all repos — grayfriday @ 093273323af9011bf873707b22b7659703557904

blackfriday fork with a few changes

block.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 block-level elements.
  12//
  13
  14package blackfriday
  15
  16import (
  17	"bytes"
  18)
  19
  20// Parse block-level data.
  21// Note: this function and many that it calls assume that
  22// the input buffer ends with a newline.
  23func (p *parser) block(out *bytes.Buffer, data []byte) {
  24	if len(data) == 0 || data[len(data)-1] != '\n' {
  25		panic("block input is missing terminating newline")
  26	}
  27
  28	// this is called recursively: enforce a maximum depth
  29	if p.nesting >= p.maxNesting {
  30		return
  31	}
  32	p.nesting++
  33
  34	lastLen := 0
  35	sameLenCount := 0
  36
  37	// parse out one block-level construct at a time
  38	for len(data) > 0 {
  39		curLen := len(data)
  40		if curLen == lastLen {
  41			sameLenCount += 1
  42			if sameLenCount >= 3 {
  43				// infinity loop detection
  44				return
  45			}
  46		} else {
  47			sameLenCount = 0
  48		}
  49		lastLen = curLen
  50
  51		// prefixed header:
  52		//
  53		// # Header 1
  54		// ## Header 2
  55		// ...
  56		// ###### Header 6
  57		if p.isPrefixHeader(data) {
  58			data = data[p.prefixHeader(out, data):]
  59			continue
  60		}
  61
  62		// block of preformatted HTML:
  63		//
  64		// <div>
  65		//     ...
  66		// </div>
  67		if data[0] == '<' {
  68			if i := p.html(out, data, true); i > 0 {
  69				data = data[i:]
  70				continue
  71			}
  72		}
  73
  74		// blank lines.  note: returns the # of bytes to skip
  75		if i := p.isEmpty(data); i > 0 {
  76			data = data[i:]
  77			continue
  78		}
  79
  80		// indented code block:
  81		//
  82		//     func max(a, b int) int {
  83		//         if a > b {
  84		//             return a
  85		//         }
  86		//         return b
  87		//      }
  88		if p.codePrefix(data) > 0 {
  89			data = data[p.code(out, data):]
  90			continue
  91		}
  92
  93		// fenced code block:
  94		//
  95		// ``` go
  96		// func fact(n int) int {
  97		//     if n <= 1 {
  98		//         return n
  99		//     }
 100		//     return n * fact(n-1)
 101		// }
 102		// ```
 103		if p.flags&EXTENSION_FENCED_CODE != 0 {
 104			if i := p.fencedCode(out, data, true); i > 0 {
 105				data = data[i:]
 106				continue
 107			}
 108		}
 109
 110		// horizontal rule:
 111		//
 112		// ------
 113		// or
 114		// ******
 115		// or
 116		// ______
 117		if p.isHRule(data) {
 118			p.r.HRule(out)
 119			var i int
 120			for i = 0; data[i] != '\n'; i++ {
 121			}
 122			data = data[i:]
 123			continue
 124		}
 125
 126		// block quote:
 127		//
 128		// > A big quote I found somewhere
 129		// > on the web
 130		if p.quotePrefix(data) > 0 {
 131			data = data[p.quote(out, data):]
 132			continue
 133		}
 134
 135		// table:
 136		//
 137		// Name  | Age | Phone
 138		// ------|-----|---------
 139		// Bob   | 31  | 555-1234
 140		// Alice | 27  | 555-4321
 141		if p.flags&EXTENSION_TABLES != 0 {
 142			if i := p.table(out, data); i > 0 {
 143				data = data[i:]
 144				continue
 145			}
 146		}
 147
 148		// an itemized/unordered list:
 149		//
 150		// * Item 1
 151		// * Item 2
 152		//
 153		// also works with + or -
 154		if p.uliPrefix(data) > 0 {
 155			data = data[p.list(out, data, 0):]
 156			continue
 157		}
 158
 159		// a numbered/ordered list:
 160		//
 161		// 1. Item 1
 162		// 2. Item 2
 163		if p.oliPrefix(data) > 0 {
 164			data = data[p.list(out, data, LIST_TYPE_ORDERED):]
 165			continue
 166		}
 167
 168		// anything else must look like a normal paragraph
 169		// note: this finds underlined headers, too
 170		data = data[p.paragraph(out, data):]
