markdown.go (view raw)
1//
2// Black Friday Markdown Processor
3// Originally based on http://github.com/tanoku/upskirt
4// by Russ Ross <russ@russross.com>
5//
6
7//
8//
9// Markdown parsing and processing
10//
11//
12
13package blackfriday
14
15import (
16 "bytes"
17 "utf8"
18)
19
20// These are the supported markdown parsing extensions.
21// OR these values together to select multiple extensions.
22const (
23 EXTENSION_NO_INTRA_EMPHASIS = 1 << iota
24 EXTENSION_TABLES
25 EXTENSION_FENCED_CODE
26 EXTENSION_AUTOLINK
27 EXTENSION_STRIKETHROUGH
28 EXTENSION_LAX_HTML_BLOCKS
29 EXTENSION_SPACE_HEADERS
30)
31
32// These are the possible flag values for the link renderer.
33// Only a single one of these values will be used; they are not ORed together.
34// These are mostly of interest if you are writing a new output format.
35const (
36 LINK_TYPE_NOT_AUTOLINK = iota
37 LINK_TYPE_NORMAL
38 LINK_TYPE_EMAIL
39)
40
41// These are the possible flag values for the listitem renderer.
42// Multiple flag values may be ORed together.
43// These are mostly of interest if you are writing a new output format.
44const (
45 LIST_TYPE_ORDERED = 1 << iota
46 LIST_ITEM_CONTAINS_BLOCK
47 LIST_ITEM_END_OF_LIST
48)
49
50// These are the possible flag values for the table cell renderer.
51// Only a single one of these values will be used; they are not ORed together.
52// These are mostly of interest if you are writing a new output format.
53const (
54 TABLE_ALIGNMENT_LEFT = 1 << iota
55 TABLE_ALIGNMENT_RIGHT
56 TABLE_ALIGNMENT_CENTER = (TABLE_ALIGNMENT_LEFT | TABLE_ALIGNMENT_RIGHT)
57)
58
59// The size of a tab stop.
60const TAB_SIZE = 4
61
62// These are the tags that are recognized as HTML block tags.
63// Any of these can be included in markdown text without special escaping.
64var block_tags = map[string]bool{
65 "p": true,
66 "dl": true,
67 "h1": true,
68 "h2": true,
69 "h3": true,
70 "h4": true,
71 "h5": true,
72 "h6": true,
73 "ol": true,
74 "ul": true,
75 "del": true,
76 "div": true,
77 "ins": true,
78 "pre": true,
79 "form": true,
80 "math": true,
81 "table": true,
82 "iframe": true,
83 "script": true,
84 "fieldset": true,
85 "noscript": true,
86 "blockquote": true,
87}
88
89// This struct defines the rendering interface.
90// A series of callback functions are registered to form a complete renderer.
91// A single interface{} value field is provided, and that value is handed to
92// each callback. Leaving a field blank suppresses rendering that type of output
93// except where noted.
94//
95// This is mostly of interest if you are implementing a new rendering format.
96// Most users will use the convenience functions to fill in this structure.
97type Renderer struct {
98 // block-level callbacks---nil skips the block
99 BlockCode func(out *bytes.Buffer, text []byte, lang string, opaque interface{})
100 BlockQuote func(out *bytes.Buffer, text []byte, opaque interface{})
101 BlockHtml func(out *bytes.Buffer, text []byte, opaque interface{})
102 Header func(out *bytes.Buffer, text []byte, level int, opaque interface{})
103 HRule func(out *bytes.Buffer, opaque interface{})
104 List func(out *bytes.Buffer, text []byte, flags int, opaque interface{})
105 ListItem func(out *bytes.Buffer, text []byte, flags int, opaque interface{})
106 Paragraph func(out *bytes.Buffer, text []byte, opaque interface{})
107 Table func(out *bytes.Buffer, header []byte, body []byte, columnData []int, opaque interface{})
108 TableRow func(out *bytes.Buffer, text []byte, opaque interface{})
109 TableCell func(out *bytes.Buffer, text []byte, flags int, opaque interface{})
110
111 // Span-level callbacks---nil or return 0 prints the span verbatim
112 AutoLink func(out *bytes.Buffer, link []byte, kind int, opaque interface{}) int
113 CodeSpan func(out *bytes.Buffer, text []byte, opaque interface{}) int
114 DoubleEmphasis func(out *bytes.Buffer, text []byte, opaque interface{}) int
115 Emphasis func(out *bytes.Buffer, text []byte, opaque interface{}) int
116 Image func(out *bytes.Buffer, link []byte, title []byte, alt []byte, opaque interface{}) int
117 LineBreak func(out *bytes.Buffer, opaque interface{}) int
118 Link func(out *bytes.Buffer, link []byte, title []byte, content []byte, opaque interface{}) int
119 RawHtmlTag func(out *bytes.Buffer, tag []byte, opaque interface{}) int
120 TripleEmphasis func(out *bytes.Buffer, text []byte, opaque interface{}) int
121 StrikeThrough func(out *bytes.Buffer, text []byte, opaque interface{}) int
122
123 // Low-level callbacks---nil copies input directly into the output
124 Entity func(out *bytes.Buffer, entity []byte, opaque interface{})
125 NormalText func(out *bytes.Buffer, text []byte, opaque interface{})
126
127 // Header and footer
128 DocumentHeader func(out *bytes.Buffer, opaque interface{})
129 DocumentFooter func(out *bytes.Buffer, opaque interface{})
130
131 // User data---passed back to every callback
132 Opaque interface{}
133}
134
135type inlineParser func(out *bytes.Buffer, rndr *render, data []byte, offset int) int
136
137type render struct {
138 mk *Renderer
139 refs map[string]*reference
140 inline [256]inlineParser
141 flags uint32
142 nesting int
143 maxNesting int
144}
145
146
147//
148//
149// Public interface
150//
151//
152
153// Parse and render a block of markdown-encoded text.
