pkg/md: Support thematic breaks, ATX headings and blockquotes.

Also rearrange fields of OutputSyntax to match the order they appear in the
CommonMark spec.

pkg/md/md.go

@@ -24,11 +24,14 @@ import (
 
 // OutputSyntax specifies the output syntax.
 type OutputSyntax struct {
+	ThematicBreak  func(original string) string
+	Heading        func(level int) TagPair
 	Paragraph      TagPair
-	Code      TagPair
-	Em        TagPair
-	Strong    TagPair
-	Link      func(dest, title string) (string, string)
+	Blockquote     TagPair
+	CodeSpan       TagPair
+	Emphasis       TagPair
+	StrongEmphasis TagPair
+	Link           func(dest, title string) TagPair
 	Image          func(dest, alt, title string) string
 	Escape         func(string) string
 }
@@ -43,7 +46,6 @@ func Render(text string, syntax OutputSyntax) string {
 	p := blockParser{
 		lines:  lineSplitter{text, 0},
 		syntax: syntax,
-		blocks: []block{{typ: documentBlock}},
 	}
 	p.render()
 	return p.sb.String()
@@ -52,70 +54,97 @@ func Render(text string, syntax OutputSyntax) string {
 type blockParser struct {
 	lines      lineSplitter
 	syntax     OutputSyntax
-	blocks []block
+	containers []container
+	paragraph  []string
 	sb         strings.Builder
 }
 
+var (
+	blockquoteMarkerRegexp = regexp.MustCompile(`^ {0,3}> ?`)
+	thematicBreakRegexp    = regexp.MustCompile(`^[ \t]*((?:-[ \t]*){3,}|(?:_[ \t]*){3,}|(?:\*[ \t]*){3,})$`)
+	atxHeadingRegexp       = regexp.MustCompile(`^ *(#{1,6})(?:[ \t]|$)`)
+	atxHeadingCloserRegexp = regexp.MustCompile(`[ \t]#+[ \t]*$`)
+)
+
 func (p *blockParser) render() {
 	for p.lines.more() {
 		line := p.lines.next()
-		if line == "\n" {
-			switch p.leaf().typ {
-			case documentBlock:
-				// Nothing to do
-			case paragraphBlock:
-				p.pop()
+		i := 0
+		for i = 0; i < len(p.containers); i++ {
+			markerLen := p.containers[i].findMarker(line)
+			if markerLen == 0 {
+				break
 			}
+			line = line[markerLen:]
+		}
+
+		if m := blockquoteMarkerRegexp.FindString(line); m != "" {
+			p.popParagraph(i)
+			for m != "" {
+				p.appendContainer(container{typ: blockquoteContainer})
+				line = line[len(m):]
+				m = blockquoteMarkerRegexp.FindString(line)
+			}
+			i = len(p.containers)
+		}
+
+		if strings.Trim(line, " \t") == "" {
+			p.popParagraph(i)
+		} else if thematicBreakRegexp.MatchString(line) {
+			p.popParagraph(i)
+			p.sb.WriteString(p.syntax.ThematicBreak(line))
+			p.sb.WriteByte('\n')
+		} else if m := atxHeadingRegexp.FindStringSubmatchIndex(line); m != nil {
+			p.popParagraph(i)
+			// ATX headings always span one line only, so render it right away
+			// without pushing a node.
+			openerStart, openerEnd := m[2], m[3]
+			opener := line[openerStart:openerEnd]
+			line = strings.TrimRight(line[openerEnd:], " \t")
+			if closer := atxHeadingCloserRegexp.FindString(line); closer != "" {
+				line = line[:len(line)-len(closer)]
+			}
+			p.renderLeaf(p.syntax.Heading(len(opener)), strings.Trim(line, " \t"))
 		} else {
-			switch p.leaf().typ {
-			case documentBlock:
-				p.push(paragraphBlock).text.WriteString(line)
-			case paragraphBlock:
-				p.leaf().text.WriteString(line)
+			if len(p.paragraph) == 0 {
+				p.popParagraph(i)
+			}
+			p.addParagraphLine(line)
 		}
 	}
-	}
-	for len(p.blocks) > 0 {
-		p.pop()
-	}
+	p.popParagraph(0)
 }
 