 171	}
 172
 173	p.nesting--
 174}
 175
 176func (p *parser) isPrefixHeader(data []byte) bool {
 177	if data[0] != '#' {
 178		return false
 179	}
 180
 181	if p.flags&EXTENSION_SPACE_HEADERS != 0 {
 182		level := 0
 183		for level < 6 && data[level] == '#' {
 184			level++
 185		}
 186		if data[level] != ' ' {
 187			return false
 188		}
 189	}
 190	return true
 191}
 192
 193func (p *parser) prefixHeader(out *bytes.Buffer, data []byte) int {
 194	level := 0
 195	for level < 6 && data[level] == '#' {
 196		level++
 197	}
 198	i, end := 0, 0
 199	for i = level; data[i] == ' '; i++ {
 200	}
 201	for end = i; data[end] != '\n'; end++ {
 202	}
 203	skip := end
 204	for end > 0 && data[end-1] == '#' {
 205		end--
 206	}
 207	for end > 0 && data[end-1] == ' ' {
 208		end--
 209	}
 210	if end > i {
 211		work := func() bool {
 212			p.inline(out, data[i:end])
 213			return true
 214		}
 215		p.r.Header(out, work, level)
 216	}
 217	return skip
 218}
 219
 220func (p *parser) isUnderlinedHeader(data []byte) int {
 221	// test of level 1 header
 222	if data[0] == '=' {
 223		i := 1
 224		for data[i] == '=' {
 225			i++
 226		}
 227		for data[i] == ' ' {
 228			i++
 229		}
 230		if data[i] == '\n' {
 231			return 1
 232		} else {
 233			return 0
 234		}
 235	}
 236
 237	// test of level 2 header
 238	if data[0] == '-' {
 239		i := 1
 240		for data[i] == '-' {
 241			i++
 242		}
 243		for data[i] == ' ' {
 244			i++
 245		}
 246		if data[i] == '\n' {
 247			return 2
 248		} else {
 249			return 0
 250		}
 251	}
 252
 253	return 0
 254}
 255
 256func (p *parser) html(out *bytes.Buffer, data []byte, doRender bool) int {
 257	var i, j int
 258
 259	// identify the opening tag
 260	if data[0] != '<' {
 261		return 0
 262	}
 263	curtag, tagfound := p.htmlFindTag(data[1:])
 264
 265	// handle special cases
 266	if !tagfound {
 267		// check for an HTML comment
 268		if size := p.htmlComment(out, data, doRender); size > 0 {
 269			return size
 270		}
 271
 272		// check for an <hr> tag
 273		if size := p.htmlHr(out, data, doRender); size > 0 {
 274			return size
 275		}
 276
 277		// no special case recognized
 278		return 0
 279	}
 280
 281	// look for an unindented matching closing tag
 282	// followed by a blank line
 283	found := false
 284	/*
 285		closetag := []byte("\n</" + curtag + ">")
 286		j = len(curtag) + 1
 287		for !found {
 288			// scan for a closing tag at the beginning of a line
 289			if skip := bytes.Index(data[j:], closetag); skip >= 0 {
 290				j += skip + len(closetag)
 291			} else {
 292				break
 293			}
 294
 295			// see if it is the only thing on the line
 296			if skip := p.isEmpty(data[j:]); skip > 0 {
 297				// see if it is followed by a blank line/eof
 298				j += skip
 299				if j >= len(data) {
 300					found = true
 301					i = j
 302				} else {
 303					if skip := p.isEmpty(data[j:]); skip > 0 {
 304						j += skip
 305						found = true
 306						i = j
 307					}
 308				}
 309			}
 310		}
 311	*/
 312
 313	// if not found, try a second pass looking for indented match
 314	// but not if tag is "ins" or "del" (following original Markdown.pl)
 315	if !found && curtag != "ins" && curtag != "del" {
 316		i = 1
 317		for i < len(data) {
 318			i++
 319			for i < len(data) && !(data[i-1] == '<' && data[i] == '/') {
 320				i++
 321			}
 322
 323			if i+2+len(curtag) >= len(data) {
 324				break
 325			}
 326
 327			j = p.htmlFindEnd(curtag, data[i-1:])
 328
 329			if j > 0 {
 330				i += j - 1
 331				found = true
 332				break
 333			}
 334		}
 335	}
 336
 337	if !found {
 338		return 0
 339	}
 340
 341	// the end of the block has been found
 342	if doRender {
 343		// trim newlines
 344		end := i
