PLT-3383: image proxy support (#7991)

* image proxy support

* go vet fix, remove mistakenly added coverage file

* fix test compile error

* add validation to config settings and documentation to model functions

* add message_source field to post

utils/markdown/block_quote.go

@@ -0,0 +1,62 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+type BlockQuote struct {
+	blockBase
+	markdown string
+
+	Children []Block
+}
+
+func (b *BlockQuote) Continuation(indentation int, r Range) *continuation {
+	if indentation > 3 {
+		return nil
+	}
+	s := b.markdown[r.Position:r.End]
+	if s == "" || s[0] != '>' {
+		return nil
+	}
+	remaining := Range{r.Position + 1, r.End}
+	indentation, indentationBytes := countIndentation(b.markdown, remaining)
+	if indentation > 0 {
+		indentation--
+	}
+	return &continuation{
+		Indentation: indentation,
+		Remaining:   Range{remaining.Position + indentationBytes, remaining.End},
+	}
+}
+
+func (b *BlockQuote) AddChild(openBlocks []Block) []Block {
+	b.Children = append(b.Children, openBlocks[0])
+	return openBlocks
+}
+
+func blockQuoteStart(markdown string, indent int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	if indent > 3 {
+		return nil
+	}
+	s := markdown[r.Position:r.End]
+	if s == "" || s[0] != '>' {
+		return nil
+	}
+
+	block := &BlockQuote{
+		markdown: markdown,
+	}
+	r.Position++
+	if len(s) > 1 && s[1] == ' ' {
+		r.Position++
+	}
+
+	indent, bytes := countIndentation(markdown, r)
+
+	ret := []Block{block}
+	if descendants := blockStartOrParagraph(markdown, indent, Range{r.Position + bytes, r.End}, nil, nil); descendants != nil {
+		block.Children = append(block.Children, descendants[0])
+		ret = append(ret, descendants...)
+	}
+	return ret
+}

utils/markdown/blocks.go

@@ -0,0 +1,153 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"strings"
+)
+
+type continuation struct {
+	Indentation int
+	Remaining   Range
+}
+
+type Block interface {
+	Continuation(indentation int, r Range) *continuation
+	AddLine(indentation int, r Range) bool
+	Close()
+	AllowsBlockStarts() bool
+	HasTrailingBlankLine() bool
+}
+
+type blockBase struct{}
+
+func (*blockBase) AddLine(indentation int, r Range) bool { return false }
+func (*blockBase) Close()                                {}
+func (*blockBase) AllowsBlockStarts() bool               { return true }
+func (*blockBase) HasTrailingBlankLine() bool            { return false }
+
+type ContainerBlock interface {
+	Block
+	AddChild(openBlocks []Block) []Block
+}
+
+type Range struct {
+	Position int
+	End      int
+}
+
+func closeBlocks(blocks []Block, referenceDefinitions *[]*ReferenceDefinition) {
+	for _, block := range blocks {
+		block.Close()
+		if p, ok := block.(*Paragraph); ok && len(p.ReferenceDefinitions) > 0 {
+			*referenceDefinitions = append(*referenceDefinitions, p.ReferenceDefinitions...)
+		}
+	}
+}
+
+func ParseBlocks(markdown string, lines []Line) (*Document, []*ReferenceDefinition) {
+	document := &Document{}
+	var referenceDefinitions []*ReferenceDefinition
+
+	openBlocks := []Block{document}
+
+	for _, line := range lines {
+		r := line.Range
+		lastMatchIndex := 0
+
+		indentation, indentationBytes := countIndentation(markdown, r)
+		r = Range{r.Position + indentationBytes, r.End}
+
+		for i, block := range openBlocks {
+			if continuation := block.Continuation(indentation, r); continuation != nil {
+				indentation = continuation.Indentation
+				r = continuation.Remaining
+				additionalIndentation, additionalIndentationBytes := countIndentation(markdown, r)
+				r = Range{r.Position + additionalIndentationBytes, r.End}
+				indentation += additionalIndentation
+				lastMatchIndex = i
+			} else {
+				break
+			}
+		}
+
+		if openBlocks[lastMatchIndex].AllowsBlockStarts() {
+			if newBlocks := blockStart(markdown, indentation, r, openBlocks[:lastMatchIndex+1], openBlocks[lastMatchIndex+1:]); newBlocks != nil {
+				didAdd := false
+				for i := lastMatchIndex; i >= 0; i-- {
+					if container, ok := openBlocks[i].(ContainerBlock); ok {
+						if newBlocks := container.AddChild(newBlocks); newBlocks != nil {
+							closeBlocks(openBlocks[i+1:], &referenceDefinitions)
+							openBlocks = openBlocks[:i+1]
+							openBlocks = append(openBlocks, newBlocks...)
+							didAdd = true
+							break
+						}
+					}
+				}
+				if didAdd {
+					continue
+				}
+			}
+		}
+
+		isBlank := strings.TrimSpace(markdown[r.Position:r.End]) == ""
+		if paragraph, ok := openBlocks[len(openBlocks)-1].(*Paragraph); ok && !isBlank {
+			paragraph.Text = append(paragraph.Text, r)
+			continue
+		}
+
+		closeBlocks(openBlocks[lastMatchIndex+1:], &referenceDefinitions)
+		openBlocks = openBlocks[:lastMatchIndex+1]
+
+		if openBlocks[lastMatchIndex].AddLine(indentation, r) {
+			continue
+		}
+
+		if paragraph := newParagraph(markdown, r); paragraph != nil {
+			for i := lastMatchIndex; i >= 0; i-- {
+				if container, ok := openBlocks[i].(ContainerBlock); ok {
+					if newBlocks := container.AddChild([]Block{paragraph}); newBlocks != nil {
+						closeBlocks(openBlocks[i+1:], &referenceDefinitions)
+						openBlocks = openBlocks[:i+1]
+						openBlocks = append(openBlocks, newBlocks...)
+						break
+					}
+				}
+			}
+		}
+	}
+
+	closeBlocks(openBlocks, &referenceDefinitions)
+
+	return document, referenceDefinitions
+}
+
+func blockStart(markdown string, indentation int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	if r.Position >= r.End {
+		return nil
+	}
+
+	if start := blockQuoteStart(markdown, indentation, r, matchedBlocks, unmatchedBlocks); start != nil {
+		return start
+	} else if start := listStart(markdown, indentation, r, matchedBlocks, unmatchedBlocks); start != nil {
+		return start
+	} else if start := indentedCodeStart(markdown, indentation, r, matchedBlocks, unmatchedBlocks); start != nil {
+		return start
+	} else if start := fencedCodeStart(markdown, indentation, r, matchedBlocks, unmatchedBlocks); start != nil {
+		return start
+	}
+
+	return nil
+}
+
+func blockStartOrParagraph(markdown string, indentation int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	if start := blockStart(markdown, indentation, r, matchedBlocks, unmatchedBlocks); start != nil {
+		return start
+	}
+	if paragraph := newParagraph(markdown, r); paragraph != nil {
+		return []Block{paragraph}
+	}
+	return nil
+}

