first commit

.gitignore

@@ -0,0 +1,13 @@
+# Binaries for programs and plugins
+*.exe
+*.exe~
+*.dll
+*.so
+*.dylib
+
+# Test binary, build with `go test -c`
+*.test
+*.pprof
+
+# Output of the go coverage tool, specifically when used with LiteIDE
+*.out

Makefile

@@ -0,0 +1,7 @@
+.PHONY: test
+
+test:
+	go test -coverprofile=profile.out -coverpkg=github.com/yuin/goldmark,github.com/yuin/goldmark/ast,github.com/yuin/goldmark/extension,github.com/yuin/goldmark/extension/ast,github.com/yuin/goldmark/parser,github.com/yuin/goldmark/renderer,github.com/yuin/goldmark/renderer/html,github.com/yuin/goldmark/text,github.com/yuin/goldmark/util ./...
+
+cov: test
+	go tool cover -html=profile.out

README.md

@@ -0,0 +1,129 @@
+goldmark
+==========================================
+
+> A markdown parser written in Go. Easy to extend, standard compliant, well structured.
+
+goldmark is compliant to CommonMark 0.29.
+
+Motivation
+----------------------
+I need a markdown parser for Go that meets following conditions:
+
+- Easy to extend.
+    - Markdown is poor in document expressions compared with other light markup languages like restructuredText.
+    - We have extended a markdown syntax. i.e. : PHPMarkdownExtra, Github Flavored Markdown.
+- Standard compliant.
+    - Markdown has many dialects.
+    - Github Flavored Markdown is widely used and it is based on CommonMark aside from whether CommonMark is good specification or not.
+        - CommonMark is too complicated and hard to implement.
+- Well structured.
+    - AST based, and preserves source potision of nodes.
+- Written in pure Go.
+
+[golang-commonmark](https://gitlab.com/golang-commonmark/markdown) may be a good choice, but it seems copy of the [markdown-it](https://github.com/markdown-it) .
+
+[blackfriday.v2](https://github.com/russross/blackfriday/tree/v2) is a fast and widely used implementation, but it is not CommonMark compliant and can not be extended from outside of the package since it's AST is not interfaces but structs.
+
+As mentioned above, CommonMark is too complicated and hard to implement, So Markdown parsers base on CommonMark barely exist.
+
+Usage
+----------------------
+
+Convert Markdown documents with the CommonMark compliant mode:
+
+```go
+var buf bytes.Buffer
+if err := goldmark.Convert(source, &buf); err != nil {
+  panic(err)
+}
+```
+
+Customize a parser and a renderer:
+
+```go
+md := goldmark.NewMarkdown(
+          goldmark.WithExtensions(extension.GFM),
+          goldmark.WithParserOptions(
+              parser.WithHeadingID(),
+          ),
+          goldmark.WithRendererOptions(
+              html.WithSoftLineBreak(true),
+              html.WithXHTML(true),
+          ),
+      )
+var buf bytes.Buffer
+if err := md.Convert(source, &buf); err != nil {
+    panic(err)
+}
+```
+
+Parser and Renderer options
+------------------------------
+
+### Parser options
+
+| Functional option | Type | Description |
+| ----------------- | ---- | ----------- |
+| `parser.WithBlockParsers` | List of `util.PrioritizedSlice` whose elements are `parser.BlockParser` | Parsers for parsing block level elements. | 
+| `parser.WithInlineParsers` | List of `util.PrioritizedSlice` whose elements are `parser.InlineParser` | Parsers for parsing inline level elements. | 
+| `parser.WithParagraphTransformers` | List of `util.PrioritizedSlice` whose elements are `parser.ParagraphTransformer` | Transformers for transforming paragraph nodes. | 
+| `parser.WithHeadingID` | `-` | Enables custom heading ids( `{#custom-id}` ) and auto heading ids. |
+| `parser.WithFilterTags` | `...string` | HTML tag names forbidden in HTML blocks and Raw HTMLs. |
+
+### HTML Renderer options
+
+| Functional option | Type | Description |
+| ----------------- | ---- | ----------- |
+| `html.WithWriter` | `html.Writer` | `html.Writer` for writing contents to an `io.Writer`. |
+| `html.WithSoftLineBreak` | `-` | Render new lines as `<br>` if true, otherwise `\n` .|
+| `html.WithXHTML` | `-` | Render as XHTML. |
+
+### Built-in extensions
+
+- `extension.Table`
+- `extension.Strikethrough`
+- `extension.Linkify`
+- `extension.TaskList`
+- `extension.GFM`
+  - This extension enables Table, Strikethrough, Linkify and TaskList.
+    In addition, this extension sets some tags to `parser.FilterTags` .
+
+Create extensions
+--------------------
+**TODO**
+
+See `extension` directory for examples of extensions.
+
+Summary:
+
+1. Define AST Node as a struct in which `ast.BaseBlock` or `ast.BaseInline` is embedded.
+2. Write a parser that implements `parser.BlockParser` or `parser.InlineParser`.
+3. Write a renderer that implements `renderer.NodeRenderer`.
+4. Define your goldmark extension that implements `goldmark.Extender`.
+
+Benchmark
+--------------------
+You can run this benchmark in the `_benchmark` directory.
+
+blackfriday v2 is fastest, but it is not CommonMark compiliant and not an extensible.
+
+Though goldmark builds clean extensible AST structure and get full compliance with 
+Commonmark, it is resonably fast and less memory consumption.
+
+```
+BenchmarkGoldMark-4                  200           7291402 ns/op         2259603 B/op   16867 allocs/op
+BenchmarkGolangCommonMark-4          200           7709939 ns/op         3053760 B/op   18682 allocs/op
+BenchmarkBlackFriday-4               300           5776369 ns/op         3356386 B/op   17480 allocs/op
+```
+
+Donation
+--------------------
+BTC: 1NEDSyUmo4SMTDP83JJQSWi1MvQUGGNMZB
+
+License
+--------------------
+MIT
+
+Author
+--------------------
+Yusuke Inuzuka

_benchmark/benchmark_test.go

@@ -0,0 +1,55 @@
+package main
+
+import (
+	"bytes"
+	"io/ioutil"
+	"testing"
+
+	"github.com/yuin/goldmark"
+	"github.com/yuin/goldmark/renderer/html"
+
+	"gitlab.com/golang-commonmark/markdown"
+
+	"gopkg.in/russross/blackfriday.v2"
+)
+
+func BenchmarkGoldMark(b *testing.B) {
+	b.ResetTimer()
+	source, err := ioutil.ReadFile("_data.md")
+	if err != nil {
+		panic(err)
+	}
+	markdown := goldmark.New(goldmark.WithRendererOptions(html.WithXHTML()))
+	var out bytes.Buffer
+	markdown.Convert([]byte(""), &out)
+
+	for i := 0; i < b.N; i++ {
+		out.Reset()
+		if err := markdown.Convert(source, &out); err != nil {
+			panic(err)
+		}
+	}
+}
+
+func BenchmarkGolangCommonMark(b *testing.B) {
+	b.ResetTimer()
+	source, err := ioutil.ReadFile("_data.md")
+	if err != nil {
+		panic(err)
+	}
+	md := markdown.New(markdown.XHTMLOutput(true))
+	for i := 0; i < b.N; i++ {
+		md.RenderToString(source)
+	}
+}
+
+func BenchmarkBlackFriday(b *testing.B) {
+	b.ResetTimer()
+	source, err := ioutil.ReadFile("_data.md")
+	if err != nil {
+		panic(err)
+	}
+	for i := 0; i < b.N; i++ {
+		blackfriday.Run(source)
+	}
+}

ast/ast.go

@@ -0,0 +1,342 @@
+// Package ast defines AST nodes that represent markdown elements.
+package ast
+
+import (
+	"bytes"
+	"fmt"
+	textm "github.com/yuin/goldmark/text"
+	"strings"
+)
+
+// A NodeType indicates what type a node belongs to.
+type NodeType int
+
+const (
+	// BlockNode indicates that a node is kind of block nodes.
+	BlockNode NodeType = iota + 1
+	// InlineNode indicates that a node is kind of inline nodes.
+	InlineNode
+)
+
+// A Node interface defines basic AST node functionalities.
+type Node interface {
+	// Type returns a type of this node.
+	Type() NodeType
+
+	// NextSibling returns a next sibling node of this node.
+	NextSibling() Node
+
+	// PreviousSibling returns a previous sibling node of this node.
+	PreviousSibling() Node
+
+	// Parent returns a parent node of this node.
+	Parent() Node
+
+	// SetParent sets a parent node to this node.
+	SetParent(Node)
+
+	// SetPreviousSibling sets a previous sibling node to this node.
+	SetPreviousSibling(Node)
+
+	// SetNextSibling sets a next sibling node to this node.
+	SetNextSibling(Node)
+
+	// HasChildren returns true if this node has any children, otherwise false.
+	HasChildren() bool
+
+	// ChildCount returns a total number of children.
+	ChildCount() int
+
+	// FirstChild returns a first child of this node.
+	FirstChild() Node
+
+	// LastChild returns a last child of this node.
+	LastChild() Node
+
+	// AppendChild append a node child to the tail of the children.
+	AppendChild(self, child Node)
+
+	// RemoveChild removes a node child from this node.
+	// If a node child is not children of this node, RemoveChild nothing to do.
+	RemoveChild(self, child Node)
+
+	// RemoveChildren removes all children from this node.
+	RemoveChildren(self Node)
+
+	// ReplaceChild replace a node v1 with a node insertee.
+	// If v1 is not children of this node, ReplaceChild append a insetee to the
+	// tail of the children.
+	ReplaceChild(self, v1, insertee Node)
+
+	// InsertBefore inserts a node insertee before a node v1.
+	// If v1 is not children of this node, InsertBefore append a insetee to the
+	// tail of the children.
+	InsertBefore(self, v1, insertee Node)
+
+	// InsertAfterinserts a node insertee after a node v1.
+	// If v1 is not children of this node, InsertBefore append a insetee to the
+	// tail of the children.
+	InsertAfter(self, v1, insertee Node)
+
+	// Dump dumps an AST tree structure to stdout.
+	// This function completely aimed for debugging.
+	// level is a indent level. Implementer should indent informations with
+	// 2 * level spaces.
+	Dump(source []byte, level int)
+
+	// Text returns text values of this node.
+	Text(source []byte) []byte
+
+	// HasBlankPreviousLines returns true if the row before this node is blank,
+	// otherwise false.
+	// This method is valid only for block nodes.
+	HasBlankPreviousLines() bool
+
+	// SetBlankPreviousLines sets whether the row before this node is blank.
+	// This method is valid only for block nodes.
+	SetBlankPreviousLines(v bool)
+
+	// Lines returns text segments that hold positions in a source.
+	// This method is valid only for block nodes.
+	Lines() *textm.Segments
+
+	// SetLines sets text segments that hold positions in a source.
+	// This method is valid only for block nodes.
+	SetLines(*textm.Segments)
+
+	// IsRaw returns true if contents should be rendered as 'raw' contents.
+	IsRaw() bool
+}
+
+// A BaseNode struct implements the Node interface.
+type BaseNode struct {
+	firstChild Node
+	lastChild  Node
+	parent     Node
+	next       Node
+	prev       Node
+}
+
+func ensureIsolated(v Node) {
+	if p := v.Parent(); p != nil {
+		p.RemoveChild(p, v)
+	}
+}
+
+// HasChildren implements Node.HasChildren .
+func (n *BaseNode) HasChildren() bool {
+	return n.firstChild != nil
+}
+
+// SetPreviousSibling implements Node.SetPreviousSibling .
+func (n *BaseNode) SetPreviousSibling(v Node) {
+	n.prev = v
+}
+
+// SetNextSibling implements Node.SetNextSibling .
+func (n *BaseNode) SetNextSibling(v Node) {
+	n.next = v
+}
+
+// PreviousSibling implements Node.PreviousSibling .
+func (n *BaseNode) PreviousSibling() Node {
+	return n.prev
+}
+
+// NextSibling implements Node.NextSibling .
+func (n *BaseNode) NextSibling() Node {
+	return n.next
+}
+
+// RemoveChild implements Node.RemoveChild .
+func (n *BaseNode) RemoveChild(self, v Node) {
+	if v.Parent() != self {
+		return
+	}
+	prev := v.PreviousSibling()
+	next := v.NextSibling()
+	if prev != nil {
+		prev.SetNextSibling(next)
+	} else {
+		n.firstChild = next
+	}
+	if next != nil {
+		next.SetPreviousSibling(prev)
+	} else {
+		n.lastChild = prev
+	}
+	v.SetParent(nil)
+	v.SetPreviousSibling(nil)
+	v.SetNextSibling(nil)
+}
+
+// RemoveChildren implements Node.RemoveChildren .
+func (n *BaseNode) RemoveChildren(self Node) {
+	for c := n.firstChild; c != nil; c = c.NextSibling() {
+		c.SetParent(nil)
+		c.SetPreviousSibling(nil)
+		c.SetNextSibling(nil)
+	}
+	n.firstChild = nil
+	n.lastChild = nil
+}
+
+// FirstChild implements Node.FirstChild .
+func (n *BaseNode) FirstChild() Node {
+	return n.firstChild
+}
+
+// LastChild implements Node.LastChild .
+func (n *BaseNode) LastChild() Node {
+	return n.lastChild
+}
+
+// ChildCount implements Node.ChildCount .
+func (n *BaseNode) ChildCount() int {
+	count := 0
+	for c := n.firstChild; c != nil; c = c.NextSibling() {
+		count++
+	}
+	return count
+}
+
+// Parent implements Node.Parent .
+func (n *BaseNode) Parent() Node {
+	return n.parent
+}
+
+// SetParent implements Node.SetParent .
+func (n *BaseNode) SetParent(v Node) {
+	n.parent = v
+}
+
+// AppendChild implements Node.AppendChild .
+func (n *BaseNode) AppendChild(self, v Node) {
+	ensureIsolated(v)
+	if n.firstChild == nil {
+		n.firstChild = v
+		v.SetNextSibling(nil)
+		v.SetPreviousSibling(nil)
+	} else {
+		last := n.lastChild
+		last.SetNextSibling(v)
+		v.SetPreviousSibling(last)
+	}
+	v.SetParent(self)
+	n.lastChild = v
+}
+
+// ReplaceChild implements Node.ReplaceChild .
+func (n *BaseNode) ReplaceChild(self, v1, insertee Node) {
+	n.InsertBefore(self, v1, insertee)
+	n.RemoveChild(self, v1)
+}
+
+// InsertAfter implements Node.InsertAfter .
+func (n *BaseNode) InsertAfter(self, v1, insertee Node) {
+	n.InsertBefore(self, v1.NextSibling(), insertee)
+}
+
+// InsertBefore implements Node.InsertBefore .
+func (n *BaseNode) InsertBefore(self, v1, insertee Node) {
+	if v1 == nil {
+		n.AppendChild(self, insertee)
+		return
+	}
+	ensureIsolated(insertee)
+	if v1.Parent() == self {
+		c := v1
+		prev := c.PreviousSibling()
+		if prev != nil {
+			prev.SetNextSibling(insertee)
+			insertee.SetPreviousSibling(prev)
+		} else {
+			n.firstChild = insertee
+			insertee.SetPreviousSibling(nil)
+		}
+		insertee.SetNextSibling(c)
+		c.SetPreviousSibling(insertee)
+		insertee.SetParent(self)
+	}
+}
+
+// Text implements Node.Text  .
+func (n *BaseNode) Text(source []byte) []byte {
+	var buf bytes.Buffer
+	for c := n.firstChild; c != nil; c = c.NextSibling() {
+		buf.Write(c.Text(source))
+	}
+	return buf.Bytes()
+}
+
+// DumpHelper is a helper function to implement Node.Dump.
+// name is a name of the node.
+// kv is pairs of an attribute name and an attribute value.
+// cb is a function called after wrote a name and attributes.
+func DumpHelper(v Node, source []byte, level int, name string, kv map[string]string, cb func(int)) {
+	indent := strings.Repeat("    ", level)
+	fmt.Printf("%s%s {\n", indent, name)
+	indent2 := strings.Repeat("    ", level+1)
+	if v.Type() == BlockNode {
+		fmt.Printf("%sRawText: \"", indent2)
+		for i := 0; i < v.Lines().Len(); i++ {
+			line := v.Lines().At(i)
+			fmt.Printf("%s", line.Value(source))
+		}
+		fmt.Printf("\"\n")
+		fmt.Printf("%sHasBlankPreviousLines: %v\n", indent2, v.HasBlankPreviousLines())
+	}
+	if kv != nil {
+		for name, value := range kv {
+			fmt.Printf("%s%s: %s\n", indent2, name, value)
+		}
+	}
+	if cb != nil {
+		cb(level + 1)
+	}
+	for c := v.FirstChild(); c != nil; c = c.NextSibling() {
+		c.Dump(source, level+1)
+	}
+	fmt.Printf("%s}\n", indent)
+}
+
+// WalkStatus represents a current status of the Walk function.
+type WalkStatus int
+
+const (
+	// WalkStop indicates no more walking needed.
+	WalkStop = iota + 1
+
+	// WalkSkipChildren indicates that Walk wont walk on children of current
+	// node.
+	WalkSkipChildren
+
+	// WalkContinue indicates that Walk can continue to walk.
+	WalkContinue
+)
+
+// Walker is a function that will be called when Walk find a
+// new node.
+// entering is set true before walks children, false after walked children.
+// If Walker returns error, Walk function immediately stop walking.
+type Walker func(n Node, entering bool) (WalkStatus, error)
+
+// Walk walks a AST tree by the depth first search algorighm.
+func Walk(n Node, walker Walker) error {
+	status, err := walker(n, true)
+	if err != nil || status == WalkStop {
+		return err
+	}
+	if status != WalkSkipChildren {
+		for c := n.FirstChild(); c != nil; c = c.NextSibling() {
+			if err := Walk(c, walker); err != nil {
+				return err
+			}
+		}
+	}
+	status, err = walker(n, false)
+	if err != nil || status == WalkStop {
+		return err
+	}
+	return nil
+}

ast/block.go

@@ -0,0 +1,364 @@
+package ast
+
+import (
+	"fmt"
+	textm "github.com/yuin/goldmark/text"
+	"strings"
+)
+
+// A BaseBlock struct implements the Node interface.
+type BaseBlock struct {
+	BaseNode
+	blankPreviousLines bool
+	lines              *textm.Segments
+}
+
+// Type implements Node.Type
+func (b *BaseBlock) Type() NodeType {
+	return BlockNode
+}
+
+// IsRaw implements Node.IsRaw
+func (b *BaseBlock) IsRaw() bool {
+	return false
+}
+
+// HasBlankPreviousLines implements Node.HasBlankPreviousLines.
+func (b *BaseBlock) HasBlankPreviousLines() bool {
+	return b.blankPreviousLines
+}
+
+// SetBlankPreviousLines implements Node.SetBlankPreviousLines.
+func (b *BaseBlock) SetBlankPreviousLines(v bool) {
+	b.blankPreviousLines = v
+}
+
+// Lines implements Node.Lines
+func (b *BaseBlock) Lines() *textm.Segments {
+	if b.lines == nil {
+		b.lines = textm.NewSegments()
+	}
+	return b.lines
+}
+
+// SetLines implements Node.SetLines
+func (b *BaseBlock) SetLines(v *textm.Segments) {
+	b.lines = v
+}
+
+// A Document struct is a root node of Markdown text.
+type Document struct {
+	BaseBlock
+}
+
+// Dump impelements Node.Dump .
+func (n *Document) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "Document", nil, nil)
+}
+
+// NewDocument returns a new Document node.
+func NewDocument() *Document {
+	return &Document{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// A TextBlock struct is a node whose lines
+// should be rendered without any containers.
+type TextBlock struct {
+	BaseBlock
+}
+
+// Dump impelements Node.Dump .
+func (n *TextBlock) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "TextBlock", nil, nil)
+}
+
+// NewTextBlock returns a new TextBlock node.
+func NewTextBlock() *TextBlock {
+	return &TextBlock{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// A Paragraph struct represents a paragraph of Markdown text.
+type Paragraph struct {
+	BaseBlock
+}
+
+// Dump impelements Node.Dump .
+func (n *Paragraph) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "Paragraph", nil, nil)
+}
+
+// NewParagraph returns a new Paragraph node.
+func NewParagraph() *Paragraph {
+	return &Paragraph{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// IsParagraph returns true if given node implements the Paragraph interface,
+// otherwise false.
+func IsParagraph(node Node) bool {
+	_, ok := node.(*Paragraph)
+	return ok
+}
+
+// A Heading struct represents headings like SetextHeading and ATXHeading.
+type Heading struct {
+	BaseBlock
+	// Level returns a level of this heading.
+	// This value is between 1 and 6.
+	Level int
+
+	// ID returns an ID of this heading.
+	ID []byte
+}
+
+// Dump impelements Node.Dump .
+func (n *Heading) Dump(source []byte, level int) {
+	m := map[string]string{
+		"Level": fmt.Sprintf("%d", n.Level),
+	}
+	DumpHelper(n, source, level, "Heading", m, nil)
+}
+
+// NewHeading returns a new Heading node.
+func NewHeading(level int) *Heading {
+	return &Heading{
+		BaseBlock: BaseBlock{},
+		Level:     level,
+	}
+}
+
+// A ThemanticBreak struct represents a themantic break of Markdown text.
+type ThemanticBreak struct {
+	BaseBlock
+}
+
+// Dump impelements Node.Dump .
+func (n *ThemanticBreak) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "ThemanticBreak", nil, nil)
+}
+
+// NewThemanticBreak returns a new ThemanticBreak node.
+func NewThemanticBreak() *ThemanticBreak {
+	return &ThemanticBreak{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// A CodeBlock interface represents an indented code block of Markdown text.
+type CodeBlock struct {
+	BaseBlock
+}
+
+// IsRaw implements Node.IsRaw.
+func (n *CodeBlock) IsRaw() bool {
+	return true
+}
+
+// Dump impelements Node.Dump .
+func (n *CodeBlock) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "CodeBlock", nil, nil)
+}
+
+// NewCodeBlock returns a new CodeBlock node.
+func NewCodeBlock() *CodeBlock {
+	return &CodeBlock{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// A FencedCodeBlock struct represents a fenced code block of Markdown text.
+type FencedCodeBlock struct {
+	BaseBlock
+	// Info returns a info text of this fenced code block.
+	Info *Text
+}
+
+// IsRaw implements Node.IsRaw.
+func (n *FencedCodeBlock) IsRaw() bool {
+	return true
+}
+
+// Dump impelements Node.Dump .
+func (n *FencedCodeBlock) Dump(source []byte, level int) {
+	m := map[string]string{}
+	if n.Info != nil {
+		m["Info"] = fmt.Sprintf("\"%s\"", n.Info.Text(source))
+	}
+	DumpHelper(n, source, level, "FencedCodeBlock", m, nil)
+}
+
+// NewFencedCodeBlock return a new FencedCodeBlock node.
+func NewFencedCodeBlock(info *Text) *FencedCodeBlock {
+	return &FencedCodeBlock{
+		BaseBlock: BaseBlock{},
+		Info:      info,
+	}
+}
+
+// A Blockquote struct represents an blockquote block of Markdown text.
+type Blockquote struct {
+	BaseBlock
+}
+
+// Dump impelements Node.Dump .
+func (n *Blockquote) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "Blockquote", nil, nil)
+}
+
+// NewBlockquote returns a new Blockquote node.
+func NewBlockquote() *Blockquote {
+	return &Blockquote{
+		BaseBlock: BaseBlock{},
+	}
+}
+
+// A List structr represents a list of Markdown text.
+type List struct {
+	BaseBlock
+
+	// Marker is a markar character like '-', '+', ')' and '.'.
+	Marker byte
+
+	// IsTight is a true if this list is a 'tight' list.
+	// See https://spec.commonmark.org/0.29/#loose for details.
+	IsTight bool
+
+	// Start is an initial number of this ordered list.
+	// If this list is not an ordered list, Start is 0.
+	Start int
+}
+
+// IsOrdered returns true if this list is an ordered list, otherwise false.
+func (l *List) IsOrdered() bool {
+	return l.Marker == '.' || l.Marker == ')'
+}
+
+// CanContinue returns true if this list can continue with
+// given mark and a list type, otherwise false.
+func (l *List) CanContinue(marker byte, isOrdered bool) bool {
+	return marker == l.Marker && isOrdered == l.IsOrdered()
+}
+
+// Dump implements Node.Dump.
+func (l *List) Dump(source []byte, level int) {
+	name := "List"
+	if l.IsOrdered() {
+		name = "OrderedList"
+	}
+	m := map[string]string{
+		"Marker": fmt.Sprintf("%c", l.Marker),
+		"Tight":  fmt.Sprintf("%v", l.IsTight),
+	}
+	if l.IsOrdered() {
+		m["Start"] = fmt.Sprintf("%d", l.Start)
+	}
+	DumpHelper(l, source, level, name, m, nil)
+}
+
+// NewList returns a new List node.
+func NewList(marker byte) *List {
+	return &List{
+		BaseBlock: BaseBlock{},
+		Marker:    marker,
+		IsTight:   true,
+	}
+}
+
+// A ListItem struct represents a list item of Markdown text.
+type ListItem struct {
+	BaseBlock
+
+	// Offset is an offset potision of this item.
+	Offset int
+}
+
+// Dump implements Node.Dump.
+func (n *ListItem) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "ListItem", nil, nil)
+}
+
+// NewListItem returns a new ListItem node.
+func NewListItem(offset int) *ListItem {
+	return &ListItem{
+		BaseBlock: BaseBlock{},
+		Offset:    offset,
+	}
+}
+
+// HTMLBlockType represents kinds of an html blocks.
+// See https://spec.commonmark.org/0.29/#html-blocks
+type HTMLBlockType int
+
+const (
+	// HTMLBlockType1 represents type 1 html blocks
+	HTMLBlockType1 = iota + 1
+	// HTMLBlockType2 represents type 2 html blocks
+	HTMLBlockType2
+	// HTMLBlockType3 represents type 3 html blocks
+	HTMLBlockType3
+	// HTMLBlockType4 represents type 4 html blocks
+	HTMLBlockType4
+	// HTMLBlockType5 represents type 5 html blocks
+	HTMLBlockType5
+	// HTMLBlockType6 represents type 6 html blocks
+	HTMLBlockType6
+	// HTMLBlockType7 represents type 7 html blocks
+	HTMLBlockType7
+)
+
+// An HTMLBlock struct represents an html block of Markdown text.
+type HTMLBlock struct {
+	BaseBlock
+
+	// Type is a type of this html block.
+	HTMLBlockType HTMLBlockType
+
+	// ClosureLine is a line that closes this html block.
+	ClosureLine textm.Segment
+}
+
+// IsRaw implements Node.IsRaw.
+func (n *HTMLBlock) IsRaw() bool {
+	return true
+}
+
+// HasClosure returns true if this html block has a closure line,
+// otherwise false.
+func (n *HTMLBlock) HasClosure() bool {
+	return n.ClosureLine.Start >= 0
+}
+
+// Dump implements Node.Dump.
+func (n *HTMLBlock) Dump(source []byte, level int) {
+	indent := strings.Repeat("    ", level)
+	fmt.Printf("%s%s {\n", indent, "HTMLBlock")
+	indent2 := strings.Repeat("    ", level+1)
+	fmt.Printf("%sRawText: \"", indent2)
+	for i := 0; i < n.Lines().Len(); i++ {
+		s := n.Lines().At(i)
+		fmt.Print(string(source[s.Start:s.Stop]))
+	}
+	fmt.Printf("\"\n")
+	for c := n.FirstChild(); c != nil; c = c.NextSibling() {
+		c.Dump(source, level+1)
+	}
+	if n.HasClosure() {
+		cl := n.ClosureLine
+		fmt.Printf("%sClosure: \"%s\"\n", indent2, string(cl.Value(source)))
+	}
+	fmt.Printf("%s}\n", indent)
+}
+
+// NewHTMLBlock returns a new HTMLBlock node.
+func NewHTMLBlock(typ HTMLBlockType) *HTMLBlock {
+	return &HTMLBlock{
+		BaseBlock:     BaseBlock{},
+		HTMLBlockType: typ,
+		ClosureLine:   textm.NewSegment(-1, -1),
+	}
+}