 345		for end > 0 && data[end-1] == '\n' {
 346			end--
 347		}
 348		p.r.BlockHtml(out, data[:end])
 349	}
 350
 351	return i
 352}
 353
 354// HTML comment, lax form
 355func (p *parser) htmlComment(out *bytes.Buffer, data []byte, doRender bool) int {
 356	if data[0] != '<' || data[1] != '!' || data[2] != '-' || data[3] != '-' {
 357		return 0
 358	}
 359
 360	i := 5
 361
 362	// scan for an end-of-comment marker, across lines if necessary
 363	for i < len(data) && !(data[i-2] == '-' && data[i-1] == '-' && data[i] == '>') {
 364		i++
 365	}
 366	i++
 367
 368	// no end-of-comment marker
 369	if i >= len(data) {
 370		return 0
 371	}
 372
 373	// needs to end with a blank line
 374	if j := p.isEmpty(data[i:]); j > 0 {
 375		size := i + j
 376		if doRender {
 377			// trim trailing newlines
 378			end := size
 379			for end > 0 && data[end-1] == '\n' {
 380				end--
 381			}
 382			p.r.BlockHtml(out, data[:end])
 383		}
 384		return size
 385	}
 386
 387	return 0
 388}
 389
 390// HR, which is the only self-closing block tag considered
 391func (p *parser) htmlHr(out *bytes.Buffer, data []byte, doRender bool) int {
 392	if data[0] != '<' || (data[1] != 'h' && data[1] != 'H') || (data[2] != 'r' && data[2] != 'R') {
 393		return 0
 394	}
 395	if data[3] != ' ' && data[3] != '/' && data[3] != '>' {
 396		// not an <hr> tag after all; at least not a valid one
 397		return 0
 398	}
 399
 400	i := 3
 401	for data[i] != '>' && data[i] != '\n' {
 402		i++
 403	}
 404
 405	if data[i] == '>' {
 406		i++
 407		if j := p.isEmpty(data[i:]); j > 0 {
 408			size := i + j
 409			if doRender {
 410				// trim newlines
 411				end := size
 412				for end > 0 && data[end-1] == '\n' {
 413					end--
 414				}
 415				p.r.BlockHtml(out, data[:end])
 416			}
 417			return size
 418		}
 419	}
 420
 421	return 0
 422}
 423
 424func (p *parser) htmlFindTag(data []byte) (string, bool) {
 425	i := 0
 426	for isalnum(data[i]) {
 427		i++
 428	}
 429	key := string(data[:i])
 430	if blockTags[key] {
 431		return key, true
 432	}
 433	return "", false
 434}
 435
 436func (p *parser) htmlFindEnd(tag string, data []byte) int {
 437	// assume data[0] == '<' && data[1] == '/' already tested
 438
 439	// check if tag is a match
 440	closetag := []byte("</" + tag + ">")
 441	if !bytes.HasPrefix(data, closetag) {
 442		return 0
 443	}
 444	i := len(closetag)
 445
 446	// check that the rest of the line is blank
 447	skip := 0
 448	if skip = p.isEmpty(data[i:]); skip == 0 {
 449		return 0
 450	}
 451	i += skip
 452	skip = 0
 453
 454	if i >= len(data) {
 455		return i
 456	}
 457
 458	if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
 459		return i
 460	}
 461	if skip = p.isEmpty(data[i:]); skip == 0 {
 462		// following line must be blank
 463		return 0
 464	}
 465
 466	return i + skip
 467}
 468
 469func (p *parser) isEmpty(data []byte) int {
 470	// it is okay to call isEmpty on an empty buffer
 471	if len(data) == 0 {
 472		return 0
 473	}
 474
 475	var i int
 476	for i = 0; i < len(data) && data[i] != '\n'; i++ {
 477		if data[i] != ' ' && data[i] != '\t' {
 478			return 0
 479		}
 480	}
 481	return i + 1
 482}
 483
 484func (p *parser) isHRule(data []byte) bool {
 485	i := 0
 486
 487	// skip up to three spaces
 488	for i < 3 && data[i] == ' ' {
 489		i++
 490	}
 491
 492	// look at the hrule char
 493	if data[i] != '*' && data[i] != '-' && data[i] != '_' {
 494		return false
 495	}
 496	c := data[i]
 497
 498	// the whole line must be the char or whitespace
 499	n := 0
 500	for data[i] != '\n' {
 501		switch {
 502		case data[i] == c:
 503			n++
 504		case data[i] != ' ':
 505			return false
 506		}
 507		i++
 508	}
 509
 510	return n >= 3
 511}
 512
