4
 73	TAB_SIZE_EIGHT   = 8
 74)
 75
 76// These are the tags that are recognized as HTML block tags.
 77// Any of these can be included in markdown text without special escaping.
 78var blockTags = map[string]bool{
 79	"p":          true,
 80	"dl":         true,
 81	"h1":         true,
 82	"h2":         true,
 83	"h3":         true,
 84	"h4":         true,
 85	"h5":         true,
 86	"h6":         true,
 87	"ol":         true,
 88	"ul":         true,
 89	"del":        true,
 90	"div":        true,
 91	"ins":        true,
 92	"pre":        true,
 93	"form":       true,
 94	"math":       true,
 95	"table":      true,
 96	"iframe":     true,
 97	"script":     true,
 98	"fieldset":   true,
 99	"noscript":   true,
100	"blockquote": true,
101}
102
103// Renderer is the rendering interface.
104// This is mostly of interest if you are implementing a new rendering format.
105//
106// When a byte slice is provided, it contains the (rendered) contents of the
107// element.
108//
109// When a callback is provided instead, it will write the contents of the
110// respective element directly to the output buffer and return true on success.
111// If the callback returns false, the rendering function should reset the
112// output buffer as though it had never been called.
113//
114// Currently Html and Latex implementations are provided
115type Renderer interface {
116	// block-level callbacks
117	BlockCode(out *bytes.Buffer, text []byte, lang string)
118	BlockQuote(out *bytes.Buffer, text []byte)
119	BlockHtml(out *bytes.Buffer, text []byte)
120	Header(out *bytes.Buffer, text func() bool, level int)
121	HRule(out *bytes.Buffer)
122	List(out *bytes.Buffer, text func() bool, flags int)
123	ListItem(out *bytes.Buffer, text []byte, flags int)
124	Paragraph(out *bytes.Buffer, text func() bool)
125	Table(out *bytes.Buffer, header []byte, body []byte, columnData []int)
126	TableRow(out *bytes.Buffer, text []byte)
127	TableCell(out *bytes.Buffer, text []byte, flags int)
128
129	// Span-level callbacks
130	AutoLink(out *bytes.Buffer, link []byte, kind int)
131	CodeSpan(out *bytes.Buffer, text []byte)
132	DoubleEmphasis(out *bytes.Buffer, text []byte)
133	Emphasis(out *bytes.Buffer, text []byte)
134	Image(out *bytes.Buffer, link []byte, title []byte, alt []byte)
135	LineBreak(out *bytes.Buffer)
136	Link(out *bytes.Buffer, link []byte, title []byte, content []byte)
137	RawHtmlTag(out *bytes.Buffer, tag []byte)
138	TripleEmphasis(out *bytes.Buffer, text []byte)
139	StrikeThrough(out *bytes.Buffer, text []byte)
140
141	// Low-level callbacks
142	Entity(out *bytes.Buffer, entity []byte)
143	NormalText(out *bytes.Buffer, text []byte)
144
145	// Header and footer
146	DocumentHeader(out *bytes.Buffer)
147	DocumentFooter(out *bytes.Buffer)
148}
149
150// Callback functions for inline parsing. One such function is defined
151// for each character that triggers a response when parsing inline data.
152type inlineParser func(p *parser, out *bytes.Buffer, data []byte, offset int) int
153
154// Parser holds runtime state used by the parser.
155// This is constructed by the Markdown function.
156type parser struct {
157	r              Renderer
158	refs           map[string]*reference
159	inlineCallback [256]inlineParser
160	flags          int
161	nesting        int
162	maxNesting     int
163	insideLink     bool
164}
165
166
167//
168//
169// Public interface
170//
171//
172
173// MarkdownBasic is a convenience function for simple rendering.
174// It processes markdown input with no extensions enabled.
175func MarkdownBasic(input []byte) []byte {
176	// set up the HTML renderer
177	htmlFlags := HTML_USE_XHTML
178	renderer := HtmlRenderer(htmlFlags, "", "")
179
180	// set up the parser
181	extensions := 0
182
183	return Markdown(input, renderer, extensions)
184}
185
186// Call Markdown with most useful extensions enabled
187// MarkdownCommon is a convenience function for simple rendering.