ast/inline.go

@@ -0,0 +1,371 @@
+package ast
+
+import (
+	"fmt"
+	"strings"
+
+	textm "github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+// A BaseInline struct implements the Node interface.
+type BaseInline struct {
+	BaseNode
+}
+
+// Type implements Node.Type
+func (b *BaseInline) Type() NodeType {
+	return InlineNode
+}
+
+// IsRaw implements Node.IsRaw
+func (b *BaseInline) IsRaw() bool {
+	return false
+}
+
+// HasBlankPreviousLines implements Node.HasBlankPreviousLines.
+func (b *BaseInline) HasBlankPreviousLines() bool {
+	panic("can not call with inline nodes.")
+}
+
+// SetBlankPreviousLines implements Node.SetBlankPreviousLines.
+func (b *BaseInline) SetBlankPreviousLines(v bool) {
+	panic("can not call with inline nodes.")
+}
+
+// Lines implements Node.Lines
+func (b *BaseInline) Lines() *textm.Segments {
+	panic("can not call with inline nodes.")
+}
+
+// SetLines implements Node.SetLines
+func (b *BaseInline) SetLines(v *textm.Segments) {
+	panic("can not call with inline nodes.")
+}
+
+// A Text struct represents a textual content of the Markdown text.
+type Text struct {
+	BaseInline
+	// Segment is a position in a source text.
+	Segment textm.Segment
+
+	flags uint8
+}
+
+const (
+	textSoftLineBreak = 1 << iota
+	textHardLineBreak
+	textRaw
+)
+
+// Inline implements Inline.Inline.
+func (n *Text) Inline() {
+}
+
+// SoftLineBreak returns true if this node ends with a new line,
+// otherwise false.
+func (n *Text) SoftLineBreak() bool {
+	return n.flags&textSoftLineBreak != 0
+}
+
+// SetSoftLineBreak sets whether this node ends with a new line.
+func (n *Text) SetSoftLineBreak(v bool) {
+	if v {
+		n.flags |= textSoftLineBreak
+	} else {
+		n.flags = n.flags &^ textHardLineBreak
+	}
+}
+
+// IsRaw returns true if this text should be rendered without unescaping
+// back slash escapes and resolving references.
+func (n *Text) IsRaw() bool {
+	return n.flags&textRaw != 0
+}
+
+// SetRaw sets whether this text should be rendered as raw contents.
+func (n *Text) SetRaw(v bool) {
+	if v {
+		n.flags |= textRaw
+	} else {
+		n.flags = n.flags &^ textRaw
+	}
+}
+
+// HardLineBreak returns true if this node ends with a hard line break.
+// See https://spec.commonmark.org/0.29/#hard-line-breaks for details.
+func (n *Text) HardLineBreak() bool {
+	return n.flags&textHardLineBreak != 0
+}
+
+// SetHardLineBreak sets whether this node ends with a hard line break.
+func (n *Text) SetHardLineBreak(v bool) {
+	if v {
+		n.flags |= textHardLineBreak
+	} else {
+		n.flags = n.flags &^ textHardLineBreak
+	}
+}
+
+// Merge merges a Node n into this node.
+// Merge returns true if given node has been merged, otherwise false.
+func (n *Text) Merge(node Node, source []byte) bool {
+	t, ok := node.(*Text)
+	if !ok {
+		return false
+	}
+	if n.Segment.Stop != t.Segment.Start || t.Segment.Padding != 0 || source[n.Segment.Stop-1] == '\n' || t.IsRaw() != n.IsRaw() {
+		return false
+	}
+	n.Segment.Stop = t.Segment.Stop
+	n.SetSoftLineBreak(t.SoftLineBreak())
+	n.SetHardLineBreak(t.HardLineBreak())
+	return true
+}
+
+// Text implements Node.Text.
+func (n *Text) Text(source []byte) []byte {
+	return n.Segment.Value(source)
+}
+
+// Dump implements Node.Dump.
+func (n *Text) Dump(source []byte, level int) {
+	fmt.Printf("%sText: \"%s\"\n", strings.Repeat("    ", level), strings.TrimRight(string(n.Text(source)), "\n"))
+}
+
+// NewText returns a new Text node.
+func NewText() *Text {
+	return &Text{
+		BaseInline: BaseInline{},
+	}
+}
+
+// NewTextSegment returns a new Text node with given source potision.
+func NewTextSegment(v textm.Segment) *Text {
+	return &Text{
+		BaseInline: BaseInline{},
+		Segment:    v,
+	}
+}
+
+// NewRawTextSegment returns a new Text node with given source position.
+// The new node should be rendered as raw contents.
+func NewRawTextSegment(v textm.Segment) *Text {
+	t := &Text{
+		BaseInline: BaseInline{},
+		Segment:    v,
+	}
+	t.SetRaw(true)
+	return t
+}
+
+// MergeOrAppendTextSegment merges a given s into the last child of the parent if
+// it can be merged, otherwise creates a new Text node and appends it to after current
+// last child.
+func MergeOrAppendTextSegment(parent Node, s textm.Segment) {
+	last := parent.LastChild()
+	t, ok := last.(*Text)
+	if ok && t.Segment.Stop == s.Start && !t.SoftLineBreak() {
+		ts := t.Segment
+		t.Segment = ts.WithStop(s.Stop)
+	} else {
+		parent.AppendChild(parent, NewTextSegment(s))
+	}
+}
+
+// MergeOrReplaceTextSegment merges a given s into a previous sibling of the node n
+// if a previous sibling of the node n is *Text, otherwise replaces Node n with s.
+func MergeOrReplaceTextSegment(parent Node, n Node, s textm.Segment) {
+	prev := n.PreviousSibling()
+	if t, ok := prev.(*Text); ok && t.Segment.Stop == s.Start && !t.SoftLineBreak() {
+		t.Segment = t.Segment.WithStop(s.Stop)
+		parent.RemoveChild(parent, n)
+	} else {
+		parent.ReplaceChild(parent, n, NewTextSegment(s))
+	}
+}
+
+// A CodeSpan struct represents a code span of Markdown text.
+type CodeSpan struct {
+	BaseInline
+}
+
+// Inline implements Inline.Inline .
+func (n *CodeSpan) Inline() {
+}
+
+// IsBlank returns true if this node consists of spaces, otherwise false.
+func (n *CodeSpan) IsBlank(source []byte) bool {
+	for c := n.FirstChild(); c != nil; c = c.NextSibling() {
+		text := c.(*Text).Segment
+		if !util.IsBlank(text.Value(source)) {
+			return false
+		}
+	}
+	return true
+}
+
+// Dump implements Node.Dump
+func (n *CodeSpan) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "CodeSpan", nil, nil)
+}
+
+// NewCodeSpan returns a new CodeSpan node.
+func NewCodeSpan() *CodeSpan {
+	return &CodeSpan{
+		BaseInline: BaseInline{},
+	}
+}
+
+// An Emphasis struct represents an emphasis of Markdown text.
+type Emphasis struct {
+	BaseInline
+
+	// Level is a level of the emphasis.
+	Level int
+}
+
+// Inline implements Inline.Inline.
+func (n *Emphasis) Inline() {
+}
+
+// Dump implements Node.Dump.
+func (n *Emphasis) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, fmt.Sprintf("Emphasis(%d)", n.Level), nil, nil)
+}
+
+// NewEmphasis returns a new Emphasis node with given level.
+func NewEmphasis(level int) *Emphasis {
+	return &Emphasis{
+		BaseInline: BaseInline{},
+		Level:      level,
+	}
+}
+
+type baseLink struct {
+	BaseInline
+
+	// Destination is a destination(URL) of this link.
+	Destination []byte
+
+	// Title is a title of this link.
+	Title []byte
+}
+
+// Inline implements Inline.Inline.
+func (n *baseLink) Inline() {
+}
+
+func (n *baseLink) Dump(source []byte, level int) {
+	m := map[string]string{}
+	m["Destination"] = string(n.Destination)
+	m["Title"] = string(n.Title)
+	DumpHelper(n, source, level, "Link", m, nil)
+}
+
+// A Link struct represents a link of the Markdown text.
+type Link struct {
+	baseLink
+}
+
+// NewLink returns a new Link node.
+func NewLink() *Link {
+	c := &Link{
+		baseLink: baseLink{
+			BaseInline: BaseInline{},
+		},
+	}
+	return c
+}
+
+// An Image struct represents an image of the Markdown text.
+type Image struct {
+	baseLink
+}
+
+// Dump implements Node.Dump.
+func (n *Image) Dump(source []byte, level int) {
+	m := map[string]string{}
+	m["Destination"] = string(n.Destination)
+	m["Title"] = string(n.Title)
+	DumpHelper(n, source, level, "Image", m, nil)
+}
+
+// NewImage returns a new Image node.
+func NewImage(link *Link) *Image {
+	c := &Image{
+		baseLink: baseLink{
+			BaseInline: BaseInline{},
+		},
+	}
+	c.Destination = link.Destination
+	c.Title = link.Title
+	for n := link.FirstChild(); n != nil; {
+		next := n.NextSibling()
+		link.RemoveChild(link, n)
+		c.AppendChild(c, n)
+		n = next
+	}
+
+	return c
+}
+
+// AutoLinkType defines kind of auto links.
+type AutoLinkType int
+
+const (
+	// AutoLinkEmail indicates that an autolink is an email address.
+	AutoLinkEmail = iota + 1
+	// AutoLinkURL indicates that an autolink is a generic URL.
+	AutoLinkURL
+)
+
+// An AutoLink struct represents an autolink of the Markdown text.
+type AutoLink struct {
+	BaseInline
+	// Value is a link text of this node.
+	Value *Text
+	// Type is a type of this autolink.
+	AutoLinkType AutoLinkType
+}
+
+// Inline implements Inline.Inline.
+func (n *AutoLink) Inline() {}
+
+// Dump implenets Node.Dump
+func (n *AutoLink) Dump(source []byte, level int) {
+	segment := n.Value.Segment
+	m := map[string]string{
+		"Value": string(segment.Value(source)),
+	}
+	DumpHelper(n, source, level, "AutoLink", m, nil)
+}
+
+// NewAutoLink returns a new AutoLink node.
+func NewAutoLink(typ AutoLinkType, value *Text) *AutoLink {
+	return &AutoLink{
+		BaseInline:   BaseInline{},
+		Value:        value,
+		AutoLinkType: typ,
+	}
+}
+
+// A RawHTML struct represents an inline raw HTML of the Markdown text.
+type RawHTML struct {
+	BaseInline
+}
+
+// Inline implements Inline.Inline.
+func (n *RawHTML) Inline() {}
+
+// Dump implements Node.Dump.
+func (n *RawHTML) Dump(source []byte, level int) {
+	DumpHelper(n, source, level, "RawHTML", nil, nil)
+}
+
+// NewRawHTML returns a new RawHTML node.
+func NewRawHTML() *RawHTML {
+	return &RawHTML{
+		BaseInline: BaseInline{},
+	}
+}

commonmark_test.go

@@ -0,0 +1,53 @@
+package goldmark
+
+import (
+	"bytes"
+	"encoding/json"
+	"io/ioutil"
+	"testing"
+
+	"github.com/yuin/goldmark/renderer/html"
+)
+
+type commonmarkSpecTestCase struct {
+	Markdown  string `json:"markdown"`
+	HTML      string `json:"html"`
+	Example   int    `json:"example"`
+	StartLine int    `json:"start_line"`
+	EndLine   int    `json:"end_line"`
+	Section   string `json:"section"`
+}
+
+func TestSpec(t *testing.T) {
+	bs, err := ioutil.ReadFile("_test/spec.json")
+	if err != nil {
+		panic(err)
+	}
+	var testCases []commonmarkSpecTestCase
+	if err := json.Unmarshal(bs, &testCases); err != nil {
+		panic(err)
+	}
+	markdown := New(WithRendererOptions(html.WithXHTML()))
+	for _, testCase := range testCases {
+		var out bytes.Buffer
+		if err := markdown.Convert([]byte(testCase.Markdown), &out); err != nil {
+			panic(err)
+		}
+		if !bytes.Equal(bytes.TrimSpace(out.Bytes()), bytes.TrimSpace([]byte(testCase.HTML))) {
+			format := `============= case %d ================
+Markdown:
+-----------
+%s
+
+Expected:
+----------
+%s
+
+Actual
+---------
+%s
+`
+			t.Errorf(format, testCase.Example, testCase.Markdown, testCase.HTML, out.Bytes())
+		}
+	}
+}

extension/ast/strikethrough.go

@@ -0,0 +1,23 @@
+// Package ast defines AST nodes that represents extension's elements
+package ast
+
+import (
+	gast "github.com/yuin/goldmark/ast"
+)
+
+// A Strikethrough struct represents a strikethrough of GFM text.
+type Strikethrough struct {
+	gast.BaseInline
+}
+
+func (n *Strikethrough) Inline() {
+}
+
+func (n *Strikethrough) Dump(source []byte, level int) {
+	gast.DumpHelper(n, source, level, "Strikethrough", nil, nil)
+}
+
+// NewStrikethrough returns a new Strikethrough node.
+func NewStrikethrough() *Strikethrough {
+	return &Strikethrough{}
+}

extension/ast/table.go

@@ -0,0 +1,113 @@
+package ast
+
+import (
+	"fmt"
+	gast "github.com/yuin/goldmark/ast"
+	"strings"
+)
+
+// Alignment is a text alignment of table cells.
+type Alignment int
+
+const (
+	// AlignLeft indicates text should be left justified.
+	AlignLeft Alignment = iota + 1
+
+	// AlignRight indicates text should be right justified.
+	AlignRight
+
+	// AlignCenter indicates text should be centered.
+	AlignCenter
+
+	// AlignNone indicates text should be aligned by default manner.
+	AlignNone
+)
+
+func (a Alignment) String() string {
+	switch a {
+	case AlignLeft:
+		return "left"
+	case AlignRight:
+		return "right"
+	case AlignCenter:
+		return "center"
+	case AlignNone:
+		return "none"
+	}
+	return ""
+}
+
+// A Table struct represents a table of Markdown(GFM) text.
+type Table struct {
+	gast.BaseBlock
+
+	// Alignments returns alignments of the columns.
+	Alignments []Alignment
+}
+
+// Dump implements Node.Dump
+func (n *Table) Dump(source []byte, level int) {
+	gast.DumpHelper(n, source, level, "Table", nil, func(level int) {
+		indent := strings.Repeat("    ", level)
+		fmt.Printf("%sAlignments {\n", indent)
+		for i, alignment := range n.Alignments {
+			indent2 := strings.Repeat("    ", level+1)
+			fmt.Printf("%s%s", indent2, alignment.String())
+			if i != len(n.Alignments)-1 {
+				fmt.Println("")
+			}
+		}
+		fmt.Printf("\n%s}\n", indent)
+	})
+}
+
+// NewTable returns a new Table node.
+func NewTable() *Table {
+	return &Table{
+		Alignments: []Alignment{},
+	}
+}
+
+// A TableRow struct represents a table row of Markdown(GFM) text.
+type TableRow struct {
+	gast.BaseBlock
+	Alignments []Alignment
+}
+
+// Dump implements Node.Dump.
+func (n *TableRow) Dump(source []byte, level int) {
+	gast.DumpHelper(n, source, level, "TableRow", nil, nil)
+}
+
+// NewTableRow returns a new TableRow node.
+func NewTableRow(alignments []Alignment) *TableRow {
+	return &TableRow{}
+}
+
+// A TableHeader struct represents a table header of Markdown(GFM) text.
+type TableHeader struct {
+	*TableRow
+}
+
+// NewTableHeader returns a new TableHeader node.
+func NewTableHeader(row *TableRow) *TableHeader {
+	return &TableHeader{row}
+}
+
+// A TableCell struct represents a table cell of a Markdown(GFM) text.
+type TableCell struct {
+	gast.BaseBlock
+	Alignment Alignment
+}
+
+// Dump implements Node.Dump.
+func (n *TableCell) Dump(source []byte, level int) {
+	gast.DumpHelper(n, source, level, "TableCell", nil, nil)
+}
+
+// NewTableCell returns a new TableCell node.
+func NewTableCell() *TableCell {
+	return &TableCell{
+		Alignment: AlignNone,
+	}
+}

extension/ast/tasklist.go

@@ -0,0 +1,27 @@
+package ast
+
+import (
+	"fmt"
+	gast "github.com/yuin/goldmark/ast"
+)
+
+// A TaskCheckBox struct represents a checkbox of a task list.
+type TaskCheckBox struct {
+	gast.BaseInline
+	IsChecked bool
+}
+
+// Dump impelemtns Node.Dump.
+func (n *TaskCheckBox) Dump(source []byte, level int) {
+	m := map[string]string{
+		"Checked": fmt.Sprintf("%v", n.IsChecked),
+	}
+	gast.DumpHelper(n, source, level, "TaskCheckBox", m, nil)
+}
+
+// NewTaskCheckBox returns a new TaskCheckBox node.
+func NewTaskCheckBox(checked bool) *TaskCheckBox {
+	return &TaskCheckBox{
+		IsChecked: checked,
+	}
+}

extension/gfm.go

@@ -0,0 +1,32 @@
+package extension
+
+import (
+	"github.com/yuin/goldmark"
+	"github.com/yuin/goldmark/parser"
+)
+
+type gfm struct {
+}
+
+// GFM is an extension that provides Github Flavored markdown functionalities.
+var GFM = &gfm{}
+
+var filterTags = []string{
+	"title",
+	"textarea",
+	"style",
+	"xmp",
+	"iframe",
+	"noembed",
+	"noframes",
+	"script",
+	"plaintext",
+}
+
+func (e *gfm) Extend(m goldmark.Markdown) {
+	m.Parser().AddOption(parser.WithFilterTags(filterTags...))
+	Linkify.Extend(m)
+	Table.Extend(m)
+	Strikethrough.Extend(m)
+	TaskList.Extend(m)
+}

extension/linkify.go

@@ -0,0 +1,125 @@
+package extension
+
+import (
+	"bytes"
+	"github.com/yuin/goldmark"
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+)
+
+var wwwURLRegxp = regexp.MustCompile(`^www\.[-a-zA-Z0-9@:%._\+~#=]{2,256}\.[a-z]{2,6}\b(?:[-a-zA-Z0-9@:%_\+.~#?&//=\(\);]*)`)
+
+var urlRegexp = regexp.MustCompile(`^(?:http|https|ftp):\/\/(?:www\.)?[-a-zA-Z0-9@:%._\+~#=]{2,256}\.[a-z]{2,6}\b([-a-zA-Z0-9@:%_\+.~#?&//=\(\);]*)`)
+
+var emailRegexp = regexp.MustCompile(`^[a-zA-Z0-9\.\-_\+]+@([a-zA-Z0-9\.\-_]+)`)
+
+type linkifyParser struct {
+}
+
+var defaultLinkifyParser = &linkifyParser{}
+
+// NewLinkifyParser return a new InlineParser can parse
+// text that seems like a URL.
+func NewLinkifyParser() parser.InlineParser {
+	return defaultLinkifyParser
+}
+
+func (s *linkifyParser) Trigger() []byte {
+	// ' ' indicates any white spaces and a line head
+	return []byte{' ', '*', '_', '~', '('}
+}
+
+func (s *linkifyParser) Parse(parent ast.Node, block text.Reader, pc parser.Context) ast.Node {
+	line, segment := block.PeekLine()
+	consumes := 0
+	start := segment.Start
+	c := line[0]
+	// advance if current position is not a line head.
+	if c == ' ' || c == '*' || c == '_' || c == '~' || c == '(' {
+		consumes++
+		start++
+		line = line[1:]
+	}
+
+	var m []int
+	typ := ast.AutoLinkType(ast.AutoLinkEmail)
+	typ = ast.AutoLinkURL
+	m = urlRegexp.FindSubmatchIndex(line)
+	if m == nil {
+		m = wwwURLRegxp.FindSubmatchIndex(line)
+	}
+	if m != nil {
+		lastChar := line[m[1]-1]
+		if lastChar == '.' {
+			m[1]--
+		} else if lastChar == ')' {
+			closing := 0
+			for i := m[1] - 1; i >= m[0]; i-- {
+				if line[i] == ')' {
+					closing++
+				} else if line[i] == '(' {
+					closing--
+				}
+			}
+			if closing > 0 {
+				m[1]--
+			}
+		} else if lastChar == ';' {
+			i := m[1] - 2
+			for ; i >= m[0]; i-- {
+				if util.IsAlphaNumeric(line[i]) {
+					continue
+				}
+				break
+			}
+			if i != m[1]-2 {
+				if line[i] == '&' {
+					m[1] -= m[1] - i
+				}
+			}
+		}
+	}
+	if m == nil {
+		typ = ast.AutoLinkEmail
+		m = emailRegexp.FindSubmatchIndex(line)
+		if m == nil || bytes.IndexByte(line[m[2]:m[3]], '.') < 0 {
+			return nil
+		}
+		lastChar := line[m[1]-1]
+		if lastChar == '.' {
+			m[1]--
+		} else if lastChar == '-' || lastChar == '_' {
+			return nil
+		}
+	}
+	if m == nil {
+		return nil
+	}
+	if consumes != 0 {
+		s := segment.WithStop(segment.Start + 1)
+		ast.MergeOrAppendTextSegment(parent, s)
+	}
+	consumes += m[1]
+	block.Advance(consumes)
+	n := ast.NewTextSegment(text.NewSegment(start, start+m[1]))
+	return ast.NewAutoLink(typ, n)
+}
+
+func (s *linkifyParser) CloseBlock(parent ast.Node, pc parser.Context) {
+	// nothing to do
+}
+
+type linkify struct {
+}
+
+// Linkify is an extension that allow you to parse text that seems like a URL.
+var Linkify = &linkify{}
+
+func (e *linkify) Extend(m goldmark.Markdown) {
+	m.Parser().AddOption(parser.WithInlineParsers(
+		util.Prioritized(NewLinkifyParser(), 999),
+	))
+}

extension/strikethrough.go

@@ -0,0 +1,111 @@
+package extension
+
+import (
+	"github.com/yuin/goldmark"
+	gast "github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/extension/ast"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/renderer"
+	"github.com/yuin/goldmark/renderer/html"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type strikethroughDelimiterProcessor struct {
+}
+
+func (p *strikethroughDelimiterProcessor) IsDelimiter(b byte) bool {
+	return b == '~'
+}
+
+func (p *strikethroughDelimiterProcessor) CanOpenCloser(opener, closer *parser.Delimiter) bool {
+	return opener.Char == closer.Char
+}
+
+func (p *strikethroughDelimiterProcessor) OnMatch(consumes int) gast.Node {
+	return ast.NewStrikethrough()
+}
+
+var defaultStrikethroughDelimiterProcessor = &strikethroughDelimiterProcessor{}
+
+type strikethroughParser struct {
+}
+
+var defaultStrikethroughParser = &strikethroughParser{}
+
+// NewStrikethroughParser return a new InlineParser that parses
+// strikethrough expressions.
+func NewStrikethroughParser() parser.InlineParser {
+	return defaultStrikethroughParser
+}
+
+func (s *strikethroughParser) Trigger() []byte {
+	return []byte{'~'}
+}
+
+func (s *strikethroughParser) Parse(parent gast.Node, block text.Reader, pc parser.Context) gast.Node {
+	before := block.PrecendingCharacter()
+	line, segment := block.PeekLine()
+	node := parser.ScanDelimiter(line, before, 2, defaultStrikethroughDelimiterProcessor)
+	if node == nil {
+		return nil
+	}
+	node.Segment = segment.WithStop(segment.Start + node.OriginalLength)
+	block.Advance(node.OriginalLength)
+	pc.PushDelimiter(node)
+	return node
+}
+
+func (s *strikethroughParser) CloseBlock(parent gast.Node, pc parser.Context) {
+	// nothing to do
+}
+
+// StrikethroughHTMLRenderer is a renderer.NodeRenderer implementation that
+// renders Strikethrough nodes.
+type StrikethroughHTMLRenderer struct {
+	html.Config
+}
+
+// NewStrikethroughHTMLRenderer returns a new StrikethroughHTMLRenderer.
+func NewStrikethroughHTMLRenderer(opts ...html.Option) renderer.NodeRenderer {
+	r := &StrikethroughHTMLRenderer{
+		Config: html.NewConfig(),
+	}
+	for _, opt := range opts {
+		opt.SetHTMLOption(&r.Config)
+	}
+	return r
+}
+
+// Render implements renderer.NodeRenderer.Render.
+func (r *StrikethroughHTMLRenderer) Render(writer util.BufWriter, source []byte, n gast.Node, entering bool) (gast.WalkStatus, error) {
+	switch node := n.(type) {
+	case *ast.Strikethrough:
+		return r.renderStrikethrough(writer, source, node, entering), nil
+	}
+	return gast.WalkContinue, renderer.NotSupported
+}
+
+func (r *StrikethroughHTMLRenderer) renderStrikethrough(w util.BufWriter, source []byte, n *ast.Strikethrough, entering bool) gast.WalkStatus {
+	if entering {
+		w.WriteString("<del>")
+	} else {
+		w.WriteString("</del>")
+	}
+	return gast.WalkContinue
+}
+
+type strikethrough struct {
+}
+
+// Strikethrough is an extension that allow you to use strikethrough expression like '~~text~~' .
+var Strikethrough = &strikethrough{}
+
+func (e *strikethrough) Extend(m goldmark.Markdown) {
+	m.Parser().AddOption(parser.WithInlineParsers(
+		util.Prioritized(NewStrikethroughParser(), 500),
+	))
+	m.Renderer().AddOption(renderer.WithNodeRenderers(
+		util.Prioritized(NewStrikethroughHTMLRenderer(), 500),
+	))
+}

