all repos — grayfriday @ ef087889f49d63f4ada853a7b255072f85ec970f

blackfriday fork with a few changes

inline.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// Functions to parse inline elements.
  12//
  13
  14package blackfriday
  15
  16import (
  17	"bytes"
  18	"regexp"
  19	"strconv"
  20)
  21
  22var (
  23	urlRe    = `((https?|ftp):\/\/|\/)[-A-Za-z0-9+&@#\/%?=~_|!:,.;\(\)]+`
  24	anchorRe = regexp.MustCompile(`^(<a\shref="` + urlRe + `"(\stitle="[^"<>]+")?\s?>` + urlRe + `<\/a>)`)
  25)
  26
  27// Functions to parse text within a block
  28// Each function returns the number of chars taken care of
  29// data is the complete block being rendered
  30// offset is the number of valid chars before the current cursor
  31
  32func (p *parser) inline(out *bytes.Buffer, data []byte) {
  33	// this is called recursively: enforce a maximum depth
  34	if p.nesting >= p.maxNesting {
  35		return
  36	}
  37	p.nesting++
  38
  39	i, end := 0, 0
  40	for i < len(data) {
  41		// copy inactive chars into the output
  42		for end < len(data) && p.inlineCallback[data[end]] == nil {
  43			end++
  44		}
  45
  46		p.r.NormalText(out, data[i:end])
  47
  48		if end >= len(data) {
  49			break
  50		}
  51		i = end
  52
  53		// call the trigger
  54		handler := p.inlineCallback[data[end]]
  55		if consumed := handler(p, out, data, i); consumed == 0 {
  56			// no action from the callback; buffer the byte for later
  57			end = i + 1
  58		} else {
  59			// skip past whatever the callback used
  60			i += consumed
  61			end = i
  62		}
  63	}
  64
  65	p.nesting--
  66}
  67
  68// single and double emphasis parsing
  69func emphasis(p *parser, out *bytes.Buffer, data []byte, offset int) int {
  70	data = data[offset:]
  71	c := data[0]
  72	ret := 0
  73
  74	if len(data) > 2 && data[1] != c {
  75		// whitespace cannot follow an opening emphasis;
  76		// strikethrough only takes two characters '~~'
  77		if c == '~' || isspace(data[1]) {
  78			return 0
  79		}
  80		if ret = helperEmphasis(p, out, data[1:], c); ret == 0 {
  81			return 0
  82		}
  83
  84		return ret + 1
  85	}
  86
  87	if len(data) > 3 && data[1] == c && data[2] != c {
  88		if isspace(data[2]) {
  89			return 0
  90		}
  91		if ret = helperDoubleEmphasis(p, out, data[2:], c); ret == 0 {
  92			return 0
  93		}
  94
  95		return ret + 2
  96	}
  97
  98	if len(data) > 4 && data[1] == c && data[2] == c && data[3] != c {
  99		if c == '~' || isspace(data[3]) {
 100			return 0
 101		}
 102		if ret = helperTripleEmphasis(p, out, data, 3, c); ret == 0 {
 103			return 0
 104		}
 105
 106		return ret + 3
 107	}
 108
 109	return 0
 110}
 111
 112func codeSpan(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 113	data = data[offset:]
 114
 115	nb := 0
 116
 117	// count the number of backticks in the delimiter
 118	for nb < len(data) && data[nb] == '`' {
 119		nb++
 120	}
 121
 122	// find the next delimiter
 123	i, end := 0, 0
 124	for end = nb; end < len(data) && i < nb; end++ {
 125		if data[end] == '`' {
 126			i++
 127		} else {
 128			i = 0
 129		}
 130	}
 131
 132	// no matching delimiter?