188// It processes markdown input with common extensions enabled, including:
189//
190// * Smartypants processing with smart fractions and LaTeX dashes
191//
192// * Intra-word emphasis supression
193//
194// * Tables
195//
196// * Fenced code blocks
197//
198// * Autolinking
199//
200// * Strikethrough support
201//
202// * Strict header parsing
203func MarkdownCommon(input []byte) []byte {
204	// set up the HTML renderer
205	htmlFlags := 0
206	htmlFlags |= HTML_USE_XHTML
207	htmlFlags |= HTML_USE_SMARTYPANTS
208	htmlFlags |= HTML_SMARTYPANTS_FRACTIONS
209	htmlFlags |= HTML_SMARTYPANTS_LATEX_DASHES
210	renderer := HtmlRenderer(htmlFlags, "", "")
211
212	// set up the parser
213	extensions := 0
214	extensions |= EXTENSION_NO_INTRA_EMPHASIS
215	extensions |= EXTENSION_TABLES
216	extensions |= EXTENSION_FENCED_CODE
217	extensions |= EXTENSION_AUTOLINK
218	extensions |= EXTENSION_STRIKETHROUGH
219	extensions |= EXTENSION_SPACE_HEADERS
220
221	return Markdown(input, renderer, extensions)
222}
223
224// Markdown is the main rendering function.
225// It parses and renders a block of markdown-encoded text.
226// The supplied Renderer is used to format the output, and extensions dictates
227// which non-standard extensions are enabled.
228//
229// To use the supplied Html or LaTeX renderers, see HtmlRenderer and
230// LatexRenderer, respectively.
231func Markdown(input []byte, renderer Renderer, extensions int) []byte {
232	// no point in parsing if we can't render
233	if renderer == nil {
234		return nil
235	}
236
237	// fill in the render structure
238	p := new(parser)
239	p.r = renderer
240	p.flags = extensions
241	p.refs = make(map[string]*reference)
242	p.maxNesting = 16
243	p.insideLink = false
244
245	// register inline parsers
246	p.inlineCallback['*'] = emphasis
247	p.inlineCallback['_'] = emphasis
248	if extensions&EXTENSION_STRIKETHROUGH != 0 {
249		p.inlineCallback['~'] = emphasis
250	}
251	p.inlineCallback['`'] = codeSpan
252	p.inlineCallback['\n'] = lineBreak
253	p.inlineCallback['['] = link
254	p.inlineCallback['<'] = leftAngle
255	p.inlineCallback['\\'] = escape
256	p.inlineCallback['&'] = entity
257
258	if extensions&EXTENSION_AUTOLINK != 0 {
259		p.inlineCallback[':'] = autoLink
260	}
261
262	first := firstPass(p, input)
263	second := secondPass(p, first)
264
265	return second
266}
267
268// first pass:
269// - extract references
270// - expand tabs
271// - normalize newlines
272// - copy everything else
273func firstPass(p *parser, input []byte) []byte {
274	var out bytes.Buffer
275	tabSize := TAB_SIZE_DEFAULT
276	if p.flags&EXTENSION_TAB_SIZE_EIGHT != 0 {
277		tabSize = TAB_SIZE_EIGHT
278	}
279	beg, end := 0, 0
280	for beg < len(input) { // iterate over lines
281		if end = isReference(p, input[beg:]); end > 0 {
282			beg += end
283		} else { // skip to the next line
284			end = beg
285			for end < len(input) && input[end] != '\n' && input[end] != '\r' {
286				end++
287			}
288
289			// add the line body if present
290			if end > beg {
291				expandTabs(&out, input[beg:end], tabSize)
292			}
293			out.WriteByte('\n')
294
295			if end < len(input) && input[end] == '\r' {
296				end++
297			}
298			if end < len(input) && input[end] == '\n' {
299				end++
300			}
301
302			beg = end
303		}
304	}
305
306	// empty input?
307	if out.Len() == 0 {
308		out.WriteByte('\n')
309	}
310
311	return out.Bytes()
312}
313
314// second pass: actual rendering
315func secondPass(p *parser, input []byte) []byte {
316	var output bytes.Buffer
317
318	p.r.DocumentHeader(&output)
319	p.block(&output, input)
320	p.r.DocumentFooter(&output)
321
322	if p.nesting != 0 {
323		panic("Nesting level did not end at zero")
324	}
325
326	return output.Bytes()
327}
328
329
330//
331// Link references
332//
333// This section implements support for references that (usually) appear
334// as footnotes in a document, and can be referenced anywhere in the document.
335// The basic format is:
336//
337//    [1]: http://www.google.com/ "Google"
338//    [2]: http://www.github.com/ "Github"
339//
340// Anywhere in the document, the reference can be linked by referring to its
341// label, i.e., 1 and 2 in this example, as in:
342//
343//    This library is hosted on [Github][2], a git hosting site.
344
345// References are parsed and stored in this struct.
346type reference struct {
347	link  []byte
348	title []byte
349}
350
351// Check whether or not data starts with a reference link.
352// If so, it is parsed and stored in the list of references
353// (in the render struct).