extension/table.go

@@ -0,0 +1,233 @@
+package extension
+
+import (
+	"bytes"
+	"fmt"
+	"regexp"
+
+	"github.com/yuin/goldmark"
+	gast "github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/extension/ast"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/renderer"
+	"github.com/yuin/goldmark/renderer/html"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+var tableDelimRegexp = regexp.MustCompile(`^[\s\-\|\:]+$`)
+var tableDelimLeft = regexp.MustCompile(`^\s*\:\-+\s*$`)
+var tableDelimRight = regexp.MustCompile(`^\s*\-+\:\s*$`)
+var tableDelimCenter = regexp.MustCompile(`^\s*\:\-+\:\s*$`)
+var tableDelimNone = regexp.MustCompile(`^\s*\-+\s*$`)
+
+type tableParagraphTransformer struct {
+}
+
+var defaultTableParagraphTransformer = &tableParagraphTransformer{}
+
+// NewTableParagraphTransformer returns  a new ParagraphTransformer
+// that can transform pargraphs into tables.
+func NewTableParagraphTransformer() parser.ParagraphTransformer {
+	return defaultTableParagraphTransformer
+}
+
+func (b *tableParagraphTransformer) Transform(node *gast.Paragraph, pc parser.Context) {
+	lines := node.Lines()
+	if lines.Len() < 2 {
+		return
+	}
+	alignments := b.parseDelimiter(lines.At(1), pc)
+	if alignments == nil {
+		return
+	}
+	header := b.parseRow(lines.At(0), alignments, pc)
+	if header == nil || len(alignments) != header.ChildCount() {
+		return
+	}
+	table := ast.NewTable()
+	table.Alignments = alignments
+	table.AppendChild(table, ast.NewTableHeader(header))
+	if lines.Len() > 2 {
+		for i := 2; i < lines.Len(); i++ {
+			table.AppendChild(table, b.parseRow(lines.At(i), alignments, pc))
+		}
+	}
+	node.Parent().InsertBefore(node.Parent(), node, table)
+	node.Parent().RemoveChild(node.Parent(), node)
+	return
+}
+
+func (b *tableParagraphTransformer) parseRow(segment text.Segment, alignments []ast.Alignment, pc parser.Context) *ast.TableRow {
+	line := segment.Value(pc.Source())
+	pos := 0
+	pos += util.TrimLeftSpaceLength(line)
+	limit := len(line)
+	limit -= util.TrimRightSpaceLength(line)
+	row := ast.NewTableRow(alignments)
+	if len(line) > 0 && line[pos] == '|' {
+		pos++
+	}
+	if len(line) > 0 && line[limit-1] == '|' {
+		limit--
+	}
+	for i := 0; pos < limit; i++ {
+		closure := util.FindClosure(line[pos:], byte(0), '|', true, false)
+		if closure < 0 {
+			closure = len(line[pos:])
+		}
+		node := ast.NewTableCell()
+		segment := text.NewSegment(segment.Start+pos, segment.Start+pos+closure)
+		segment = segment.TrimLeftSpace(pc.Source())
+		segment = segment.TrimRightSpace(pc.Source())
+		node.Lines().Append(segment)
+		node.Alignment = alignments[i]
+		row.AppendChild(row, node)
+		pos += closure + 1
+	}
+	return row
+}
+
+func (b *tableParagraphTransformer) parseDelimiter(segment text.Segment, pc parser.Context) []ast.Alignment {
+	line := segment.Value(pc.Source())
+	if !tableDelimRegexp.Match(line) {
+		return nil
+	}
+	cols := bytes.Split(line, []byte{'|'})
+	if util.IsBlank(cols[0]) {
+		cols = cols[1:]
+	}
+	if len(cols) > 0 && util.IsBlank(cols[len(cols)-1]) {
+		cols = cols[:len(cols)-1]
+	}
+
+	var alignments []ast.Alignment
+	for _, col := range cols {
+		if tableDelimLeft.Match(col) {
+			if alignments == nil {
+				alignments = []ast.Alignment{}
+			}
+			alignments = append(alignments, ast.AlignLeft)
+		} else if tableDelimRight.Match(col) {
+			if alignments == nil {
+				alignments = []ast.Alignment{}
+			}
+			alignments = append(alignments, ast.AlignRight)
+		} else if tableDelimCenter.Match(col) {
+			if alignments == nil {
+				alignments = []ast.Alignment{}
+			}
+			alignments = append(alignments, ast.AlignCenter)
+		} else if tableDelimNone.Match(col) {
+			if alignments == nil {
+				alignments = []ast.Alignment{}
+			}
+			alignments = append(alignments, ast.AlignNone)
+		} else {
+			return nil
+		}
+	}
+	return alignments
+}
+
+// TableHTMLRenderer is a renderer.NodeRenderer implementation that
+// renders Table nodes.
+type TableHTMLRenderer struct {
+	html.Config
+}
+
+// NewTableHTMLRenderer returns a new TableHTMLRenderer.
+func NewTableHTMLRenderer(opts ...html.Option) renderer.NodeRenderer {
+	r := &TableHTMLRenderer{
+		Config: html.NewConfig(),
+	}
+	for _, opt := range opts {
+		opt.SetHTMLOption(&r.Config)
+	}
+	return r
+}
+
+// Render implements renderer.Renderer.Render.
+func (r *TableHTMLRenderer) Render(writer util.BufWriter, source []byte, n gast.Node, entering bool) (gast.WalkStatus, error) {
+	switch node := n.(type) {
+	case *ast.Table:
+		return r.renderTable(writer, source, node, entering), nil
+	case *ast.TableHeader:
+		return r.renderTableHeader(writer, source, node, entering), nil
+	case *ast.TableRow:
+		return r.renderTableRow(writer, source, node, entering), nil
+	case *ast.TableCell:
+		return r.renderTableCell(writer, source, node, entering), nil
+	}
+	return gast.WalkContinue, renderer.NotSupported
+}
+
+func (r *TableHTMLRenderer) renderTable(w util.BufWriter, source []byte, n *ast.Table, entering bool) gast.WalkStatus {
+	if entering {
+		w.WriteString("<table>\n")
+	} else {
+		w.WriteString("</table>\n")
+	}
+	return gast.WalkContinue
+}
+
+func (r *TableHTMLRenderer) renderTableHeader(w util.BufWriter, source []byte, n *ast.TableHeader, entering bool) gast.WalkStatus {
+	if entering {
+		w.WriteString("<thead>\n")
+		w.WriteString("<tr>\n")
+	} else {
+		w.WriteString("</tr>\n")
+		w.WriteString("</thead>\n")
+		if n.NextSibling() != nil {
+			w.WriteString("<tbody>\n")
+		}
+		if n.Parent().LastChild() == n {
+			w.WriteString("</tbody>\n")
+		}
+	}
+	return gast.WalkContinue
+}
+
+func (r *TableHTMLRenderer) renderTableRow(w util.BufWriter, source []byte, n *ast.TableRow, entering bool) gast.WalkStatus {
+	if entering {
+		w.WriteString("<tr>\n")
+	} else {
+		w.WriteString("</tr>\n")
+		if n.Parent().LastChild() == n {
+			w.WriteString("</tbody>\n")
+		}
+	}
+	return gast.WalkContinue
+}
+
+func (r *TableHTMLRenderer) renderTableCell(w util.BufWriter, source []byte, n *ast.TableCell, entering bool) gast.WalkStatus {
+	tag := "td"
+	if n.Parent().Parent().FirstChild() == n.Parent() {
+		tag = "th"
+	}
+	if entering {
+		align := ""
+		if n.Alignment != ast.AlignNone {
+			align = fmt.Sprintf(` align="%s"`, n.Alignment.String())
+		}
+		fmt.Fprintf(w, "<%s%s>", tag, align)
+	} else {
+		fmt.Fprintf(w, "</%s>\n", tag)
+	}
+	return gast.WalkContinue
+}
+
+type table struct {
+}
+
+// Table is an extension that allow you to use GFM tables .
+var Table = &table{}
+
+func (e *table) Extend(m goldmark.Markdown) {
+	m.Parser().AddOption(parser.WithParagraphTransformers(
+		util.Prioritized(NewTableParagraphTransformer(), 200),
+	))
+	m.Renderer().AddOption(renderer.WithNodeRenderers(
+		util.Prioritized(NewTableHTMLRenderer(), 500),
+	))
+}

extension/tasklist.go

@@ -0,0 +1,118 @@
+package extension
+
+import (
+	"github.com/yuin/goldmark"
+	gast "github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/extension/ast"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/renderer"
+	"github.com/yuin/goldmark/renderer/html"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+)
+
+var taskListRegexp = regexp.MustCompile(`^\[([\sxX])\]\s*`)
+
+type taskCheckBoxParser struct {
+}
+
+var defaultTaskCheckBoxParser = &taskCheckBoxParser{}
+
+// NewTaskCheckBoxParser returns a new  InlineParser that can parse
+// checkboxes in list items.
+// This parser must take precedence over the parser.LinkParser.
+func NewTaskCheckBoxParser() parser.InlineParser {
+	return defaultTaskCheckBoxParser
+}
+
+func (s *taskCheckBoxParser) Trigger() []byte {
+	return []byte{'['}
+}
+
+func (s *taskCheckBoxParser) Parse(parent gast.Node, block text.Reader, pc parser.Context) gast.Node {
+	// Given AST structure must be like
+	// - List
+	//   - ListItem         : parent.Parent
+	//     - TextBlock      : parent
+	//       (current line)
+	if parent.Parent() == nil || parent.Parent().FirstChild() != parent {
+		return nil
+	}
+
+	if _, ok := parent.Parent().(*gast.ListItem); !ok {
+		return nil
+	}
+	line, _ := block.PeekLine()
+	m := taskListRegexp.FindSubmatchIndex(line)
+	if m == nil {
+		return nil
+	}
+	value := line[m[2]:m[3]][0]
+	block.Advance(m[1])
+	checked := value == 'x' || value == 'X'
+	return ast.NewTaskCheckBox(checked)
+}
+
+func (s *taskCheckBoxParser) CloseBlock(parent gast.Node, pc parser.Context) {
+	// nothing to do
+}
+
+// TaskCheckBoxHTMLRenderer is a renderer.NodeRenderer implementation that
+// renders checkboxes in list items.
+type TaskCheckBoxHTMLRenderer struct {
+	html.Config
+}
+
+// NewTaskCheckBoxHTMLRenderer returns a new TaskCheckBoxHTMLRenderer.
+func NewTaskCheckBoxHTMLRenderer(opts ...html.Option) renderer.NodeRenderer {
+	r := &TaskCheckBoxHTMLRenderer{
+		Config: html.NewConfig(),
+	}
+	for _, opt := range opts {
+		opt.SetHTMLOption(&r.Config)
+	}
+	return r
+}
+
+// Render implements renderer.NodeRenderer.Render.
+func (r *TaskCheckBoxHTMLRenderer) Render(writer util.BufWriter, source []byte, n gast.Node, entering bool) (gast.WalkStatus, error) {
+	switch node := n.(type) {
+	case *ast.TaskCheckBox:
+		return r.renderTaskCheckBox(writer, source, node, entering), nil
+	}
+	return gast.WalkContinue, renderer.NotSupported
+}
+
+func (r *TaskCheckBoxHTMLRenderer) renderTaskCheckBox(w util.BufWriter, source []byte, n *ast.TaskCheckBox, entering bool) gast.WalkStatus {
+	if !entering {
+		return gast.WalkContinue
+	}
+
+	if n.IsChecked {
+		w.WriteString(`<input checked="" disabled="" type="checkbox"`)
+	} else {
+		w.WriteString(`<input disabled="" type="checkbox"`)
+	}
+	if r.XHTML {
+		w.WriteString(" />")
+	} else {
+		w.WriteString(">")
+	}
+	return gast.WalkContinue
+}
+
+type taskList struct {
+}
+
+// TaskList is an extension that allow you to use GFM task lists.
+var TaskList = &taskList{}
+
+func (e *taskList) Extend(m goldmark.Markdown) {
+	m.Parser().AddOption(parser.WithInlineParsers(
+		util.Prioritized(NewTaskCheckBoxParser(), 0),
+	))
+	m.Renderer().AddOption(renderer.WithNodeRenderers(
+		util.Prioritized(NewTaskCheckBoxHTMLRenderer(), 500),
+	))
+}

go.mod

@@ -0,0 +1 @@
+module github.com/yuin/goldmark

markdown.go

@@ -0,0 +1,144 @@
+// Package goldmark implements functions to convert markdown text to a desired format.
+package goldmark
+
+import (
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/renderer"
+	"github.com/yuin/goldmark/renderer/html"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"io"
+)
+
+// DefaultParser returns a new Parser that is configured by default values.
+func DefaultParser() parser.Parser {
+	return parser.NewParser(parser.WithBlockParsers(parser.DefaultBlockParsers()...),
+		parser.WithInlineParsers(parser.DefaultInlineParsers()...),
+		parser.WithParagraphTransformers(parser.DefaultParagraphTransformers()...),
+	)
+}
+
+// DefaultRenderer returns a new Renderer that is configured by default values.
+func DefaultRenderer() renderer.Renderer {
+	return renderer.NewRenderer(renderer.WithNodeRenderers(util.Prioritized(html.NewRenderer(), 1000)))
+}
+
+var defaultMarkdown = New()
+
+// Convert interprets a UTF-8 bytes source in Markdown and
+// write rendered contents to a writer w.
+func Convert(source []byte, w io.Writer) error {
+	return defaultMarkdown.Convert(source, w)
+}
+
+// A Markdown interface offers functions to convert Markdown text to
+// a desired format.
+type Markdown interface {
+	// Convert interprets a UTF-8 bytes source in Markdown and write rendered
+	// contents to a writer w.
+	Convert(source []byte, writer io.Writer) error
+
+	// Parser returns a Parser that will be used for conversion.
+	Parser() parser.Parser
+
+	// SetParser sets a Parser to this object.
+	SetParser(parser.Parser)
+
+	// Parser returns a Renderer that will be used for conversion.
+	Renderer() renderer.Renderer
+
+	// SetRenderer sets a Renderer to this object.
+	SetRenderer(renderer.Renderer)
+}
+
+// Option is a functional option type for Markdown objects.
+type Option func(*markdown)
+
+// WithExtensions adds extensions.
+func WithExtensions(ext ...Extender) Option {
+	return func(m *markdown) {
+		m.extensions = append(m.extensions, ext...)
+	}
+}
+
+// WithParser allows you to override the default parser.
+func WithParser(p parser.Parser) Option {
+	return func(m *markdown) {
+		m.parser = p
+	}
+}
+
+// WithParserOptions applies options for the parser.
+func WithParserOptions(opts ...parser.Option) Option {
+	return func(m *markdown) {
+		for _, opt := range opts {
+			m.parser.AddOption(opt)
+		}
+	}
+}
+
+// WithRenderer allows you to override the default renderer.
+func WithRenderer(r renderer.Renderer) Option {
+	return func(m *markdown) {
+		m.renderer = r
+	}
+}
+
+// WithRendererOptions applies options for the renderer.
+func WithRendererOptions(opts ...renderer.Option) Option {
+	return func(m *markdown) {
+		for _, opt := range opts {
+			m.renderer.AddOption(opt)
+		}
+	}
+}
+
+type markdown struct {
+	parser     parser.Parser
+	renderer   renderer.Renderer
+	extensions []Extender
+}
+
+// New returns a new Markdown with given options.
+func New(options ...Option) Markdown {
+	md := &markdown{
+		parser:     DefaultParser(),
+		renderer:   DefaultRenderer(),
+		extensions: []Extender{},
+	}
+	for _, opt := range options {
+		opt(md)
+	}
+	for _, e := range md.extensions {
+		e.Extend(md)
+	}
+	return md
+}
+
+func (m *markdown) Convert(source []byte, writer io.Writer) error {
+	reader := text.NewReader(source)
+	doc, _ := m.parser.Parse(reader)
+	return m.renderer.Render(writer, reader.Source(), doc)
+}
+
+func (m *markdown) Parser() parser.Parser {
+	return m.parser
+}
+
+func (m *markdown) SetParser(v parser.Parser) {
+	m.parser = v
+}
+
+func (m *markdown) Renderer() renderer.Renderer {
+	return m.renderer
+}
+
+func (m *markdown) SetRenderer(v renderer.Renderer) {
+	m.renderer = v
+}
+
+// An Extender interface is used for extending Markdown.
+type Extender interface {
+	// Extend extends the Markdown.
+	Extend(Markdown)
+}

parser/atx_heading.go

@@ -0,0 +1,146 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+)
+
+// A HeadingConfig struct is a data structure that holds configuration of the renderers related to headings.
+type HeadingConfig struct {
+	HeadingID bool
+}
+
+// SetOption implements SetOptioner.
+func (b *HeadingConfig) SetOption(name OptionName, value interface{}) {
+	switch name {
+	case HeadingID:
+		b.HeadingID = true
+	}
+}
+
+// A HeadingOption interface sets options for heading parsers.
+type HeadingOption interface {
+	SetHeadingOption(*HeadingConfig)
+}
+
+// HeadingID is an option name that enables custom and auto IDs for headings.
+var HeadingID OptionName = "HeadingID"
+
+type withHeadingID struct {
+}
+
+func (o *withHeadingID) SetConfig(c *Config) {
+	c.Options[HeadingID] = true
+}
+
+func (o *withHeadingID) SetHeadingOption(p *HeadingConfig) {
+	p.HeadingID = true
+}
+
+// WithHeadingID is a functional option that enables custom heading ids and
+// auto generated heading ids.
+func WithHeadingID() interface {
+	Option
+	HeadingOption
+} {
+	return &withHeadingID{}
+}
+
+var atxHeadingRegexp = regexp.MustCompile(`^[ ]{0,3}(#{1,6})(?:\s+(.*?)\s*([\s]#+\s*)?)?\n?$`)
+
+type atxHeadingParser struct {
+	HeadingConfig
+}
+
+// NewATXHeadingParser return a new BlockParser that can parse ATX headings.
+func NewATXHeadingParser(opts ...HeadingOption) BlockParser {
+	p := &atxHeadingParser{}
+	for _, o := range opts {
+		o.SetHeadingOption(&p.HeadingConfig)
+	}
+	return p
+}
+
+func (b *atxHeadingParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	line, segment := reader.PeekLine()
+	pos := pc.BlockOffset()
+	i := pos
+	for ; i < len(line) && line[i] == '#'; i++ {
+	}
+	level := i - pos
+	if i == pos || level > 6 {
+		return nil, NoChildren
+	}
+	l := util.TrimLeftSpaceLength(line[i:])
+	if l == 0 {
+		return nil, NoChildren
+	}
+	start := i + l
+	stop := len(line) - util.TrimRightSpaceLength(line)
+	if stop <= start { // empty headings like '##[space]'
+		stop = start + 1
+	} else {
+		i = stop - 1
+		for ; line[i] == '#' && i >= start; i-- {
+		}
+		if i != stop-1 && !util.IsSpace(line[i]) {
+			i = stop - 1
+		}
+		i++
+		stop = i
+	}
+
+	node := ast.NewHeading(level)
+	if len(util.TrimRight(line[start:stop], []byte{'#'})) != 0 { // empty heading like '### ###'
+		node.Lines().Append(text.NewSegment(segment.Start+start, segment.Start+stop))
+	}
+	return node, NoChildren
+}
+
+func (b *atxHeadingParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	return Close
+}
+
+func (b *atxHeadingParser) Close(node ast.Node, pc Context) {
+	if !b.HeadingID {
+		return
+	}
+	parseOrGenerateHeadingID(node.(*ast.Heading), pc)
+}
+
+func (b *atxHeadingParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *atxHeadingParser) CanAcceptIndentedLine() bool {
+	return false
+}
+
+var headingIDRegexp = regexp.MustCompile(`^(.*[^\\])({#([^}]+)}\s*)\n?$`)
+var headingIDMap = NewContextKey()
+
+func parseOrGenerateHeadingID(node *ast.Heading, pc Context) {
+	existsv := pc.Get(headingIDMap)
+	var exists map[string]bool
+	if existsv == nil {
+		exists = map[string]bool{}
+		pc.Set(headingIDMap, exists)
+	} else {
+		exists = existsv.(map[string]bool)
+	}
+	lastIndex := node.Lines().Len() - 1
+	lastLine := node.Lines().At(lastIndex)
+	line := lastLine.Value(pc.Source())
+	m := headingIDRegexp.FindSubmatchIndex(line)
+	var headingID []byte
+	if m != nil {
+		headingID = line[m[6]:m[7]]
+		lastLine.Stop -= m[5] - m[4]
+		node.Lines().Set(lastIndex, lastLine)
+	} else {
+		headingID = util.GenerateLinkID(line, exists)
+	}
+	node.ID = headingID
+}

parser/auto_link.go

@@ -0,0 +1,46 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"regexp"
+)
+
+type autoLinkParser struct {
+}
+
+var defaultAutoLinkParser = &autoLinkParser{}
+
+// NewAutoLinkParser returns a new InlineParser that parses autolinks
+// surrounded by '<' and '>' .
+func NewAutoLinkParser() InlineParser {
+	return defaultAutoLinkParser
+}
+
+func (s *autoLinkParser) Trigger() []byte {
+	return []byte{'<'}
+}
+
+var emailAutoLinkRegexp = regexp.MustCompile(`^<([a-zA-Z0-9.!#$%&'*+\/=?^_` + "`" + `{|}~-]+@[a-zA-Z0-9](?:[a-zA-Z0-9-]{0,61}[a-zA-Z0-9])?(?:\.[a-zA-Z0-9](?:[a-zA-Z0-9-]{0,61}[a-zA-Z0-9])?)*)>`)
+
+var autoLinkRegexp = regexp.MustCompile(`(?i)^<[A-Za-z][A-Za-z0-9.+-]{1,31}:[^<>\x00-\x20]*>`)
+
+func (s *autoLinkParser) Parse(parent ast.Node, block text.Reader, pc Context) ast.Node {
+	line, segment := block.PeekLine()
+	match := emailAutoLinkRegexp.FindSubmatchIndex(line)
+	typ := ast.AutoLinkType(ast.AutoLinkEmail)
+	if match == nil {
+		match = autoLinkRegexp.FindSubmatchIndex(line)
+		typ = ast.AutoLinkURL
+	}
+	if match == nil {
+		return nil
+	}
+	value := ast.NewTextSegment(text.NewSegment(segment.Start+1, segment.Start+match[1]-1))
+	block.Advance(match[1])
+	return ast.NewAutoLink(typ, value)
+}
+
+func (s *autoLinkParser) CloseBlock(parent ast.Node, pc Context) {
+	// nothing to do
+}

parser/blockquote.go

@@ -0,0 +1,65 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type blockquoteParser struct {
+}
+
+var defaultBlockquoteParser = &blockquoteParser{}
+
+// NewBlockquoteParser returns a new BlockParser that
+// parses blockquotes.
+func NewBlockquoteParser() BlockParser {
+	return defaultBlockquoteParser
+}
+
+func (b *blockquoteParser) process(reader text.Reader) bool {
+	line, _ := reader.PeekLine()
+	w, pos := util.IndentWidth(line, 0)
+	if w > 3 || pos >= len(line) || line[pos] != '>' {
+		return false
+	}
+	pos++
+	if pos >= len(line) || line[pos] == '\n' {
+		reader.Advance(pos)
+		return true
+	}
+	if line[pos] == ' ' || line[pos] == '\t' {
+		pos++
+	}
+	reader.Advance(pos)
+	if line[pos-1] == '\t' {
+		reader.SetPadding(2)
+	}
+	return true
+}
+
+func (b *blockquoteParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	if b.process(reader) {
+		return ast.NewBlockquote(), HasChildren
+	}
+	return nil, NoChildren
+}
+
+func (b *blockquoteParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	if b.process(reader) {
+		return Continue | HasChildren
+	}
+	return Close
+}
+
+func (b *blockquoteParser) Close(node ast.Node, pc Context) {
+	// nothing to do
+}
+
+func (b *blockquoteParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *blockquoteParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/code_block.go

@@ -0,0 +1,75 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type codeBlockParser struct {
+}
+
+// CodeBlockParser is a BlockParser implementation that parses indented code blocks.
+var defaultCodeBlockParser = &codeBlockParser{}
+
+// NewCodeBlockParser returns a new BlockParser that
+// parses code blocks.
+func NewCodeBlockParser() BlockParser {
+	return defaultCodeBlockParser
+}
+
+func (b *codeBlockParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	line, segment := reader.PeekLine()
+	pos, padding := util.IndentPosition(line, reader.LineOffset(), 4)
+	if pos < 0 {
+		return nil, NoChildren
+	}
+	node := ast.NewCodeBlock()
+	reader.AdvanceAndSetPadding(pos, padding)
+	_, segment = reader.PeekLine()
+	node.Lines().Append(segment)
+	reader.Advance(segment.Len() - 1)
+	return node, NoChildren
+
+}
+
+func (b *codeBlockParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	line, segment := reader.PeekLine()
+	if util.IsBlank(line) {
+		node.Lines().Append(segment.TrimLeftSpaceWidth(4, reader.Source()))
+		return Continue | NoChildren
+	}
+	pos, padding := util.IndentPosition(line, reader.LineOffset(), 4)
+	if pos < 0 {
+		return Close
+	}
+	reader.AdvanceAndSetPadding(pos, padding)
+	_, segment = reader.PeekLine()
+	node.Lines().Append(segment)
+	reader.Advance(segment.Len() - 1)
+	return Continue | NoChildren
+}
+
+func (b *codeBlockParser) Close(node ast.Node, pc Context) {
+	// trim trailing blank lines
+	lines := node.Lines()
+	length := lines.Len() - 1
+	source := pc.Source()
+	for {
+		line := lines.At(length)
+		if util.IsBlank(line.Value(source)) {
+			length--
+		} else {
+			break
+		}
+	}
+	lines.SetSliced(0, length+1)
+}
+
+func (b *codeBlockParser) CanInterruptParagraph() bool {
+	return false
+}
+
+func (b *codeBlockParser) CanAcceptIndentedLine() bool {
+	return true
+}

parser/code_span.go

@@ -0,0 +1,87 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type codeSpanParser struct {
+}
+
+var defaultCodeSpanParser = &codeSpanParser{}
+
+// NewCodeSpanParser return a new InlineParser that parses inline codes
+// surrounded by '`' .
+func NewCodeSpanParser() InlineParser {
+	return defaultCodeSpanParser
+}
+
+func (s *codeSpanParser) Trigger() []byte {
+	return []byte{'`'}
+}
+
+func (s *codeSpanParser) Parse(parent ast.Node, block text.Reader, pc Context) ast.Node {
+	line, startSegment := block.PeekLine()
+	opener := 0
+	for ; opener < len(line) && line[opener] == '`'; opener++ {
+	}
+	block.Advance(opener)
+	l, pos := block.Position()
+	node := ast.NewCodeSpan()
+	for {
+		line, segment := block.PeekLine()
+		if line == nil {
+			block.SetPosition(l, pos)
+			return ast.NewTextSegment(startSegment.WithStop(startSegment.Start + opener))
+		}
+		for i := 0; i < len(line); i++ {
+			c := line[i]
+			if c == '`' {
+				oldi := i
+				for ; i < len(line) && line[i] == '`'; i++ {
+				}
+				closure := i - oldi
+				if closure == opener && (i+1 >= len(line) || line[i+1] != '`') {
+					segment := segment.WithStop(segment.Start + i - closure)
+					if !segment.IsEmpty() {
+						node.AppendChild(node, ast.NewRawTextSegment(segment))
+					}
+					block.Advance(i)
+					goto end
+				}
+			}
+		}
+		if !util.IsBlank(line) {
+			node.AppendChild(node, ast.NewRawTextSegment(segment))
+		}
+		block.AdvanceLine()
+	}
+end:
+	if !node.IsBlank(pc.Source()) {
+		// trim first halfspace and last halfspace
+		segment := node.FirstChild().(*ast.Text).Segment
+		shouldTrimmed := true
+		if !(!segment.IsEmpty() && pc.Source()[segment.Start] == ' ') {
+			shouldTrimmed = false
+		}
+		segment = node.LastChild().(*ast.Text).Segment
+		if !(!segment.IsEmpty() && pc.Source()[segment.Stop-1] == ' ') {
+			shouldTrimmed = false
+		}
+		if shouldTrimmed {
+			t := node.FirstChild().(*ast.Text)
+			segment := t.Segment
+			t.Segment = segment.WithStart(segment.Start + 1)
+			t = node.LastChild().(*ast.Text)
+			segment = node.LastChild().(*ast.Text).Segment
+			t.Segment = segment.WithStop(segment.Stop - 1)
+		}
+
+	}
+	return node
+}
+
+func (s *codeSpanParser) CloseBlock(parent ast.Node, pc Context) {
+	// nothing to do
+}