-func (p *blockParser) push(typ blockType) *block {
-	switch typ {
-	case paragraphBlock:
-		p.sb.WriteString(p.syntax.Paragraph.Start)
-	}
-	p.blocks = append(p.blocks, block{typ: typ})
-	return p.leaf()
+func (p *blockParser) appendContainer(c container) {
+	p.containers = append(p.containers, c)
+	p.sb.WriteString(c.tagPair(&p.syntax).Start)
+	p.sb.WriteByte('\n')
 }
 
-func (p *blockParser) leaf() *block { return &p.blocks[len(p.blocks)-1] }
+func (p *blockParser) addParagraphLine(line string) {
+	p.paragraph = append(p.paragraph, line)
+}
 
-func (p *blockParser) pop() {
-	leaf := p.leaf()
-	switch leaf.typ {
-	case paragraphBlock:
-		text := strings.Trim(strings.TrimSuffix(leaf.text.String(), "\n"), " \t")
-		p.sb.WriteString(renderInline(text, p.syntax))
-		p.sb.WriteString(p.syntax.Paragraph.End)
+func (p *blockParser) popParagraph(i int) {
+	if len(p.paragraph) > 0 {
+		text := strings.Trim(strings.Join(p.paragraph, "\n"), " \t")
+		p.renderLeaf(p.syntax.Paragraph, text)
+		p.paragraph = p.paragraph[:0]
+	}
+	for j := len(p.containers) - 1; j >= i; j-- {
+		p.sb.WriteString(p.containers[i].tagPair(&p.syntax).End)
 		p.sb.WriteByte('\n')
 	}
-	p.blocks = p.blocks[:len(p.blocks)-1]
+	p.containers = p.containers[:i]
 }
 
-type block struct {
-	typ  blockType
-	text strings.Builder
+func (p *blockParser) renderLeaf(tags TagPair, content string) {
+	p.sb.WriteString(tags.Start)
+	p.sb.WriteString(renderInline(content, p.syntax))
+	p.sb.WriteString(tags.End)
+	p.sb.WriteByte('\n')
 }
 
-type blockType uint
-
-const (
-	documentBlock blockType = iota
-	paragraphBlock
-)
-
-// Splits a string into lines, preserving the trailing newlines.
 type lineSplitter struct {
 	text string
 	pos  int
@@ -133,7 +162,33 @@ func (s *lineSplitter) next() string {
 		return s.text[begin:]
 	}
 	s.pos += delta + 1
-	return s.text[begin:s.pos]
+	return s.text[begin : s.pos-1]
+}
+
+type container struct {
+	typ containerType
+}
+
+type containerType uint8
+
+const (
+	blockquoteContainer containerType = iota
+)
+
+func (c container) findMarker(line string) int {
+	switch c.typ {
+	case blockquoteContainer:
+		return len(blockquoteMarkerRegexp.FindString(line))
+	}
+	panic("unreachable")
+}
+
+func (c container) tagPair(syntax *OutputSyntax) TagPair {
+	switch c.typ {
+	case blockquoteContainer:
+		return syntax.Blockquote
+	}
+	panic("unreachable")
 }
 