utils/markdown/document.go

@@ -0,0 +1,22 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+type Document struct {
+	blockBase
+
+	Children []Block
+}
+
+func (b *Document) Continuation(indentation int, r Range) *continuation {
+	return &continuation{
+		Indentation: indentation,
+		Remaining:   r,
+	}
+}
+
+func (b *Document) AddChild(openBlocks []Block) []Block {
+	b.Children = append(b.Children, openBlocks[0])
+	return openBlocks
+}

utils/markdown/fenced_code.go

@@ -0,0 +1,112 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"strings"
+)
+
+type FencedCodeLine struct {
+	Indentation int
+	Range       Range
+}
+
+type FencedCode struct {
+	blockBase
+	markdown           string
+	didSeeClosingFence bool
+
+	Indentation  int
+	OpeningFence Range
+	RawInfo      Range
+	RawCode      []FencedCodeLine
+}
+
+func (b *FencedCode) Code() (result string) {
+	for _, code := range b.RawCode {
+		result += strings.Repeat(" ", code.Indentation) + b.markdown[code.Range.Position:code.Range.End]
+	}
+	return
+}
+
+func (b *FencedCode) Info() string {
+	return Unescape(b.markdown[b.RawInfo.Position:b.RawInfo.End])
+}
+
+func (b *FencedCode) Continuation(indentation int, r Range) *continuation {
+	if b.didSeeClosingFence {
+		return nil
+	}
+	return &continuation{
+		Indentation: indentation,
+		Remaining:   r,
+	}
+}
+
+func (b *FencedCode) AddLine(indentation int, r Range) bool {
+	s := b.markdown[r.Position:r.End]
+	if indentation <= 3 && strings.HasPrefix(s, b.markdown[b.OpeningFence.Position:b.OpeningFence.End]) {
+		suffix := strings.TrimSpace(s[b.OpeningFence.End-b.OpeningFence.Position:])
+		isClosingFence := true
+		for _, c := range suffix {
+			if c != rune(s[0]) {
+				isClosingFence = false
+				break
+			}
+		}
+		if isClosingFence {
+			b.didSeeClosingFence = true
+			return true
+		}
+	}
+
+	if indentation >= b.Indentation {
+		indentation -= b.Indentation
+	} else {
+		indentation = 0
+	}
+
+	b.RawCode = append(b.RawCode, FencedCodeLine{
+		Indentation: indentation,
+		Range:       r,
+	})
+	return true
+}
+
+func (b *FencedCode) AllowsBlockStarts() bool {
+	return false
+}
+
+func fencedCodeStart(markdown string, indentation int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	s := markdown[r.Position:r.End]
+
+	if !strings.HasPrefix(s, "```") && !strings.HasPrefix(s, "~~~") {
+		return nil
+	}
+
+	fenceCharacter := rune(s[0])
+	fenceLength := 3
+	for _, c := range s[3:] {
+		if c == fenceCharacter {
+			fenceLength++
+		} else {
+			break
+		}
+	}
+
+	for i := r.Position + fenceLength; i < r.End; i++ {
+		if markdown[i] == '`' {
+			return nil
+		}
+	}
+
+	return []Block{
+		&FencedCode{
+			markdown:     markdown,
+			Indentation:  indentation,
+			RawInfo:      trimRightSpace(markdown, Range{r.Position + fenceLength, r.End}),
+			OpeningFence: Range{r.Position, r.Position + fenceLength},
+		},
+	}
+}

utils/markdown/html.go

@@ -0,0 +1,186 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"fmt"
+	"strings"
+)
+
+var htmlEscaper = strings.NewReplacer(
+	`&`, "&",
+	`<`, "&lt;",
+	`>`, "&gt;",
+	`"`, "&quot;",
+)
+
+// RenderHTML produces HTML with the same behavior as the example renderer used in the CommonMark
+// reference materials except for one slight difference: for brevity, no unnecessary whitespace is
+// inserted between elements. The output is not defined by the CommonMark spec, and it exists
+// primarily as an aid in testing.
+func RenderHTML(markdown string) string {
+	return RenderBlockHTML(Parse(markdown))
+}
+
+func RenderBlockHTML(block Block, referenceDefinitions []*ReferenceDefinition) (result string) {
+	return renderBlockHTML(block, referenceDefinitions, false)
+}
+
+func renderBlockHTML(block Block, referenceDefinitions []*ReferenceDefinition, isTightList bool) (result string) {
+	switch v := block.(type) {
+	case *Document:
+		for _, block := range v.Children {
+			result += RenderBlockHTML(block, referenceDefinitions)
+		}
+	case *Paragraph:
+		if len(v.Text) == 0 {
+			return
+		}
+		if !isTightList {
+			result += "<p>"
+		}
+		for _, inline := range v.ParseInlines(referenceDefinitions) {
+			result += RenderInlineHTML(inline)
+		}
+		if !isTightList {
+			result += "</p>"
+		}
+	case *List:
+		if v.IsOrdered {
+			if v.OrderedStart != 1 {
+				result += fmt.Sprintf(`<ol start="%v">`, v.OrderedStart)
+			} else {
+				result += "<ol>"
+			}
+		} else {
+			result += "<ul>"
+		}
+		for _, block := range v.Children {
+			result += renderBlockHTML(block, referenceDefinitions, !v.IsLoose)
+		}
+		if v.IsOrdered {
+			result += "</ol>"
+		} else {
+			result += "</ul>"
+		}
+	case *ListItem:
+		result += "<li>"
+		for _, block := range v.Children {
+			result += renderBlockHTML(block, referenceDefinitions, isTightList)
+		}
+		result += "</li>"
+	case *BlockQuote:
+		result += "<blockquote>"
+		for _, block := range v.Children {
+			result += RenderBlockHTML(block, referenceDefinitions)
+		}
+		result += "</blockquote>"
+	case *FencedCode:
+		if info := v.Info(); info != "" {
+			language := strings.Fields(info)[0]
+			result += `<pre><code class="language-` + htmlEscaper.Replace(language) + `">`
+		} else {
+			result += "<pre><code>"
+		}
+		result += htmlEscaper.Replace(v.Code()) + "</code></pre>"
+	case *IndentedCode:
+		result += "<pre><code>" + htmlEscaper.Replace(v.Code()) + "</code></pre>"
+	default:
+		panic(fmt.Sprintf("missing case for type %T", v))
+	}
+	return
+}
+
+func escapeURL(url string) (result string) {
+	for i := 0; i < len(url); {
+		switch b := url[i]; b {
+		case ';', '/', '?', ':', '@', '&', '=', '+', '$', ',', '-', '_', '.', '!', '~', '*', '\'', '(', ')', '#':
+			result += string(b)
+			i++
+		default:
+			if b == '%' && i+2 < len(url) && isHexByte(url[i+1]) && isHexByte(url[i+2]) {
+				result += url[i : i+3]
+				i += 3
+			} else if (b >= 'a' && b <= 'z') || (b >= 'A' && b <= 'Z') || (b >= '0' && b <= '9') {
+				result += string(b)
+				i++
+			} else {
+				result += fmt.Sprintf("%%%0X", b)
+				i++
+			}
+		}
+	}
+	return
+}
+
+func RenderInlineHTML(inline Inline) (result string) {
+	switch v := inline.(type) {
+	case *Text:
+		return htmlEscaper.Replace(v.Text)
+	case *HardLineBreak:
+		return "<br />"
+	case *SoftLineBreak:
+		return "\n"
+	case *CodeSpan:
+		return "<code>" + htmlEscaper.Replace(v.Code) + "</code>"
+	case *InlineImage:
+		result += `<img src="` + htmlEscaper.Replace(escapeURL(v.Destination())) + `" alt="` + htmlEscaper.Replace(renderImageAltText(v.Children)) + `"`
+		if title := v.Title(); title != "" {
+			result += ` title="` + htmlEscaper.Replace(title) + `"`
+		}
+		result += ` />`
+	case *ReferenceImage:
+		result += `<img src="` + htmlEscaper.Replace(escapeURL(v.Destination())) + `" alt="` + htmlEscaper.Replace(renderImageAltText(v.Children)) + `"`
+		if title := v.Title(); title != "" {
+			result += ` title="` + htmlEscaper.Replace(title) + `"`
+		}
+		result += ` />`
+	case *InlineLink:
+		result += `<a href="` + htmlEscaper.Replace(escapeURL(v.Destination())) + `"`
+		if title := v.Title(); title != "" {
+			result += ` title="` + htmlEscaper.Replace(title) + `"`
+		}
+		result += `>`
+		for _, inline := range v.Children {
+			result += RenderInlineHTML(inline)
+		}
+		result += "</a>"
+	case *ReferenceLink:
+		result += `<a href="` + htmlEscaper.Replace(escapeURL(v.Destination())) + `"`
+		if title := v.Title(); title != "" {
+			result += ` title="` + htmlEscaper.Replace(title) + `"`
+		}
+		result += `>`
+		for _, inline := range v.Children {
+			result += RenderInlineHTML(inline)
+		}
+		result += "</a>"
+	default:
+		panic(fmt.Sprintf("missing case for type %T", v))
+	}
+	return
+}
+
+func renderImageAltText(children []Inline) (result string) {
+	for _, inline := range children {
+		result += renderImageChildAltText(inline)
+	}
+	return
+}
+
+func renderImageChildAltText(inline Inline) (result string) {
+	switch v := inline.(type) {
+	case *Text:
+		return v.Text
+	case *InlineImage:
+		for _, inline := range v.Children {
+			result += renderImageChildAltText(inline)
+		}
+	case *InlineLink:
+		for _, inline := range v.Children {
+			result += renderImageChildAltText(inline)
+		}
+	}
+	return
+}

