all repos — grayfriday @ 2a18706ca4952462e699c51b746e188970933d8d

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