 133	if i < nb && end >= len(data) {
 134		return 0
 135	}
 136
 137	// trim outside whitespace
 138	fBegin := nb
 139	for fBegin < end && data[fBegin] == ' ' {
 140		fBegin++
 141	}
 142
 143	fEnd := end - nb
 144	for fEnd > fBegin && data[fEnd-1] == ' ' {
 145		fEnd--
 146	}
 147
 148	// render the code span
 149	if fBegin != fEnd {
 150		p.r.CodeSpan(out, data[fBegin:fEnd])
 151	}
 152
 153	return end
 154
 155}
 156
 157// newline preceded by two spaces becomes <br>
 158// newline without two spaces works when EXTENSION_HARD_LINE_BREAK is enabled
 159func lineBreak(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 160	// remove trailing spaces from out
 161	outBytes := out.Bytes()
 162	end := len(outBytes)
 163	eol := end
 164	for eol > 0 && outBytes[eol-1] == ' ' {
 165		eol--
 166	}
 167	out.Truncate(eol)
 168
 169	precededByTwoSpaces := offset >= 2 && data[offset-2] == ' ' && data[offset-1] == ' '
 170	precededByBackslash := offset >= 1 && data[offset-1] == '\\' // see http://spec.commonmark.org/0.18/#example-527
 171	precededByBackslash = precededByBackslash && p.flags&BackslashLineBreak != 0
 172
 173	// should there be a hard line break here?
 174	if p.flags&HardLineBreak == 0 && !precededByTwoSpaces && !precededByBackslash {
 175		return 0
 176	}
 177
 178	if precededByBackslash && eol > 0 {
 179		out.Truncate(eol - 1)
 180	}
 181	p.r.LineBreak(out)
 182	return 1
 183}
 184
 185type linkType int
 186
 187const (
 188	linkNormal linkType = iota
 189	linkImg
 190	linkDeferredFootnote
 191	linkInlineFootnote
 192)
 193
 194func isReferenceStyleLink(data []byte, pos int, t linkType) bool {
 195	if t == linkDeferredFootnote {
 196		return false
 197	}
 198	return pos < len(data)-1 && data[pos] == '[' && data[pos+1] != '^'
 199}
 200
 201func maybeImage(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 202	if offset < len(data)-1 && data[offset+1] == '[' {
 203		return link(p, out, data, offset)
 204	}
 205	return 0
 206}
 207
 208func maybeInlineFootnote(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 209	if offset < len(data)-1 && data[offset+1] == '[' {
 210		return link(p, out, data, offset)
 211	}
 212	return 0
 213}
 214
 215// '[': parse a link or an image or a footnote
 216func link(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 217	// no links allowed inside regular links, footnote, and deferred footnotes
 218	if p.insideLink && (offset > 0 && data[offset-1] == '[' || len(data)-1 > offset && data[offset+1] == '^') {
 219		return 0
 220	}
 221
 222	var t linkType
 223	switch {
 224	// special case: ![^text] == deferred footnote (that follows something with
 225	// an exclamation point)
 226	case p.flags&Footnotes != 0 && len(data)-1 > offset && data[offset+1] == '^':
 227		t = linkDeferredFootnote
 228	// ![alt] == image
 229	case offset >= 0 && data[offset] == '!':
 230		t = linkImg
 231		offset += 1
 232	// ^[text] == inline footnote
 233	// [^refId] == deferred footnote
 234	case p.flags&Footnotes != 0:
 235		if offset >= 0 && data[offset] == '^' {
 236			t = linkInlineFootnote
 237			offset += 1
 238		} else if len(data)-1 > offset && data[offset+1] == '^' {
 239			t = linkDeferredFootnote
 240		}
 241	// [text] == regular link
 242	default:
 243		t = linkNormal
 244	}
 245
 246	data = data[offset:]
 247
 248	var (
 249		i                       = 1
 250		noteId                  int
 251		title, link, altContent []byte
 252		textHasNl               = false
 253	)
 254
 255	if t == linkDeferredFootnote {
 256		i++
 257	}
 258
 259	// look for the matching closing bracket
 260	for level := 1; level > 0 && i < len(data); i++ {
 261		switch {
 262		case data[i] == '\n':
 263			textHasNl = true
 264
 265		case data[i-1] == '\\':
 266			continue
 267
 268		case data[i] == '[':
 269			level++
 270
 271		case data[i] == ']':
 272			level--
 273			if level <= 0 {
 274				i-- // compensate for extra i++ in for loop
 275			}
 276		}
 277	}
 278
 279	if i >= len(data) {
 280		return 0
 281	}
 282
 283	txtE := i
 284	i++
 285
 286	// skip any amount of whitespace or newline