354// Returns the number of bytes to skip to move past it,
355// or zero if the first line is not a reference.
356func isReference(p *parser, data []byte) int {
357	// up to 3 optional leading spaces
358	if len(data) < 4 {
359		return 0
360	}
361	i := 0
362	for i < 3 && data[i] == ' ' {
363		i++
364	}
365
366	// id part: anything but a newline between brackets
367	if data[i] != '[' {
368		return 0
369	}
370	i++
371	idOffset := i
372	for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
373		i++
374	}
375	if i >= len(data) || data[i] != ']' {
376		return 0
377	}
378	idEnd := i
379
380	// spacer: colon (space | tab)* newline? (space | tab)*
381	i++
382	if i >= len(data) || data[i] != ':' {
383		return 0
384	}
385	i++
386	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
387		i++
388	}
389	if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
390		i++
391		if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
392			i++
393		}
394	}
395	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
396		i++
397	}
398	if i >= len(data) {
399		return 0
400	}
401
402	// link: whitespace-free sequence, optionally between angle brackets
403	if data[i] == '<' {
404		i++
405	}
406	linkOffset := i
407	for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
408		i++
409	}
410	linkEnd := i
411	if data[linkOffset] == '<' && data[linkEnd-1] == '>' {
412		linkOffset++
413		linkEnd--
414	}
415
416	// optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
417	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
418		i++
419	}
420	if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
421		return 0
422	}
423
424	// compute end-of-line
425	lineEnd := 0
426	if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
427		lineEnd = i
428	}
429	if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
430		lineEnd++
431	}
432
433	// optional (space|tab)* spacer after a newline
434	if lineEnd > 0 {
435		i = lineEnd + 1
436		for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
437			i++
438		}
439	}
440
441	// optional title: any non-newline sequence enclosed in '"() alone on its line
442	titleOffset, titleEnd := 0, 0
443	if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
444		i++
445		titleOffset = i
446
447		// look for EOL
448		for i < len(data) && data[i] != '\n' && data[i] != '\r' {
449			i++
450		}
451		if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
452			titleEnd = i + 1
453		} else {
454			titleEnd = i
455		}
456
457		// step back
458		i--
459		for i > titleOffset && (data[i] == ' ' || data[i] == '\t') {
460			i--
461		}
462		if i > titleOffset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
463			lineEnd = titleEnd
464			titleEnd = i
465		}
466	}
467	if lineEnd == 0 { // garbage after the link
468		return 0
469	}
470
471	// a valid ref has been found
472
473	// id matches are case-insensitive
474	id := string(bytes.ToLower(data[idOffset:idEnd]))
475	p.refs[id] = &reference{
476		link:  data[linkOffset:linkEnd],
477		title: data[titleOffset:titleEnd],
478	}
479
480	return lineEnd
481}
482
483
484//
485//
486// Miscellaneous helper functions
487//
488//
489
490
491// Test if a character is a punctuation symbol.
492// Taken from a private function in regexp in the stdlib.
493func ispunct(c byte) bool {
494	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
495		if c == r {
496			return true
497		}
498	}
499	return false
500}
501
502// Test if a character is a whitespace character.
503func isspace(c byte) bool {
504	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
505}
506
507// Test if a character is a letter or a digit.
508// TODO: check when this is looking for ASCII alnum and when it should use unicode
509func isalnum(c byte) bool {
510	return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
511}
512
513// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
514// always ends output with a newline
515func expandTabs(out *bytes.Buffer, line []byte, tabSize int) {
516	// first, check for common cases: no tabs, or only tabs at beginning of line
517	i, prefix := 0, 0
518	slowcase := false
519	for i = 0; i < len(line); i++ {
520		if line[i] == '\t' {
521			if prefix == i {
522				prefix++
523			} else {
524				slowcase = true
525				break
526			}
527		}
528	}
529
530	// no need to decode runes if all tabs are at the beginning of the line
531	if !slowcase {
532		for i = 0; i < prefix*tabSize; i++ {
533			out.WriteByte(' ')
534		}
535		out.Write(line[prefix:])
536		return
537	}
538
539	// the slow case: we need to count runes to figure out how
540	// many spaces to insert for each tab
541	column := 0
542	i = 0
543	for i < len(line) {
544		start := i
545		for i < len(line) && line[i] != '\t' {
546			_, size := utf8.DecodeRune(line[i:])
547			i += size
548			column++
549		}
550
551		if i > start {
552			out.Write(line[start:i])
553		}
554
555		if i >= len(line) {
556			break
557		}
558
559		for {
560			out.WriteByte(' ')
561			column++
562			if column%tabSize == 0 {
563				break
564			}
565		}
566
567		i++
568	}
569}