parser/delimiter.go

@@ -0,0 +1,232 @@
+package parser
+
+import (
+	"fmt"
+	"strings"
+	"unicode"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+// A DelimiterProcessor interface provides a set of functions about
+// Deliiter nodes.
+type DelimiterProcessor interface {
+	// IsDelimiter returns true if given character is a delimiter, otherwise false.
+	IsDelimiter(byte) bool
+
+	// CanOpenCloser returns true if given opener can close given closer, otherwise false.
+	CanOpenCloser(opener, closer *Delimiter) bool
+
+	// OnMatch will be called when new matched delimiter found.
+	// OnMatch should return a new Node correspond to the matched delimiter.
+	OnMatch(consumes int) ast.Node
+}
+
+// A Delimiter struct represents a delimiter like '*' of the Markdown text.
+type Delimiter struct {
+	ast.BaseInline
+
+	Segment text.Segment
+
+	// CanOpen is set true if this delimiter can open a span for a new node.
+	// See https://spec.commonmark.org/0.29/#can-open-emphasis for details.
+	CanOpen bool
+
+	// CanClose is set true if this delimiter can close a span for a new node.
+	// See https://spec.commonmark.org/0.29/#can-open-emphasis for details.
+	CanClose bool
+
+	// Length is a remaining length of this delmiter.
+	Length int
+
+	// OriginalLength is a original length of this delimiter.
+	OriginalLength int
+
+	// Char is a character of this delimiter.
+	Char byte
+
+	// PreviousDelimiter is a previous sibling delimiter node of this delimiter.
+	PreviousDelimiter *Delimiter
+
+	// NextDelimiter is a next sibling delimiter node of this delimiter.
+	NextDelimiter *Delimiter
+
+	// Processor is a DelimiterProcessor associated with this delimiter.
+	Processor DelimiterProcessor
+}
+
+// Inline implements Inline.Inline.
+func (d *Delimiter) Inline() {}
+
+// Dump implements Node.Dump.
+func (d *Delimiter) Dump(source []byte, level int) {
+	fmt.Printf("%sDelimiter: \"%s\"\n", strings.Repeat("    ", level), string(d.Text(source)))
+}
+
+// Text implements Node.Text
+func (d *Delimiter) Text(source []byte) []byte {
+	return d.Segment.Value(source)
+}
+
+// ConsumeCharacters consumes delimiters.
+func (d *Delimiter) ConsumeCharacters(n int) {
+	d.Length -= n
+	d.Segment = d.Segment.WithStop(d.Segment.Start + d.Length)
+}
+
+// CalcComsumption calculates how many characters should be used for opening
+// a new span correspond to given closer.
+func (d *Delimiter) CalcComsumption(closer *Delimiter) int {
+	if (d.CanClose || closer.CanOpen) && (d.OriginalLength+closer.OriginalLength)%3 == 0 && closer.OriginalLength%3 != 0 {
+		return 0
+	}
+	if d.Length >= 2 && closer.Length >= 2 {
+		return 2
+	}
+	return 1
+}
+
+// NewDelimiter returns a new Delimiter node.
+func NewDelimiter(canOpen, canClose bool, length int, char byte, processor DelimiterProcessor) *Delimiter {
+	c := &Delimiter{
+		BaseInline:        ast.BaseInline{},
+		CanOpen:           canOpen,
+		CanClose:          canClose,
+		Length:            length,
+		OriginalLength:    length,
+		Char:              char,
+		PreviousDelimiter: nil,
+		NextDelimiter:     nil,
+		Processor:         processor,
+	}
+	return c
+}
+
+// ScanDelimiter scans a delimiter by given DelimiterProcessor.
+func ScanDelimiter(line []byte, before rune, min int, processor DelimiterProcessor) *Delimiter {
+	i := 0
+	c := line[i]
+	j := i
+	if !processor.IsDelimiter(c) {
+		return nil
+	}
+	for ; j < len(line) && c == line[j]; j++ {
+	}
+	if (j - i) >= min {
+		after := rune(' ')
+		if j != len(line) {
+			after = util.ToRune(line, j)
+		}
+
+		isLeft, isRight, canOpen, canClose := false, false, false, false
+		beforeIsPunctuation := unicode.IsPunct(before)
+		beforeIsWhitespace := unicode.IsSpace(before)
+		afterIsPunctuation := unicode.IsPunct(after)
+		afterIsWhitespace := unicode.IsSpace(after)
+
+		isLeft = !afterIsWhitespace &&
+			(!afterIsPunctuation || beforeIsWhitespace || beforeIsPunctuation)
+		isRight = !beforeIsWhitespace &&
+			(!beforeIsPunctuation || afterIsWhitespace || afterIsPunctuation)
+
+		if line[i] == '_' {
+			canOpen = isLeft && (!isRight || beforeIsPunctuation)
+			canClose = isRight && (!isLeft || afterIsPunctuation)
+		} else {
+			canOpen = isLeft
+			canClose = isRight
+		}
+		return NewDelimiter(canOpen, canClose, j-i, c, processor)
+	}
+	return nil
+}
+
+// ProcessDelimiters processes the delimiter list in the context.
+// Processing will be stop when reaching the bottom.
+//
+// If you implement an inline parser that can have other inline nodes as
+// children, you should call this function when nesting span has closed.
+func ProcessDelimiters(bottom ast.Node, pc Context) {
+	if pc.LastDelimiter() == nil {
+		return
+	}
+	var closer *Delimiter
+	if bottom != nil {
+		for c := pc.LastDelimiter().PreviousSibling(); c != nil; {
+			if d, ok := c.(*Delimiter); ok {
+				closer = d
+			}
+			prev := c.PreviousSibling()
+			if prev == bottom {
+				break
+			}
+			c = prev
+		}
+	} else {
+		closer = pc.FirstDelimiter()
+	}
+	if closer == nil {
+		pc.ClearDelimiters(bottom)
+		return
+	}
+	for closer != nil {
+		if !closer.CanClose {
+			closer = closer.NextDelimiter
+			continue
+		}
+		consume := 0
+		found := false
+		maybeOpener := false
+		var opener *Delimiter
+		for opener = closer.PreviousDelimiter; opener != nil; opener = opener.PreviousDelimiter {
+			if opener.CanOpen && opener.Processor.CanOpenCloser(opener, closer) {
+				maybeOpener = true
+				consume = opener.CalcComsumption(closer)
+				if consume > 0 {
+					found = true
+					break
+				}
+			}
+		}
+		if !found {
+			if !maybeOpener && !closer.CanOpen {
+				pc.RemoveDelimiter(closer)
+			}
+			closer = closer.NextDelimiter
+			continue
+		}
+		opener.ConsumeCharacters(consume)
+		closer.ConsumeCharacters(consume)
+
+		node := opener.Processor.OnMatch(consume)
+
+		parent := opener.Parent()
+		child := opener.NextSibling()
+
+		for child != nil && child != closer {
+			next := child.NextSibling()
+			node.AppendChild(node, child)
+			child = next
+		}
+		parent.InsertAfter(parent, opener, node)
+
+		for c := opener.NextDelimiter; c != nil && c != closer; {
+			next := c.NextDelimiter
+			pc.RemoveDelimiter(c)
+			c = next
+		}
+
+		if opener.Length == 0 {
+			pc.RemoveDelimiter(opener)
+		}
+
+		if closer.Length == 0 {
+			next := closer.NextDelimiter
+			pc.RemoveDelimiter(closer)
+			closer = next
+		}
+	}
+	pc.ClearDelimiters(bottom)
+}

parser/emphasis.go

@@ -0,0 +1,54 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+)
+
+type emphasisDelimiterProcessor struct {
+}
+
+func (p *emphasisDelimiterProcessor) IsDelimiter(b byte) bool {
+	return b == '*' || b == '_'
+}
+
+func (p *emphasisDelimiterProcessor) CanOpenCloser(opener, closer *Delimiter) bool {
+	return opener.Char == closer.Char
+}
+
+func (p *emphasisDelimiterProcessor) OnMatch(consumes int) ast.Node {
+	return ast.NewEmphasis(consumes)
+}
+
+var defaultEmphasisDelimiterProcessor = &emphasisDelimiterProcessor{}
+
+type emphasisParser struct {
+}
+
+var defaultEmphasisParser = &emphasisParser{}
+
+// NewEmphasisParser return a new InlineParser that parses emphasises.
+func NewEmphasisParser() InlineParser {
+	return defaultEmphasisParser
+}
+
+func (s *emphasisParser) Trigger() []byte {
+	return []byte{'*', '_'}
+}
+
+func (s *emphasisParser) Parse(parent ast.Node, block text.Reader, pc Context) ast.Node {
+	before := block.PrecendingCharacter()
+	line, segment := block.PeekLine()
+	node := ScanDelimiter(line, before, 1, defaultEmphasisDelimiterProcessor)
+	if node == nil {
+		return nil
+	}
+	node.Segment = segment.WithStop(segment.Start + node.OriginalLength)
+	block.Advance(node.OriginalLength)
+	pc.PushDelimiter(node)
+	return node
+}
+
+func (s *emphasisParser) CloseBlock(parent ast.Node, pc Context) {
+	// nothing to do
+}

parser/fcode_block.go

@@ -0,0 +1,96 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type fencedCodeBlockParser struct {
+}
+
+var defaultFencedCodeBlockParser = &fencedCodeBlockParser{}
+
+// NewFencedCodeBlockParser returns a new BlockParser that
+// parses fenced code blocks.
+func NewFencedCodeBlockParser() BlockParser {
+	return defaultFencedCodeBlockParser
+}
+
+type fenceData struct {
+	char   byte
+	indent int
+	length int
+}
+
+var fencedCodeBlockInfoKey = NewContextKey()
+
+func (b *fencedCodeBlockParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	line, segment := reader.PeekLine()
+	pos := pc.BlockOffset()
+	if line[pos] != '`' && line[pos] != '~' {
+		return nil, NoChildren
+	}
+	findent := pos
+	fenceChar := line[pos]
+	i := pos
+	for ; i < len(line) && line[i] == fenceChar; i++ {
+	}
+	oFenceLength := i - pos
+	if oFenceLength < 3 {
+		return nil, NoChildren
+	}
+	var info *ast.Text
+	if i < len(line)-1 {
+		rest := line[i:]
+		left := util.TrimLeftSpaceLength(rest)
+		right := util.TrimRightSpaceLength(rest)
+		infoStart, infoStop := segment.Start+i+left, segment.Stop-right
+		value := rest[left : len(rest)-right]
+		if fenceChar == '`' && bytes.IndexByte(value, '`') > -1 {
+			return nil, NoChildren
+		} else if infoStart != infoStop {
+			info = ast.NewTextSegment(text.NewSegment(infoStart, infoStop))
+		}
+	}
+	pc.Set(fencedCodeBlockInfoKey, &fenceData{fenceChar, findent, oFenceLength})
+	node := ast.NewFencedCodeBlock(info)
+	return node, NoChildren
+
+}
+
+func (b *fencedCodeBlockParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	line, segment := reader.PeekLine()
+	fdata := pc.Get(fencedCodeBlockInfoKey).(*fenceData)
+	w, pos := util.IndentWidth(line, 0)
+	if w < 4 {
+		i := pos
+		for ; i < len(line) && line[i] == fdata.char; i++ {
+		}
+		length := i - pos
+		if length >= fdata.length && util.IsBlank(line[i:]) {
+			reader.Advance(segment.Stop - segment.Start - 1 - segment.Padding)
+			return Close
+		}
+	}
+
+	pos, padding := util.DedentPosition(line, fdata.indent)
+	seg := text.NewSegmentPadding(segment.Start+pos, segment.Stop, padding)
+	node.Lines().Append(seg)
+	reader.AdvanceAndSetPadding(segment.Stop-segment.Start-pos-1, padding)
+	return Continue | NoChildren
+}
+
+func (b *fencedCodeBlockParser) Close(node ast.Node, pc Context) {
+	pc.Set(fencedCodeBlockInfoKey, nil)
+}
+
+func (b *fencedCodeBlockParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *fencedCodeBlockParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/html_block.go

@@ -0,0 +1,278 @@
+package parser
+
+import (
+	"bytes"
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+	"strings"
+)
+
+// An HTMLConfig struct is a data structure that holds configuration of the renderers related to raw htmls.
+type HTMLConfig struct {
+	FilterTags map[string]bool
+}
+
+// SetOption implements SetOptioner.
+func (b *HTMLConfig) SetOption(name OptionName, value interface{}) {
+	switch name {
+	case FilterTags:
+		b.FilterTags = value.(map[string]bool)
+	}
+}
+
+// A HTMLOption interface sets options for the raw HTML parsers.
+type HTMLOption interface {
+	SetHTMLOption(*HTMLConfig)
+}
+
+// FilterTags is an otpion name that specify forbidden tag names.
+const FilterTags OptionName = "FilterTags"
+
+type withFilterTags struct {
+	value map[string]bool
+}
+
+func (o *withFilterTags) SetConfig(c *Config) {
+	c.Options[FilterTags] = o.value
+}
+
+func (o *withFilterTags) SetHTMLOption(p *HTMLConfig) {
+	p.FilterTags = o.value
+}
+
+// WithFilterTags is a functional otpion that specify forbidden tag names.
+func WithFilterTags(names ...string) interface {
+	Option
+	HTMLOption
+} {
+	m := map[string]bool{}
+	for _, name := range names {
+		m[name] = true
+	}
+	return &withFilterTags{m}
+}
+
+var allowedBlockTags = map[string]bool{
+	"address":    true,
+	"article":    true,
+	"aside":      true,
+	"base":       true,
+	"basefont":   true,
+	"blockquote": true,
+	"body":       true,
+	"caption":    true,
+	"center":     true,
+	"col":        true,
+	"colgroup":   true,
+	"dd":         true,
+	"details":    true,
+	"dialog":     true,
+	"dir":        true,
+	"div":        true,
+	"dl":         true,
+	"dt":         true,
+	"fieldset":   true,
+	"figcaption": true,
+	"figure":     true,
+	"footer":     true,
+	"form":       true,
+	"frame":      true,
+	"frameset":   true,
+	"h1":         true,
+	"h2":         true,
+	"h3":         true,
+	"h4":         true,
+	"h5":         true,
+	"h6":         true,
+	"head":       true,
+	"header":     true,
+	"hr":         true,
+	"html":       true,
+	"iframe":     true,
+	"legend":     true,
+	"li":         true,
+	"link":       true,
+	"main":       true,
+	"menu":       true,
+	"menuitem":   true,
+	"meta":       true,
+	"nav":        true,
+	"noframes":   true,
+	"ol":         true,
+	"optgroup":   true,
+	"option":     true,
+	"p":          true,
+	"param":      true,
+	"section":    true,
+	"source":     true,
+	"summary":    true,
+	"table":      true,
+	"tbody":      true,
+	"td":         true,
+	"tfoot":      true,
+	"th":         true,
+	"thead":      true,
+	"title":      true,
+	"tr":         true,
+	"track":      true,
+	"ul":         true,
+}
+
+var htmlBlockType1OpenRegexp = regexp.MustCompile(`(?i)^[ ]{0,3}<(script|pre|style)(?:\s.*|>.*|/>.*|)\n?$`)
+var htmlBlockType1CloseRegexp = regexp.MustCompile(`(?i)^[ ]{0,3}(?:[^ ].*|)</(?:script|pre|style)>.*`)
+
+var htmlBlockType2OpenRegexp = regexp.MustCompile(`^[ ]{0,3}<!\-\-`)
+var htmlBlockType2Close = []byte{'-', '-', '>'}
+
+var htmlBlockType3OpenRegexp = regexp.MustCompile(`^[ ]{0,3}<\?`)
+var htmlBlockType3Close = []byte{'?', '>'}
+
+var htmlBlockType4OpenRegexp = regexp.MustCompile(`^[ ]{0,3}<![A-Z]+.*\n?$`)
+var htmlBlockType4Close = []byte{'>'}
+
+var htmlBlockType5OpenRegexp = regexp.MustCompile(`^[ ]{0,3}<\!\[CDATA\[`)
+var htmlBlockType5Close = []byte{']', ']', '>'}
+
+var htmlBlockType6Regexp = regexp.MustCompile(`^[ ]{0,3}</?([a-zA-Z0-9]+)(?:\s.*|>.*|/>.*|)\n?$`)
+
+var htmlBlockType7Regexp = regexp.MustCompile(`^[ ]{0,3}<(/)?([a-zA-Z0-9]+)(` + attributePattern + `*)(:?>|/>)\s*\n?$`)
+
+type htmlBlockParser struct {
+	HTMLConfig
+}
+
+// NewHTMLBlockParser return a new BlockParser that can parse html
+// blocks.
+func NewHTMLBlockParser(opts ...HTMLOption) BlockParser {
+	p := &htmlBlockParser{}
+	for _, o := range opts {
+		o.SetHTMLOption(&p.HTMLConfig)
+	}
+	return p
+}
+
+func (b *htmlBlockParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	var node *ast.HTMLBlock
+	line, segment := reader.PeekLine()
+	last := pc.LastOpenedBlock().Node
+	if pos := pc.BlockOffset(); line[pos] != '<' {
+		return nil, NoChildren
+	}
+
+	tagName := ""
+	if m := htmlBlockType1OpenRegexp.FindSubmatchIndex(line); m != nil {
+		tagName = string(line[m[2]:m[3]])
+		node = ast.NewHTMLBlock(ast.HTMLBlockType1)
+	} else if htmlBlockType2OpenRegexp.Match(line) {
+		node = ast.NewHTMLBlock(ast.HTMLBlockType2)
+	} else if htmlBlockType3OpenRegexp.Match(line) {
+		node = ast.NewHTMLBlock(ast.HTMLBlockType3)
+	} else if htmlBlockType4OpenRegexp.Match(line) {
+		node = ast.NewHTMLBlock(ast.HTMLBlockType4)
+	} else if htmlBlockType5OpenRegexp.Match(line) {
+		node = ast.NewHTMLBlock(ast.HTMLBlockType5)
+	} else if match := htmlBlockType7Regexp.FindSubmatchIndex(line); match != nil {
+		isCloseTag := match[2] > -1 && bytes.Equal(line[match[2]:match[3]], []byte("/"))
+		hasAttr := match[6] != match[7]
+		tagName = strings.ToLower(string(line[match[4]:match[5]]))
+		_, ok := allowedBlockTags[strings.ToLower(string(tagName))]
+		if ok {
+			node = ast.NewHTMLBlock(ast.HTMLBlockType6)
+		} else if tagName != "script" && tagName != "style" && tagName != "pre" && !ast.IsParagraph(last) && !(isCloseTag && hasAttr) { // type 7 can not interrupt paragraph
+			node = ast.NewHTMLBlock(ast.HTMLBlockType7)
+		}
+	}
+	if node == nil {
+		if match := htmlBlockType6Regexp.FindSubmatchIndex(line); match != nil {
+			tagName = string(line[match[2]:match[3]])
+			_, ok := allowedBlockTags[strings.ToLower(tagName)]
+			if ok {
+				node = ast.NewHTMLBlock(ast.HTMLBlockType6)
+			}
+		}
+	}
+	if node != nil {
+		if b.FilterTags != nil {
+			if _, ok := b.FilterTags[tagName]; ok {
+				return nil, NoChildren
+			}
+		}
+		reader.Advance(segment.Len() - 1)
+		node.Lines().Append(segment)
+		return node, NoChildren
+	}
+	return nil, NoChildren
+}
+
+func (b *htmlBlockParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	htmlBlock := node.(*ast.HTMLBlock)
+	lines := htmlBlock.Lines()
+	line, segment := reader.PeekLine()
+	var closurePattern []byte
+
+	switch htmlBlock.HTMLBlockType {
+	case ast.HTMLBlockType1:
+		if lines.Len() == 1 {
+			firstLine := lines.At(0)
+			if htmlBlockType1CloseRegexp.Match(firstLine.Value(reader.Source())) {
+				return Close
+			}
+		}
+		if htmlBlockType1CloseRegexp.Match(line) {
+			htmlBlock.ClosureLine = segment
+			reader.Advance(segment.Len() - 1)
+			return Close
+		}
+	case ast.HTMLBlockType2:
+		closurePattern = htmlBlockType2Close
+		fallthrough
+	case ast.HTMLBlockType3:
+		if closurePattern == nil {
+			closurePattern = htmlBlockType3Close
+		}
+		fallthrough
+	case ast.HTMLBlockType4:
+		if closurePattern == nil {
+			closurePattern = htmlBlockType4Close
+		}
+		fallthrough
+	case ast.HTMLBlockType5:
+		if closurePattern == nil {
+			closurePattern = htmlBlockType5Close
+		}
+
+		if lines.Len() == 1 {
+			firstLine := lines.At(0)
+			if bytes.Contains(firstLine.Value(reader.Source()), closurePattern) {
+				return Close
+			}
+		}
+		if bytes.Contains(line, closurePattern) {
+			htmlBlock.ClosureLine = segment
+			reader.Advance(segment.Len() - 1)
+			return Close
+		}
+
+	case ast.HTMLBlockType6, ast.HTMLBlockType7:
+		if util.IsBlank(line) {
+			return Close
+		}
+	}
+	node.Lines().Append(segment)
+	reader.Advance(segment.Len() - 1)
+	return Continue | NoChildren
+}
+
+func (b *htmlBlockParser) Close(node ast.Node, pc Context) {
+	// nothing to do
+}
+
+func (b *htmlBlockParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *htmlBlockParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/link.go

@@ -0,0 +1,366 @@
+package parser
+
+import (
+	"fmt"
+	"regexp"
+	"strings"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+var linkLabelStateKey = NewContextKey()
+
+type linkLabelState struct {
+	ast.BaseInline
+
+	Segment text.Segment
+
+	IsImage bool
+
+	Prev *linkLabelState
+
+	Next *linkLabelState
+
+	First *linkLabelState
+
+	Last *linkLabelState
+}
+
+func newLinkLabelState(segment text.Segment, isImage bool) *linkLabelState {
+	return &linkLabelState{
+		Segment: segment,
+		IsImage: isImage,
+	}
+}
+
+func (s *linkLabelState) Text(source []byte) []byte {
+	return s.Segment.Value(source)
+}
+
+func (s *linkLabelState) Dump(source []byte, level int) {
+	fmt.Printf("%slinkLabelState: \"%s\"\n", strings.Repeat("    ", level), s.Text(source))
+}
+
+func pushLinkLabelState(pc Context, v *linkLabelState) {
+	tlist := pc.Get(linkLabelStateKey)
+	var list *linkLabelState
+	if tlist == nil {
+		list = v
+		v.First = v
+		v.Last = v
+		pc.Set(linkLabelStateKey, list)
+	} else {
+		list = tlist.(*linkLabelState)
+		l := list.Last
+		list.Last = v
+		l.Next = v
+		v.Prev = l
+	}
+}
+
+func removeLinkLabelState(pc Context, d *linkLabelState) {
+	tlist := pc.Get(linkLabelStateKey)
+	var list *linkLabelState
+	if tlist == nil {
+		return
+	}
+	list = tlist.(*linkLabelState)
+
+	if d.Prev == nil {
+		list = d.Next
+		if list != nil {
+			list.First = d
+			list.Last = d.Last
+			list.Prev = nil
+			pc.Set(linkLabelStateKey, list)
+		} else {
+			pc.Set(linkLabelStateKey, nil)
+		}
+	} else {
+		d.Prev.Next = d.Next
+		if d.Next != nil {
+			d.Next.Prev = d.Prev
+		}
+	}
+	if list != nil && d.Next == nil {
+		list.Last = d.Prev
+	}
+	d.Next = nil
+	d.Prev = nil
+	d.First = nil
+	d.Last = nil
+}
+
+type linkParser struct {
+}
+
+var defaultLinkParser = &linkParser{}
+
+// NewLinkParser return a new InlineParser that parses links.
+func NewLinkParser() InlineParser {
+	return defaultLinkParser
+}
+
+func (s *linkParser) Trigger() []byte {
+	return []byte{'!', '[', ']'}
+}
+
+var linkDestinationRegexp = regexp.MustCompile(`\s*([^\s].+)`)
+var linkTitleRegexp = regexp.MustCompile(`\s+(\)|["'\(].+)`)
+var linkBottom = NewContextKey()
+
+func (s *linkParser) Parse(parent ast.Node, block text.Reader, pc Context) ast.Node {
+	line, segment := block.PeekLine()
+	if line[0] == '!' && len(line) > 1 && line[1] == '[' {
+		block.Advance(1)
+		pc.Set(linkBottom, pc.LastDelimiter())
+		return processLinkLabelOpen(block, segment.Start+1, true, pc)
+	}
+	if line[0] == '[' {
+		pc.Set(linkBottom, pc.LastDelimiter())
+		return processLinkLabelOpen(block, segment.Start, false, pc)
+	}
+
+	// line[0] == ']'
+	tlist := pc.Get(linkLabelStateKey)
+	if tlist == nil {
+		return nil
+	}
+	last := tlist.(*linkLabelState).Last
+	if last == nil {
+		return nil
+	}
+	block.Advance(1)
+	removeLinkLabelState(pc, last)
+	if s.containsLink(last) { // a link in a link text is not allowed
+		ast.MergeOrReplaceTextSegment(last.Parent(), last, last.Segment)
+		return nil
+	}
+	labelValue := block.Value(text.NewSegment(last.Segment.Start+1, segment.Start))
+	if util.IsBlank(labelValue) && !last.IsImage {
+		ast.MergeOrReplaceTextSegment(last.Parent(), last, last.Segment)
+		return nil
+	}
+
+	c := block.Peek()
+	l, pos := block.Position()
+	var link *ast.Link
+	var hasValue bool
+	if c == '(' { // normal link
+		link = s.parseLink(parent, last, block, pc)
+	} else if c == '[' { // reference link
+		link, hasValue = s.parseReferenceLink(parent, last, block, pc)
+		if link == nil && hasValue {
+			ast.MergeOrReplaceTextSegment(last.Parent(), last, last.Segment)
+			return nil
+		}
+	}
+
+	if link == nil {
+		// maybe shortcut reference link
+		block.SetPosition(l, pos)
+		ssegment := text.NewSegment(last.Segment.Stop, segment.Start)
+		maybeReference := block.Value(ssegment)
+		ref, ok := pc.Reference(util.ToLinkReference(maybeReference))
+		if !ok {
+			ast.MergeOrReplaceTextSegment(last.Parent(), last, last.Segment)
+			return nil
+		}
+		link = ast.NewLink()
+		s.processLinkLabel(parent, link, last, pc)
+		link.Title = ref.Title()
+		link.Destination = ref.Destination()
+	}
+	if last.IsImage {
+		last.Parent().RemoveChild(last.Parent(), last)
+		return ast.NewImage(link)
+	}
+	last.Parent().RemoveChild(last.Parent(), last)
+	return link
+}
+
+func (s *linkParser) containsLink(last *linkLabelState) bool {
+	if last.IsImage {
+		return false
+	}
+	var c ast.Node
+	for c = last; c != nil; c = c.NextSibling() {
+		if _, ok := c.(*ast.Link); ok {
+			return true
+		}
+	}
+	return false
+}
+
+func processLinkLabelOpen(block text.Reader, pos int, isImage bool, pc Context) *linkLabelState {
+	start := pos
+	if isImage {
+		start--
+	}
+	state := newLinkLabelState(text.NewSegment(start, pos+1), isImage)
+	pushLinkLabelState(pc, state)
+	block.Advance(1)
+	return state
+}
+
+func (s *linkParser) processLinkLabel(parent ast.Node, link *ast.Link, last *linkLabelState, pc Context) {
+	var bottom ast.Node
+	if v := pc.Get(linkBottom); v != nil {
+		bottom = v.(ast.Node)
+	}
+	pc.Set(linkBottom, nil)
+	ProcessDelimiters(bottom, pc)
+	for c := last.NextSibling(); c != nil; {
+		next := c.NextSibling()
+		parent.RemoveChild(parent, c)
+		link.AppendChild(link, c)
+		c = next
+	}
+}
+
+func (s *linkParser) parseReferenceLink(parent ast.Node, last *linkLabelState, block text.Reader, pc Context) (*ast.Link, bool) {
+	_, orgpos := block.Position()
+	block.Advance(1) // skip '['
+	line, segment := block.PeekLine()
+	endIndex := util.FindClosure(line, '[', ']', false, true)
+	if endIndex < 0 {
+		return nil, false
+	}
+
+	block.Advance(endIndex + 1)
+	ssegment := segment.WithStop(segment.Start + endIndex)
+	maybeReference := block.Value(ssegment)
+	if util.IsBlank(maybeReference) { // collapsed reference link
+		ssegment = text.NewSegment(last.Segment.Stop, orgpos.Start-1)
+		maybeReference = block.Value(ssegment)
+	}
+
+	ref, ok := pc.Reference(util.ToLinkReference(maybeReference))
+	if !ok {
+		return nil, true
+	}
+
+	link := ast.NewLink()
+	s.processLinkLabel(parent, link, last, pc)
+	link.Title = ref.Title()
+	link.Destination = ref.Destination()
+	return link, true
+}
+
+func (s *linkParser) parseLink(parent ast.Node, last *linkLabelState, block text.Reader, pc Context) *ast.Link {
+	block.Advance(1) // skip '('
+	block.SkipSpaces()
+	var title []byte
+	var destination []byte
+	var ok bool
+	if block.Peek() == ')' { // empty link like '[link]()'
+		block.Advance(1)
+	} else {
+		destination, ok = parseLinkDestination(block)
+		if !ok {
+			return nil
+		}
+		block.SkipSpaces()
+		if block.Peek() == ')' {
+			block.Advance(1)
+		} else {
+			title, ok = parseLinkTitle(block)
+			if !ok {
+				return nil
+			}
+			block.SkipSpaces()
+			if block.Peek() == ')' {
+				block.Advance(1)
+			} else {
+				return nil
+			}
+		}
+	}
+
+	link := ast.NewLink()
+	s.processLinkLabel(parent, link, last, pc)
+	link.Destination = destination
+	link.Title = title
+	return link
+}
+
+func parseLinkDestination(block text.Reader) ([]byte, bool) {
+	block.SkipSpaces()
+	line, _ := block.PeekLine()
+	buf := []byte{}
+	if block.Peek() == '<' {
+		i := 1
+		for i < len(line) {
+			c := line[i]
+			if c == '\\' && i < len(line)-1 && util.IsPunct(line[i+1]) {
+				buf = append(buf, '\\', line[i+1])
+				i += 2
+				continue
+			} else if c == '>' {
+				block.Advance(i + 1)
+				return line[1:i], true
+			}
+			buf = append(buf, c)
+			i++
+		}
+		return nil, false
+	}
+	opened := 0
+	i := 0
+	for i < len(line) {
+		c := line[i]
+		if c == '\\' && i < len(line)-1 && util.IsPunct(line[i+1]) {
+			buf = append(buf, '\\', line[i+1])
+			i += 2
+			continue
+		} else if c == '(' {
+			opened++
+		} else if c == ')' {
+			opened--
+			if opened < 0 {
+				break
+			}
+		} else if util.IsSpace(c) {
+			break
+		}
+		buf = append(buf, c)
+		i++
+	}
+	block.Advance(i)
+	return line[:i], len(line[:i]) != 0
+}
+
+func parseLinkTitle(block text.Reader) ([]byte, bool) {
+	block.SkipSpaces()
+	opener := block.Peek()
+	if opener != '"' && opener != '\'' && opener != '(' {
+		return nil, false
+	}
+	closer := opener
+	if opener == '(' {
+		closer = ')'
+	}
+	line, _ := block.PeekLine()
+	pos := util.FindClosure(line[1:], opener, closer, false, true)
+	if pos < 0 {
+		return nil, false
+	}
+	pos += 2 // opener + closer
+	block.Advance(pos)
+	return line[1 : pos-1], true
+}
+
+func (s *linkParser) CloseBlock(parent ast.Node, pc Context) {
+	tlist := pc.Get(linkLabelStateKey)
+	if tlist == nil {
+		return
+	}
+	for s := tlist.(*linkLabelState); s != nil; {
+		next := s.Next
+		removeLinkLabelState(pc, s)
+		s.Parent().ReplaceChild(s.Parent(), s, ast.NewTextSegment(s.Segment))
+		s = next
+	}
+}