 287	// (this is much more lax than original markdown syntax)
 288	for i < len(data) && isspace(data[i]) {
 289		i++
 290	}
 291
 292	// inline style link
 293	switch {
 294	case i < len(data) && data[i] == '(':
 295		// skip initial whitespace
 296		i++
 297
 298		for i < len(data) && isspace(data[i]) {
 299			i++
 300		}
 301
 302		linkB := i
 303
 304		// look for link end: ' " )
 305	findlinkend:
 306		for i < len(data) {
 307			switch {
 308			case data[i] == '\\':
 309				i += 2
 310
 311			case data[i] == ')' || data[i] == '\'' || data[i] == '"':
 312				break findlinkend
 313
 314			default:
 315				i++
 316			}
 317		}
 318
 319		if i >= len(data) {
 320			return 0
 321		}
 322		linkE := i
 323
 324		// look for title end if present
 325		titleB, titleE := 0, 0
 326		if data[i] == '\'' || data[i] == '"' {
 327			i++
 328			titleB = i
 329
 330		findtitleend:
 331			for i < len(data) {
 332				switch {
 333				case data[i] == '\\':
 334					i += 2
 335
 336				case data[i] == ')':
 337					break findtitleend
 338
 339				default:
 340					i++
 341				}
 342			}
 343
 344			if i >= len(data) {
 345				return 0
 346			}
 347
 348			// skip whitespace after title
 349			titleE = i - 1
 350			for titleE > titleB && isspace(data[titleE]) {
 351				titleE--
 352			}
 353
 354			// check for closing quote presence
 355			if data[titleE] != '\'' && data[titleE] != '"' {
 356				titleB, titleE = 0, 0
 357				linkE = i
 358			}
 359		}
 360
 361		// remove whitespace at the end of the link
 362		for linkE > linkB && isspace(data[linkE-1]) {
 363			linkE--
 364		}
 365
 366		// remove optional angle brackets around the link
 367		if data[linkB] == '<' {
 368			linkB++
 369		}
 370		if data[linkE-1] == '>' {
 371			linkE--
 372		}
 373
 374		// build escaped link and title
 375		if linkE > linkB {
 376			link = data[linkB:linkE]
 377		}
 378
 379		if titleE > titleB {
 380			title = data[titleB:titleE]
 381		}
 382
 383		i++
 384
 385	// reference style link
 386	case isReferenceStyleLink(data, i, t):
 387		var id []byte
 388		altContentConsidered := false
 389
 390		// look for the id
 391		i++
 392		linkB := i
 393		for i < len(data) && data[i] != ']' {
 394			i++
 395		}
 396		if i >= len(data) {
 397			return 0
 398		}
 399		linkE := i
 400
 401		// find the reference
 402		if linkB == linkE {
 403			if textHasNl {
 404				var b bytes.Buffer
 405
 406				for j := 1; j < txtE; j++ {
 407					switch {
 408					case data[j] != '\n':
 409						b.WriteByte(data[j])
 410					case data[j-1] != ' ':
 411						b.WriteByte(' ')
 412					}
 413				}
 414
 415				id = b.Bytes()
 416			} else {
 417				id = data[1:txtE]
 418				altContentConsidered = true
 419			}
 420		} else {
 421			id = data[linkB:linkE]
 422		}
 423
 424		// find the reference with matching id
 425		lr, ok := p.getRef(string(id))
 426		if !ok {
 427			return 0
 428		}
 429
 430		// keep link and title from reference
 431		link = lr.link
 432		title = lr.title
 433		if altContentConsidered {
 434			altContent = lr.text
 435		}
 436		i++
 437
 438	// shortcut reference style link or reference or inline footnote
 439	default:
 440		var id []byte
 441
 442		// craft the id
 443		if textHasNl {
 444			var b bytes.Buffer
 445
 446			for j := 1; j < txtE; j++ {
 447				switch {
 448				case data[j] != '\n':
 449					b.WriteByte(data[j])
 450				case data[j-1] != ' ':
 451					b.WriteByte(' ')
 452				}
 453			}
 454
 455			id = b.Bytes()
 456		} else {
 457			if t == linkDeferredFootnote {
 458				id = data[2:txtE] // get rid of the ^
 459			} else {
 460				id = data[1:txtE]
 461			}
 462		}
 463
 464		if t == linkInlineFootnote {
 465			// create a new reference
 466			noteId = len(p.notes) + 1
 467
 468			var fragment []byte
 469			if len(id) > 0 {
 470				if len(id) < 16 {
 471					fragment = make([]byte, len(id))
 472				} else {
 473					fragment = make([]byte, 16)
 474				}
 475				copy(fragment, slugify(id))
 476			} else {
 477				fragment = append([]byte("footnote-"), []byte(strconv.Itoa(noteId))...)