 513func (p *parser) isFencedCode(data []byte, syntax **string, oldmarker string) (skip int, marker string) {
 514	i, size := 0, 0
 515	skip = 0
 516
 517	// skip up to three spaces
 518	for i < 3 && data[i] == ' ' {
 519		i++
 520	}
 521
 522	// check for the marker characters: ~ or `
 523	if data[i] != '~' && data[i] != '`' {
 524		return
 525	}
 526
 527	c := data[i]
 528
 529	// the whole line must be the same char or whitespace
 530	for data[i] == c {
 531		size++
 532		i++
 533	}
 534
 535	// the marker char must occur at least 3 times
 536	if size < 3 {
 537		return
 538	}
 539	marker = string(data[i-size : i])
 540
 541	// if this is the end marker, it must match the beginning marker
 542	if oldmarker != "" && marker != oldmarker {
 543		return
 544	}
 545
 546	if syntax != nil {
 547		syn := 0
 548
 549		for data[i] == ' ' {
 550			i++
 551		}
 552
 553		syntaxStart := i
 554
 555		if data[i] == '{' {
 556			i++
 557			syntaxStart++
 558
 559			for data[i] != '}' && data[i] != '\n' {
 560				syn++
 561				i++
 562			}
 563
 564			if data[i] != '}' {
 565				return
 566			}
 567
 568			// strip all whitespace at the beginning and the end
 569			// of the {} block
 570			for syn > 0 && isspace(data[syntaxStart]) {
 571				syntaxStart++
 572				syn--
 573			}
 574
 575			for syn > 0 && isspace(data[syntaxStart+syn-1]) {
 576				syn--
 577			}
 578
 579			i++
 580		} else {
 581			for !isspace(data[i]) {
 582				syn++
 583				i++
 584			}
 585		}
 586
 587		language := string(data[syntaxStart : syntaxStart+syn])
 588		*syntax = &language
 589	}
 590
 591	for data[i] == ' ' {
 592		i++
 593	}
 594	if data[i] != '\n' {
 595		return
 596	}
 597
 598	skip = i + 1
 599	return
 600}
 601
 602func (p *parser) fencedCode(out *bytes.Buffer, data []byte, doRender bool) int {
 603	var lang *string
 604	beg, marker := p.isFencedCode(data, &lang, "")
 605	if beg == 0 || beg >= len(data) {
 606		return 0
 607	}
 608
 609	var work bytes.Buffer
 610
 611	for {
 612		// safe to assume beg < len(data)
 613
 614		// check for the end of the code block
 615		fenceEnd, _ := p.isFencedCode(data[beg:], nil, marker)
 616		if fenceEnd != 0 {
 617			beg += fenceEnd
 618			break
 619		}
 620
 621		// copy the current line
 622		end := beg
 623		for data[end] != '\n' {
 624			end++
 625		}
 626		end++
 627
 628		// did we reach the end of the buffer without a closing marker?
 629		if end >= len(data) {
 630			return 0
 631		}
 632
 633		// verbatim copy to the working buffer
 634		if doRender {
 635			work.Write(data[beg:end])
 636		}
 637		beg = end
 638	}
 639
 640	syntax := ""
 641	if lang != nil {
 642		syntax = *lang
 643	}
 644
 645	if doRender {
 646		p.r.BlockCode(out, work.Bytes(), syntax)
 647	}
 648
 649	return beg
 650}
 651
 652func (p *parser) table(out *bytes.Buffer, data []byte) int {
 653	var header bytes.Buffer
 654	i, columns := p.tableHeader(&header, data)
 655	if i == 0 {
 656		return 0
 657	}
 658
 659	var body bytes.Buffer
 660
 661	for i < len(data) {
 662		pipes, rowStart := 0, i
 663		for ; data[i] != '\n'; i++ {
 664			if data[i] == '|' {
 665				pipes++
 666			}
 667		}
 668
 669		if pipes == 0 {
 670			i = rowStart
 671			break
 672		}
 673
 674		// include the newline in data sent to tableRow
 675		i++
 676		p.tableRow(&body, data[rowStart:i], columns, false)
 677	}
 678
 679	p.r.Table(out, header.Bytes(), body.Bytes(), columns)
 680
 681	return i
 682}
 683
 684// check if the specified position is preceeded by an odd number of backslashes
 685func isBackslashEscaped(data []byte, i int) bool {
 686	backslashes := 0
 687	for i-backslashes-1 >= 0 && data[i-backslashes-1] == '\\' {
 688		backslashes++
 689	}
 690	return backslashes&1 == 1
