all repos — grayfriday @ 55cd82008e9b35b9a03a80e06d5a4c4601320211

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