parser/link_ref.go

@@ -0,0 +1,163 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type linkReferenceParagraphTransformer struct {
+}
+
+// LinkReferenceParagraphTransformer is a ParagraphTransformer implementation
+// that parses and extracts link reference from paragraphs.
+var LinkReferenceParagraphTransformer = &linkReferenceParagraphTransformer{}
+
+func (p *linkReferenceParagraphTransformer) Transform(node *ast.Paragraph, pc Context) {
+	lines := node.Lines()
+	block := text.NewBlockReader(pc.Source(), lines)
+	removes := [][2]int{}
+	for {
+		start, end := parseLinkReferenceDefinition(block, pc)
+		if start > -1 {
+			if start == end {
+				end++
+			}
+			removes = append(removes, [2]int{start, end})
+			continue
+		}
+		break
+	}
+
+	offset := 0
+	for _, remove := range removes {
+		if lines.Len() == 0 {
+			break
+		}
+		s := lines.Sliced(remove[1]-offset, lines.Len())
+		lines.SetSliced(0, remove[0]-offset)
+		lines.AppendAll(s)
+		offset = remove[1]
+	}
+
+	if lines.Len() == 0 {
+		t := ast.NewTextBlock()
+		t.SetBlankPreviousLines(node.HasBlankPreviousLines())
+		node.Parent().ReplaceChild(node.Parent(), node, t)
+		return
+	}
+
+	node.SetLines(lines)
+}
+
+func parseLinkReferenceDefinition(block text.Reader, pc Context) (int, int) {
+	block.SkipSpaces()
+	line, segment := block.PeekLine()
+	if line == nil {
+		return -1, -1
+	}
+	startLine, _ := block.Position()
+	width, pos := util.IndentWidth(line, 0)
+	if width > 3 {
+		return -1, -1
+	}
+	if width != 0 {
+		pos++
+	}
+	if line[pos] != '[' {
+		return -1, -1
+	}
+	open := segment.Start + pos + 1
+	closes := -1
+	block.Advance(pos + 1)
+	for {
+		line, segment = block.PeekLine()
+		if line == nil {
+			return -1, -1
+		}
+		closure := util.FindClosure(line, '[', ']', false, false)
+		if closure > -1 {
+			closes = segment.Start + closure
+			next := closure + 1
+			if next >= len(line) || line[next] != ':' {
+				return -1, -1
+			}
+			block.Advance(next + 1)
+			break
+		}
+		block.AdvanceLine()
+	}
+	if closes < 0 {
+		return -1, -1
+	}
+	label := block.Value(text.NewSegment(open, closes))
+	if util.IsBlank(label) {
+		return -1, -1
+	}
+	block.SkipSpaces()
+	destination, ok := parseLinkDestination(block)
+	if !ok {
+		return -1, -1
+	}
+	line, segment = block.PeekLine()
+	isNewLine := line == nil || util.IsBlank(line)
+
+	endLine, _ := block.Position()
+	_, spaces, _ := block.SkipSpaces()
+	opener := block.Peek()
+	if opener != '"' && opener != '\'' && opener != '(' {
+		if !isNewLine {
+			return -1, -1
+		}
+		ref := NewReference(label, destination, nil)
+		pc.AddReference(ref)
+		return startLine, endLine + 1
+	}
+	if spaces == 0 {
+		return -1, -1
+	}
+	block.Advance(1)
+	open = -1
+	closes = -1
+	closer := opener
+	if opener == '(' {
+		closer = ')'
+	}
+	for {
+		line, segment = block.PeekLine()
+		if line == nil {
+			return -1, -1
+		}
+		if open < 0 {
+			open = segment.Start
+		}
+		closure := util.FindClosure(line, opener, closer, false, true)
+		if closure > -1 {
+			closes = segment.Start + closure
+			block.Advance(closure + 1)
+			break
+		}
+		block.AdvanceLine()
+	}
+	if closes < 0 {
+		return -1, -1
+	}
+
+	line, segment = block.PeekLine()
+	if line != nil && !util.IsBlank(line) {
+		if !isNewLine {
+			return -1, -1
+		}
+		title := block.Value(text.NewSegment(open, closes))
+		ref := NewReference(label, destination, title)
+		pc.AddReference(ref)
+		return startLine, endLine
+	}
+
+	title := block.Value(text.NewSegment(open, closes))
+
+	endLine, _ = block.Position()
+	ref := NewReference(label, destination, title)
+	pc.AddReference(ref)
+	return startLine, endLine + 1
+}

parser/list.go

@@ -0,0 +1,241 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"strconv"
+)
+
+type listItemType int
+
+const (
+	notList listItemType = iota
+	bulletList
+	orderedList
+)
+
+// Same as
+// `^(([ ]*)([\-\*\+]))(\s+.*)?\n?$`.FindSubmatchIndex or
+// `^(([ ]*)(\d{1,9}[\.\)]))(\s+.*)?\n?$`.FindSubmatchIndex
+func parseListItem(line []byte) ([6]int, listItemType) {
+	i := 0
+	l := len(line)
+	ret := [6]int{}
+	for ; i < l && line[i] == ' '; i++ {
+		c := line[i]
+		if c == '\t' {
+			return ret, notList
+		}
+	}
+	if i > 3 {
+		return ret, notList
+	}
+	ret[0] = 0
+	ret[1] = i
+	ret[2] = i
+	var typ listItemType
+	if i < l && line[i] == '-' || line[i] == '*' || line[i] == '+' {
+		i++
+		ret[3] = i
+		typ = bulletList
+	} else if i < l {
+		for ; i < l && util.IsNumeric(line[i]); i++ {
+		}
+		ret[3] = i
+		if ret[3] == ret[2] || ret[3]-ret[2] > 9 {
+			return ret, notList
+		}
+		if i < l && line[i] == '.' || line[i] == ')' {
+			i++
+			ret[3] = i
+		} else {
+			return ret, notList
+		}
+		typ = orderedList
+	} else {
+		return ret, notList
+	}
+	if line[i] != '\n' {
+		w, _ := util.IndentWidth(line[i:], 0)
+		if w == 0 {
+			return ret, notList
+		}
+	}
+	ret[4] = i
+	ret[5] = len(line)
+	if line[ret[5]-1] == '\n' && line[i] != '\n' {
+		ret[5]--
+	}
+	return ret, typ
+}
+
+func matchesListItem(source []byte, strict bool) ([6]int, listItemType) {
+	m, typ := parseListItem(source)
+	if typ != notList && (!strict || strict && m[1] < 4) {
+		return m, typ
+	}
+	return m, notList
+}
+
+func calcListOffset(source []byte, match [6]int) int {
+	offset := 0
+	if util.IsBlank(source[match[4]:]) { // list item starts with a blank line
+		offset = 1
+	} else {
+		offset, _ = util.IndentWidth(source[match[4]:], match[2])
+		if offset > 4 { // offseted codeblock
+			offset = 1
+		}
+	}
+	return offset
+}
+
+func lastOffset(node ast.Node) int {
+	lastChild := node.LastChild()
+	if lastChild != nil {
+		return lastChild.(*ast.ListItem).Offset
+	}
+	return 0
+}
+
+type listParser struct {
+}
+
+var defaultListParser = &listParser{}
+
+// NewListParser returns a new BlockParser that
+// parses lists.
+// This parser must take predecence over the ListItemParser.
+func NewListParser() BlockParser {
+	return defaultListParser
+}
+
+func (b *listParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	last := pc.LastOpenedBlock().Node
+	if _, lok := last.(*ast.List); lok || pc.Get(skipListParser) != nil {
+		pc.Set(skipListParser, nil)
+		return nil, NoChildren
+	}
+	line, _ := reader.PeekLine()
+	match, typ := matchesListItem(line, true)
+	if typ == notList {
+		return nil, NoChildren
+	}
+	start := -1
+	if typ == orderedList {
+		number := line[match[2] : match[3]-1]
+		start, _ = strconv.Atoi(string(number))
+	}
+
+	if ast.IsParagraph(last) && last.Parent() == parent {
+		// we allow only lists starting with 1 to interrupt paragraphs.
+		if typ == orderedList && start != 1 {
+			return nil, NoChildren
+		}
+		//an empty list item cannot interrupt a paragraph:
+		if match[5]-match[4] == 1 {
+			return nil, NoChildren
+		}
+	}
+
+	marker := line[match[3]-1]
+	node := ast.NewList(marker)
+	if start > -1 {
+		node.Start = start
+	}
+	return node, HasChildren
+}
+
+func (b *listParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	list := node.(*ast.List)
+	line, _ := reader.PeekLine()
+	if util.IsBlank(line) {
+		// A list item can begin with at most one blank line
+		if node.ChildCount() == 1 && node.LastChild().ChildCount() == 0 {
+			return Close
+		}
+		return Continue | HasChildren
+	}
+	// Themantic Breaks take predecence over lists
+	if isThemanticBreak(line) {
+		isHeading := false
+		last := pc.LastOpenedBlock().Node
+		if ast.IsParagraph(last) {
+			c, ok := matchesSetextHeadingBar(line)
+			if ok && c == '-' {
+				isHeading = true
+			}
+		}
+		if !isHeading {
+			return Close
+		}
+	}
+
+	// "offset" means a width that bar indicates.
+	//    -  aaaaaaaa
+	// |----|
+	//
+	// If the indent is less than the last offset like
+	// - a
+	//  - b          <--- current line
+	// it maybe a new child of the list.
+	offset := lastOffset(node)
+	indent, _ := util.IndentWidth(line, 0)
+
+	if indent < offset {
+		if indent < 4 {
+			match, typ := matchesListItem(line, false) // may have a leading spaces more than 3
+			if typ != notList && match[1]-offset < 4 {
+				marker := line[match[3]-1]
+				if !list.CanContinue(marker, typ == orderedList) {
+					return Close
+				}
+				return Continue | HasChildren
+			}
+		}
+		return Close
+	}
+	return Continue | HasChildren
+}
+
+func (b *listParser) Close(node ast.Node, pc Context) {
+	list := node.(*ast.List)
+
+	for c := node.FirstChild(); c != nil && list.IsTight; c = c.NextSibling() {
+		if c.FirstChild() != nil && c.FirstChild() != c.LastChild() {
+			for c1 := c.FirstChild().NextSibling(); c1 != nil; c1 = c1.NextSibling() {
+				if bl, ok := c1.(ast.Node); ok && bl.HasBlankPreviousLines() {
+					list.IsTight = false
+					break
+				}
+			}
+		}
+		if c != node.FirstChild() {
+			if bl, ok := c.(ast.Node); ok && bl.HasBlankPreviousLines() {
+				list.IsTight = false
+			}
+		}
+	}
+
+	if list.IsTight {
+		for child := node.FirstChild(); child != nil; child = child.NextSibling() {
+			for gc := child.FirstChild(); gc != nil; gc = gc.NextSibling() {
+				paragraph, ok := gc.(*ast.Paragraph)
+				if ok {
+					textBlock := ast.NewTextBlock()
+					textBlock.SetLines(paragraph.Lines())
+					child.ReplaceChild(child, paragraph, textBlock)
+				}
+			}
+		}
+	}
+}
+
+func (b *listParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *listParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/list_item.go

@@ -0,0 +1,81 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type listItemParser struct {
+}
+
+var defaultListItemParser = &listItemParser{}
+
+// NewListItemParser returns a new BlockParser that
+// parses list items.
+func NewListItemParser() BlockParser {
+	return defaultListItemParser
+}
+
+var skipListParser = NewContextKey()
+var skipListParserValue interface{} = true
+
+func (b *listItemParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	list, lok := parent.(*ast.List)
+	if !lok { // list item must be a child of a list
+		return nil, NoChildren
+	}
+	offset := lastOffset(list)
+	line, _ := reader.PeekLine()
+	match, typ := matchesListItem(line, false)
+	if typ == notList {
+		return nil, NoChildren
+	}
+	if match[1]-offset > 3 {
+		return nil, NoChildren
+	}
+	itemOffset := calcListOffset(line, match)
+	node := ast.NewListItem(match[3] + itemOffset)
+	if match[5]-match[4] == 1 {
+		return node, NoChildren
+	}
+
+	pos, padding := util.IndentPosition(line[match[4]:], match[4], itemOffset)
+	child := match[3] + pos
+	reader.AdvanceAndSetPadding(child, padding)
+	return node, HasChildren
+}
+
+func (b *listItemParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	line, _ := reader.PeekLine()
+	if util.IsBlank(line) {
+		return Continue | HasChildren
+	}
+
+	indent, _ := util.IndentWidth(line, reader.LineOffset())
+	offset := lastOffset(node.Parent())
+	if indent < offset && indent < 4 {
+		_, typ := matchesListItem(line, true)
+		// new list item found
+		if typ != notList {
+			pc.Set(skipListParser, skipListParserValue)
+		}
+		return Close
+	}
+	pos, padding := util.IndentPosition(line, reader.LineOffset(), offset)
+	reader.AdvanceAndSetPadding(pos, padding)
+
+	return Continue | HasChildren
+}
+
+func (b *listItemParser) Close(node ast.Node, pc Context) {
+	// nothing to do
+}
+
+func (b *listItemParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *listItemParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/paragraph.go

@@ -0,0 +1,62 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+)
+
+type paragraphParser struct {
+}
+
+var defaultParagraphParser = &paragraphParser{}
+
+// NewParagraphParser returns a new BlockParser that
+// parses paragraphs.
+func NewParagraphParser() BlockParser {
+	return defaultParagraphParser
+}
+
+func (b *paragraphParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	_, segment := reader.PeekLine()
+	segment = segment.TrimLeftSpace(reader.Source())
+	if segment.IsEmpty() {
+		return nil, NoChildren
+	}
+	node := ast.NewParagraph()
+	node.Lines().Append(segment)
+	reader.Advance(segment.Len() - 1)
+	return node, NoChildren
+}
+
+func (b *paragraphParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	_, segment := reader.PeekLine()
+	segment = segment.TrimLeftSpace(reader.Source())
+	if segment.IsEmpty() {
+		return Close
+	}
+	node.Lines().Append(segment)
+	reader.Advance(segment.Len() - 1)
+	return Continue | NoChildren
+}
+
+func (b *paragraphParser) Close(node ast.Node, pc Context) {
+	lines := node.Lines()
+	if lines.Len() != 0 {
+		// trim trailing spaces
+		length := lines.Len()
+		lastLine := node.Lines().At(length - 1)
+		node.Lines().Set(length-1, lastLine.TrimRightSpace(pc.Source()))
+	}
+	if lines.Len() == 0 {
+		node.Parent().RemoveChild(node.Parent(), node)
+		return
+	}
+}
+
+func (b *paragraphParser) CanInterruptParagraph() bool {
+	return false
+}
+
+func (b *paragraphParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/parser.go