 478			}
 479
 480			ref := &reference{
 481				noteId:   noteId,
 482				hasBlock: false,
 483				link:     fragment,
 484				title:    id,
 485			}
 486
 487			p.notes = append(p.notes, ref)
 488
 489			link = ref.link
 490			title = ref.title
 491		} else {
 492			// find the reference with matching id
 493			lr, ok := p.getRef(string(id))
 494			if !ok {
 495				return 0
 496			}
 497
 498			if t == linkDeferredFootnote {
 499				lr.noteId = len(p.notes) + 1
 500				p.notes = append(p.notes, lr)
 501			}
 502
 503			// keep link and title from reference
 504			link = lr.link
 505			// if inline footnote, title == footnote contents
 506			title = lr.title
 507			noteId = lr.noteId
 508		}
 509
 510		// rewind the whitespace
 511		i = txtE + 1
 512	}
 513
 514	// build content: img alt is escaped, link content is parsed
 515	var content bytes.Buffer
 516	if txtE > 1 {
 517		if t == linkImg {
 518			content.Write(data[1:txtE])
 519		} else {
 520			// links cannot contain other links, so turn off link parsing temporarily
 521			insideLink := p.insideLink
 522			p.insideLink = true
 523			p.inline(&content, data[1:txtE])
 524			p.insideLink = insideLink
 525		}
 526	}
 527
 528	var uLink []byte
 529	if t == linkNormal || t == linkImg {
 530		if len(link) > 0 {
 531			var uLinkBuf bytes.Buffer
 532			unescapeText(&uLinkBuf, link)
 533			uLink = uLinkBuf.Bytes()
 534		}
 535
 536		// links need something to click on and somewhere to go
 537		if len(uLink) == 0 || (t == linkNormal && content.Len() == 0) {
 538			return 0
 539		}
 540	}
 541
 542	// call the relevant rendering function
 543	switch t {
 544	case linkNormal:
 545		if len(altContent) > 0 {
 546			p.r.Link(out, uLink, title, altContent)
 547		} else {
 548			p.r.Link(out, uLink, title, content.Bytes())
 549		}
 550
 551	case linkImg:
 552		p.r.Image(out, uLink, title, content.Bytes())
 553		i += 1
 554
 555	case linkInlineFootnote:
 556		p.r.FootnoteRef(out, link, noteId)
 557		i += 1
 558
 559	case linkDeferredFootnote:
 560		p.r.FootnoteRef(out, link, noteId)
 561
 562	default:
 563		return 0
 564	}
 565
 566	return i
 567}
 568
 569func (p *parser) inlineHtmlComment(out *bytes.Buffer, data []byte) int {
 570	if len(data) < 5 {
 571		return 0
 572	}
 573	if data[0] != '<' || data[1] != '!' || data[2] != '-' || data[3] != '-' {
 574		return 0
 575	}
 576	i := 5
 577	// scan for an end-of-comment marker, across lines if necessary
 578	for i < len(data) && !(data[i-2] == '-' && data[i-1] == '-' && data[i] == '>') {
 579		i++
 580	}
 581	// no end-of-comment marker
 582	if i >= len(data) {
 583		return 0
 584	}
 585	return i + 1
 586}
 587
 588// '<' when tags or autolinks are allowed
 589func leftAngle(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 590	data = data[offset:]
 591	altype := LinkTypeNotAutolink
 592	end := tagLength(data, &altype)
 593	if size := p.inlineHtmlComment(out, data); size > 0 {
 594		end = size
 595	}
 596	if end > 2 {
 597		if altype != LinkTypeNotAutolink {
 598			var uLink bytes.Buffer
 599			unescapeText(&uLink, data[1:end+1-2])
 600			if uLink.Len() > 0 {
 601				p.r.AutoLink(out, uLink.Bytes(), altype)
 602			}
 603		} else {
 604			p.r.RawHtmlTag(out, data[:end])