 691}
 692
 693func (p *parser) tableHeader(out *bytes.Buffer, data []byte) (size int, columns []int) {
 694	i := 0
 695	colCount := 1
 696	for i = 0; data[i] != '\n'; i++ {
 697		if data[i] == '|' && !isBackslashEscaped(data, i) {
 698			colCount++
 699		}
 700	}
 701
 702	// doesn't look like a table header
 703	if colCount == 1 {
 704		return
 705	}
 706
 707	// include the newline in the data sent to tableRow
 708	header := data[:i+1]
 709
 710	// column count ignores pipes at beginning or end of line
 711	if data[0] == '|' {
 712		colCount--
 713	}
 714	if i > 2 && data[i-1] == '|' && !isBackslashEscaped(data, i-1) {
 715		colCount--
 716	}
 717
 718	columns = make([]int, colCount)
 719
 720	// move on to the header underline
 721	i++
 722	if i >= len(data) {
 723		return
 724	}
 725
 726	if data[i] == '|' && !isBackslashEscaped(data, i) {
 727		i++
 728	}
 729	for data[i] == ' ' {
 730		i++
 731	}
 732
 733	// each column header is of form: / *:?-+:? *|/ with # dashes + # colons >= 3
 734	// and trailing | optional on last column
 735	col := 0
 736	for data[i] != '\n' {
 737		dashes := 0
 738
 739		if data[i] == ':' {
 740			i++
 741			columns[col] |= TABLE_ALIGNMENT_LEFT
 742			dashes++
 743		}
 744		for data[i] == '-' {
 745			i++
 746			dashes++
 747		}
 748		if data[i] == ':' {
 749			i++
 750			columns[col] |= TABLE_ALIGNMENT_RIGHT
 751			dashes++
 752		}
 753		for data[i] == ' ' {
 754			i++
 755		}
 756
 757		// end of column test is messy
 758		switch {
 759		case dashes < 3:
 760			// not a valid column
 761			return
 762
 763		case data[i] == '|' && !isBackslashEscaped(data, i):
 764			// marker found, now skip past trailing whitespace
 765			col++
 766			i++
 767			for data[i] == ' ' {
 768				i++
 769			}
 770
 771			// trailing junk found after last column
 772			if col >= colCount && data[i] != '\n' {
 773				return
 774			}
 775
 776		case (data[i] != '|' || isBackslashEscaped(data, i)) && col+1 < colCount:
 777			// something else found where marker was required
 778			return
 779
 780		case data[i] == '\n':
 781			// marker is optional for the last column
 782			col++
 783
 784		default:
 785			// trailing junk found after last column
 786			return
 787		}
 788	}
 789	if col != colCount {
 790		return
 791	}
 792
 793	p.tableRow(out, header, columns, true)
 794	size = i + 1
 795	return
 796}
 797
 798func (p *parser) tableRow(out *bytes.Buffer, data []byte, columns []int, header bool) {
 799	i, col := 0, 0
 800	var rowWork bytes.Buffer
 801
 802	if data[i] == '|' && !isBackslashEscaped(data, i) {
 803		i++
 804	}
 805
 806	for col = 0; col < len(columns) && i < len(data); col++ {
 807		for data[i] == ' ' {
 808			i++
 809		}
 810
 811		cellStart := i
 812
 813		for (data[i] != '|' || isBackslashEscaped(data, i)) && data[i] != '\n' {
 814			i++
 815		}
 816
 817		cellEnd := i
 818
 819		// skip the end-of-cell marker, possibly taking us past end of buffer
 820		i++
 821
 822		for cellEnd > cellStart && data[cellEnd-1] == ' ' {
 823			cellEnd--
 824		}
 825
 826		var cellWork bytes.Buffer
 827		p.inline(&cellWork, data[cellStart:cellEnd])
 828
 829		if header {
 830			p.r.TableHeaderCell(&rowWork, cellWork.Bytes(), columns[col])
 831		} else {
 832			p.r.TableCell(&rowWork, cellWork.Bytes(), columns[col])
 833		}
 834	}
 835
 836	// pad it out with empty columns to get the right number
 837	for ; col < len(columns); col++ {
 838		if header {
 839			p.r.TableHeaderCell(&rowWork, nil, columns[col])
 840		} else {
 841			p.r.TableCell(&rowWork, nil, columns[col])
 842		}
 843	}
 844
 845	// silently ignore rows with too many cells
 846
 847	p.r.TableRow(out, rowWork.Bytes())
 848}
 849
 850// returns blockquote prefix length