utils/markdown/indented_code.go

@@ -0,0 +1,98 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"strings"
+)
+
+type IndentedCodeLine struct {
+	Indentation int
+	Range       Range
+}
+
+type IndentedCode struct {
+	blockBase
+	markdown string
+
+	RawCode []IndentedCodeLine
+}
+
+func (b *IndentedCode) Code() (result string) {
+	for _, code := range b.RawCode {
+		result += strings.Repeat(" ", code.Indentation) + b.markdown[code.Range.Position:code.Range.End]
+	}
+	return
+}
+
+func (b *IndentedCode) Continuation(indentation int, r Range) *continuation {
+	if indentation >= 4 {
+		return &continuation{
+			Indentation: indentation - 4,
+			Remaining:   r,
+		}
+	}
+	s := b.markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		return &continuation{
+			Remaining: r,
+		}
+	}
+	return nil
+}
+
+func (b *IndentedCode) AddLine(indentation int, r Range) bool {
+	b.RawCode = append(b.RawCode, IndentedCodeLine{
+		Indentation: indentation,
+		Range:       r,
+	})
+	return true
+}
+
+func (b *IndentedCode) Close() {
+	for {
+		last := b.RawCode[len(b.RawCode)-1]
+		s := b.markdown[last.Range.Position:last.Range.End]
+		if strings.TrimRight(s, "\r\n") == "" {
+			b.RawCode = b.RawCode[:len(b.RawCode)-1]
+		} else {
+			break
+		}
+	}
+}
+
+func (b *IndentedCode) AllowsBlockStarts() bool {
+	return false
+}
+
+func indentedCodeStart(markdown string, indentation int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	if len(unmatchedBlocks) > 0 {
+		if _, ok := unmatchedBlocks[len(unmatchedBlocks)-1].(*Paragraph); ok {
+			return nil
+		}
+	} else if len(matchedBlocks) > 0 {
+		if _, ok := matchedBlocks[len(matchedBlocks)-1].(*Paragraph); ok {
+			return nil
+		}
+	}
+
+	if indentation < 4 {
+		return nil
+	}
+
+	s := markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		return nil
+	}
+
+	return []Block{
+		&IndentedCode{
+			markdown: markdown,
+			RawCode: []IndentedCodeLine{{
+				Indentation: indentation - 4,
+				Range:       r,
+			}},
+		},
+	}
+}