 605		}
 606	}
 607
 608	return end
 609}
 610
 611// '\\' backslash escape
 612var escapeChars = []byte("\\`*_{}[]()#+-.!:|&<>~")
 613
 614func escape(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 615	data = data[offset:]
 616
 617	if len(data) > 1 {
 618		if bytes.IndexByte(escapeChars, data[1]) < 0 {
 619			return 0
 620		}
 621
 622		p.r.NormalText(out, data[1:2])
 623	}
 624
 625	return 2
 626}
 627
 628func unescapeText(ob *bytes.Buffer, src []byte) {
 629	i := 0
 630	for i < len(src) {
 631		org := i
 632		for i < len(src) && src[i] != '\\' {
 633			i++
 634		}
 635
 636		if i > org {
 637			ob.Write(src[org:i])
 638		}
 639
 640		if i+1 >= len(src) {
 641			break
 642		}
 643
 644		ob.WriteByte(src[i+1])
 645		i += 2
 646	}
 647}
 648
 649// '&' escaped when it doesn't belong to an entity
 650// valid entities are assumed to be anything matching &#?[A-Za-z0-9]+;
 651func entity(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 652	data = data[offset:]
 653
 654	end := 1
 655
 656	if end < len(data) && data[end] == '#' {
 657		end++
 658	}
 659
 660	for end < len(data) && isalnum(data[end]) {
 661		end++
 662	}
 663
 664	if end < len(data) && data[end] == ';' {
 665		end++ // real entity
 666	} else {
 667		return 0 // lone '&'
 668	}
 669
 670	p.r.Entity(out, data[:end])
 671
 672	return end
 673}
 674
 675func linkEndsWithEntity(data []byte, linkEnd int) bool {
 676	entityRanges := htmlEntity.FindAllIndex(data[:linkEnd], -1)
 677	return entityRanges != nil && entityRanges[len(entityRanges)-1][1] == linkEnd
 678}
 679
 680func autoLink(p *parser, out *bytes.Buffer, data []byte, offset int) int {
 681	// quick check to rule out most false hits on ':'
 682	if p.insideLink || len(data) < offset+3 || data[offset+1] != '/' || data[offset+2] != '/' {
 683		return 0
 684	}
 685
 686	// Now a more expensive check to see if we're not inside an anchor element
 687	anchorStart := offset
 688	offsetFromAnchor := 0
 689	for anchorStart > 0 && data[anchorStart] != '<' {
 690		anchorStart--
 691		offsetFromAnchor++
 692	}
 693
 694	anchorStr := anchorRe.Find(data[anchorStart:])
 695	if anchorStr != nil {
 696		out.Write(anchorStr[offsetFromAnchor:])
 697		return len(anchorStr) - offsetFromAnchor
 698	}
 699
 700	// scan backward for a word boundary
 701	rewind := 0
 702	for offset-rewind > 0 && rewind <= 7 && isletter(data[offset-rewind-1]) {
 703		rewind++
 704	}
 705	if rewind > 6 { // longest supported protocol is "mailto" which has 6 letters
 706		return 0
 707	}
 708
 709	origData := data
 710	data = data[offset-rewind:]
 711
 712	if !isSafeLink(data) {
 713		return 0
 714	}
 715
 716	linkEnd := 0
 717	for linkEnd < len(data) && !isEndOfLink(data[linkEnd]) {
 718		linkEnd++
 719	}
 720
 721	// Skip punctuation at the end of the link
 722	if (data[linkEnd-1] == '.' || data[linkEnd-1] == ',') && data[linkEnd-2] != '\\' {
 723		linkEnd--
 724	}
 725
 726	// But don't skip semicolon if it's a part of escaped entity:
 727	if data[linkEnd-1] == ';' && data[linkEnd-2] != '\\' && !linkEndsWithEntity(data, linkEnd) {
 728		linkEnd--
 729	}
 730
 731	// See if the link finishes with a punctuation sign that can be closed.