@@ -0,0 +1,987 @@
+// Package parser contains stuff that are related to parsing a Markdown text.
+package parser
+
+import (
+	"fmt"
+	"strings"
+	"sync"
+	"sync/atomic"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+// A Reference interface represents a link reference in Markdown text.
+type Reference interface {
+	// String implements Stringer.
+	String() string
+
+	// Label returns a label of the reference.
+	Label() []byte
+
+	// Destination returns a destination(URL) of the reference.
+	Destination() []byte
+
+	// Title returns a title of the reference.
+	Title() []byte
+}
+
+type reference struct {
+	label       []byte
+	destination []byte
+	title       []byte
+}
+
+// NewReference returns a new Reference.
+func NewReference(label, destination, title []byte) Reference {
+	return &reference{label, destination, title}
+}
+
+func (r *reference) Label() []byte {
+	return r.label
+}
+
+func (r *reference) Destination() []byte {
+	return r.destination
+}
+
+func (r *reference) Title() []byte {
+	return r.title
+}
+
+func (r *reference) String() string {
+	return fmt.Sprintf("Reference{Label:%s, Destination:%s, Title:%s}", r.label, r.destination, r.title)
+}
+
+// ContextKey is a key that is used to set arbitary values to the context.
+type ContextKey int32
+
+// New returns a new ContextKey value.
+func (c *ContextKey) New() ContextKey {
+	return ContextKey(atomic.AddInt32((*int32)(c), 1))
+}
+
+var contextKey ContextKey
+
+// NewContextKey return a new ContextKey value.
+func NewContextKey() ContextKey {
+	return contextKey.New()
+}
+
+// A Context interface holds a information that are necessary to parse
+// Markdown text.
+type Context interface {
+	// String implements Stringer.
+	String() string
+
+	// Source returns a source of Markdown text.
+	Source() []byte
+
+	// Get returns a value associated with given key.
+	Get(ContextKey) interface{}
+
+	// Set sets given value to the context.
+	Set(ContextKey, interface{})
+
+	// AddReference adds given reference to this context.
+	AddReference(Reference)
+
+	// Reference returns (a reference, true) if a reference associated with
+	// given label exists, otherwise (nil, false).
+	Reference(label string) (Reference, bool)
+
+	// References returns a list of references.
+	References() []Reference
+
+	// BlockOffset returns a first non-space character position on current line.
+	// This value is valid only for BlockParser.Open.
+	BlockOffset() int
+
+	// BlockOffset sets a first non-space character position on current line.
+	// This value is valid only for BlockParser.Open.
+	SetBlockOffset(int)
+
+	// FirstDelimiter returns a first delimiter of the current delimiter list.
+	FirstDelimiter() *Delimiter
+
+	// LastDelimiter returns a last delimiter of the current delimiter list.
+	LastDelimiter() *Delimiter
+
+	// PushDelimiter appends given delimiter to the tail of the current
+	// delimiter list.
+	PushDelimiter(delimiter *Delimiter)
+
+	// RemoveDelimiter removes given delimiter from the current delimiter list.
+	RemoveDelimiter(d *Delimiter)
+
+	// ClearDelimiters clears the current delimiter list.
+	ClearDelimiters(bottom ast.Node)
+
+	// OpenedBlocks returns a list of nodes that are currently in parsing.
+	OpenedBlocks() []Block
+
+	// SetOpenedBlocks sets a list of nodes that are currently in parsing.
+	SetOpenedBlocks([]Block)
+
+	// LastOpenedBlock returns a last node that is currently in parsing.
+	LastOpenedBlock() Block
+
+	// SetLastOpenedBlock sets a last node that is currently in parsing.
+	SetLastOpenedBlock(Block)
+}
+
+// A Result interface holds a result of parsing Markdown text.
+type Result interface {
+	// Reference returns (a reference, true) if a reference associated with
+	// given label exists, otherwise (nil, false).
+	Reference(label string) (Reference, bool)
+}
+
+type parseContext struct {
+	store           []interface{}
+	source          []byte
+	refs            map[string]Reference
+	blockOffset     int
+	delimiters      *Delimiter
+	lastDelimiter   *Delimiter
+	openedBlocks    []Block
+	lastOpenedBlock Block
+}
+
+func newContext(source []byte) Context {
+	return &parseContext{
+		store:           make([]interface{}, contextKey+1),
+		source:          source,
+		refs:            map[string]Reference{},
+		blockOffset:     0,
+		delimiters:      nil,
+		lastDelimiter:   nil,
+		openedBlocks:    []Block{},
+		lastOpenedBlock: Block{},
+	}
+}
+
+func (p *parseContext) Get(key ContextKey) interface{} {
+	return p.store[key]
+}
+
+func (p *parseContext) Set(key ContextKey, value interface{}) {
+	p.store[key] = value
+}
+
+func (p *parseContext) BlockOffset() int {
+	return p.blockOffset
+}
+
+func (p *parseContext) SetBlockOffset(v int) {
+	p.blockOffset = v
+}
+
+func (p *parseContext) Source() []byte {
+	return p.source
+}
+
+func (p *parseContext) LastDelimiter() *Delimiter {
+	return p.lastDelimiter
+}
+
+func (p *parseContext) FirstDelimiter() *Delimiter {
+	return p.delimiters
+}
+
+func (p *parseContext) PushDelimiter(d *Delimiter) {
+	if p.delimiters == nil {
+		p.delimiters = d
+		p.lastDelimiter = d
+	} else {
+		l := p.lastDelimiter
+		p.lastDelimiter = d
+		l.NextDelimiter = d
+		d.PreviousDelimiter = l
+	}
+}
+
+func (p *parseContext) RemoveDelimiter(d *Delimiter) {
+	if d.PreviousDelimiter == nil {
+		p.delimiters = d.NextDelimiter
+	} else {
+		d.PreviousDelimiter.NextDelimiter = d.NextDelimiter
+		if d.NextDelimiter != nil {
+			d.NextDelimiter.PreviousDelimiter = d.PreviousDelimiter
+		}
+	}
+	if d.NextDelimiter == nil {
+		p.lastDelimiter = d.PreviousDelimiter
+	}
+	if p.delimiters != nil {
+		p.delimiters.PreviousDelimiter = nil
+	}
+	if p.lastDelimiter != nil {
+		p.lastDelimiter.NextDelimiter = nil
+	}
+	d.NextDelimiter = nil
+	d.PreviousDelimiter = nil
+	if d.Length != 0 {
+		ast.MergeOrReplaceTextSegment(d.Parent(), d, d.Segment)
+	} else {
+		d.Parent().RemoveChild(d.Parent(), d)
+	}
+}
+
+func (p *parseContext) ClearDelimiters(bottom ast.Node) {
+	if p.lastDelimiter == nil {
+		return
+	}
+	var c ast.Node
+	for c = p.lastDelimiter; c != nil && c != bottom; {
+		prev := c.PreviousSibling()
+		if d, ok := c.(*Delimiter); ok {
+			p.RemoveDelimiter(d)
+		}
+		c = prev
+	}
+}
+
+func (p *parseContext) AddReference(ref Reference) {
+	key := util.ToLinkReference(ref.Label())
+	if _, ok := p.refs[key]; !ok {
+		p.refs[key] = ref
+	}
+}
+
+func (p *parseContext) Reference(label string) (Reference, bool) {
+	v, ok := p.refs[label]
+	return v, ok
+}
+
+func (p *parseContext) References() []Reference {
+	ret := make([]Reference, 0, len(p.refs))
+	for _, v := range p.refs {
+		ret = append(ret, v)
+	}
+	return ret
+}
+
+func (p *parseContext) String() string {
+	refs := []string{}
+	for _, r := range p.refs {
+		refs = append(refs, r.String())
+	}
+
+	return fmt.Sprintf("Context{Store:%#v, Refs:%s}", p.store, strings.Join(refs, ","))
+}
+
+func (p *parseContext) OpenedBlocks() []Block {
+	return p.openedBlocks
+}
+
+func (p *parseContext) SetOpenedBlocks(v []Block) {
+	p.openedBlocks = v
+}
+
+func (p *parseContext) LastOpenedBlock() Block {
+	return p.lastOpenedBlock
+}
+
+func (p *parseContext) SetLastOpenedBlock(v Block) {
+	p.lastOpenedBlock = v
+}
+
+// State represents parser's state.
+// State is designed to use as a bit flag.
+type State int
+
+const (
+	none State = 1 << iota
+
+	// Continue indicates parser can continue parsing.
+	Continue
+
+	// Close indicates parser cannot parse anymore.
+	Close
+
+	// HasChildren indicates parser may have child blocks.
+	HasChildren
+
+	// NoChildren indicates parser does not have child blocks.
+	NoChildren
+)
+
+// A Config struct is a data structure that holds configuration of the Parser.
+type Config struct {
+	Options               map[OptionName]interface{}
+	BlockParsers          util.PrioritizedSlice /*<BlockParser>*/
+	InlineParsers         util.PrioritizedSlice /*<InlineParser>*/
+	ParagraphTransformers util.PrioritizedSlice /*<ParagraphTransformer>*/
+	ASTTransformers       util.PrioritizedSlice /*<ASTTransformer>*/
+}
+
+// NewConfig returns a new Config.
+func NewConfig() *Config {
+	return &Config{
+		Options:               map[OptionName]interface{}{},
+		BlockParsers:          util.PrioritizedSlice{},
+		InlineParsers:         util.PrioritizedSlice{},
+		ParagraphTransformers: util.PrioritizedSlice{},
+		ASTTransformers:       util.PrioritizedSlice{},
+	}
+}
+
+// An Option interface is a functional option type for the Parser.
+type Option interface {
+	SetConfig(*Config)
+}
+
+// OptionName is a name of parser options.
+type OptionName string
+
+// A Parser interface parses Markdown text into AST nodes.
+type Parser interface {
+	// Parse parses given Markdown text into AST nodes.
+	Parse(reader text.Reader) (ast.Node, Result)
+
+	// AddOption adds given option to thie parser.
+	AddOption(Option)
+}
+
+// A SetOptioner interface sets given option to the object.
+type SetOptioner interface {
+	// SetOption sets given option to the object.
+	// Unacceptable options may be passed.
+	// Thus implementations must ignore unacceptable options.
+	SetOption(name OptionName, value interface{})
+}
+
+// A BlockParser interface parses a block level element like Paragraph, List,
+// Blockquote etc.
+type BlockParser interface {
+	// Open parses the current line and returns a result of parsing.
+	//
+	// Open must not parse beyond the current line.
+	// If Open has been able to parse the current line, Open must advance a reader
+	// position by consumed byte length.
+	//
+	// If Open has not been able to parse the current line, Open should returns
+	// (nil, NoChildren). If Open has been able to parse the current line, Open
+	// should returns a new Block node and returns HasChildren or NoChildren.
+	Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State)
+
+	// Continue parses the current line and returns a result of parsing.
+	//
+	// Continue must not parse beyond the current line.
+	// If Continue has been able to parse the current line, Continue must advance
+	// a reader position by consumed byte length.
+	//
+	// If Continue has not been able to parse the current line, Continue should
+	// returns Close. If Continue has been able to parse the current line,
+	// Continue should returns (Continue | NoChildren) or
+	// (Continue | HasChildren)
+	Continue(node ast.Node, reader text.Reader, pc Context) State
+
+	// Close will be called when the parser returns Close.
+	Close(node ast.Node, pc Context)
+
+	// CanInterruptParagraph returns true if the parser can interrupt pargraphs,
+	// otherwise false.
+	CanInterruptParagraph() bool
+
+	// CanAcceptIndentedLine returns true if the parser can open new node when
+	// given line is being indented more than 3 spaces.
+	CanAcceptIndentedLine() bool
+}
+
+// An InlineParser interface parses an inline level element like CodeSpan, Link etc.
+type InlineParser interface {
+	// Trigger returns a list of characters that triggers Parse method of
+	// this parser.
+	// Trigger characters must be a punctuation or a halfspace.
+	// Halfspaces triggers this parser when character is any spaces characters or
+	// a head of line
+	Trigger() []byte
+
+	// Parse parse given block into an inline node.
+	//
+	// Parse can parse beyond the current line.
+	// If Parse has been able to parse the current line, it must advance a reader
+	// position by consumed byte length.
+	Parse(parent ast.Node, block text.Reader, pc Context) ast.Node
+
+	// CloseBlock will be called when a block is closed.
+	CloseBlock(parent ast.Node, pc Context)
+}
+
+// A ParagraphTransformer transforms parsed Paragraph nodes.
+// For example, link references are searched in parsed Paragraphs.
+type ParagraphTransformer interface {
+	// Transform transforms given paragraph.
+	Transform(node *ast.Paragraph, pc Context)
+}
+
+// ASTTransformer transforms entire Markdown document AST tree.
+type ASTTransformer interface {
+	// Transform transforms given AST tree.
+	Transform(node *ast.Document, pc Context)
+}
+
+// DefaultBlockParsers returns a new list of default BlockParsers.
+// Priorities of default BlockParsers are:
+//
+//     SetextHeadingParser, 100
+//     ThemanticBreakParser, 200
+//     ListParser, 300
+//     ListItemParser, 400
+//     CodeBlockParser, 500
+//     ATXHeadingParser, 600
+//     FencedCodeBlockParser, 700
+//     BlockquoteParser, 800
+//     HTMLBlockParser, 900
+//     ParagraphParser, 1000
+func DefaultBlockParsers() []util.PrioritizedValue {
+	return []util.PrioritizedValue{
+		util.Prioritized(NewSetextHeadingParser(), 100),
+		util.Prioritized(NewThemanticBreakParser(), 200),
+		util.Prioritized(NewListParser(), 300),
+		util.Prioritized(NewListItemParser(), 400),
+		util.Prioritized(NewCodeBlockParser(), 500),
+		util.Prioritized(NewATXHeadingParser(), 600),
+		util.Prioritized(NewFencedCodeBlockParser(), 700),
+		util.Prioritized(NewBlockquoteParser(), 800),
+		util.Prioritized(NewHTMLBlockParser(), 900),
+		util.Prioritized(NewParagraphParser(), 1000),
+	}
+}
+
+// DefaultInlineParsers returns a new list of default InlineParsers.
+// Priorities of default InlineParsers are:
+//
+//     CodeSpanParser, 100
+//     LinkParser, 200
+//     AutoLinkParser, 300
+//     RawHTMLParser, 400
+//     EmphasisParser, 500
+func DefaultInlineParsers() []util.PrioritizedValue {
+	return []util.PrioritizedValue{
+		util.Prioritized(NewCodeSpanParser(), 100),
+		util.Prioritized(NewLinkParser(), 200),
+		util.Prioritized(NewAutoLinkParser(), 300),
+		util.Prioritized(NewRawHTMLParser(), 400),
+		util.Prioritized(NewEmphasisParser(), 500),
+	}
+}
+
+// DefaultParagraphTransformers returns a new list of default ParagraphTransformers.
+// Priorities of default ParagraphTransformers are:
+//
+//     LinkReferenceParagraphTransformer, 100
+func DefaultParagraphTransformers() []util.PrioritizedValue {
+	return []util.PrioritizedValue{
+		util.Prioritized(LinkReferenceParagraphTransformer, 100),
+	}
+}
+
+// A Block struct holds a node and correspond parser pair.
+type Block struct {
+	// Node is a BlockNode.
+	Node ast.Node
+	// Parser is a BlockParser.
+	Parser BlockParser
+}
+
+type parser struct {
+	options               map[OptionName]interface{}
+	blockParsers          []BlockParser
+	inlineParsers         [256][]InlineParser
+	inlineParsersList     []InlineParser
+	paragraphTransformers []ParagraphTransformer
+	astTransformers       []ASTTransformer
+	config                *Config
+	initSync              sync.Once
+}
+
+type withBlockParsers struct {
+	value []util.PrioritizedValue
+}
+
+func (o *withBlockParsers) SetConfig(c *Config) {
+	c.BlockParsers = append(c.BlockParsers, o.value...)
+}
+
+// WithBlockParsers is a functional option that allow you to add
+// BlockParsers to the parser.
+func WithBlockParsers(bs ...util.PrioritizedValue) Option {
+	return &withBlockParsers{bs}
+}
+
+type withInlineParsers struct {
+	value []util.PrioritizedValue
+}
+
+func (o *withInlineParsers) SetConfig(c *Config) {
+	c.InlineParsers = append(c.InlineParsers, o.value...)
+}
+
+// WithInlineParsers is a functional option that allow you to add
+// InlineParsers to the parser.
+func WithInlineParsers(bs ...util.PrioritizedValue) Option {
+	return &withInlineParsers{bs}
+}
+
+type withParagraphTransformers struct {
+	value []util.PrioritizedValue
+}
+
+func (o *withParagraphTransformers) SetConfig(c *Config) {
+	c.ParagraphTransformers = append(c.ParagraphTransformers, o.value...)
+}
+
+// WithParagraphTransformers is a functional option that allow you to add
+// ParagraphTransformers to the parser.
+func WithParagraphTransformers(ps ...util.PrioritizedValue) Option {
+	return &withParagraphTransformers{ps}
+}
+
+type withASTTransformers struct {
+	value []util.PrioritizedValue
+}
+
+func (o *withASTTransformers) SetConfig(c *Config) {
+	c.ASTTransformers = append(c.ASTTransformers, o.value...)
+}
+
+// WithASTTransformers is a functional option that allow you to add
+// ASTTransformers to the parser.
+func WithASTTransformers(ps ...util.PrioritizedValue) Option {
+	return &withASTTransformers{ps}
+}
+
+type withOption struct {
+	name  OptionName
+	value interface{}
+}
+
+func (o *withOption) SetConfig(c *Config) {
+	c.Options[o.name] = o.value
+}
+
+// WithOption is a functional option that allow you to set
+// an arbitary option to the parser.
+func WithOption(name OptionName, value interface{}) Option {
+	return &withOption{name, value}
+}
+
+// NewParser returns a new Parser with given options.
+func NewParser(options ...Option) Parser {
+	config := NewConfig()
+	for _, opt := range options {
+		opt.SetConfig(config)
+	}
+
+	p := &parser{
+		options: map[OptionName]interface{}{},
+		config:  config,
+	}
+
+	return p
+}
+
+func (p *parser) AddOption(o Option) {
+	o.SetConfig(p.config)
+}
+
+func (p *parser) addBlockParser(v util.PrioritizedValue, options map[OptionName]interface{}) {
+	bp, ok := v.Value.(BlockParser)
+	if !ok {
+		panic(fmt.Sprintf("%v is not a BlockParser", v.Value))
+	}
+	so, ok := v.Value.(SetOptioner)
+	if ok {
+		for oname, ovalue := range options {
+			so.SetOption(oname, ovalue)
+		}
+	}
+	p.blockParsers = append(p.blockParsers, bp)
+}
+
+func (p *parser) addInlineParser(v util.PrioritizedValue, options map[OptionName]interface{}) {
+	ip, ok := v.Value.(InlineParser)
+	if !ok {
+		panic(fmt.Sprintf("%v is not a InlineParser", v.Value))
+	}
+	tcs := ip.Trigger()
+	so, ok := v.Value.(SetOptioner)
+	if ok {
+		for oname, ovalue := range options {
+			so.SetOption(oname, ovalue)
+		}
+	}
+	p.inlineParsersList = append(p.inlineParsersList, ip)
+	for _, tc := range tcs {
+		if p.inlineParsers[tc] == nil {
+			p.inlineParsers[tc] = []InlineParser{}
+		}
+		p.inlineParsers[tc] = append(p.inlineParsers[tc], ip)
+	}
+}
+
+func (p *parser) addParagraphTransformer(v util.PrioritizedValue, options map[OptionName]interface{}) {
+	pt, ok := v.Value.(ParagraphTransformer)
+	if !ok {
+		panic(fmt.Sprintf("%v is not a ParagraphTransformer", v.Value))
+	}
+	so, ok := v.Value.(SetOptioner)
+	if ok {
+		for oname, ovalue := range options {
+			so.SetOption(oname, ovalue)
+		}
+	}
+	p.paragraphTransformers = append(p.paragraphTransformers, pt)
+}
+
+func (p *parser) addASTTransformer(v util.PrioritizedValue, options map[OptionName]interface{}) {
+	at, ok := v.Value.(ASTTransformer)
+	if !ok {
+		panic(fmt.Sprintf("%v is not a ASTTransformer", v.Value))
+	}
+	so, ok := v.Value.(SetOptioner)
+	if ok {
+		for oname, ovalue := range options {
+			so.SetOption(oname, ovalue)
+		}
+	}
+	p.astTransformers = append(p.astTransformers, at)
+}
+
+func (p *parser) Parse(reader text.Reader) (ast.Node, Result) {
+	p.initSync.Do(func() {
+		p.config.BlockParsers.Sort()
+		for _, v := range p.config.BlockParsers {
+			p.addBlockParser(v, p.config.Options)
+		}
+		p.config.InlineParsers.Sort()
+		for _, v := range p.config.InlineParsers {
+			p.addInlineParser(v, p.config.Options)
+		}
+		p.config.ParagraphTransformers.Sort()
+		for _, v := range p.config.ParagraphTransformers {
+			p.addParagraphTransformer(v, p.config.Options)
+		}
+		p.config.ASTTransformers.Sort()
+		for _, v := range p.config.ASTTransformers {
+			p.addASTTransformer(v, p.config.Options)
+		}
+		p.config = nil
+	})
+	root := ast.NewDocument()
+	pc := newContext(reader.Source())
+	p.parseBlocks(root, reader, pc)
+	blockReader := text.NewBlockReader(reader.Source(), nil)
+	p.walkBlock(root, func(node ast.Node) {
+		p.parseBlock(blockReader, node, pc)
+	})
+	for _, at := range p.astTransformers {
+		at.Transform(root, pc)
+	}
+	//root.Dump(reader.Source(), 0)
+	return root, pc
+}
+
+func (p *parser) transformParagraph(node *ast.Paragraph, pc Context) {
+	for _, pt := range p.paragraphTransformers {
+		pt.Transform(node, pc)
+		if node.Parent() == nil {
+			break
+		}
+	}
+}
+
+func (p *parser) closeBlocks(from, to int, pc Context) {
+	blocks := pc.OpenedBlocks()
+	last := pc.LastOpenedBlock()
+	for i := from; i >= to; i-- {
+		node := blocks[i].Node
+		if node.Parent() != nil {
+			blocks[i].Parser.Close(blocks[i].Node, pc)
+			paragraph, ok := node.(*ast.Paragraph)
+			if ok && node.Parent() != nil {
+				p.transformParagraph(paragraph, pc)
+			}
+		}
+	}
+	if from == len(blocks)-1 {
+		blocks = blocks[0:to]
+	} else {
+		blocks = append(blocks[0:to], blocks[from+1:]...)
+	}
+	l := len(blocks)
+	if l == 0 {
+		last.Node = nil
+	} else {
+		last = blocks[l-1]
+	}
+	pc.SetOpenedBlocks(blocks)
+	pc.SetLastOpenedBlock(last)
+}
+
+type blockOpenResult int
+
+const (
+	paragraphContinuation blockOpenResult = iota + 1
+	newBlocksOpened
+	noBlocksOpened
+)
+
+func (p *parser) openBlocks(parent ast.Node, blankLine bool, reader text.Reader, pc Context) blockOpenResult {
+	result := blockOpenResult(noBlocksOpened)
+	continuable := false
+	lastBlock := pc.LastOpenedBlock()
+	if lastBlock.Node != nil {
+		continuable = ast.IsParagraph(lastBlock.Node)
+	}
+retry:
+	shouldPeek := true
+	var currentLineNum int
+	var w int
+	var pos int
+	var line []byte
+	for _, bp := range p.blockParsers {
+		if shouldPeek {
+			currentLineNum, _ = reader.Position()
+			line, _ = reader.PeekLine()
+			w, pos = util.IndentWidth(line, 0)
+			pc.SetBlockOffset(pos)
+			shouldPeek = false
+			if line == nil || line[0] == '\n' {
+				break
+			}
+		}
+		if continuable && result == noBlocksOpened && !bp.CanInterruptParagraph() {
+			continue
+		}
+		if w > 3 && !bp.CanAcceptIndentedLine() {
+			continue
+		}
+		last := pc.LastOpenedBlock().Node
+		node, state := bp.Open(parent, reader, pc)
+		if l, _ := reader.Position(); l != currentLineNum {
+			panic("BlockParser.Open must not advance position beyond the current line")
+		}
+		if node != nil {
+			shouldPeek = true
+			node.SetBlankPreviousLines(blankLine)
+			if last != nil && last.Parent() == nil {
+				lastPos := len(pc.OpenedBlocks()) - 1
+				p.closeBlocks(lastPos, lastPos, pc)
+			}
+			parent.AppendChild(parent, node)
+			result = newBlocksOpened
+			be := Block{node, bp}
+			pc.SetOpenedBlocks(append(pc.OpenedBlocks(), be))
+			pc.SetLastOpenedBlock(be)
+			if state == HasChildren {
+				parent = node
+				goto retry // try child block
+			}
+			break // no children, can not open more blocks on this line
+		}
+	}
+	if result == noBlocksOpened && continuable {
+		state := lastBlock.Parser.Continue(lastBlock.Node, reader, pc)
+		if state&Continue != 0 {
+			result = paragraphContinuation
+		}
+	}
+	return result
+}
+
+type lineStat struct {
+	lineNum int
+	level   int
+	isBlank bool
+}
+
+func isBlankLine(lineNum, level int, stats []lineStat) ([]lineStat, bool) {
+	ret := false
+	for i := len(stats) - 1 - level; i >= 0; i-- {
+		s := stats[i]
+		if s.lineNum == lineNum && s.level == level {
+			ret = s.isBlank
+			continue
+		}
+		if s.lineNum < lineNum {
+			return stats[i:], ret
+		}
+	}
+	return stats[0:0], ret
+}
+
+func (p *parser) parseBlocks(parent ast.Node, reader text.Reader, pc Context) {
+	pc.SetLastOpenedBlock(Block{})
+	pc.SetOpenedBlocks([]Block{})
+	blankLines := make([]lineStat, 0, 64)
+	isBlank := false
+	for { // process blocks separated by blank lines
+		_, lines, ok := reader.SkipBlankLines()
+		if !ok {
+			return
+		}
+		// first, we try to open blocks
+		if p.openBlocks(parent, lines != 0, reader, pc) != newBlocksOpened {
+			return
+		}
+		lineNum, _ := reader.Position()
+		for i := 0; i < len(pc.OpenedBlocks()); i++ {
+			blankLines = append(blankLines, lineStat{lineNum - 1, i, lines != 0})
+		}
+		reader.AdvanceLine()
+		for len(pc.OpenedBlocks()) != 0 { // process opened blocks line by line
+			lastIndex := len(pc.OpenedBlocks()) - 1
+			for i := 0; i < len(pc.OpenedBlocks()); i++ {
+				be := pc.OpenedBlocks()[i]
+				line, _ := reader.PeekLine()
+				if line == nil {
+					p.closeBlocks(lastIndex, 0, pc)
+					reader.AdvanceLine()
+					return
+				}
+				lineNum, _ := reader.Position()
+				blankLines = append(blankLines, lineStat{lineNum, i, util.IsBlank(line)})
+				// If node is a paragraph, p.openBlocks determines whether it is continuable.
+				// So we do not process paragraphs here.
+				if !ast.IsParagraph(be.Node) {
+					state := be.Parser.Continue(be.Node, reader, pc)
+					if state&Continue != 0 {
+						// When current node is a container block and has no children,
+						// we try to open new child nodes
+						if state&HasChildren != 0 && i == lastIndex {
+							blankLines, isBlank = isBlankLine(lineNum-1, i, blankLines)
+							p.openBlocks(be.Node, isBlank, reader, pc)
+							break
+						}
+						continue
+					}
+				}
+				// current node may be closed or lazy continuation
+				blankLines, isBlank = isBlankLine(lineNum-1, i, blankLines)
+				thisParent := parent
+				if i != 0 {
+					thisParent = pc.OpenedBlocks()[i-1].Node
+				}
+				result := p.openBlocks(thisParent, isBlank, reader, pc)
+				if result != paragraphContinuation {
+					p.closeBlocks(lastIndex, i, pc)
+				}
+				break
+			}
+
+			reader.AdvanceLine()
+		}
+	}
+}
+
+func (p *parser) walkBlock(block ast.Node, cb func(node ast.Node)) {
+	for c := block.FirstChild(); c != nil; c = c.NextSibling() {
+		p.walkBlock(c, cb)
+	}
+	cb(block)
+}
+
+func (p *parser) parseBlock(block text.BlockReader, parent ast.Node, pc Context) {
+	if parent.IsRaw() {
+		return
+	}
+	escaped := false
+	source := block.Source()
+	block.Reset(parent.Lines())
+	for {
+	retry:
+		line, _ := block.PeekLine()
+		if line == nil {
+			break
+		}
+		lineLength := len(line)
+		l, startPosition := block.Position()
+		n := 0
+		softLinebreak := false
+		for i := 0; i < lineLength; i++ {
+			c := line[i]
+			if c == '\n' {
+				softLinebreak = true
+				break
+			}
+			isSpace := util.IsSpace(c)
+			isPunct := util.IsPunct(c)
+			if (isPunct && !escaped) || isSpace || i == 0 {
+				parserChar := c
+				if isSpace || (i == 0 && !isPunct) {
+					parserChar = ' '
+				}
+				ips := p.inlineParsers[parserChar]
+				if ips != nil {
+					block.Advance(n)
+					n = 0
+					savedLine, savedPosition := block.Position()
+					if i != 0 {
+						_, currentPosition := block.Position()
+						ast.MergeOrAppendTextSegment(parent, startPosition.Between(currentPosition))
+						_, startPosition = block.Position()
+					}
+					var inlineNode ast.Node
+					for _, ip := range ips {
+						inlineNode = ip.Parse(parent, block, pc)
+						if inlineNode != nil {
+							break
+						}
+						block.SetPosition(savedLine, savedPosition)
+					}
+					if inlineNode != nil {
+						parent.AppendChild(parent, inlineNode)
+						goto retry
+					}
+				}
+			}
+			if escaped {
+				escaped = false
+				n++
+				continue
+			}
+
+			if c == '\\' {
+				escaped = true
+				n++
+				continue
+			}
+
+			escaped = false
+			n++
+		}
+		if n != 0 {
+			block.Advance(n)
+		}
+		currentL, currentPosition := block.Position()
+		if l != currentL {
+			continue
+		}
+		diff := startPosition.Between(currentPosition)
+		stop := diff.Stop
+		hardlineBreak := false
+		if lineLength > 2 && line[lineLength-2] == '\\' && softLinebreak { // ends with \\n
+			stop--
+			hardlineBreak = true
+		} else if lineLength > 3 && line[lineLength-3] == ' ' && line[lineLength-2] == ' ' && softLinebreak { // ends with [space][space]\n
+			hardlineBreak = true
+		}
+		rest := diff.WithStop(stop)
+		text := ast.NewTextSegment(rest.TrimRightSpace(source))
+		text.SetSoftLineBreak(softLinebreak)
+		text.SetHardLineBreak(hardlineBreak)
+		parent.AppendChild(parent, text)
+		block.AdvanceLine()
+	}
+
+	ProcessDelimiters(nil, pc)
+	for _, ip := range p.inlineParsersList {
+		ip.CloseBlock(parent, pc)
+	}
+
+}

parser/raw_html.go

@@ -0,0 +1,126 @@
+package parser
+
+import (
+	"bytes"
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+)
+
+type rawHTMLParser struct {
+	HTMLConfig
+}
+
+// NewRawHTMLParser return a new InlineParser that can parse
+// inline htmls
+func NewRawHTMLParser(opts ...HTMLOption) InlineParser {
+	p := &rawHTMLParser{}
+	for _, o := range opts {
+		o.SetHTMLOption(&p.HTMLConfig)
+	}
+	return p
+}
+
+func (s *rawHTMLParser) Trigger() []byte {
+	return []byte{'<'}
+}
+
+func (s *rawHTMLParser) Parse(parent ast.Node, block text.Reader, pc Context) ast.Node {
+	line, _ := block.PeekLine()
+	if len(line) > 1 && util.IsAlphaNumeric(line[1]) {
+		return s.parseMultiLineRegexp(openTagRegexp, block, pc)
+	}
+	if len(line) > 2 && line[1] == '/' && util.IsAlphaNumeric(line[2]) {
+		return s.parseMultiLineRegexp(closeTagRegexp, block, pc)
+	}
+	if bytes.HasPrefix(line, []byte("<!--")) {
+		return s.parseMultiLineRegexp(commentRegexp, block, pc)
+	}
+	if bytes.HasPrefix(line, []byte("<?")) {
+		return s.parseSingleLineRegexp(processingInstructionRegexp, block, pc)
+	}
+	if len(line) > 2 && line[1] == '!' && line[2] >= 'A' && line[2] <= 'Z' {
+		return s.parseSingleLineRegexp(declRegexp, block, pc)
+	}
+	if bytes.HasPrefix(line, []byte("<![CDATA[")) {
+		return s.parseMultiLineRegexp(cdataRegexp, block, pc)
+	}
+	return nil
+}
+
+var tagnamePattern = `([A-Za-z][A-Za-z0-9-]*)`
+var attributePattern = `(?:\s+[a-zA-Z_:][a-zA-Z0-9:._-]*(?:\s*=\s*(?:[^\"'=<>` + "`" + `\x00-\x20]+|'[^']*'|"[^"]*"))?)`
+var openTagRegexp = regexp.MustCompile("^<" + tagnamePattern + attributePattern + `*\s*/?>`)
+var closeTagRegexp = regexp.MustCompile("^</" + tagnamePattern + `\s*>`)
+var commentRegexp = regexp.MustCompile(`^<!---->|<!--(?:-?[^>-])(?:-?[^-])*-->`)
+var processingInstructionRegexp = regexp.MustCompile(`^(?:<\?).*?(?:\?>)`)
+var declRegexp = regexp.MustCompile(`^<![A-Z]+\s+[^>]*>`)
+var cdataRegexp = regexp.MustCompile(`<!\[CDATA\[[\s\S]*?\]\]>`)
+
+func (s *rawHTMLParser) parseSingleLineRegexp(reg *regexp.Regexp, block text.Reader, pc Context) ast.Node {
+	line, segment := block.PeekLine()
+	match := reg.FindSubmatchIndex(line)
+	if match == nil {
+		return nil
+	}
+	node := ast.NewRawHTML()
+	node.AppendChild(node, ast.NewRawTextSegment(segment.WithStop(segment.Start+match[1])))
+	block.Advance(match[1])
+	return node
+}
+
+var dummyMatch = [][]byte{}
+
+func (s *rawHTMLParser) parseMultiLineRegexp(reg *regexp.Regexp, block text.Reader, pc Context) ast.Node {
+	sline, ssegment := block.Position()
+	var m [][]byte
+	if s.FilterTags != nil {
+		m = block.FindSubMatch(reg)
+	} else {
+		if block.Match(reg) {
+			m = dummyMatch
+		}
+	}
+
+	if m != nil {
+		if s.FilterTags != nil {
+			tagName := string(m[1])
+			if _, ok := s.FilterTags[tagName]; ok {
+				return nil
+			}
+		}
+		node := ast.NewRawHTML()
+		eline, esegment := block.Position()
+		block.SetPosition(sline, ssegment)
+		for {
+			line, segment := block.PeekLine()
+			if line == nil {
+				break
+			}
+			l, _ := block.Position()
+			start := segment.Start
+			if l == sline {
+				start = ssegment.Start
+			}
+			end := segment.Stop
+			if l == eline {
+				end = esegment.Start
+			}
+
+			node.AppendChild(node, ast.NewRawTextSegment(text.NewSegment(start, end)))
+			if l == eline {
+				block.Advance(end - start)
+				break
+			} else {
+				block.AdvanceLine()
+			}
+		}
+		return node
+	}
+	return nil
+}
+
+func (s *rawHTMLParser) CloseBlock(parent ast.Node, pc Context) {
+	// nothing to do
+}

parser/setext_headings.go

@@ -0,0 +1,109 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+var temporaryParagraphKey = NewContextKey()
+
+type setextHeadingParser struct {
+	HeadingConfig
+}
+
+func matchesSetextHeadingBar(line []byte) (byte, bool) {
+	start := 0
+	end := len(line)
+	space := util.TrimLeftLength(line, []byte{' '})
+	if space > 3 {
+		return 0, false
+	}
+	start += space
+	level1 := util.TrimLeftLength(line[start:end], []byte{'='})
+	c := byte('=')
+	var level2 int
+	if level1 == 0 {
+		level2 = util.TrimLeftLength(line[start:end], []byte{'-'})
+		c = '-'
+	}
+	end -= util.TrimRightSpaceLength(line[start:end])
+	if !((level1 > 0 && start+level1 == end) || (level2 > 0 && start+level2 == end)) {
+		return 0, false
+	}
+	return c, true
+}
+
+// NewSetextHeadingParser return a new BlockParser that can parse Setext headings.
+func NewSetextHeadingParser(opts ...HeadingOption) BlockParser {
+	p := &setextHeadingParser{}
+	for _, o := range opts {
+		o.SetHeadingOption(&p.HeadingConfig)
+	}
+	return p
+}
+
+func (b *setextHeadingParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	last := pc.LastOpenedBlock().Node
+	if last == nil {
+		return nil, NoChildren
+	}
+	paragraph, ok := last.(*ast.Paragraph)
+	if !ok || paragraph.Parent() != parent {
+		return nil, NoChildren
+	}
+	line, segment := reader.PeekLine()
+	c, ok := matchesSetextHeadingBar(line)
+	if !ok {
+		return nil, NoChildren
+	}
+	level := 1
+	if c == '-' {
+		level = 2
+	}
+	node := ast.NewHeading(level)
+	node.Lines().Append(segment)
+	pc.Set(temporaryParagraphKey, paragraph)
+	return node, NoChildren
+}
+
+func (b *setextHeadingParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	return Close
+}
+
+func (b *setextHeadingParser) Close(node ast.Node, pc Context) {
+	heading := node.(*ast.Heading)
+	segment := node.Lines().At(0)
+	heading.Lines().Clear()
+	tmp := pc.Get(temporaryParagraphKey).(*ast.Paragraph)
+	pc.Set(temporaryParagraphKey, nil)
+	if tmp.Lines().Len() == 0 {
+		next := heading.NextSibling()
+		segment = segment.TrimLeftSpace(pc.Source())
+		if next == nil || !ast.IsParagraph(next) {
+			para := ast.NewParagraph()
+			para.Lines().Append(segment)
+			heading.Parent().InsertAfter(heading.Parent(), heading, para)
+		} else {
+			next.(ast.Node).Lines().Unshift(segment)
+		}
+		heading.Parent().RemoveChild(heading.Parent(), heading)
+	} else {
+		heading.SetLines(tmp.Lines())
+		heading.SetBlankPreviousLines(tmp.HasBlankPreviousLines())
+		tmp.Parent().RemoveChild(tmp.Parent(), tmp)
+	}
+
+	if !b.HeadingID {
+		return
+	}
+	parseOrGenerateHeadingID(heading, pc)
+}
+
+func (b *setextHeadingParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *setextHeadingParser) CanAcceptIndentedLine() bool {
+	return false
+}

parser/themantic_break.go

@@ -0,0 +1,71 @@
+package parser
+
+import (
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+)
+
+type themanticBreakParser struct {
+}
+
+var defaultThemanticBreakParser = &themanticBreakParser{}
+
+// NewThemanticBreakParser returns a new BlockParser that
+// parses themantic breaks.
+func NewThemanticBreakParser() BlockParser {
+	return defaultThemanticBreakParser
+}
+
+func isThemanticBreak(line []byte) bool {
+	w, pos := util.IndentWidth(line, 0)
+	if w > 3 {
+		return false
+	}
+	mark := byte(0)
+	count := 0
+	for i := pos; i < len(line); i++ {
+		c := line[i]
+		if util.IsSpace(c) {
+			continue
+		}
+		if mark == 0 {
+			mark = c
+			count = 1
+			if mark == '*' || mark == '-' || mark == '_' {
+				continue
+			}
+			return false
+		}
+		if c != mark {
+			return false
+		}
+		count++
+	}
+	return count > 2
+}
+
+func (b *themanticBreakParser) Open(parent ast.Node, reader text.Reader, pc Context) (ast.Node, State) {
+	line, segment := reader.PeekLine()
+	if isThemanticBreak(line) {
+		reader.Advance(segment.Len() - 1)
+		return ast.NewThemanticBreak(), NoChildren
+	}
+	return nil, NoChildren
+}
+
+func (b *themanticBreakParser) Continue(node ast.Node, reader text.Reader, pc Context) State {
+	return Close
+}
+
+func (b *themanticBreakParser) Close(node ast.Node, pc Context) {
+	// nothing to do
+}
+
+func (b *themanticBreakParser) CanInterruptParagraph() bool {
+	return true
+}
+
+func (b *themanticBreakParser) CanAcceptIndentedLine() bool {
+	return false
+}