utils/markdown/inlines.go

@@ -0,0 +1,489 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"container/list"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+type Inline interface {
+	IsInline() bool
+}
+
+type inlineBase struct{}
+
+func (inlineBase) IsInline() bool { return true }
+
+type Text struct {
+	inlineBase
+
+	Text string
+}
+
+type CodeSpan struct {
+	inlineBase
+
+	Code string
+}
+
+type HardLineBreak struct {
+	inlineBase
+}
+
+type SoftLineBreak struct {
+	inlineBase
+}
+
+type InlineLinkOrImage struct {
+	inlineBase
+
+	Children []Inline
+
+	RawDestination Range
+
+	markdown string
+	rawTitle string
+}
+
+func (i *InlineLinkOrImage) Destination() string {
+	return Unescape(i.markdown[i.RawDestination.Position:i.RawDestination.End])
+}
+
+func (i *InlineLinkOrImage) Title() string {
+	return Unescape(i.rawTitle)
+}
+
+type InlineLink struct {
+	InlineLinkOrImage
+}
+
+type InlineImage struct {
+	InlineLinkOrImage
+}
+
+type ReferenceLinkOrImage struct {
+	inlineBase
+	*ReferenceDefinition
+
+	Children []Inline
+}
+
+type ReferenceLink struct {
+	ReferenceLinkOrImage
+}
+
+type ReferenceImage struct {
+	ReferenceLinkOrImage
+}
+
+type delimiterType int
+
+const (
+	linkOpeningDelimiter delimiterType = iota
+	imageOpeningDelimiter
+)
+
+type delimiter struct {
+	Type       delimiterType
+	IsInactive bool
+	TextNode   int
+	Range      Range
+}
+
+type inlineParser struct {
+	markdown             string
+	ranges               []Range
+	referenceDefinitions []*ReferenceDefinition
+
+	raw            string
+	position       int
+	inlines        []Inline
+	delimiterStack *list.List
+}
+
+func newInlineParser(markdown string, ranges []Range, referenceDefinitions []*ReferenceDefinition) *inlineParser {
+	return &inlineParser{
+		markdown:             markdown,
+		ranges:               ranges,
+		referenceDefinitions: referenceDefinitions,
+		delimiterStack:       list.New(),
+	}
+}
+
+func (p *inlineParser) parseBackticks() {
+	count := 1
+	for i := p.position + 1; i < len(p.raw) && p.raw[i] == '`'; i++ {
+		count++
+	}
+	opening := p.raw[p.position : p.position+count]
+	search := p.position + count
+	for search < len(p.raw) {
+		end := strings.Index(p.raw[search:], opening)
+		if end == -1 {
+			break
+		}
+		if search+end+count < len(p.raw) && p.raw[search+end+count] == '`' {
+			search += end + count
+			for search < len(p.raw) && p.raw[search] == '`' {
+				search++
+			}
+			continue
+		}
+		code := strings.Join(strings.Fields(p.raw[p.position+count:search+end]), " ")
+		p.position = search + end + count
+		p.inlines = append(p.inlines, &CodeSpan{
+			Code: code,
+		})
+		return
+	}
+	p.position += len(opening)
+	p.inlines = append(p.inlines, &Text{
+		Text: opening,
+	})
+}
+
+func (p *inlineParser) parseLineEnding() {
+	if p.position >= 1 && p.raw[p.position-1] == '\t' {
+		p.inlines = append(p.inlines, &HardLineBreak{})
+	} else if p.position >= 2 && p.raw[p.position-1] == ' ' && (p.raw[p.position-2] == '\t' || p.raw[p.position-1] == ' ') {
+		p.inlines = append(p.inlines, &HardLineBreak{})
+	} else {
+		p.inlines = append(p.inlines, &SoftLineBreak{})
+	}
+	p.position++
+	if p.position < len(p.raw) && p.raw[p.position] == '\n' {
+		p.position++
+	}
+}
+
+func (p *inlineParser) parseEscapeCharacter() {
+	if p.position+1 < len(p.raw) && isEscapableByte(p.raw[p.position+1]) {
+		p.inlines = append(p.inlines, &Text{
+			Text: string(p.raw[p.position+1]),
+		})
+		p.position += 2
+	} else {
+		p.inlines = append(p.inlines, &Text{
+			Text: `\`,
+		})
+		p.position++
+	}
+}
+
+func (p *inlineParser) parseText() {
+	if next := strings.IndexAny(p.raw[p.position:], "\r\n\\`&![]"); next == -1 {
+		p.inlines = append(p.inlines, &Text{
+			Text: strings.TrimRightFunc(p.raw[p.position:], isWhitespace),
+		})
+		p.position = len(p.raw)
+	} else {
+		if p.raw[p.position+next] == '\r' || p.raw[p.position+next] == '\n' {
+			p.inlines = append(p.inlines, &Text{
+				Text: strings.TrimRightFunc(p.raw[p.position:p.position+next], isWhitespace),
+			})
+		} else {
+			p.inlines = append(p.inlines, &Text{
+				Text: p.raw[p.position : p.position+next],
+			})
+		}
+		p.position += next
+	}
+}
+
+func (p *inlineParser) parseLinkOrImageDelimiter() {
+	if p.raw[p.position] == '[' {
+		p.inlines = append(p.inlines, &Text{
+			Text: "[",
+		})
+		p.delimiterStack.PushBack(&delimiter{
+			Type:     linkOpeningDelimiter,
+			TextNode: len(p.inlines) - 1,
+			Range:    Range{p.position, p.position + 1},
+		})
+		p.position++
+	} else if p.raw[p.position] == '!' && p.position+1 < len(p.raw) && p.raw[p.position+1] == '[' {
+		p.inlines = append(p.inlines, &Text{
+			Text: "![",
+		})
+		p.delimiterStack.PushBack(&delimiter{
+			Type:     imageOpeningDelimiter,
+			TextNode: len(p.inlines) - 1,
+			Range:    Range{p.position, p.position + 2},
+		})
+		p.position += 2
+	} else {
+		p.inlines = append(p.inlines, &Text{
+			Text: "!",
+		})
+		p.position++
+	}
+}
+
+func (p *inlineParser) peekAtInlineLinkDestinationAndTitle(position int) (destination, title Range, end int, ok bool) {
+	if position >= len(p.raw) || p.raw[position] != '(' {
+		return
+	}
+	position++
+
+	destinationStart := nextNonWhitespace(p.raw, position)
+	if destinationStart >= len(p.raw) {
+		return
+	} else if p.raw[destinationStart] == ')' {
+		return Range{destinationStart, destinationStart}, Range{destinationStart, destinationStart}, destinationStart + 1, true
+	}
+
+	destination, end, ok = parseLinkDestination(p.raw, destinationStart)
+	if !ok {
+		return
+	}
+	position = end
+
+	if position < len(p.raw) && isWhitespaceByte(p.raw[position]) {
+		titleStart := nextNonWhitespace(p.raw, position)
+		if titleStart >= len(p.raw) {
+			return
+		} else if p.raw[titleStart] == ')' {
+			return destination, Range{titleStart, titleStart}, titleStart + 1, true
+		}
+
+		title, end, ok = parseLinkTitle(p.raw, titleStart)
+		if !ok {
+			return
+		}