 732	var copen byte
 733	switch data[linkEnd-1] {
 734	case '"':
 735		copen = '"'
 736	case '\'':
 737		copen = '\''
 738	case ')':
 739		copen = '('
 740	case ']':
 741		copen = '['
 742	case '}':
 743		copen = '{'
 744	default:
 745		copen = 0
 746	}
 747
 748	if copen != 0 {
 749		bufEnd := offset - rewind + linkEnd - 2
 750
 751		openDelim := 1
 752
 753		/* Try to close the final punctuation sign in this same line;
 754		 * if we managed to close it outside of the URL, that means that it's
 755		 * not part of the URL. If it closes inside the URL, that means it
 756		 * is part of the URL.
 757		 *
 758		 * Examples:
 759		 *
 760		 *      foo http://www.pokemon.com/Pikachu_(Electric) bar
 761		 *              => http://www.pokemon.com/Pikachu_(Electric)
 762		 *
 763		 *      foo (http://www.pokemon.com/Pikachu_(Electric)) bar
 764		 *              => http://www.pokemon.com/Pikachu_(Electric)
 765		 *
 766		 *      foo http://www.pokemon.com/Pikachu_(Electric)) bar
 767		 *              => http://www.pokemon.com/Pikachu_(Electric))
 768		 *
 769		 *      (foo http://www.pokemon.com/Pikachu_(Electric)) bar
 770		 *              => foo http://www.pokemon.com/Pikachu_(Electric)
 771		 */
 772
 773		for bufEnd >= 0 && origData[bufEnd] != '\n' && openDelim != 0 {
 774			if origData[bufEnd] == data[linkEnd-1] {
 775				openDelim++
 776			}
 777
 778			if origData[bufEnd] == copen {
 779				openDelim--
 780			}
 781
 782			bufEnd--
 783		}
 784
 785		if openDelim == 0 {
 786			linkEnd--
 787		}
 788	}
 789
 790	// we were triggered on the ':', so we need to rewind the output a bit
 791	if out.Len() >= rewind {
 792		out.Truncate(len(out.Bytes()) - rewind)
 793	}
 794
 795	var uLink bytes.Buffer
 796	unescapeText(&uLink, data[:linkEnd])
 797
 798	if uLink.Len() > 0 {
 799		p.r.AutoLink(out, uLink.Bytes(), LinkTypeNormal)
 800	}
 801
 802	return linkEnd - rewind
 803}
 804
 805func isEndOfLink(char byte) bool {
 806	return isspace(char) || char == '<'
 807}
 808
 809var validUris = [][]byte{[]byte("http://"), []byte("https://"), []byte("ftp://"), []byte("mailto://")}
 810var validPaths = [][]byte{[]byte("/"), []byte("./"), []byte("../")}
 811
 812func isSafeLink(link []byte) bool {
 813	for _, path := range validPaths {
 814		if len(link) >= len(path) && bytes.Equal(link[:len(path)], path) {
 815			if len(link) == len(path) {
 816				return true
 817			} else if isalnum(link[len(path)]) {
 818				return true
 819			}
 820		}
 821	}
 822
 823	for _, prefix := range validUris {
 824		// TODO: handle unicode here
 825		// case-insensitive prefix test
 826		if len(link) > len(prefix) && bytes.Equal(bytes.ToLower(link[:len(prefix)]), prefix) && isalnum(link[len(prefix)]) {
 827			return true
 828		}
 829	}
 830
 831	return false
 832}
 833
 834// return the length of the given tag, or 0 is it's not valid
 835func tagLength(data []byte, autolink *LinkType) int {
 836	var i, j int
 837
 838	// a valid tag can't be shorter than 3 chars
 839	if len(data) < 3 {
 840		return 0
 841	}
 842
 843	// begins with a '<' optionally followed by '/', followed by letter or number
 844	if data[0] != '<' {
 845		return 0
 846	}
 847	if data[1] == '/' {
 848		i = 2