renderer/html/html.go

@@ -0,0 +1,588 @@
+package html
+
+import (
+	"bytes"
+	"fmt"
+	"strconv"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/renderer"
+	"github.com/yuin/goldmark/util"
+)
+
+// A Config struct has configurations for the HTML based renderers.
+type Config struct {
+	Writer        Writer
+	SoftLineBreak bool
+	XHTML         bool
+}
+
+// NewConfig returns a new Config with defaults.
+func NewConfig() Config {
+	return Config{
+		Writer:        DefaultWriter,
+		SoftLineBreak: false,
+		XHTML:         false,
+	}
+}
+
+// SetOption implements renderer.NodeRenderer.SetOption.
+func (c *Config) SetOption(name renderer.OptionName, value interface{}) {
+	switch name {
+	case SoftLineBreak:
+		c.SoftLineBreak = value.(bool)
+	case XHTML:
+		c.XHTML = value.(bool)
+	case TextWriter:
+		c.Writer = value.(Writer)
+	}
+}
+
+// An Option interface sets options for HTML based renderers.
+type Option interface {
+	SetHTMLOption(*Config)
+}
+
+// TextWriter is an option name used in WithWriter.
+const TextWriter renderer.OptionName = "Writer"
+
+type withWriter struct {
+	value Writer
+}
+
+func (o *withWriter) SetConfig(c *renderer.Config) {
+	c.Options[TextWriter] = o.value
+}
+
+func (o *withWriter) SetHTMLOption(c *Config) {
+	c.Writer = o.value
+}
+
+// WithWriter is a functional option that allow you to set given writer to
+// the renderer.
+func WithWriter(writer Writer) interface {
+	renderer.Option
+	Option
+} {
+	return &withWriter{writer}
+}
+
+// SoftLineBreak is an option name used in WithSoftLineBreak.
+const SoftLineBreak renderer.OptionName = "SoftLineBreak"
+
+type withSoftLineBreak struct {
+}
+
+func (o *withSoftLineBreak) SetConfig(c *renderer.Config) {
+	c.Options[SoftLineBreak] = true
+}
+
+func (o *withSoftLineBreak) SetHTMLOption(c *Config) {
+	c.SoftLineBreak = true
+}
+
+// WithSoftLineBreak is a functional option that indicates whether softline breaks
+// should be rendered as '<br>'.
+func WithSoftLineBreak() interface {
+	renderer.Option
+	Option
+} {
+	return &withSoftLineBreak{}
+}
+
+// XHTML is an option name used in WithXHTML.
+const XHTML renderer.OptionName = "XHTML"
+
+type withXHTML struct {
+}
+
+func (o *withXHTML) SetConfig(c *renderer.Config) {
+	c.Options[XHTML] = true
+}
+
+func (o *withXHTML) SetHTMLOption(c *Config) {
+	c.XHTML = true
+}
+
+// WithXHTML is a functional option indicates that nodes should be rendered in
+// xhtml instead of HTML5.
+func WithXHTML() interface {
+	Option
+	renderer.Option
+} {
+	return &withXHTML{}
+}
+
+// A Renderer struct is an implementation of renderer.NodeRenderer that renders
+// nodes as (X)HTML.
+type Renderer struct {
+	Config
+}
+
+// NewRenderer returns a new Renderer with given options.
+func NewRenderer(opts ...Option) renderer.NodeRenderer {
+	r := &Renderer{
+		Config: NewConfig(),
+	}
+
+	for _, opt := range opts {
+		opt.SetHTMLOption(&r.Config)
+	}
+	return r
+}
+
+// Render implements renderer.NodeRenderer.Render.
+func (r *Renderer) Render(writer util.BufWriter, source []byte, n ast.Node, entering bool) (ast.WalkStatus, error) {
+	switch node := n.(type) {
+
+	// blocks
+
+	case *ast.Document:
+		return r.renderDocument(writer, source, node, entering), nil
+	case *ast.Heading:
+		return r.renderHeading(writer, source, node, entering), nil
+	case *ast.Blockquote:
+		return r.renderBlockquote(writer, source, node, entering), nil
+	case *ast.CodeBlock:
+		return r.renderCodeBlock(writer, source, node, entering), nil
+	case *ast.FencedCodeBlock:
+		return r.renderFencedCodeBlock(writer, source, node, entering), nil
+	case *ast.HTMLBlock:
+		return r.renderHTMLBlock(writer, source, node, entering), nil
+	case *ast.List:
+		return r.renderList(writer, source, node, entering), nil
+	case *ast.ListItem:
+		return r.renderListItem(writer, source, node, entering), nil
+	case *ast.Paragraph:
+		return r.renderParagraph(writer, source, node, entering), nil
+	case *ast.TextBlock:
+		return r.renderTextBlock(writer, source, node, entering), nil
+	case *ast.ThemanticBreak:
+		return r.renderThemanticBreak(writer, source, node, entering), nil
+	// inlines
+
+	case *ast.AutoLink:
+		return r.renderAutoLink(writer, source, node, entering), nil
+	case *ast.CodeSpan:
+		return r.renderCodeSpan(writer, source, node, entering), nil
+	case *ast.Emphasis:
+		return r.renderEmphasis(writer, source, node, entering), nil
+	case *ast.Image:
+		return r.renderImage(writer, source, node, entering), nil
+	case *ast.Link:
+		return r.renderLink(writer, source, node, entering), nil
+	case *ast.RawHTML:
+		return r.renderRawHTML(writer, source, node, entering), nil
+	case *ast.Text:
+		return r.renderText(writer, source, node, entering), nil
+	}
+	return ast.WalkContinue, renderer.NotSupported
+}
+func (r *Renderer) writeLines(w util.BufWriter, source []byte, n ast.Node) {
+	l := n.Lines().Len()
+	for i := 0; i < l; i++ {
+		line := n.Lines().At(i)
+		r.Writer.RawWrite(w, line.Value(source))
+	}
+}
+
+func (r *Renderer) renderDocument(w util.BufWriter, source []byte, n *ast.Document, entering bool) ast.WalkStatus {
+	// nothing to do
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderHeading(w util.BufWriter, source []byte, n *ast.Heading, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<h")
+		w.WriteByte("0123456"[n.Level])
+		if n.ID != nil {
+			w.WriteString(` id="`)
+			w.Write(n.ID)
+			w.WriteByte('"')
+		}
+		w.WriteByte('>')
+	} else {
+		w.WriteString("</h")
+		w.WriteByte("0123456"[n.Level])
+		w.WriteString(">\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderBlockquote(w util.BufWriter, source []byte, n *ast.Blockquote, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<blockquote>\n")
+	} else {
+		w.WriteString("</blockquote>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderCodeBlock(w util.BufWriter, source []byte, n *ast.CodeBlock, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<pre><code>")
+		r.writeLines(w, source, n)
+	} else {
+		w.WriteString("</code></pre>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderFencedCodeBlock(w util.BufWriter, source []byte, n *ast.FencedCodeBlock, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<pre><code")
+		if n.Info != nil {
+			segment := n.Info.Segment
+			info := segment.Value(source)
+			i := 0
+			for ; i < len(info); i++ {
+				if info[i] == ' ' {
+					break
+				}
+			}
+			language := info[:i]
+			w.WriteString(" class=\"language-")
+			r.Writer.Write(w, language)
+			w.WriteString("\"")
+		}
+		w.WriteByte('>')
+		r.writeLines(w, source, n)
+	} else {
+		w.WriteString("</code></pre>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderHTMLBlock(w util.BufWriter, source []byte, n *ast.HTMLBlock, entering bool) ast.WalkStatus {
+	if entering {
+		l := n.Lines().Len()
+		for i := 0; i < l; i++ {
+			line := n.Lines().At(i)
+			w.Write(line.Value(source))
+		}
+	} else {
+		if n.HasClosure() {
+			closure := n.ClosureLine
+			w.Write(closure.Value(source))
+		}
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderList(w util.BufWriter, source []byte, n *ast.List, entering bool) ast.WalkStatus {
+	tag := "ul"
+	if n.IsOrdered() {
+		tag = "ol"
+	}
+	if entering {
+		w.WriteByte('<')
+		w.WriteString(tag)
+		if n.IsOrdered() && n.Start != 1 {
+			fmt.Fprintf(w, " start=\"%d\">\n", n.Start)
+		} else {
+			w.WriteString(">\n")
+		}
+	} else {
+		w.WriteString("</")
+		w.WriteString(tag)
+		w.WriteString(">\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderListItem(w util.BufWriter, source []byte, n *ast.ListItem, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<li>")
+		fc := n.FirstChild()
+		if fc != nil {
+			if _, ok := fc.(*ast.TextBlock); !ok {
+				w.WriteByte('\n')
+			}
+		}
+	} else {
+		w.WriteString("</li>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderParagraph(w util.BufWriter, source []byte, n *ast.Paragraph, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<p>")
+	} else {
+		w.WriteString("</p>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderTextBlock(w util.BufWriter, source []byte, n *ast.TextBlock, entering bool) ast.WalkStatus {
+	if !entering {
+		if _, ok := n.NextSibling().(ast.Node); ok && n.FirstChild() != nil {
+			w.WriteByte('\n')
+		}
+		return ast.WalkContinue
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderThemanticBreak(w util.BufWriter, source []byte, n *ast.ThemanticBreak, entering bool) ast.WalkStatus {
+	if !entering {
+		return ast.WalkContinue
+	}
+	if r.XHTML {
+		w.WriteString("<hr />\n")
+	} else {
+		w.WriteString("<hr>\n")
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderAutoLink(w util.BufWriter, source []byte, n *ast.AutoLink, entering bool) ast.WalkStatus {
+	if !entering {
+		return ast.WalkContinue
+	}
+	w.WriteString(`<a href="`)
+	segment := n.Value.Segment
+	value := segment.Value(source)
+	if n.AutoLinkType == ast.AutoLinkEmail && !bytes.HasPrefix(bytes.ToLower(value), []byte("mailto:")) {
+		w.WriteString("mailto:")
+	}
+	w.Write(util.EscapeHTML(util.URLEscape(value, false)))
+	w.WriteString(`">`)
+	w.Write(util.EscapeHTML(value))
+	w.WriteString(`</a>`)
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderCodeSpan(w util.BufWriter, source []byte, n *ast.CodeSpan, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<code>")
+		for c := n.FirstChild(); c != nil; c = c.NextSibling() {
+			segment := c.(*ast.Text).Segment
+			value := segment.Value(source)
+			if bytes.HasSuffix(value, []byte("\n")) {
+				r.Writer.RawWrite(w, value[:len(value)-1])
+				if c != n.LastChild() {
+					r.Writer.RawWrite(w, []byte(" "))
+				}
+			} else {
+				r.Writer.RawWrite(w, value)
+			}
+		}
+		return ast.WalkSkipChildren
+	}
+	w.WriteString("</code>")
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderEmphasis(w util.BufWriter, source []byte, n *ast.Emphasis, entering bool) ast.WalkStatus {
+	tag := "em"
+	if n.Level == 2 {
+		tag = "strong"
+	}
+	if entering {
+		w.WriteByte('<')
+		w.WriteString(tag)
+		w.WriteByte('>')
+	} else {
+		w.WriteString("</")
+		w.WriteString(tag)
+		w.WriteByte('>')
+	}
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderLink(w util.BufWriter, source []byte, n *ast.Link, entering bool) ast.WalkStatus {
+	if entering {
+		w.WriteString("<a href=\"")
+		w.Write(util.EscapeHTML(util.URLEscape(n.Destination, true)))
+		w.WriteByte('"')
+		if n.Title != nil {
+			w.WriteString(` title="`)
+			r.Writer.Write(w, n.Title)
+			w.WriteByte('"')
+		}
+		w.WriteByte('>')
+	} else {
+		w.WriteString("</a>")
+	}
+	return ast.WalkContinue
+}
+func (r *Renderer) renderImage(w util.BufWriter, source []byte, n *ast.Image, entering bool) ast.WalkStatus {
+	if !entering {
+		return ast.WalkContinue
+	}
+	w.WriteString("<img src=\"")
+	w.Write(util.EscapeHTML(util.URLEscape(n.Destination, true)))
+	w.WriteString(`" alt="`)
+	w.Write(n.Text(source))
+	w.WriteByte('"')
+	if n.Title != nil {
+		w.WriteString(` title="`)
+		r.Writer.Write(w, n.Title)
+		w.WriteByte('"')
+	}
+	if r.XHTML {
+		w.WriteString(" />")
+	} else {
+		w.WriteString(">")
+	}
+	return ast.WalkSkipChildren
+}
+
+func (r *Renderer) renderRawHTML(w util.BufWriter, source []byte, n *ast.RawHTML, entering bool) ast.WalkStatus {
+	return ast.WalkContinue
+}
+
+func (r *Renderer) renderText(w util.BufWriter, source []byte, n *ast.Text, entering bool) ast.WalkStatus {
+	if !entering {
+		return ast.WalkContinue
+	}
+	segment := n.Segment
+	if n.IsRaw() {
+		w.Write(segment.Value(source))
+	} else {
+		r.Writer.Write(w, segment.Value(source))
+		if n.HardLineBreak() || (n.SoftLineBreak() && r.SoftLineBreak) {
+			if r.XHTML {
+				w.WriteString("<br />\n")
+			} else {
+				w.WriteString("<br>\n")
+			}
+		} else if n.SoftLineBreak() {
+			w.WriteByte('\n')
+		}
+	}
+	return ast.WalkContinue
+}
+
+func readWhile(source []byte, index [2]int, pred func(byte) bool) (int, bool) {
+	j := index[0]
+	ok := false
+	for ; j < index[1]; j++ {
+		c1 := source[j]
+		if pred(c1) {
+			ok = true
+			continue
+		}
+		break
+	}
+	return j, ok
+}
+
+// A Writer interface wirtes textual contents to a writer.
+type Writer interface {
+	// Write writes given source to writer with resolving references and unescaping
+	// backslash escaped characters.
+	Write(writer util.BufWriter, source []byte)
+
+	// RawWrite wirtes given source to writer without resolving references and
+	// unescaping backslash escaped characters.
+	RawWrite(writer util.BufWriter, source []byte)
+}
+
+type defaultWriter struct {
+}
+
+var htmlEscaleTable = [256][]byte{nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, []byte("&quot;"), nil, nil, nil, []byte("&amp;"), nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, []byte("&lt;"), nil, []byte("&gt;"), nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil}
+
+func escapeRune(writer util.BufWriter, r rune) {
+	if r < 256 {
+		v := htmlEscaleTable[byte(r)]
+		if v != nil {
+			writer.Write(v)
+			return
+		}
+	}
+	writer.WriteRune(util.ToValidRune(r))
+}
+
+func (d *defaultWriter) RawWrite(writer util.BufWriter, source []byte) {
+	n := 0
+	l := len(source)
+	for i := 0; i < l; i++ {
+		v := htmlEscaleTable[source[i]]
+		if v != nil {
+			writer.Write(source[i-n : i])
+			n = 0
+			writer.Write(v)
+			continue
+		}
+		n++
+	}
+	if n != 0 {
+		writer.Write(source[l-n:])
+	}
+}
+
+func (d *defaultWriter) Write(writer util.BufWriter, source []byte) {
+	escaped := false
+	ok := false
+	limit := len(source)
+	n := 0
+	for i := 0; i < limit; i++ {
+		c := source[i]
+		if escaped {
+			if util.IsPunct(c) {
+				d.RawWrite(writer, source[n:i-1])
+				n = i
+				escaped = false
+				continue
+			}
+		}
+		if c == '&' {
+			pos := i
+			next := i + 1
+			if next < limit && source[next] == '#' {
+				nnext := next + 1
+				nc := source[nnext]
+				// code point like #x22;
+				if nnext < limit && nc == 'x' || nc == 'X' {
+					start := nnext + 1
+					i, ok = readWhile(source, [2]int{start, limit}, util.IsHexDecimal)
+					if ok && i < limit && source[i] == ';' {
+						v, _ := strconv.ParseUint(util.BytesToReadOnlyString(source[start:i]), 16, 32)
+						d.RawWrite(writer, source[n:pos])
+						n = i + 1
+						escapeRune(writer, rune(v))
+						continue
+					}
+					// code point like #1234;
+				} else if nc >= '0' && nc <= '9' {
+					start := nnext
+					i, ok = readWhile(source, [2]int{start, limit}, util.IsNumeric)
+					if ok && i < limit && i-start < 8 && source[i] == ';' {
+						v, _ := strconv.ParseUint(util.BytesToReadOnlyString(source[start:i]), 0, 32)
+						d.RawWrite(writer, source[n:pos])
+						n = i + 1
+						escapeRune(writer, rune(v))
+						continue
+					}
+				}
+			} else {
+				start := next
+				i, ok = readWhile(source, [2]int{start, limit}, util.IsAlphaNumeric)
+				// entity reference
+				if ok && i < limit && source[i] == ';' {
+					name := util.BytesToReadOnlyString(source[start:i])
+					entity, ok := util.LookUpHTML5EntityByName(name)
+					if ok {
+						d.RawWrite(writer, source[n:pos])
+						n = i + 1
+						d.RawWrite(writer, entity.Characters)
+						continue
+					}
+				}
+			}
+			i = next - 1
+		}
+		if c == '\\' {
+			escaped = true
+			continue
+		}
+		escaped = false
+	}
+	d.RawWrite(writer, source[n:len(source)])
+}
+
+// DefaultWriter is a default implementation of the Writer.
+var DefaultWriter = &defaultWriter{}

renderer/renderer.go

@@ -0,0 +1,161 @@
+// Package renderer	renders given AST to certain formats.
+package renderer
+
+import (
+	"bufio"
+	"io"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/util"
+
+	"sync"
+)
+
+// A Config struct is a data structure that holds configuration of the Renderer.
+type Config struct {
+	Options       map[OptionName]interface{}
+	NodeRenderers util.PrioritizedSlice
+}
+
+// NewConfig returns a new Config
+func NewConfig() *Config {
+	return &Config{
+		Options:       map[OptionName]interface{}{},
+		NodeRenderers: util.PrioritizedSlice{},
+	}
+}
+
+type notSupported struct {
+}
+
+func (e *notSupported) Error() string {
+	return "not supported by this parser"
+}
+
+// NotSupported indicates given node can not be rendered by this NodeRenderer.
+var NotSupported = &notSupported{}
+
+// An OptionName is a name of the option.
+type OptionName string
+
+// An Option interface is a functional option type for the Renderer.
+type Option interface {
+	SetConfig(*Config)
+}
+
+type withNodeRenderers struct {
+	value []util.PrioritizedValue
+}
+
+func (o *withNodeRenderers) SetConfig(c *Config) {
+	c.NodeRenderers = append(c.NodeRenderers, o.value...)
+}
+
+// WithNodeRenderers is a functional option that allow you to add
+// NodeRenderers to the renderer.
+func WithNodeRenderers(ps ...util.PrioritizedValue) Option {
+	return &withNodeRenderers{ps}
+}
+
+type withOption struct {
+	name  OptionName
+	value interface{}
+}
+
+func (o *withOption) SetConfig(c *Config) {
+	c.Options[o.name] = o.value
+}
+
+// WithOption is a functional option that allow you to set
+// an arbitary option to the parser.
+func WithOption(name OptionName, value interface{}) Option {
+	return &withOption{name, value}
+}
+
+// A SetOptioner interface sets given option to the object.
+type SetOptioner interface {
+	// SetOption sets given option to the object.
+	// Unacceptable options may be passed.
+	// Thus implementations must ignore unacceptable options.
+	SetOption(name OptionName, value interface{})
+}
+
+// A NodeRenderer interface renders given AST node to given writer.
+type NodeRenderer interface {
+	// Render renders given AST node to given writer.
+	Render(writer util.BufWriter, source []byte, n ast.Node, entering bool) (ast.WalkStatus, error)
+}
+
+// A Renderer interface renders given AST node to given
+// writer with given Renderer.
+type Renderer interface {
+	Render(w io.Writer, source []byte, n ast.Node) error
+
+	// AddOption adds given option to thie parser.
+	AddOption(Option)
+}
+
+type renderer struct {
+	config        *Config
+	options       map[OptionName]interface{}
+	nodeRenderers []NodeRenderer
+	initSync      sync.Once
+}
+
+// NewRenderer returns a new Renderer with given options.
+func NewRenderer(options ...Option) Renderer {
+	config := NewConfig()
+	for _, opt := range options {
+		opt.SetConfig(config)
+	}
+
+	r := &renderer{
+		options: map[OptionName]interface{}{},
+		config:  config,
+	}
+
+	return r
+}
+
+func (r *renderer) AddOption(o Option) {
+	o.SetConfig(r.config)
+}
+
+// Render renders given AST node to given writer with given Renderer.
+func (r *renderer) Render(w io.Writer, source []byte, n ast.Node) error {
+	r.initSync.Do(func() {
+		r.options = r.config.Options
+		r.config.NodeRenderers.Sort()
+		r.nodeRenderers = make([]NodeRenderer, 0, len(r.config.NodeRenderers))
+		for _, v := range r.config.NodeRenderers {
+			nr, _ := v.Value.(NodeRenderer)
+			if se, ok := v.Value.(SetOptioner); ok {
+				for oname, ovalue := range r.options {
+					se.SetOption(oname, ovalue)
+				}
+			}
+			r.nodeRenderers = append(r.nodeRenderers, nr)
+		}
+		r.config = nil
+	})
+	writer, ok := w.(util.BufWriter)
+	if !ok {
+		writer = bufio.NewWriter(w)
+	}
+	err := ast.Walk(n, func(n ast.Node, entering bool) (ast.WalkStatus, error) {
+		var s ast.WalkStatus
+		var err error
+		for _, nr := range r.nodeRenderers {
+			s, err = nr.Render(writer, source, n, entering)
+			if err == NotSupported {
+				continue
+			}
+			break
+		}
+		return s, err
+	})
+	if err != nil {
+		return err
+	}
+	return writer.Flush()
+}

text/reader.go

@@ -0,0 +1,492 @@
+package text
+
+import (
+	"github.com/yuin/goldmark/util"
+	"io"
+	"regexp"
+	"unicode/utf8"
+)
+
+const invalidValue = -1
+
+// EOF indicates the end of file.
+const EOF = byte(0xff)
+
+// A Reader interface provides abstracted method for reading text.
+type Reader interface {
+	io.RuneReader
+
+	// Source returns a source of the reader.
+	Source() []byte
+
+	// Peek returns a byte at current position without advancing the internal pointer.
+	Peek() byte
+
+	// PeekLine returns the current line without advancing the internal pointer.
+	PeekLine() ([]byte, Segment)
+
+	// PrecendingCharacter returns a character just before current internal pointer.
+	PrecendingCharacter() rune
+
+	// Value returns a value of given segment.
+	Value(Segment) []byte
+
+	// LineOffset returns a distance from the line head to current position.
+	LineOffset() int
+
+	// Position returns current line number and position.
+	Position() (int, Segment)
+
+	// SetPosition sets current line number and position.
+	SetPosition(int, Segment)
+
+	// SetPadding sets padding to the reader.
+	SetPadding(int)
+
+	// Advance advances the internal pointer.
+	Advance(int)
+
+	// AdvanceAndSetPadding advances the internal pointer and add padding to the
+	// reader.
+	AdvanceAndSetPadding(int, int)
+
+	// AdvanceLine advances the internal pointer to the next line head.
+	AdvanceLine()
+
+	// SkipSpaces skips space characters and returns a non-blank line.
+	// If it reaches EOF, returns false.
+	SkipSpaces() (Segment, int, bool)
+
+	// SkipSpaces skips blank lines and returns a non-blank line.
+	// If it reaches EOF, returns false.
+	SkipBlankLines() (Segment, int, bool)
+
+	// Match performs regular expression matching to current line.
+	Match(reg *regexp.Regexp) bool
+
+	// Match performs regular expression searching to current line.
+	FindSubMatch(reg *regexp.Regexp) [][]byte
+}
+
+type reader struct {
+	source       []byte
+	sourceLength int
+	line         int
+	peekedLine   []byte
+	pos          Segment
+	head         int
+}
+
+// NewReader return a new Reader that can read UTF-8 bytes .
+func NewReader(source []byte) Reader {
+	r := &reader{
+		source:       source,
+		sourceLength: len(source),
+		line:         -1,
+		head:         0,
+	}
+	r.AdvanceLine()
+	return r
+}
+
+func (r *reader) Source() []byte {
+	return r.source
+}
+
+func (r *reader) Value(seg Segment) []byte {
+	return seg.Value(r.source)
+}
+
+func (r *reader) Peek() byte {
+	if r.pos.Start >= 0 && r.pos.Start < r.sourceLength {
+		if r.pos.Padding != 0 {
+			return space[0]
+		}
+		return r.source[r.pos.Start]
+	}
+	return EOF
+}
+
+func (r *reader) PeekLine() ([]byte, Segment) {
+	if r.pos.Start >= 0 && r.pos.Start < r.sourceLength {
+		if r.peekedLine == nil {
+			r.peekedLine = r.pos.Value(r.Source())
+		}
+		return r.peekedLine, r.pos
+	}
+	return nil, r.pos
+}
+
+// io.RuneReader interface
+func (r *reader) ReadRune() (rune, int, error) {
+	return readRuneReader(r)
+}
+
+func (r *reader) LineOffset() int {
+	v := r.pos.Start - r.head
+	if r.pos.Padding > 0 {
+		v += util.TabWidth(v) - r.pos.Padding
+	}
+	return v
+}
+
+func (r *reader) PrecendingCharacter() rune {
+	if r.pos.Start <= 0 {
+		if r.pos.Padding != 0 {
+			return rune(' ')
+		}
+		return rune('\n')
+	}
+	i := r.pos.Start - 1
+	for ; i >= 0; i-- {
+		if utf8.RuneStart(r.source[i]) {
+			break
+		}
+	}
+	rn, _ := utf8.DecodeRune(r.source[i:])
+	return rn
+}
+
+func (r *reader) Advance(n int) {
+	if n < len(r.peekedLine) && r.pos.Padding == 0 {
+		r.pos.Start += n
+		r.peekedLine = nil
+		return
+	}
+	r.peekedLine = nil
+	l := r.sourceLength
+	for ; n > 0 && r.pos.Start < l; n-- {
+		if r.pos.Padding != 0 {
+			r.pos.Padding--
+			continue
+		}
+		if r.source[r.pos.Start] == '\n' {
+			r.AdvanceLine()
+			continue
+		}
+		r.pos.Start++
+	}
+}
+
+func (r *reader) AdvanceAndSetPadding(n, padding int) {
+	r.Advance(n)
+	if padding > r.pos.Padding {
+		r.SetPadding(padding)
+	}
+}
+
+func (r *reader) AdvanceLine() {
+	r.peekedLine = nil
+	r.pos.Start = r.pos.Stop
+	r.head = r.pos.Start
+	if r.pos.Start < 0 {
+		return
+	}
+	r.pos.Stop = invalidValue
+	for i := r.pos.Start; i < r.sourceLength; i++ {
+		c := r.source[i]
+		if c == '\n' {
+			r.pos.Stop = i + 1
+			break
+		}
+	}
+	r.line++
+	r.pos.Padding = 0
+}
+
+func (r *reader) Position() (int, Segment) {
+	return r.line, r.pos
+}
+
+func (r *reader) SetPosition(line int, pos Segment) {
+	r.line = line
+	r.pos = pos
+}
+
+func (r *reader) SetPadding(v int) {
+	r.pos.Padding = v
+}
+
+func (r *reader) SkipSpaces() (Segment, int, bool) {
+	return skipSpacesReader(r)
+}
+
+func (r *reader) SkipBlankLines() (Segment, int, bool) {
+	return skipBlankLinesReader(r)
+}
+
+func (r *reader) Match(reg *regexp.Regexp) bool {
+	return matchReader(r, reg)
+}
+
+func (r *reader) FindSubMatch(reg *regexp.Regexp) [][]byte {
+	return findSubMatchReader(r, reg)
+}
+
+// A BlockReader interface is a reader that is optimized for Blocks.
+type BlockReader interface {
+	Reader
+	Reset(segment *Segments)
+}
+
+type blockReader struct {
+	source         []byte
+	segments       *Segments
+	segmentsLength int
+	line           int
+	pos            Segment
+	head           int
+	last           int
+}
+
+// NewBlockReader returns a new BlockReader.
+func NewBlockReader(source []byte, segments *Segments) BlockReader {
+	r := &blockReader{
+		source: source,
+	}
+	if segments != nil {
+		r.Reset(segments)
+	}
+	return r
+}
+
+// Reset resets current state and sets new segments to the reader.
+func (r *blockReader) Reset(segments *Segments) {
+	r.segments = segments
+	r.segmentsLength = segments.Len()
+	r.line = -1
+	r.head = 0
+	r.last = 0
+	r.pos.Start = -1
+	r.pos.Stop = -1
+	r.pos.Padding = 0
+	if r.segmentsLength > 0 {
+		last := r.segments.At(r.segmentsLength - 1)
+		r.last = last.Stop
+	}
+	r.AdvanceLine()
+}
+
+func (r *blockReader) Source() []byte {
+	return r.source
+}
+
+func (r *blockReader) Value(seg Segment) []byte {
+	line := r.segmentsLength - 1
+	ret := make([]byte, 0, seg.Stop-seg.Start+1)
+	for ; line >= 0; line-- {
+		if seg.Start >= r.segments.At(line).Start {
+			break
+		}
+	}
+	i := seg.Start
+	for ; line < r.segmentsLength; line++ {
+		s := r.segments.At(line)
+		if i < 0 {
+			i = s.Start
+		}
+		ret = s.ConcatPadding(ret)
+		for ; i < seg.Stop && i < s.Stop; i++ {
+			ret = append(ret, r.source[i])
+		}
+		i = -1
+		if s.Stop > seg.Stop {
+			break
+		}
+	}
+	return ret
+}
+
+// io.RuneReader interface
+func (r *blockReader) ReadRune() (rune, int, error) {
+	return readRuneReader(r)
+}
+
+func (r *blockReader) PrecendingCharacter() rune {
+	if r.pos.Padding != 0 {
+		return rune(' ')
+	}
+	if r.pos.Start <= 0 {
+		return rune('\n')
+	}
+	i := r.pos.Start - 1
+	for ; i >= 0; i-- {
+		if utf8.RuneStart(r.source[i]) {
+			break
+		}
+	}
+	rn, _ := utf8.DecodeRune(r.source[i:])
+	return rn
+}
+
+func (r *blockReader) LineOffset() int {
+	v := r.pos.Start - r.head
+	if r.pos.Padding > 0 {
+		v += util.TabWidth(v) - r.pos.Padding
+	}
+	return v
+}
+
+func (r *blockReader) Peek() byte {
+	if r.line < r.segmentsLength && r.pos.Start >= 0 && r.pos.Start < r.last {
+		if r.pos.Padding != 0 {
+			return space[0]
+		}
+		return r.source[r.pos.Start]
+	}
+	return EOF
+}
+
+func (r *blockReader) PeekLine() ([]byte, Segment) {
+	if r.line < r.segmentsLength && r.pos.Start >= 0 && r.pos.Start < r.last {
+		return r.pos.Value(r.source), r.pos
+	}
+	return nil, r.pos
+}
+
+func (r *blockReader) Advance(n int) {
+	if n < r.pos.Stop-r.pos.Start && r.pos.Padding == 0 {
+		r.pos.Start += n
+		return
+	}
+
+	for ; n > 0; n-- {
+		if r.pos.Padding != 0 {
+			r.pos.Padding--
+			continue
+		}
+		if r.pos.Start >= r.pos.Stop-1 && r.pos.Stop < r.last {
+			r.AdvanceLine()
+			continue
+		}
+		r.pos.Start++
+	}
+}
+
+func (r *blockReader) AdvanceAndSetPadding(n, padding int) {
+	r.Advance(n)
+	if padding > r.pos.Padding {
+		r.SetPadding(padding)
+	}
+}
+
+func (r *blockReader) AdvanceLine() {
+	r.SetPosition(r.line+1, NewSegment(invalidValue, invalidValue))
+	r.head = r.pos.Start
+}
+
+func (r *blockReader) Position() (int, Segment) {
+	return r.line, r.pos
+}
+
+func (r *blockReader) SetPosition(line int, pos Segment) {
+	r.line = line
+	if pos.Start == invalidValue {
+		if r.line < r.segmentsLength {
+			r.pos = r.segments.At(line)
+		}
+	} else {
+		r.pos = pos
+	}
+}
+
+func (r *blockReader) SetPadding(v int) {
+	r.pos.Padding = v
+}
+
+func (r *blockReader) SkipSpaces() (Segment, int, bool) {
+	return skipSpacesReader(r)
+}
+
+func (r *blockReader) SkipBlankLines() (Segment, int, bool) {
+	return skipBlankLinesReader(r)
+}
+
+func (r *blockReader) Match(reg *regexp.Regexp) bool {
+	return matchReader(r, reg)
+}
+
+func (r *blockReader) FindSubMatch(reg *regexp.Regexp) [][]byte {
+	return findSubMatchReader(r, reg)
+}
+
+func skipBlankLinesReader(r Reader) (Segment, int, bool) {
+	lines := 0
+	for {
+		line, seg := r.PeekLine()
+		if line == nil {
+			return seg, lines, false
+		}
+		if util.IsBlank(line) {
+			lines++
+			r.AdvanceLine()
+		} else {
+			return seg, lines, true
+		}
+	}
+}
+
+func skipSpacesReader(r Reader) (Segment, int, bool) {
+	chars := 0
+	for {
+		line, segment := r.PeekLine()
+		if line == nil {
+			return segment, chars, false
+		}
+		for i, c := range line {
+			if util.IsSpace(c) {
+				chars++
+				r.Advance(1)
+				continue
+			}
+			return segment.WithStart(segment.Start + i + 1), chars, true
+		}
+	}
+}
+
+func matchReader(r Reader, reg *regexp.Regexp) bool {
+	oldline, oldseg := r.Position()
+	match := reg.FindReaderSubmatchIndex(r)
+	r.SetPosition(oldline, oldseg)
+	if match == nil {
+		return false
+	}
+	r.Advance(match[1] - match[0])
+	return true
+}
+
+func findSubMatchReader(r Reader, reg *regexp.Regexp) [][]byte {
+	oldline, oldseg := r.Position()
+	match := reg.FindReaderSubmatchIndex(r)
+	r.SetPosition(oldline, oldseg)
+	if match == nil {
+		return nil
+	}
+	runes := make([]rune, 0, match[1]-match[0])
+	for i := 0; i < match[1]; {
+		r, size, _ := readRuneReader(r)
+		i += size
+		runes = append(runes, r)
+	}
+	result := [][]byte{}
+	for i := 0; i < len(match); i += 2 {
+		result = append(result, []byte(string(runes[match[i]:match[i+1]])))
+	}
+
+	r.SetPosition(oldline, oldseg)
+	r.Advance(match[1] - match[0])
+	return result
+}
+
+func readRuneReader(r Reader) (rune, int, error) {
+	line, _ := r.PeekLine()
+	if line == nil {
+		return 0, 0, io.EOF
+	}
+	rn, size := utf8.DecodeRune(line)
+	if rn == utf8.RuneError {
+		return 0, 0, io.EOF
+	}
+	r.Advance(size)
+	return rn, size, nil
+}