+		position = end
+	}
+
+	closingPosition := nextNonWhitespace(p.raw, position)
+	if closingPosition >= len(p.raw) || p.raw[closingPosition] != ')' {
+		return Range{}, Range{}, 0, false
+	}
+
+	return destination, title, closingPosition + 1, true
+}
+
+func (p *inlineParser) referenceDefinition(label string) *ReferenceDefinition {
+	clean := strings.Join(strings.Fields(label), " ")
+	for _, d := range p.referenceDefinitions {
+		if strings.EqualFold(clean, strings.Join(strings.Fields(d.Label()), " ")) {
+			return d
+		}
+	}
+	return nil
+}
+
+func (p *inlineParser) lookForLinkOrImage() {
+	for element := p.delimiterStack.Back(); element != nil; element = element.Prev() {
+		d := element.Value.(*delimiter)
+		if d.Type != imageOpeningDelimiter && d.Type != linkOpeningDelimiter {
+			continue
+		}
+		if d.IsInactive {
+			p.delimiterStack.Remove(element)
+			break
+		}
+
+		var inline Inline
+
+		if destination, title, next, ok := p.peekAtInlineLinkDestinationAndTitle(p.position + 1); ok {
+			destinationMarkdownPosition := relativeToAbsolutePosition(p.ranges, destination.Position)
+			linkOrImage := InlineLinkOrImage{
+				Children:       append([]Inline(nil), p.inlines[d.TextNode+1:]...),
+				RawDestination: Range{destinationMarkdownPosition, destinationMarkdownPosition + destination.End - destination.Position},
+				markdown:       p.markdown,
+				rawTitle:       p.raw[title.Position:title.End],
+			}
+			if d.Type == imageOpeningDelimiter {
+				inline = &InlineImage{linkOrImage}
+			} else {
+				inline = &InlineLink{linkOrImage}
+			}
+			p.position = next
+		} else {
+			referenceLabel := ""
+			label, next, hasLinkLabel := parseLinkLabel(p.raw, p.position+1)
+			if hasLinkLabel && label.End > label.Position {
+				referenceLabel = p.raw[label.Position:label.End]
+			} else {
+				referenceLabel = p.raw[d.Range.End:p.position]
+				if !hasLinkLabel {
+					next = p.position + 1
+				}
+			}
+			if referenceLabel != "" {
+				if reference := p.referenceDefinition(referenceLabel); reference != nil {
+					linkOrImage := ReferenceLinkOrImage{
+						ReferenceDefinition: reference,
+						Children:            append([]Inline(nil), p.inlines[d.TextNode+1:]...),
+					}
+					if d.Type == imageOpeningDelimiter {
+						inline = &ReferenceImage{linkOrImage}
+					} else {
+						inline = &ReferenceLink{linkOrImage}
+					}
+					p.position = next
+				}
+			}
+		}
+
+		if inline != nil {
+			if d.Type == imageOpeningDelimiter {
+				p.inlines = append(p.inlines[:d.TextNode], inline)
+			} else {
+				p.inlines = append(p.inlines[:d.TextNode], inline)
+				for element := element.Prev(); element != nil; element = element.Prev() {
+					if d := element.Value.(*delimiter); d.Type == linkOpeningDelimiter {
+						d.IsInactive = true
+					}
+				}
+			}
+			p.delimiterStack.Remove(element)
+			return
+		} else {
+			p.delimiterStack.Remove(element)
+			break
+		}
+	}
+	p.inlines = append(p.inlines, &Text{
+		Text: "]",
+	})
+	p.position++
+}
+
+func CharacterReference(ref string) string {
+	if ref == "" {
+		return ""
+	}
+	if ref[0] == '#' {
+		if len(ref) < 2 {
+			return ""
+		}
+		n := 0
+		if ref[1] == 'X' || ref[1] == 'x' {
+			if len(ref) < 3 {
+				return ""
+			}
+			for i := 2; i < len(ref); i++ {
+				if i > 9 {
+					return ""
+				}
+				d := ref[i]
+				switch {
+				case d >= '0' && d <= '9':
+					n = n*16 + int(d-'0')
+				case d >= 'a' && d <= 'f':
+					n = n*16 + 10 + int(d-'a')
+				case d >= 'A' && d <= 'F':
+					n = n*16 + 10 + int(d-'A')
+				default:
+					return ""
+				}
+			}
+		} else {
+			for i := 1; i < len(ref); i++ {
+				if i > 8 || ref[i] < '0' || ref[i] > '9' {
+					return ""
+				}
+				n = n*10 + int(ref[i]-'0')
+			}
+		}
+		c := rune(n)
+		if c == '\u0000' || !utf8.ValidRune(c) {
+			return string(unicode.ReplacementChar)
+		}
+		return string(c)
+	}
+	if entity, ok := htmlEntities[ref]; ok {
+		return entity
+	}
+	return ""
+}
+
+func (p *inlineParser) parseCharacterReference() {
+	p.position++
+	if semicolon := strings.IndexByte(p.raw[p.position:], ';'); semicolon == -1 {
+		p.inlines = append(p.inlines, &Text{
+			Text: "&",
+		})
+	} else if s := CharacterReference(p.raw[p.position : p.position+semicolon]); s != "" {
+		p.position += semicolon + 1
+		p.inlines = append(p.inlines, &Text{
+			Text: s,
+		})
+	} else {
+		p.inlines = append(p.inlines, &Text{
+			Text: "&",
+		})
+	}
+}
+
+func (p *inlineParser) Parse() []Inline {
+	for _, r := range p.ranges {
+		p.raw += p.markdown[r.Position:r.End]
+	}
+
+	for p.position < len(p.raw) {
+		c, _ := utf8.DecodeRuneInString(p.raw[p.position:])
+
+		switch c {
+		case '\r', '\n':
+			p.parseLineEnding()
+		case '\\':
+			p.parseEscapeCharacter()
+		case '`':
+			p.parseBackticks()
+		case '&':
+			p.parseCharacterReference()
+		case '!', '[':
+			p.parseLinkOrImageDelimiter()
+		case ']':
+			p.lookForLinkOrImage()
+		default:
+			p.parseText()
+		}
+	}
+
+	return p.inlines
+}
+
+func ParseInlines(markdown string, ranges []Range, referenceDefinitions []*ReferenceDefinition) (inlines []Inline) {
+	return newInlineParser(markdown, ranges, referenceDefinitions).Parse()
+}
+
+func Unescape(markdown string) string {
+	ret := ""
+
+	position := 0
+	for position < len(markdown) {
+		c, cSize := utf8.DecodeRuneInString(markdown[position:])
+
+		switch c {
+		case '\\':
+			if position+1 < len(markdown) && isEscapableByte(markdown[position+1]) {
+				ret += string(markdown[position+1])
+				position += 2
+			} else {
+				ret += `\`
+				position++
+			}
+		case '&':
+			position++
+			if semicolon := strings.IndexByte(markdown[position:], ';'); semicolon == -1 {
+				ret += "&"
+			} else if s := CharacterReference(markdown[position : position+semicolon]); s != "" {
+				position += semicolon + 1
+				ret += s
+			} else {
+				ret += "&"
+			}
+		default:
+			ret += string(c)
+			position += cSize
+		}
+	}
+
+	return ret
+}