 849	} else {
 850		i = 1
 851	}
 852
 853	if !isalnum(data[i]) {
 854		return 0
 855	}
 856
 857	// scheme test
 858	*autolink = LinkTypeNotAutolink
 859
 860	// try to find the beginning of an URI
 861	for i < len(data) && (isalnum(data[i]) || data[i] == '.' || data[i] == '+' || data[i] == '-') {
 862		i++
 863	}
 864
 865	if i > 1 && i < len(data) && data[i] == '@' {
 866		if j = isMailtoAutoLink(data[i:]); j != 0 {
 867			*autolink = LinkTypeEmail
 868			return i + j
 869		}
 870	}
 871
 872	if i > 2 && i < len(data) && data[i] == ':' {
 873		*autolink = LinkTypeNormal
 874		i++
 875	}
 876
 877	// complete autolink test: no whitespace or ' or "
 878	switch {
 879	case i >= len(data):
 880		*autolink = LinkTypeNotAutolink
 881	case *autolink != 0:
 882		j = i
 883
 884		for i < len(data) {
 885			if data[i] == '\\' {
 886				i += 2
 887			} else if data[i] == '>' || data[i] == '\'' || data[i] == '"' || isspace(data[i]) {
 888				break
 889			} else {
 890				i++
 891			}
 892
 893		}
 894
 895		if i >= len(data) {
 896			return 0
 897		}
 898		if i > j && data[i] == '>' {
 899			return i + 1
 900		}
 901
 902		// one of the forbidden chars has been found
 903		*autolink = LinkTypeNotAutolink
 904	}
 905
 906	// look for something looking like a tag end
 907	for i < len(data) && data[i] != '>' {
 908		i++
 909	}
 910	if i >= len(data) {
 911		return 0
 912	}
 913	return i + 1
 914}
 915
 916// look for the address part of a mail autolink and '>'
 917// this is less strict than the original markdown e-mail address matching
 918func isMailtoAutoLink(data []byte) int {
 919	nb := 0
 920
 921	// address is assumed to be: [-@._a-zA-Z0-9]+ with exactly one '@'
 922	for i := 0; i < len(data); i++ {
 923		if isalnum(data[i]) {
 924			continue
 925		}
 926
 927		switch data[i] {
 928		case '@':
 929			nb++
 930
 931		case '-', '.', '_':
 932			break
 933
 934		case '>':
 935			if nb == 1 {
 936				return i + 1
 937			} else {
 938				return 0
 939			}
 940		default:
 941			return 0
 942		}
 943	}
 944
 945	return 0
 946}
 947
 948// look for the next emph char, skipping other constructs
 949func helperFindEmphChar(data []byte, c byte) int {
 950	i := 0
 951
 952	for i < len(data) {
 953		for i < len(data) && data[i] != c && data[i] != '`' && data[i] != '[' {
 954			i++
 955		}
 956		if i >= len(data) {
 957			return 0
 958		}
 959		// do not count escaped chars
 960		if i != 0 && data[i-1] == '\\' {
 961			i++
 962			continue
 963		}
 964		if data[i] == c {
 965			return i
 966		}
 967
 968		if data[i] == '`' {
 969			// skip a code span
 970			tmpI := 0
 971			i++
 972			for i < len(data) && data[i] != '`' {
 973				if tmpI == 0 && data[i] == c {
 974					tmpI = i
 975				}
 976				i++
 977			}
 978			if i >= len(data) {
 979				return tmpI
 980			}
 981			i++
 982		} else if data[i] == '[' {
 983			// skip a link
 984			tmpI := 0
 985			i++
 986			for i < len(data) && data[i] != ']' {
 987				if tmpI == 0 && data[i] == c {
 988					tmpI = i
 989				}
 990				i++
 991			}
 992			i++
 993			for i < len(data) && (data[i] == ' ' || data[i] == '\n') {
 994				i++
 995			}
 996			if i >= len(data) {
 997				return tmpI
 998			}