 type buffer struct {
@@ -313,9 +368,9 @@ func (p *inlineParser) render() {
 			}
 			unlink(opener)
 			if opener.typ == '[' {
-				start, end := p.syntax.Link(dest, title)
-				p.buf.pieces[opener.bufIdx] = piece{appendMarkup: []string{start}}
-				p.buf.push(piece{appendMarkup: []string{end}})
+				tags := p.syntax.Link(dest, title)
+				p.buf.pieces[opener.bufIdx] = piece{appendMarkup: []string{tags.Start}}
+				p.buf.push(piece{appendMarkup: []string{tags.End}})
 			} else {
 				var altBuilder strings.Builder
 				for _, piece := range p.buf.pieces[opener.bufIdx+1:] {
@@ -363,9 +418,9 @@ func (p *inlineParser) render() {
 				continue
 			}
 			p.buf.push(piece{
-				prependMarkup: []string{p.syntax.Code.Start},
+				prependMarkup: []string{p.syntax.CodeSpan.Start},
 				text:          p.syntax.Escape(normalizeCodeSpanContent(p.text[p.pos:closer])),
-				appendMarkup:  []string{p.syntax.Code.End}})
+				appendMarkup:  []string{p.syntax.CodeSpan.End}})
 			p.pos = closer + (p.pos - begin)
 		case '<':
 			if p.pos == len(p.text) {
@@ -441,11 +496,11 @@ func (p *inlineParser) render() {
 						if email {
 							dest = "mailto:" + dest
 						}
-						start, end := p.syntax.Link(dest, "")
+						tags := p.syntax.Link(dest, "")
 						p.buf.push(piece{
-							prependMarkup: []string{start},
+							prependMarkup: []string{tags.Start},
 							text:          text,
-							appendMarkup:  []string{end},
+							appendMarkup:  []string{tags.End},
 						})
 						continue
 					}
@@ -467,6 +522,7 @@ func (p *inlineParser) render() {
 			}
 			parseText()
 		case '\n':
+			if len(p.buf.pieces) > 0 {
 				last := &p.buf.pieces[len(p.buf.pieces)-1]
 				if last.prependMarkup == nil && last.appendMarkup == nil {
 					if p.pos == len(p.text) {
@@ -485,6 +541,7 @@ func (p *inlineParser) render() {
 						}
 					}
 				}
+			}
 			p.buf.push(piece{text: "\n"})
 			for p.pos < len(p.text) && p.text[p.pos] == ' ' {
 				p.pos++
@@ -530,14 +587,14 @@ func (p *inlineParser) processEmphasis(bottom *delim) {
 		strong := len(openerPiece.text) >= 2 && len(closerPiece.text) >= 2
 		if strong {
 			openerPiece.text = openerPiece.text[2:]
-			openerPiece.appendMarkup = append(openerPiece.appendMarkup, p.syntax.Strong.Start)
+			openerPiece.appendMarkup = append(openerPiece.appendMarkup, p.syntax.StrongEmphasis.Start)
 			closerPiece.text = closerPiece.text[2:]
-			closerPiece.prependMarkup = append(closerPiece.prependMarkup, p.syntax.Strong.End)
+			closerPiece.prependMarkup = append(closerPiece.prependMarkup, p.syntax.StrongEmphasis.End)
 		} else {
 			openerPiece.text = openerPiece.text[1:]
-			openerPiece.appendMarkup = append(openerPiece.appendMarkup, p.syntax.Em.Start)
+			openerPiece.appendMarkup = append(openerPiece.appendMarkup, p.syntax.Emphasis.Start)
 			closerPiece.text = closerPiece.text[1:]
-			closerPiece.prependMarkup = append(closerPiece.prependMarkup, p.syntax.Em.End)
+			closerPiece.prependMarkup = append(closerPiece.prependMarkup, p.syntax.Emphasis.End)
 		}
 		opener.next = closer
 		closer.prev = opener

pkg/md/md_test.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"regexp"
+	"strconv"
 	"strings"
 	"testing"
 
@@ -12,18 +13,52 @@ import (
 	"src.elv.sh/pkg/must"
 )
 
-//go:embed spec.json
-var specJSON []byte
-
-var spec []struct {
+type testCase struct {
 	Markdown string `json:"markdown"`
 	HTML     string `json:"html"`
 	Example  int    `json:"example"`
 	Section  string `json:"section"`
+	Name     string
+}
+
+//go:embed spec.json
+var specJSON []byte
+
+var testCases []testCase
+
+var additionalCases = []testCase{
+	{
+		Markdown: `> a
+>> b
+`,
+		HTML: `<blockquote>
+<p>a</p>
+<blockquote>
+<p>b</p>
+</blockquote>
+</blockquote>
+`,
+		Name: "Increasing blockquote level",
+	},
+	{
+		Markdown: `>> a
+>
+> b
+`,
+		HTML: `<blockquote>
+<blockquote>
+<p>a</p>
+</blockquote>
+<p>b</p>
+</blockquote>
+`,
+		Name: "Reducing blockquote level",
+	},
 }
 