utils/markdown/inspect.go

@@ -0,0 +1,78 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+// Inspect traverses the markdown tree in depth-first order. If f returns true, Inspect invokes f
+// recursively for each child of the block or inline, followed by a call of f(nil).
+func Inspect(markdown string, f func(interface{}) bool) {
+	document, referenceDefinitions := Parse(markdown)
+	InspectBlock(document, func(block Block) bool {
+		if !f(block) {
+			return false
+		}
+		switch v := block.(type) {
+		case *Paragraph:
+			for _, inline := range v.ParseInlines(referenceDefinitions) {
+				InspectInline(inline, func(inline Inline) bool {
+					return f(inline)
+				})
+			}
+		}
+		return true
+	})
+}
+
+// InspectBlock traverses the blocks in depth-first order, starting with block. If f returns true,
+// InspectBlock invokes f recursively for each child of the block, followed by a call of f(nil).
+func InspectBlock(block Block, f func(Block) bool) {
+	if !f(block) {
+		return
+	}
+	switch v := block.(type) {
+	case *Document:
+		for _, child := range v.Children {
+			InspectBlock(child, f)
+		}
+	case *List:
+		for _, child := range v.Children {
+			InspectBlock(child, f)
+		}
+	case *ListItem:
+		for _, child := range v.Children {
+			InspectBlock(child, f)
+		}
+	case *BlockQuote:
+		for _, child := range v.Children {
+			InspectBlock(child, f)
+		}
+	}
+	f(nil)
+}
+
+// InspectInline traverses the blocks in depth-first order, starting with block. If f returns true,
+// InspectInline invokes f recursively for each child of the block, followed by a call of f(nil).
+func InspectInline(inline Inline, f func(Inline) bool) {
+	if !f(inline) {
+		return
+	}
+	switch v := inline.(type) {
+	case *InlineImage:
+		for _, child := range v.Children {
+			InspectInline(child, f)
+		}
+	case *InlineLink:
+		for _, child := range v.Children {
+			InspectInline(child, f)
+		}
+	case *ReferenceImage:
+		for _, child := range v.Children {
+			InspectInline(child, f)
+		}
+	case *ReferenceLink:
+		for _, child := range v.Children {
+			InspectInline(child, f)
+		}
+	}
+	f(nil)
+}

utils/markdown/inspect_test.go

@@ -0,0 +1,54 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestInspect(t *testing.T) {
+	markdown := `
+[foo]: bar
+- a
+  > [![]()]()
+  > [![foo]][foo]
+- d
+`
+
+	visited := []string{}
+	level := 0
+	Inspect(markdown, func(blockOrInline interface{}) bool {
+		if blockOrInline == nil {
+			level--
+		} else {
+			visited = append(visited, strings.Repeat(" ", level*4)+strings.TrimPrefix(fmt.Sprintf("%T", blockOrInline), "*markdown."))
+			level++
+		}
+		return true
+	})
+
+	assert.Equal(t, []string{
+		"Document",
+		"    Paragraph",
+		"    List",
+		"        ListItem",
+		"            Paragraph",
+		"                Text",
+		"            BlockQuote",
+		"                Paragraph",
+		"                    InlineLink",
+		"                        InlineImage",
+		"                    SoftLineBreak",
+		"                    ReferenceLink",
+		"                        ReferenceImage",
+		"                            Text",
+		"        ListItem",
+		"            Paragraph",
+		"                Text",
+	}, visited)
+}

utils/markdown/lines.go

@@ -0,0 +1,27 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+type Line struct {
+	Range
+}
+
+func ParseLines(markdown string) (lines []Line) {
+	lineStartPosition := 0
+	isAfterCarriageReturn := false
+	for position, r := range markdown {
+		if r == '\n' {
+			lines = append(lines, Line{Range{lineStartPosition, position + 1}})
+			lineStartPosition = position + 1
+		} else if isAfterCarriageReturn {
+			lines = append(lines, Line{Range{lineStartPosition, position}})
+			lineStartPosition = position
+		}
+		isAfterCarriageReturn = r == '\r'
+	}
+	if lineStartPosition < len(markdown) {
+		lines = append(lines, Line{Range{lineStartPosition, len(markdown)}})
+	}
+	return
+}

utils/markdown/lines_test.go

@@ -0,0 +1,36 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestParseLines(t *testing.T) {
+	assert.Equal(t, []Line{
+		{Range{0, 4}}, {Range{4, 7}},
+	}, ParseLines("foo\nbar"))
+
+	assert.Equal(t, []Line{
+		{Range{0, 5}}, {Range{5, 8}},
+	}, ParseLines("foo\r\nbar"))
+
+	assert.Equal(t, []Line{
+		{Range{0, 4}}, {Range{4, 6}}, {Range{6, 9}},
+	}, ParseLines("foo\r\r\nbar"))
+
+	assert.Equal(t, []Line{
+		{Range{0, 4}},
+	}, ParseLines("foo\n"))
+
+	assert.Equal(t, []Line{
+		{Range{0, 4}},
+	}, ParseLines("foo\r"))
+
+	assert.Equal(t, []Line{
+		{Range{0, 5}},
+	}, ParseLines("foo\r\n"))
+}

utils/markdown/links.go

@@ -0,0 +1,130 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"unicode/utf8"
+)
+
+func parseLinkDestination(markdown string, position int) (raw Range, next int, ok bool) {
+	if position >= len(markdown) {
+		return
+	}
+
+	if markdown[position] == '<' {
+		isEscaped := false
+
+		for offset, c := range []byte(markdown[position+1:]) {
+			if isEscaped {
+				isEscaped = false
+				if isEscapableByte(c) {
+					continue
+				}
+			}
+
+			if c == '\\' {
+				isEscaped = true
+			} else if c == '<' {
+				break
+			} else if c == '>' {
+				return Range{position + 1, position + 1 + offset}, position + 1 + offset + 1, true
+			} else if isWhitespaceByte(c) {
+				break
+			}
+		}
+	}
+
+	openCount := 0
+	isEscaped := false
+	for offset, c := range []byte(markdown[position:]) {
+		if isEscaped {
+			isEscaped = false
+			if isEscapableByte(c) {
+				continue
+			}
+		}
+
+		switch c {
+		case '\\':
+			isEscaped = true
+		case '(':
+			openCount++
+		case ')':
+			if openCount < 1 {
+				return Range{position, position + offset}, position + offset, true
+			}
+			openCount--
+		default:
+			if isWhitespaceByte(c) {
+				return Range{position, position + offset}, position + offset, true
+			}
+		}
+	}
+	return Range{position, len(markdown)}, len(markdown), true
+}
+
+func parseLinkTitle(markdown string, position int) (raw Range, next int, ok bool) {
+	if position >= len(markdown) {
+		return
+	}
+
+	originalPosition := position
+
+	var closer byte
+	switch markdown[position] {
+	case '"', '\'':
+		closer = markdown[position]
+	case '(':
+		closer = ')'
+	default:
+		return
+	}
+	position++
+
+	for position < len(markdown) {
+		switch markdown[position] {
+		case '\\':
+			position++
+			if position < len(markdown) && isEscapableByte(markdown[position]) {
+				position++
+			}
+		case closer:
+			return Range{originalPosition + 1, position}, position + 1, true
+		default:
+			position++
+		}
+	}
+
+	return
+}
+
+func parseLinkLabel(markdown string, position int) (raw Range, next int, ok bool) {
+	if position >= len(markdown) || markdown[position] != '[' {
+		return
+	}
+
+	originalPosition := position
+	position++
+
+	for position < len(markdown) {
+		switch markdown[position] {
+		case '\\':
+			position++
+			if position < len(markdown) && isEscapableByte(markdown[position]) {
+				position++
+			}
+		case '[':
+			return
+		case ']':
+			if position-originalPosition >= 1000 && utf8.RuneCountInString(markdown[originalPosition:position]) >= 1000 {
+				return
+			}
+			return Range{originalPosition + 1, position}, position + 1, true
+		default:
+			position++
+		}
+	}
+
+	return
+}