 851func (p *parser) quotePrefix(data []byte) int {
 852	i := 0
 853	for i < 3 && data[i] == ' ' {
 854		i++
 855	}
 856	if data[i] == '>' {
 857		if data[i+1] == ' ' {
 858			return i + 2
 859		}
 860		return i + 1
 861	}
 862	return 0
 863}
 864
 865// parse a blockquote fragment
 866func (p *parser) quote(out *bytes.Buffer, data []byte) int {
 867	var raw bytes.Buffer
 868	beg, end := 0, 0
 869	for beg < len(data) {
 870		end = beg
 871		for data[end] != '\n' {
 872			end++
 873		}
 874		end++
 875
 876		if pre := p.quotePrefix(data[beg:]); pre > 0 {
 877			// skip the prefix
 878			beg += pre
 879		} else if p.isEmpty(data[beg:]) > 0 &&
 880			(end >= len(data) ||
 881				(p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0)) {
 882			// blockquote ends with at least one blank line
 883			// followed by something without a blockquote prefix
 884			break
 885		}
 886
 887		// this line is part of the blockquote
 888		raw.Write(data[beg:end])
 889		beg = end
 890	}
 891
 892	var cooked bytes.Buffer
 893	p.block(&cooked, raw.Bytes())
 894	p.r.BlockQuote(out, cooked.Bytes())
 895	return end
 896}
 897
 898// returns prefix length for block code
 899func (p *parser) codePrefix(data []byte) int {
 900	if data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
 901		return 4
 902	}
 903	return 0
 904}
 905
 906func (p *parser) code(out *bytes.Buffer, data []byte) int {
 907	var work bytes.Buffer
 908
 909	i := 0
 910	for i < len(data) {
 911		beg := i
 912		for data[i] != '\n' {
 913			i++
 914		}
 915		i++
 916
 917		blankline := p.isEmpty(data[beg:i]) > 0
 918		if pre := p.codePrefix(data[beg:i]); pre > 0 {
 919			beg += pre
 920		} else if !blankline {
 921			// non-empty, non-prefixed line breaks the pre
 922			i = beg
 923			break
 924		}
 925
 926		// verbatim copy to the working buffeu
 927		if blankline {
 928			work.WriteByte('\n')
 929		} else {
 930			work.Write(data[beg:i])
 931		}
 932	}
 933
 934	// trim all the \n off the end of work
 935	workbytes := work.Bytes()
 936	eol := len(workbytes)
 937	for eol > 0 && workbytes[eol-1] == '\n' {
 938		eol--
 939	}
 940	if eol != len(workbytes) {
 941		work.Truncate(eol)
 942	}
 943
 944	work.WriteByte('\n')
 945
 946	p.r.BlockCode(out, work.Bytes(), "")
 947
 948	return i
 949}
 950
 951// returns unordered list item prefix
 952func (p *parser) uliPrefix(data []byte) int {
 953	i := 0
 954
 955	// start with up to 3 spaces
 956	for i < 3 && data[i] == ' ' {
 957		i++
 958	}
 959
 960	// need a *, +, or - followed by a space
 961	if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
 962		data[i+1] != ' ' {
 963		return 0
 964	}
 965	return i + 2
 966}
 967
 968// returns ordered list item prefix
 969func (p *parser) oliPrefix(data []byte) int {
 970	i := 0
 971
 972	// start with up to 3 spaces
 973	for i < 3 && data[i] == ' ' {
 974		i++
 975	}
 976
 977	// count the digits
 978	start := i
 979	for data[i] >= '0' && data[i] <= '9' {
 980		i++
 981	}
 982
 983	// we need >= 1 digits followed by a dot and a space
 984	if start == i || data[i] != '.' || data[i+1] != ' ' {
 985		return 0
 986	}
 987	return i + 2
 988}
 989
 990// parse ordered or unordered list block
 991func (p *parser) list(out *bytes.Buffer, data []byte, flags int) int {
 992	i := 0
 993	flags |= LIST_ITEM_BEGINNING_OF_LIST
 994	work := func() bool {
 995		for i < len(data) {
 996			skip := p.listItem(out, data[i:], &flags)
 997			i += skip
 998
 999			if skip == 0 || flags&LIST_ITEM_END_OF_LIST != 0 {
1000				break
1001			}
1002			flags &= ^LIST_ITEM_BEGINNING_OF_LIST
1003		}
1004		return true
1005	}
1006
1007	p.r.List(out, work, flags)
1008	return i
1009}
1010
1011// Parse a single list item.