 func init() {
-	must.OK(json.Unmarshal(specJSON, &spec))
+	must.OK(json.Unmarshal(specJSON, &testCases))
+	testCases = append(testCases, additionalCases...)
 }
 
 var (
@@ -37,18 +72,24 @@ var (
 )
 
 var htmlSyntax = OutputSyntax{
+	ThematicBreak: func(_ string) string { return "<hr />" },
+	Heading: func(level int) TagPair {
+		tag := "h" + strconv.Itoa(level)
+		return TagPair{Start: "<" + tag + ">", End: "</" + tag + ">"}
+	},
 	Paragraph:      TagPair{Start: "<p>", End: "</p>"},
-	Code:      TagPair{Start: "<code>", End: "</code>"},
-	Em:        TagPair{Start: "<em>", End: "</em>"},
-	Strong:    TagPair{Start: "<strong>", End: "</strong>"},
-	Link: func(dest, title string) (string, string) {
+	Blockquote:     TagPair{Start: "<blockquote>", End: "</blockquote>"},
+	CodeSpan:       TagPair{Start: "<code>", End: "</code>"},
+	Emphasis:       TagPair{Start: "<em>", End: "</em>"},
+	StrongEmphasis: TagPair{Start: "<strong>", End: "</strong>"},
+	Link: func(dest, title string) TagPair {
 		start := ""
 		if title == "" {
 			start = fmt.Sprintf(`<a href="%s">`, escapeDest(dest))
 		} else {
 			start = fmt.Sprintf(`<a href="%s" title="%s">`, escapeDest(dest), escapeHTML(title))
 		}
-		return start, "</a>"
+		return TagPair{Start: start, End: "</a>"}
 	},
 	Image: func(dest, alt, title string) string {
 		if title == "" {
@@ -61,19 +102,24 @@ var htmlSyntax = OutputSyntax{
 
 var (
 	linkRef   = regexp.MustCompile(`(^|\n)\[([^\\\[\]]|\\[\\\[\]])+\]:`)
-	listItem = regexp.MustCompile(`(^|\n)\* `)
+	listItem  = regexp.MustCompile(`(^|\n)[*-] `)
+	codeBlock = regexp.MustCompile("(^|\n)>*(```|~~~|    )")
 )
 
 func TestRender(t *testing.T) {
-	for _, tc := range spec {
-		t.Run(fmt.Sprintf("%s/%d", tc.Section, tc.Example), func(t *testing.T) {
-			if !supportedSection(tc.Section) {
+	for _, tc := range testCases {
+		name := tc.Name
+		if name == "" {
+			name = fmt.Sprintf("%s/%d", tc.Section, tc.Example)
+		}
+		t.Run(name, func(t *testing.T) {
+			if unsupportedSection(tc.Section) {
 				t.Skipf("Section %q not supported", tc.Section)
 			}
-			if strings.HasPrefix(tc.Markdown, "#") {
-				t.Skipf("Header not supported")
+			if reason := unsupportedExample(tc.Example); reason != "" {
+				t.Skipf("Example %d not supported: %s", tc.Example, reason)
 			}
-			if strings.HasPrefix(tc.Markdown, "```") || strings.HasPrefix(tc.Markdown, "~~~") || strings.HasPrefix(tc.Markdown, "    ") {
+			if codeBlock.MatchString(tc.Markdown) {
 				t.Skipf("Code block not supported")
 			}
 			if linkRef.MatchString(tc.Markdown) {
@@ -94,23 +140,29 @@ func TestRender(t *testing.T) {
 	}
 }
 
-func supportedSection(section string) bool {
+func unsupportedSection(section string) bool {
 	switch section {
 	case "Tabs",
 		"Precedence",
-		"Thematic breaks",
-		"ATX headings",
 		"Setext headings",
 		"Indented code blocks",
 		"Fenced code blocks",
 		"HTML blocks",
 		"Link reference definitions",
 		"Blank lines",
-		"Block quotes",
 		"List items",
 		"Lists":
-		return false
-	default:
 		return true
+	default:
+		return false
+	}
+}
+
+func unsupportedExample(example int) string {
+	switch example {
+	case 59:
+		return "has setext heading"
+	default:
+		return ""
 	}
 }