154// The renderer is used to format the output, and extensions dictates which
155// non-standard extensions are enabled.
156func Markdown(input []byte, renderer *Renderer, extensions uint32) []byte {
157 // no point in parsing if we can't render
158 if renderer == nil {
159 return nil
160 }
161
162 // fill in the render structure
163 rndr := new(render)
164 rndr.mk = renderer
165 rndr.flags = extensions
166 rndr.refs = make(map[string]*reference)
167 rndr.maxNesting = 16
168
169 // register inline parsers
170 if rndr.mk.Emphasis != nil || rndr.mk.DoubleEmphasis != nil || rndr.mk.TripleEmphasis != nil {
171 rndr.inline['*'] = inlineEmphasis
172 rndr.inline['_'] = inlineEmphasis
173 if extensions&EXTENSION_STRIKETHROUGH != 0 {
174 rndr.inline['~'] = inlineEmphasis
175 }
176 }
177 if rndr.mk.CodeSpan != nil {
178 rndr.inline['`'] = inlineCodeSpan
179 }
180 if rndr.mk.LineBreak != nil {
181 rndr.inline['\n'] = inlineLineBreak
182 }
183 if rndr.mk.Image != nil || rndr.mk.Link != nil {
184 rndr.inline['['] = inlineLink
185 }
186 rndr.inline['<'] = inlineLAngle
187 rndr.inline['\\'] = inlineEscape
188 rndr.inline['&'] = inlineEntity
189
190 if extensions&EXTENSION_AUTOLINK != 0 {
191 rndr.inline[':'] = inlineAutoLink
192 }
193
194 // first pass: look for references, copy everything else
195 var text bytes.Buffer
196 beg, end := 0, 0
197 for beg < len(input) { // iterate over lines
198 if end = isReference(rndr, input[beg:]); end > 0 {
199 beg += end
200 } else { // skip to the next line
201 end = beg
202 for end < len(input) && input[end] != '\n' && input[end] != '\r' {
203 end++
204 }
205
206 // add the line body if present
207 if end > beg {
208 expandTabs(&text, input[beg:end])
209 }
210
211 for end < len(input) && (input[end] == '\n' || input[end] == '\r') {
212 // add one \n per newline
213 if input[end] == '\n' || (end+1 < len(input) && input[end+1] != '\n') {
214 text.WriteByte('\n')
215 }
216 end++
217 }
218
219 beg = end
220 }
221 }
222
223 // second pass: actual rendering
224 var output bytes.Buffer
225 if rndr.mk.DocumentHeader != nil {
226 rndr.mk.DocumentHeader(&output, rndr.mk.Opaque)
227 }
228
229 if text.Len() > 0 {
230 // add a final newline if not already present
231 finalchar := text.Bytes()[text.Len()-1]
232 if finalchar != '\n' && finalchar != '\r' {
233 text.WriteByte('\n')
234 }
235 parseBlock(&output, rndr, text.Bytes())
236 }
237
238 if rndr.mk.DocumentFooter != nil {
239 rndr.mk.DocumentFooter(&output, rndr.mk.Opaque)
240 }
241
242 if rndr.nesting != 0 {
243 panic("Nesting level did not end at zero")
244 }
245
246 return output.Bytes()
247}
248
249
250//
251// Link references
252//
253// This section implements support for references that (usually) appear
254// as footnotes in a document, and can be referenced anywhere in the document.
255// The basic format is:
256//
257// [1]: http://www.google.com/ "Google"
258// [2]: http://www.github.com/ "Github"
259//
260// Anywhere in the document, the reference can be linked by referring to its
261// label, i.e., 1 and 2 in this example, as in:
262//
263// This library is hosted on [Github][2], a git hosting site.
264
265// References are parsed and stored in this struct.