1012// Assumes initial prefix is already removed if this is a sublist.
1013func (p *parser) listItem(out *bytes.Buffer, data []byte, flags *int) int {
1014	// keep track of the indentation of the first line
1015	itemIndent := 0
1016	for itemIndent < 3 && data[itemIndent] == ' ' {
1017		itemIndent++
1018	}
1019
1020	i := p.uliPrefix(data)
1021	if i == 0 {
1022		i = p.oliPrefix(data)
1023	}
1024	if i == 0 {
1025		return 0
1026	}
1027
1028	// skip leading whitespace on first line
1029	for data[i] == ' ' {
1030		i++
1031	}
1032
1033	// find the end of the line
1034	line := i
1035	for data[i-1] != '\n' {
1036		i++
1037	}
1038
1039	// get working buffer
1040	var raw bytes.Buffer
1041
1042	// put the first line into the working buffer
1043	raw.Write(data[line:i])
1044	line = i
1045
1046	// process the following lines
1047	containsBlankLine := false
1048	sublist := 0
1049
1050gatherlines:
1051	for line < len(data) {
1052		i++
1053
1054		// find the end of this line
1055		for data[i-1] != '\n' {
1056			i++
1057		}
1058
1059		// if it is an empty line, guess that it is part of this item
1060		// and move on to the next line
1061		if p.isEmpty(data[line:i]) > 0 {
1062			containsBlankLine = true
1063			line = i
1064			continue
1065		}
1066
1067		// calculate the indentation
1068		indent := 0
1069		for indent < 4 && line+indent < i && data[line+indent] == ' ' {
1070			indent++
1071		}
1072
1073		chunk := data[line+indent : i]
1074
1075		// evaluate how this line fits in
1076		switch {
1077		// is this a nested list item?
1078		case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) ||
1079			p.oliPrefix(chunk) > 0:
1080
1081			if containsBlankLine {
1082				*flags |= LIST_ITEM_CONTAINS_BLOCK
1083			}
1084
1085			// to be a nested list, it must be indented more
1086			// if not, it is the next item in the same list
1087			if indent <= itemIndent {
1088				break gatherlines
1089			}
1090
1091			// is this the first item in the the nested list?
1092			if sublist == 0 {
1093				sublist = raw.Len()
1094			}
1095
1096		// is this a nested prefix header?
1097		case p.isPrefixHeader(chunk):
1098			// if the header is not indented, it is not nested in the list
1099			// and thus ends the list
1100			if containsBlankLine && indent < 4 {
1101				*flags |= LIST_ITEM_END_OF_LIST
1102				break gatherlines
1103			}
1104			*flags |= LIST_ITEM_CONTAINS_BLOCK
1105
1106		// anything following an empty line is only part
1107		// of this item if it is indented 4 spaces
1108		// (regardless of the indentation of the beginning of the item)
1109		case containsBlankLine && indent < 4:
1110			*flags |= LIST_ITEM_END_OF_LIST
1111			break gatherlines
1112
1113		// a blank line means this should be parsed as a block
1114		case containsBlankLine:
1115			raw.WriteByte('\n')
1116			*flags |= LIST_ITEM_CONTAINS_BLOCK
1117		}
1118
1119		// if this line was preceeded by one or more blanks,
1120		// re-introduce the blank into the buffer
1121		if containsBlankLine {
1122			containsBlankLine = false
1123			raw.WriteByte('\n')
1124		}
1125
1126		// add the line into the working buffer without prefix
1127		raw.Write(data[line+indent : i])
1128
1129		line = i
1130	}
1131
1132	rawBytes := raw.Bytes()
1133
1134	// render the contents of the list item
1135	var cooked bytes.Buffer
1136	if *flags&LIST_ITEM_CONTAINS_BLOCK != 0 {
1137		// intermediate render of block li
1138		if sublist > 0 {
1139			p.block(&cooked, rawBytes[:sublist])
1140			p.block(&cooked, rawBytes[sublist:])
1141		} else {
1142			p.block(&cooked, rawBytes)
1143		}
1144	} else {
1145		// intermediate render of inline li
1146		if sublist > 0 {
1147			p.inline(&cooked, rawBytes[:sublist])