utils/markdown/list.go

@@ -0,0 +1,220 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"strings"
+)
+
+type ListItem struct {
+	blockBase
+	markdown                    string
+	hasTrailingBlankLine        bool
+	hasBlankLineBetweenChildren bool
+
+	Indentation int
+	Children    []Block
+}
+
+func (b *ListItem) Continuation(indentation int, r Range) *continuation {
+	s := b.markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		if b.Children == nil {
+			return nil
+		}
+		return &continuation{
+			Remaining: r,
+		}
+	}
+	if indentation < b.Indentation {
+		return nil
+	}
+	return &continuation{
+		Indentation: indentation - b.Indentation,
+		Remaining:   r,
+	}
+}
+
+func (b *ListItem) AddChild(openBlocks []Block) []Block {
+	b.Children = append(b.Children, openBlocks[0])
+	if b.hasTrailingBlankLine {
+		b.hasBlankLineBetweenChildren = true
+	}
+	b.hasTrailingBlankLine = false
+	return openBlocks
+}
+
+func (b *ListItem) AddLine(indentation int, r Range) bool {
+	isBlank := strings.TrimSpace(b.markdown[r.Position:r.End]) == ""
+	if isBlank {
+		b.hasTrailingBlankLine = true
+	}
+	return false
+}
+
+func (b *ListItem) HasTrailingBlankLine() bool {
+	return b.hasTrailingBlankLine || (len(b.Children) > 0 && b.Children[len(b.Children)-1].HasTrailingBlankLine())
+}
+
+func (b *ListItem) isLoose() bool {
+	if b.hasBlankLineBetweenChildren {
+		return true
+	}
+	for i, child := range b.Children {
+		if i < len(b.Children)-1 && child.HasTrailingBlankLine() {
+			return true
+		}
+	}
+	return false
+}
+
+type List struct {
+	blockBase
+	markdown                    string
+	hasTrailingBlankLine        bool
+	hasBlankLineBetweenChildren bool
+
+	IsLoose           bool
+	IsOrdered         bool
+	OrderedStart      int
+	BulletOrDelimiter byte
+	Children          []*ListItem
+}
+
+func (b *List) Continuation(indentation int, r Range) *continuation {
+	s := b.markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		return &continuation{
+			Remaining: r,
+		}
+	}
+	return &continuation{
+		Indentation: indentation,
+		Remaining:   r,
+	}
+}
+
+func (b *List) AddChild(openBlocks []Block) []Block {
+	if item, ok := openBlocks[0].(*ListItem); ok {
+		b.Children = append(b.Children, item)
+		if b.hasTrailingBlankLine {
+			b.hasBlankLineBetweenChildren = true
+		}
+		b.hasTrailingBlankLine = false
+		return openBlocks
+	} else if list, ok := openBlocks[0].(*List); ok {
+		if len(list.Children) == 1 && list.IsOrdered == b.IsOrdered && list.BulletOrDelimiter == b.BulletOrDelimiter {
+			return b.AddChild(openBlocks[1:])
+		}
+	}
+	return nil
+}
+
+func (b *List) AddLine(indentation int, r Range) bool {
+	isBlank := strings.TrimSpace(b.markdown[r.Position:r.End]) == ""
+	if isBlank {
+		b.hasTrailingBlankLine = true
+	}
+	return false
+}
+
+func (b *List) HasTrailingBlankLine() bool {
+	return b.hasTrailingBlankLine || (len(b.Children) > 0 && b.Children[len(b.Children)-1].HasTrailingBlankLine())
+}
+
+func (b *List) isLoose() bool {
+	if b.hasBlankLineBetweenChildren {
+		return true
+	}
+	for i, child := range b.Children {
+		if child.isLoose() || (i < len(b.Children)-1 && child.HasTrailingBlankLine()) {
+			return true
+		}
+	}
+	return false
+}
+
+func (b *List) Close() {
+	b.IsLoose = b.isLoose()
+}
+
+func parseListMarker(markdown string, r Range) (success, isOrdered bool, orderedStart int, bulletOrDelimiter byte, markerWidth int, remaining Range) {
+	digits := 0
+	n := 0
+	for i := r.Position; i < r.End && markdown[i] >= '0' && markdown[i] <= '9'; i++ {
+		digits++
+		n = n*10 + int(markdown[i]-'0')
+	}
+	if digits > 0 {
+		if digits > 9 || r.Position+digits >= r.End {
+			return
+		}
+		next := markdown[r.Position+digits]
+		if next != '.' && next != ')' {
+			return
+		}
+		return true, true, n, next, digits + 1, Range{r.Position + digits + 1, r.End}
+	}
+	if r.Position >= r.End {
+		return
+	}
+	next := markdown[r.Position]
+	if next != '-' && next != '+' && next != '*' {
+		return
+	}
+	return true, false, 0, next, 1, Range{r.Position + 1, r.End}
+}
+
+func listStart(markdown string, indent int, r Range, matchedBlocks, unmatchedBlocks []Block) []Block {
+	afterList := false
+	if len(matchedBlocks) > 0 {
+		_, afterList = matchedBlocks[len(matchedBlocks)-1].(*List)
+	}
+	if !afterList && indent > 3 {
+		return nil
+	}
+
+	success, isOrdered, orderedStart, bulletOrDelimiter, markerWidth, remaining := parseListMarker(markdown, r)
+	if !success {
+		return nil
+	}
+
+	isBlank := strings.TrimSpace(markdown[remaining.Position:remaining.End]) == ""
+	if len(matchedBlocks) > 0 && len(unmatchedBlocks) == 0 {
+		if _, ok := matchedBlocks[len(matchedBlocks)-1].(*Paragraph); ok {
+			if isBlank || (isOrdered && orderedStart != 1) {
+				return nil
+			}
+		}
+	}
+
+	indentAfterMarker, indentBytesAfterMarker := countIndentation(markdown, remaining)
+	if !isBlank && indentAfterMarker < 1 {
+		return nil
+	}
+
+	remaining = Range{remaining.Position + indentBytesAfterMarker, remaining.End}
+	consumedIndentAfterMarker := indentAfterMarker
+	if isBlank || indentAfterMarker >= 5 {
+		consumedIndentAfterMarker = 1
+	}
+
+	listItem := &ListItem{
+		markdown:    markdown,
+		Indentation: indent + markerWidth + consumedIndentAfterMarker,
+	}
+	list := &List{
+		markdown:          markdown,
+		IsOrdered:         isOrdered,
+		OrderedStart:      orderedStart,
+		BulletOrDelimiter: bulletOrDelimiter,
+		Children:          []*ListItem{listItem},
+	}
+	ret := []Block{list, listItem}
+	if descendants := blockStartOrParagraph(markdown, indentAfterMarker-consumedIndentAfterMarker, remaining, nil, nil); descendants != nil {
+		listItem.Children = append(listItem.Children, descendants[0])
+		ret = append(ret, descendants...)
+	}
+	return ret
+}