text/segment.go

@@ -0,0 +1,209 @@
+package text
+
+import (
+	"bytes"
+	"github.com/yuin/goldmark/util"
+)
+
+var space = []byte(" ")
+
+// A Segment struct holds information about source potisions.
+type Segment struct {
+	// Start is a start position of the segment.
+	Start int
+
+	// Stop is a stop position of the segment.
+	// This value should be excluded.
+	Stop int
+
+	// Padding is a padding length of the segment.
+	Padding int
+}
+
+// NewSegment return a new Segment.
+func NewSegment(start, stop int) Segment {
+	return Segment{
+		Start:   start,
+		Stop:    stop,
+		Padding: 0,
+	}
+}
+
+// NewSegmentPadding returns a new Segment with given padding.
+func NewSegmentPadding(start, stop, n int) Segment {
+	return Segment{
+		Start:   start,
+		Stop:    stop,
+		Padding: n,
+	}
+}
+
+// Value returns a value of the segment.
+func (t *Segment) Value(buffer []byte) []byte {
+	if t.Padding == 0 {
+		return buffer[t.Start:t.Stop]
+	}
+	result := make([]byte, 0, t.Padding+t.Stop-t.Start+1)
+	result = append(result, bytes.Repeat(space, t.Padding)...)
+	return append(result, buffer[t.Start:t.Stop]...)
+}
+
+// Len returns a length of the segment.
+func (t *Segment) Len() int {
+	return t.Stop - t.Start + t.Padding
+}
+
+// Between returns a segment between this segment and given segment.
+func (t *Segment) Between(other Segment) Segment {
+	if t.Stop != other.Stop {
+		panic("invalid state")
+	}
+	return NewSegmentPadding(
+		t.Start,
+		other.Start,
+		t.Padding-other.Padding,
+	)
+}
+
+// IsEmpty returns true if this segment is empty, otherwise false.
+func (t *Segment) IsEmpty() bool {
+	return t.Start >= t.Stop && t.Padding == 0
+}
+
+// TrimRightSpace returns a new segment by slicing off all trailing
+// space characters.
+func (t *Segment) TrimRightSpace(buffer []byte) Segment {
+	v := buffer[t.Start:t.Stop]
+	l := util.TrimRightSpaceLength(v)
+	if l == len(v) {
+		return NewSegment(t.Start, t.Start)
+	}
+	return NewSegmentPadding(t.Start, t.Stop-l, t.Padding)
+}
+
+// TrimLeftSpace returns a new segment by slicing off all leading
+// space characters including padding.
+func (t *Segment) TrimLeftSpace(buffer []byte) Segment {
+	v := buffer[t.Start:t.Stop]
+	l := util.TrimLeftSpaceLength(v)
+	return NewSegment(t.Start+l, t.Stop)
+}
+
+// TrimLeftSpaceWidth returns a new segment by slicing off leading space
+// characters until given width.
+func (t *Segment) TrimLeftSpaceWidth(width int, buffer []byte) Segment {
+	padding := t.Padding
+	for ; width > 0; width-- {
+		if padding == 0 {
+			break
+		}
+		padding--
+	}
+	if width == 0 {
+		return NewSegmentPadding(t.Start, t.Stop, padding)
+	}
+	text := buffer[t.Start:t.Stop]
+	start := t.Start
+	for _, c := range text {
+		if start >= t.Stop-1 || width <= 0 {
+			break
+		}
+		if c == ' ' {
+			width--
+		} else if c == '\t' {
+			width -= 4
+		} else {
+			break
+		}
+		start++
+	}
+	if width < 0 {
+		padding = width * -1
+	}
+	return NewSegmentPadding(start, t.Stop, padding)
+}
+
+// WithStart returns a new Segment with same value except Start.
+func (t *Segment) WithStart(v int) Segment {
+	return NewSegmentPadding(v, t.Stop, t.Padding)
+}
+
+// WithStop returns a new Segment with same value except Stop.
+func (t *Segment) WithStop(v int) Segment {
+	return NewSegmentPadding(t.Start, v, t.Padding)
+}
+
+// ConcatPadding concats the padding to given slice.
+func (t *Segment) ConcatPadding(v []byte) []byte {
+	if t.Padding > 0 {
+		return append(v, bytes.Repeat(space, t.Padding)...)
+	}
+	return v
+}
+
+// Segments is a collection of the Segment.
+type Segments struct {
+	values []Segment
+}
+
+// NewSegments return a new Segments.
+func NewSegments() *Segments {
+	return &Segments{
+		values: nil,
+	}
+}
+
+// Append appends given segment after the tail of the collection.
+func (s *Segments) Append(t Segment) {
+	if s.values == nil {
+		s.values = make([]Segment, 0, 20)
+	}
+	s.values = append(s.values, t)
+}
+
+// AppendAll appends all elements of given segments after the tail of the collection.
+func (s *Segments) AppendAll(t []Segment) {
+	if s.values == nil {
+		s.values = make([]Segment, 0, 20)
+	}
+	s.values = append(s.values, t...)
+}
+
+// Len returns the length of the collection.
+func (s *Segments) Len() int {
+	if s.values == nil {
+		return 0
+	}
+	return len(s.values)
+}
+
+// At returns a segment at given index.
+func (s *Segments) At(i int) Segment {
+	return s.values[i]
+}
+
+// Set sets given Segment.
+func (s *Segments) Set(i int, v Segment) {
+	s.values[i] = v
+}
+
+// SetSliced replace the collection with a subsliced value.
+func (s *Segments) SetSliced(lo, hi int) {
+	s.values = s.values[lo:hi]
+}
+
+// Sliced returns a subslice of the collection.
+func (s *Segments) Sliced(lo, hi int) []Segment {
+	return s.values[lo:hi]
+}
+
+// Clear delete all element of the collction.
+func (s *Segments) Clear() {
+	s.values = nil
+}
+
+// Unshift insert given Segment to head of the collection.
+func (s *Segments) Unshift(v Segment) {
+	s.values = append(s.values[0:1], s.values[0:]...)
+	s.values[0] = v
+}

util/util.go

@@ -0,0 +1,538 @@
+// Package util provides utility functions for the goldmark.
+package util
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"net/url"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+	"unicode/utf8"
+)
+
+// IsBlank returns true if given string is all space characters.
+func IsBlank(bs []byte) bool {
+	for _, b := range bs {
+		if IsSpace(b) {
+			continue
+		}
+		return false
+	}
+	return true
+}
+
+// DedentPosition dedents lines by given width.
+func DedentPosition(bs []byte, width int) (pos, padding int) {
+	i := 0
+	l := len(bs)
+	w := 0
+	for ; i < l && w < width; i++ {
+		b := bs[i]
+		if b == ' ' {
+			w++
+		} else if b == '\t' {
+			w += 4
+		} else {
+			break
+		}
+	}
+	padding = w - width
+	if padding < 0 {
+		padding = 0
+	}
+	return i, padding
+}
+
+// VisualizeSpaces visualize invisible space characters.
+func VisualizeSpaces(bs []byte) []byte {
+	bs = bytes.Replace(bs, []byte(" "), []byte("[SPACE]"), -1)
+	bs = bytes.Replace(bs, []byte("\t"), []byte("[TAB]"), -1)
+	bs = bytes.Replace(bs, []byte("\n"), []byte("[NEWLINE]\n"), -1)
+	return bs
+}
+
+// TabWidth calculates actual width of a tab at given position.
+func TabWidth(currentPos int) int {
+	return 4 - currentPos%4
+}
+
+// IndentPosition searches an indent position with given width for given line.
+// If the line contains tab characters, paddings may be not zero.
+// currentPos==0 and width==2:
+//
+//     position: 0    1
+//               [TAB]aaaa
+//     width:    1234 5678
+//
+// width=2 is in the tab character. In this case, IndentPosition returns
+// (pos=1, padding=2)
+func IndentPosition(bs []byte, currentPos, width int) (pos, padding int) {
+	w := 0
+	l := len(bs)
+	for i := 0; i < l; i++ {
+		b := bs[i]
+		if b == ' ' {
+			w++
+		} else if b == '\t' {
+			w += TabWidth(currentPos + w)
+		} else {
+			break
+		}
+		if w >= width {
+			return i + 1, w - width
+		}
+	}
+	return -1, -1
+}
+
+// IndentWidth calculate an indent width for given line.
+func IndentWidth(bs []byte, currentPos int) (width, pos int) {
+	l := len(bs)
+	for i := 0; i < l; i++ {
+		b := bs[i]
+		if b == ' ' {
+			width++
+			pos++
+		} else if b == '\t' {
+			width += TabWidth(currentPos + width)
+			pos++
+		} else {
+			break
+		}
+	}
+	return
+}
+
+// FirstNonSpacePosition returns a potisoin line that is a first nonspace
+// character.
+func FirstNonSpacePosition(bs []byte) int {
+	i := 0
+	for ; i < len(bs); i++ {
+		c := bs[i]
+		if c == ' ' || c == '\t' {
+			continue
+		}
+		if c == '\n' {
+			return -1
+		}
+		return i
+	}
+	return -1
+}
+
+// FindClosure returns a position that closes given opener.
+// If codeSpan is set true, it ignores characters in code spans.
+// If allowNesting is set true, closures correspond to nested opener will be
+// ignored.
+func FindClosure(bs []byte, opener, closure byte, codeSpan, allowNesting bool) int {
+	i := 0
+	opened := 1
+	codeSpanOpener := 0
+	for i < len(bs) {
+		c := bs[i]
+		if codeSpan && codeSpanOpener != 0 && c == '`' {
+			codeSpanCloser := 0
+			for ; i < len(bs); i++ {
+				if bs[i] == '`' {
+					codeSpanCloser++
+				} else {
+					break
+				}
+			}
+			if codeSpanCloser == codeSpanOpener {
+				codeSpanOpener = 0
+			}
+		} else if c == '\\' && i < len(bs)-1 && IsPunct(bs[i+1]) {
+			i += 2
+			continue
+		} else if codeSpan && codeSpanOpener == 0 && c == '`' {
+			for ; i < len(bs); i++ {
+				if bs[i] == '`' {
+					codeSpanOpener++
+				} else {
+					break
+				}
+			}
+		} else if (codeSpan && codeSpanOpener == 0) || !codeSpan {
+			if c == closure {
+				opened--
+				if opened == 0 {
+					return i
+				}
+			} else if c == opener {
+				if !allowNesting {
+					return -1
+				}
+				opened++
+			}
+		}
+		i++
+	}
+	return -1
+}
+
+// TrimLeft trims characters in given s from head of the source.
+// bytes.TrimLeft offers same functionalities, but bytes.TrimLeft
+// allocates new buffer for the result.
+func TrimLeft(source, b []byte) []byte {
+	i := 0
+	for ; i < len(source); i++ {
+		c := source[i]
+		found := false
+		for j := 0; j < len(b); j++ {
+			if c == b[j] {
+				found = true
+				break
+			}
+		}
+		if !found {
+			break
+		}
+	}
+	return source[i:]
+}
+
+// TrimRight trims characters in given s from tail of the source.
+func TrimRight(source, b []byte) []byte {
+	i := len(source) - 1
+	for ; i >= 0; i-- {
+		c := source[i]
+		found := false
+		for j := 0; j < len(b); j++ {
+			if c == b[j] {
+				found = true
+				break
+			}
+		}
+		if !found {
+			break
+		}
+	}
+	return source[:i+1]
+}
+
+// TrimLeftLength returns a length of leading specified characters.
+func TrimLeftLength(source, s []byte) int {
+	return len(source) - len(TrimLeft(source, s))
+}
+
+// TrimRightLength returns a length of trailing specified characters.
+func TrimRightLength(source, s []byte) int {
+	return len(source) - len(TrimRight(source, s))
+}
+
+// TrimLeftSpaceLength returns a length of leading space characters.
+func TrimLeftSpaceLength(source []byte) int {
+	return TrimLeftLength(source, spaces)
+}
+
+// TrimRightSpaceLength returns a length of trailing space characters.
+func TrimRightSpaceLength(source []byte) int {
+	return TrimRightLength(source, spaces)
+}
+
+// TrimLeftSpace returns a subslice of given string by slicing off all leading
+// space characters.
+func TrimLeftSpace(source []byte) []byte {
+	return TrimLeft(source, spaces)
+}
+
+// TrimRightSpace returns a subslice of given string by slicing off all trailing
+// space characters.
+func TrimRightSpace(source []byte) []byte {
+	return TrimRight(source, spaces)
+}
+
+// ReplaceSpaces replaces sequence of spaces with given repl.
+func ReplaceSpaces(source []byte, repl byte) []byte {
+	var ret []byte
+	start := -1
+	for i, c := range source {
+		iss := IsSpace(c)
+		if start < 0 && iss {
+			start = i
+			continue
+		} else if start >= 0 && iss {
+			continue
+		} else if start >= 0 {
+			if ret == nil {
+				ret = make([]byte, 0, len(source))
+				ret = append(ret, source[:start]...)
+			}
+			ret = append(ret, repl)
+			start = -1
+		}
+		if ret != nil {
+			ret = append(ret, c)
+		}
+	}
+	if start >= 0 && ret != nil {
+		ret = append(ret, repl)
+	}
+	if ret == nil {
+		return source
+	}
+	return ret
+}
+
+// ToRune decode given bytes start at pos and returns a rune.
+func ToRune(source []byte, pos int) rune {
+	i := pos
+	for ; i >= 0; i-- {
+		if utf8.RuneStart(source[i]) {
+			break
+		}
+	}
+	r, _ := utf8.DecodeRune(source[i:])
+	return r
+}
+
+// ToValidRune returns 0xFFFD if given rune is invalid, otherwise v.
+func ToValidRune(v rune) rune {
+	if v == 0 || !utf8.ValidRune(v) {
+		return rune(0xFFFD)
+	}
+	return v
+}
+
+// ToLinkReference convert given bytes into a valid link reference string.
+// ToLinkReference trims leading and trailing spaces and convert into lower
+// case and replace spaces with a single space character.
+func ToLinkReference(v []byte) string {
+	v = TrimLeftSpace(v)
+	v = TrimRightSpace(v)
+	return strings.ToLower(string(ReplaceSpaces(v, ' ')))
+}
+
+var escapeRegex = regexp.MustCompile(`\\.`)
+var hexRefRegex = regexp.MustCompile(`#[xX][\da-fA-F]+;`)
+var numRefRegex = regexp.MustCompile(`#\d{1,7};`)
+var entityRefRegex = regexp.MustCompile(`&([a-zA-Z\d]+);`)
+
+var entityLt = []byte("&lt;")
+var entityGt = []byte("&gt;")
+var entityAmp = []byte("&amp;")
+var entityQuot = []byte("&quot;")
+
+// EscapeHTML escapes characters that should be escaped in HTML text.
+func EscapeHTML(v []byte) []byte {
+	result := make([]byte, 0, len(v)+10)
+	for _, c := range v {
+		switch c {
+		case '<':
+			result = append(result, entityLt...)
+		case '>':
+			result = append(result, entityGt...)
+		case '&':
+			result = append(result, entityAmp...)
+		case '"':
+			result = append(result, entityQuot...)
+		default:
+			result = append(result, c)
+		}
+	}
+	return result
+}
+
+// UnescapePunctuations unescapes blackslash escaped punctuations.
+func UnescapePunctuations(v []byte) []byte {
+	return escapeRegex.ReplaceAllFunc(v, func(match []byte) []byte {
+		if IsPunct(match[1]) {
+			return []byte{match[1]}
+		}
+		return match
+	})
+}
+
+// ResolveNumericReferences resolve numeric references like '&#1234;" .
+func ResolveNumericReferences(v []byte) []byte {
+	buf := make([]byte, 6, 6)
+	v = hexRefRegex.ReplaceAllFunc(v, func(match []byte) []byte {
+		v, _ := strconv.ParseUint(string(match[2:len(match)-1]), 16, 32)
+		n := utf8.EncodeRune(buf, ToValidRune(rune(v)))
+		return buf[:n]
+	})
+	return numRefRegex.ReplaceAllFunc(v, func(match []byte) []byte {
+		v, _ := strconv.ParseUint(string(match[1:len(match)-1]), 0, 32)
+		n := utf8.EncodeRune(buf, ToValidRune(rune(v)))
+		return buf[:n]
+	})
+}
+
+// ResolveEntityNames resolve entity references like '&ouml;" .
+func ResolveEntityNames(v []byte) []byte {
+	return entityRefRegex.ReplaceAllFunc(v, func(match []byte) []byte {
+		entity, ok := LookUpHTML5EntityByName(string(match[1 : len(match)-1]))
+		if ok {
+			return entity.Characters
+		}
+		return match
+	})
+}
+
+// URLEscape escape given URL.
+// If resolveReference is set true:
+//   1. unescape punctuations
+//   2. resolve numeric references
+//   3. resolve entity references
+//
+// URL encoded values (%xx) are keeped as is.
+func URLEscape(v []byte, resolveReference bool) []byte {
+	if resolveReference {
+		v = UnescapePunctuations(v)
+		v = ResolveNumericReferences(v)
+		v = ResolveEntityNames(v)
+	}
+	result := make([]byte, 0, len(v)+10)
+	for i := 0; i < len(v); {
+		c := v[i]
+		if urlEscapeTable[c] == 1 {
+			result = append(result, c)
+			i++
+			continue
+		}
+		if c == '%' && i+2 < len(v) && IsHexDecimal(v[i+1]) && IsHexDecimal(v[i+1]) {
+			result = append(result, c, v[i+1], v[i+2])
+			i += 3
+			continue
+		}
+		u8len := utf8lenTable[c]
+		if u8len == 99 { // invalid utf8 leading byte, skip it
+			result = append(result, c)
+			i++
+			continue
+		}
+		if c == ' ' {
+			result = append(result, '%', '2', '0')
+			i++
+			continue
+		}
+		result = append(result, []byte(url.QueryEscape(string(v[i:i+int(u8len)])))...)
+		i += int(u8len)
+	}
+	return result
+}
+
+// GenerateLinkID generates an ID for links.
+func GenerateLinkID(value []byte, exists map[string]bool) []byte {
+	value = TrimLeftSpace(value)
+	value = TrimRightSpace(value)
+	result := []byte{}
+	for i := 0; i < len(value); {
+		v := value[i]
+		l := utf8lenTable[v]
+		i += int(l)
+		if l != 1 {
+			continue
+		}
+		if IsAlphaNumeric(v) {
+			result = append(result, v)
+		} else if v == ' ' {
+			result = append(result, '-')
+		}
+	}
+	if len(result) == 0 {
+		result = []byte("id")
+	}
+	if _, ok := exists[string(result)]; !ok {
+		exists[string(result)] = true
+		return result
+	}
+	for i := 1; ; i++ {
+		newResult := fmt.Sprintf("%s%d", result, i)
+		if _, ok := exists[newResult]; !ok {
+			exists[newResult] = true
+			return []byte(newResult)
+		}
+
+	}
+}
+
+var spaces = []byte(" \t\n\x0b\x0c\x0d")
+
+var spaceTable = [256]int8{0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+var punctTable = [256]int8{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+// a-zA-Z0-9, ;/?:@&=+$,-_.!~*'()#
+var urlEscapeTable = [256]int8{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+var utf8lenTable = [256]int8{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 99, 99, 99, 99, 99, 99, 99, 99}
+
+// IsPunct returns true if given character is a punctuation, otherwise false.
+func IsPunct(c byte) bool {
+	return punctTable[c] == 1
+}
+
+// IsSpace returns true if given character is a space, otherwise false.
+func IsSpace(c byte) bool {
+	return spaceTable[c] == 1
+}
+
+// IsNumeric returns true if given character is a numeric, otherwise false.
+func IsNumeric(c byte) bool {
+	return c >= '0' && c <= '9'
+}
+
+// IsHexDecimal returns true if given character is a hexdecimal, otherwise false.
+func IsHexDecimal(c byte) bool {
+	return c >= '0' && c <= '9' || c >= 'a' && c <= 'f' || c >= 'A' && c <= 'F'
+}
+
+// IsAlphaNumeric returns true if given character is a alphabet or a numeric, otherwise false.
+func IsAlphaNumeric(c byte) bool {
+	return c >= 'a' && c <= 'z' || c >= 'A' && c <= 'Z' || c >= '0' && c <= '9'
+}
+
+// A BufWriter is a subset of the bufio.Writer .
+type BufWriter interface {
+	io.Writer
+	Available() int
+	Buffered() int
+	Flush() error
+	WriteByte(c byte) error
+	WriteRune(r rune) (size int, err error)
+	WriteString(s string) (int, error)
+}
+
+// A PrioritizedValue struct holds pair of an arbitary value and a priority.
+type PrioritizedValue struct {
+	// Value is an arbitary value that you want to prioritize.
+	Value interface{}
+	// Priority is a priority of the value.
+	Priority int
+}
+
+// PrioritizedSlice is a slice of the PrioritizedValues
+type PrioritizedSlice []PrioritizedValue
+
+// Sort sorts the PrioritizedSlice in ascending order.
+func (s PrioritizedSlice) Sort() {
+	sort.Slice(s, func(i, j int) bool {
+		return s[i].Priority < s[j].Priority
+	})
+}
+
+// Remove removes given value from this slice.
+func (s PrioritizedSlice) Remove(v interface{}) PrioritizedSlice {
+	i := 0
+	found := false
+	for ; i < len(s); i++ {
+		if s[i].Value == v {
+			found = true
+			break
+		}
+	}
+	if !found {
+		return s
+	}
+	return append(s[:i], s[i+1:]...)
+}
+
+// Prioritized returns a new PrioritizedValue.
+func Prioritized(v interface{}, priority int) PrioritizedValue {
+	return PrioritizedValue{v, priority}
+}

util/util_safe.go

@@ -0,0 +1,13 @@
+// +build appengine,js
+
+package util
+
+// BytesToReadOnlyString returns a string converted from given bytes.
+func BytesToReadOnlyString(b []byte) string {
+	return string(b)
+}
+
+// StringToReadOnlyBytes returns bytes converted from given string.
+func StringToReadOnlyBytes(s string) []byte {
+	return []byte(s)
+}

util/util_unsafe.go

@@ -0,0 +1,20 @@
+// +build !appengine,!js
+
+package util
+
+import (
+	"reflect"
+	"unsafe"
+)
+
+// BytesToReadOnlyString returns a string converted from given bytes.
+func BytesToReadOnlyString(b []byte) string {
+	return *(*string)(unsafe.Pointer(&b))
+}
+
+// StringToReadOnlyBytes returns bytes converted from given string.
+func StringToReadOnlyBytes(s string) []byte {
+	sh := (*reflect.StringHeader)(unsafe.Pointer(&s))
+	bh := reflect.SliceHeader{Data: sh.Data, Len: sh.Len, Cap: sh.Len}
+	return *(*[]byte)(unsafe.Pointer(&bh))
+}