1148			p.block(&cooked, rawBytes[sublist:])
1149		} else {
1150			p.inline(&cooked, rawBytes)
1151		}
1152	}
1153
1154	// render the actual list item
1155	cookedBytes := cooked.Bytes()
1156	parsedEnd := len(cookedBytes)
1157
1158	// strip trailing newlines
1159	for parsedEnd > 0 && cookedBytes[parsedEnd-1] == '\n' {
1160		parsedEnd--
1161	}
1162	p.r.ListItem(out, cookedBytes[:parsedEnd], *flags)
1163
1164	return line
1165}
1166
1167// render a single paragraph that has already been parsed out
1168func (p *parser) renderParagraph(out *bytes.Buffer, data []byte) {
1169	if len(data) == 0 {
1170		return
1171	}
1172
1173	// trim leading spaces
1174	beg := 0
1175	for data[beg] == ' ' {
1176		beg++
1177	}
1178
1179	// trim trailing newline
1180	end := len(data) - 1
1181
1182	// trim trailing spaces
1183	for end > beg && data[end-1] == ' ' {
1184		end--
1185	}
1186
1187	work := func() bool {
1188		p.inline(out, data[beg:end])
1189		return true
1190	}
1191	p.r.Paragraph(out, work)
1192}
1193
1194func (p *parser) paragraph(out *bytes.Buffer, data []byte) int {
1195	// prev: index of 1st char of previous line
1196	// line: index of 1st char of current line
1197	// i: index of cursor/end of current line
1198	var prev, line, i int
1199
1200	// keep going until we find something to mark the end of the paragraph
1201	for i < len(data) {
1202		// mark the beginning of the current line
1203		prev = line
1204		current := data[i:]
1205		line = i
1206
1207		// did we find a blank line marking the end of the paragraph?
1208		if n := p.isEmpty(current); n > 0 {
1209			p.renderParagraph(out, data[:i])
1210			return i + n
1211		}
1212
1213		// an underline under some text marks a header, so our paragraph ended on prev line
1214		if i > 0 {
1215			if level := p.isUnderlinedHeader(current); level > 0 {
1216				// render the paragraph
1217				p.renderParagraph(out, data[:prev])
1218
1219				// ignore leading and trailing whitespace
1220				eol := i - 1
1221				for prev < eol && data[prev] == ' ' {
1222					prev++
1223				}
1224				for eol > prev && data[eol-1] == ' ' {
1225					eol--
1226				}
1227
1228				// render the header
1229				// this ugly double closure avoids forcing variables onto the heap
1230				work := func(o *bytes.Buffer, pp *parser, d []byte) func() bool {
1231					return func() bool {
1232						pp.inline(o, d)
1233						return true
1234					}
1235				}(out, p, data[prev:eol])
1236				p.r.Header(out, work, level)
1237
1238				// find the end of the underline
1239				for data[i] != '\n' {
1240					i++
1241				}
1242				return i
1243			}
1244		}
1245
1246		// if the next line starts a block of HTML, then the paragraph ends here
1247		if p.flags&EXTENSION_LAX_HTML_BLOCKS != 0 {
1248			if data[i] == '<' && p.html(out, current, false) > 0 {
1249				// rewind to before the HTML block
1250				p.renderParagraph(out, data[:i])
1251				return i
1252			}
1253		}
1254
1255		// if there's a prefixed header or a horizontal rule after this, paragraph is over
1256		if p.isPrefixHeader(current) || p.isHRule(current) {
1257			p.renderParagraph(out, data[:i])
1258			return i
1259		}
1260
1261		// if there's a list after this, paragraph is over
1262		if p.flags&EXTENSION_NO_EMPTY_LINE_BEFORE_BLOCK != 0 {
1263			if p.uliPrefix(current) != 0 ||
1264				p.oliPrefix(current) != 0 ||
1265				p.quotePrefix(current) != 0 ||
1266				p.codePrefix(current) != 0 {
1267				p.renderParagraph(out, data[:i])
1268				return i
1269			}
1270		}
1271
1272		// otherwise, scan to the beginning of the next line
1273		for data[i] != '\n' {
1274			i++
1275		}
1276		i++
1277	}
1278
1279	p.renderParagraph(out, data[:i])
1280	return i
1281}