266type reference struct {
267 link []byte
268 title []byte
269}
270
271// Check whether or not data starts with a reference link.
272// If so, it is parsed and stored in the list of references
273// (in the render struct).
274// Returns the number of bytes to skip to move past it, or zero
275// if there is the first line is not a reference.
276func isReference(rndr *render, data []byte) int {
277 // up to 3 optional leading spaces
278 if len(data) < 4 {
279 return 0
280 }
281 i := 0
282 for i < 3 && data[i] == ' ' {
283 i++
284 }
285 if data[i] == ' ' {
286 return 0
287 }
288
289 // id part: anything but a newline between brackets
290 if data[i] != '[' {
291 return 0
292 }
293 i++
294 id_offset := i
295 for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
296 i++
297 }
298 if i >= len(data) || data[i] != ']' {
299 return 0
300 }
301 id_end := i
302
303 // spacer: colon (space | tab)* newline? (space | tab)*
304 i++
305 if i >= len(data) || data[i] != ':' {
306 return 0
307 }
308 i++
309 for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
310 i++
311 }
312 if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
313 i++
314 if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
315 i++
316 }
317 }
318 for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
319 i++
320 }
321 if i >= len(data) {
322 return 0
323 }
324
325 // link: whitespace-free sequence, optionally between angle brackets
326 if data[i] == '<' {
327 i++
328 }
329 link_offset := i
330 for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
331 i++
332 }
333 link_end := i
334 if data[link_offset] == '<' && data[link_end-1] == '>' {
335 link_offset++
336 link_end--
337 }
338
339 // optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
340 for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
341 i++
342 }
343 if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
344 return 0
345 }
346
347 // compute end-of-line
348 line_end := 0
349 if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
350 line_end = i
351 }
352 if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
353 line_end++
354 }
355
356 // optional (space|tab)* spacer after a newline
357 if line_end > 0 {
358 i = line_end + 1
359 for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
360 i++
361 }
362 }
363
364 // optional title: any non-newline sequence enclosed in '"() alone on its line
365 title_offset, title_end := 0, 0
366 if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
367 i++
368 title_offset = i
369
370 // look for EOL
371 for i < len(data) && data[i] != '\n' && data[i] != '\r' {
372 i++
373 }
374 if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
375 title_end = i + 1
376 } else {
377 title_end = i
378 }
379
380 // step back
381 i--
382 for i > title_offset && (data[i] == ' ' || data[i] == '\t') {
383 i--
384 }
385 if i > title_offset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
386 line_end = title_end
387 title_end = i
388 }
389 }
390 if line_end == 0 { // garbage after the link
391 return 0
392 }
393
394 // a valid ref has been found
395 if rndr == nil {
396 return line_end
397 }
398
399 // id matches are case-insensitive
400 id := string(bytes.ToLower(data[id_offset:id_end]))
401 rndr.refs[id] = &reference{
402 link: data[link_offset:link_end],
403 title: data[title_offset:title_end],
404 }
405
406 return line_end
407}
408
409
410//
411//
412// Miscellaneous helper functions
413//
414//
415
416
417// Test if a character is a punctuation symbol.
418// Taken from a private function in regexp in the stdlib.
419func ispunct(c byte) bool {
420 for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
421 if c == r {
422 return true
423 }
424 }
425 return false
426}
427
428// Test if a character is a whitespace character.
429func isspace(c byte) bool {
430 return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
431}
432
433// Test if a character is a letter or a digit.
434// TODO: check when this is looking for ASCII alnum and when it should use unicode
435func isalnum(c byte) bool {
436 return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
437}
438
439// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
440func expandTabs(out *bytes.Buffer, line []byte) {
441 // first, check for common cases: no tabs, or only tabs at beginning of line
442 i, prefix := 0, 0
443 slowcase := false
444 for i = 0; i < len(line); i++ {
445 if line[i] == '\t' {
446 if prefix == i {
447 prefix++
448 } else {
449 slowcase = true
450 break
451 }
452 }
453 }
454
455 // no need to decode runes if all tabs are at the beginning of the line
456 if !slowcase {
457 for i = 0; i < prefix*TAB_SIZE; i++ {
458 out.WriteByte(' ')
459 }
460 out.Write(line[prefix:])
461 return
462 }
463
464 // the slow case: we need to count runes to figure out how
465 // many spaces to insert for each tab
466 column := 0
467 i = 0
468 for i < len(line) {
469 start := i
470 for i < len(line) && line[i] != '\t' {
471 _, size := utf8.DecodeRune(line[i:])
472 i += size
473 column++
474 }
475
476 if i > start {
477 out.Write(line[start:i])
478 }
479
480 if i >= len(line) {
481 break
482 }
483
484 for {
485 out.WriteByte(' ')
486 column++
487 if column%TAB_SIZE == 0 {
488 break
489 }
490 }
491
492 i++
493 }
494}