 999			if data[i] != '[' && data[i] != '(' { // not a link
1000				if tmpI > 0 {
1001					return tmpI
1002				} else {
1003					continue
1004				}
1005			}
1006			cc := data[i]
1007			i++
1008			for i < len(data) && data[i] != cc {
1009				if tmpI == 0 && data[i] == c {
1010					return i
1011				}
1012				i++
1013			}
1014			if i >= len(data) {
1015				return tmpI
1016			}
1017			i++
1018		}
1019	}
1020	return 0
1021}
1022
1023func helperEmphasis(p *parser, out *bytes.Buffer, data []byte, c byte) int {
1024	i := 0
1025
1026	// skip one symbol if coming from emph3
1027	if len(data) > 1 && data[0] == c && data[1] == c {
1028		i = 1
1029	}
1030
1031	for i < len(data) {
1032		length := helperFindEmphChar(data[i:], c)
1033		if length == 0 {
1034			return 0
1035		}
1036		i += length
1037		if i >= len(data) {
1038			return 0
1039		}
1040
1041		if i+1 < len(data) && data[i+1] == c {
1042			i++
1043			continue
1044		}
1045
1046		if data[i] == c && !isspace(data[i-1]) {
1047
1048			if p.flags&NoIntraEmphasis != 0 {
1049				if !(i+1 == len(data) || isspace(data[i+1]) || ispunct(data[i+1])) {
1050					continue
1051				}
1052			}
1053
1054			var work bytes.Buffer
1055			p.inline(&work, data[:i])
1056			p.r.Emphasis(out, work.Bytes())
1057			return i + 1
1058		}
1059	}
1060
1061	return 0
1062}
1063
1064func helperDoubleEmphasis(p *parser, out *bytes.Buffer, data []byte, c byte) int {
1065	i := 0
1066
1067	for i < len(data) {
1068		length := helperFindEmphChar(data[i:], c)
1069		if length == 0 {
1070			return 0
1071		}
1072		i += length
1073
1074		if i+1 < len(data) && data[i] == c && data[i+1] == c && i > 0 && !isspace(data[i-1]) {
1075			var work bytes.Buffer
1076			p.inline(&work, data[:i])
1077
1078			if work.Len() > 0 {
1079				// pick the right renderer
1080				if c == '~' {
1081					p.r.StrikeThrough(out, work.Bytes())
1082				} else {
1083					p.r.DoubleEmphasis(out, work.Bytes())
1084				}
1085			}
1086			return i + 2
1087		}
1088		i++
1089	}
1090	return 0
1091}
1092
1093func helperTripleEmphasis(p *parser, out *bytes.Buffer, data []byte, offset int, c byte) int {
1094	i := 0
1095	origData := data
1096	data = data[offset:]
1097
1098	for i < len(data) {
1099		length := helperFindEmphChar(data[i:], c)
1100		if length == 0 {
1101			return 0
1102		}
1103		i += length
1104
1105		// skip whitespace preceded symbols
1106		if data[i] != c || isspace(data[i-1]) {
1107			continue
1108		}
1109
1110		switch {
1111		case i+2 < len(data) && data[i+1] == c && data[i+2] == c:
1112			// triple symbol found
1113			var work bytes.Buffer
1114
1115			p.inline(&work, data[:i])
1116			if work.Len() > 0 {
1117				p.r.TripleEmphasis(out, work.Bytes())
1118			}
1119			return i + 3
1120		case (i+1 < len(data) && data[i+1] == c):
1121			// double symbol found, hand over to emph1
1122			length = helperEmphasis(p, out, origData[offset-2:], c)
1123			if length == 0 {
1124				return 0
1125			} else {
1126				return length - 2
1127			}
1128		default:
1129			// single symbol found, hand over to emph2
1130			length = helperDoubleEmphasis(p, out, origData[offset-1:], c)
1131			if length == 0 {
1132				return 0
1133			} else {
1134				return length - 1
1135			}
1136		}
1137	}
1138	return 0
1139}