utils/markdown/markdown.go

@@ -0,0 +1,132 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+// This package implements a parser for the subset of the CommonMark spec necessary for us to do
+// server-side processing. It is not a full implementation and lacks many features. But it is
+// complete enough to efficiently and accurately allow us to do what we need to like rewrite image
+// URLs for proxying.
+package markdown
+
+import (
+	"strings"
+)
+
+func isEscapable(c rune) bool {
+	return c > ' ' && (c < '0' || (c > '9' && (c < 'A' || (c > 'Z' && (c < 'a' || (c > 'z' && c <= '~'))))))
+}
+
+func isEscapableByte(c byte) bool {
+	return isEscapable(rune(c))
+}
+
+func isWhitespace(c rune) bool {
+	switch c {
+	case ' ', '\t', '\n', '\u000b', '\u000c', '\r':
+		return true
+	default:
+		return false
+	}
+}
+
+func isWhitespaceByte(c byte) bool {
+	return isWhitespace(rune(c))
+}
+
+func isHex(c rune) bool {
+	return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F')
+}
+
+func isHexByte(c byte) bool {
+	return isHex(rune(c))
+}
+
+func nextNonWhitespace(markdown string, position int) int {
+	for offset, c := range []byte(markdown[position:]) {
+		if !isWhitespaceByte(c) {
+			return position + offset
+		}
+	}
+	return len(markdown)
+}
+
+func nextLine(markdown string, position int) (linePosition int, skippedNonWhitespace bool) {
+	for i := position; i < len(markdown); i++ {
+		c := markdown[i]
+		if c == '\r' {
+			if i+1 < len(markdown) && markdown[i+1] == '\n' {
+				return i + 2, skippedNonWhitespace
+			}
+			return i + 1, skippedNonWhitespace
+		} else if c == '\n' {
+			return i + 1, skippedNonWhitespace
+		} else if !isWhitespaceByte(c) {
+			skippedNonWhitespace = true
+		}
+	}
+	return len(markdown), skippedNonWhitespace
+}
+
+func countIndentation(markdown string, r Range) (spaces, bytes int) {
+	for i := r.Position; i < r.End; i++ {
+		if markdown[i] == ' ' {
+			spaces++
+			bytes++
+		} else if markdown[i] == '\t' {
+			spaces += 4
+			bytes++
+		} else {
+			break
+		}
+	}
+	return
+}
+
+func trimLeftSpace(markdown string, r Range) Range {
+	s := markdown[r.Position:r.End]
+	trimmed := strings.TrimLeftFunc(s, isWhitespace)
+	return Range{r.Position, r.End - (len(s) - len(trimmed))}
+}
+
+func trimRightSpace(markdown string, r Range) Range {
+	s := markdown[r.Position:r.End]
+	trimmed := strings.TrimRightFunc(s, isWhitespace)
+	return Range{r.Position, r.End - (len(s) - len(trimmed))}
+}
+
+func relativeToAbsolutePosition(ranges []Range, position int) int {
+	rem := position
+	for _, r := range ranges {
+		l := r.End - r.Position
+		if rem < l {
+			return r.Position + rem
+		}
+		rem -= l
+	}
+	if len(ranges) == 0 {
+		return 0
+	}
+	return ranges[len(ranges)-1].End
+}
+
+func trimBytesFromRanges(ranges []Range, bytes int) (result []Range) {
+	rem := bytes
+	for _, r := range ranges {
+		if rem == 0 {
+			result = append(result, r)
+			continue
+		}
+		l := r.End - r.Position
+		if rem < l {
+			result = append(result, Range{r.Position + rem, r.End})
+			rem = 0
+			continue
+		}
+		rem -= l
+	}
+	return
+}
+
+func Parse(markdown string) (*Document, []*ReferenceDefinition) {
+	lines := ParseLines(markdown)
+	return ParseBlocks(markdown, lines)
+}

utils/markdown/paragraph.go

@@ -0,0 +1,71 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+import (
+	"strings"
+)
+
+type Paragraph struct {
+	blockBase
+	markdown string
+
+	Text                 []Range
+	ReferenceDefinitions []*ReferenceDefinition
+}
+
+func (b *Paragraph) ParseInlines(referenceDefinitions []*ReferenceDefinition) []Inline {
+	return ParseInlines(b.markdown, b.Text, referenceDefinitions)
+}
+
+func (b *Paragraph) Continuation(indentation int, r Range) *continuation {
+	s := b.markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		return nil
+	}
+	return &continuation{
+		Indentation: indentation,
+		Remaining:   r,
+	}
+}
+
+func (b *Paragraph) Close() {
+	for {
+		for i := 0; i < len(b.Text); i++ {
+			b.Text[i] = trimLeftSpace(b.markdown, b.Text[i])
+			if b.Text[i].Position < b.Text[i].End {
+				break
+			}
+		}
+
+		if len(b.Text) == 0 || b.Text[0].Position < b.Text[0].End && b.markdown[b.Text[0].Position] != '[' {
+			break
+		}
+
+		definition, remaining := parseReferenceDefinition(b.markdown, b.Text)
+		if definition == nil {
+			break
+		}
+		b.ReferenceDefinitions = append(b.ReferenceDefinitions, definition)
+		b.Text = remaining
+	}
+
+	for i := len(b.Text) - 1; i >= 0; i-- {
+		b.Text[i] = trimRightSpace(b.markdown, b.Text[i])
+		if b.Text[i].Position < b.Text[i].End {
+			break
+		}
+	}
+}
+
+func newParagraph(markdown string, r Range) *Paragraph {
+	s := markdown[r.Position:r.End]
+	if strings.TrimSpace(s) == "" {
+		return nil
+	}
+	return &Paragraph{
+		markdown: markdown,
+		Text:     []Range{r},
+	}
+}

utils/markdown/reference_definition.go

@@ -0,0 +1,75 @@
+// Copyright (c) 2017-present Mattermost, Inc. All Rights Reserved.
+// See License.txt for license information.
+
+package markdown
+
+type ReferenceDefinition struct {
+	RawDestination Range
+
+	markdown string
+	rawLabel string
+	rawTitle string
+}
+
+func (d *ReferenceDefinition) Destination() string {
+	return Unescape(d.markdown[d.RawDestination.Position:d.RawDestination.End])
+}
+
+func (d *ReferenceDefinition) Label() string {
+	return d.rawLabel
+}
+
+func (d *ReferenceDefinition) Title() string {
+	return Unescape(d.rawTitle)
+}
+
+func parseReferenceDefinition(markdown string, ranges []Range) (*ReferenceDefinition, []Range) {
+	raw := ""
+	for _, r := range ranges {
+		raw += markdown[r.Position:r.End]
+	}
+
+	label, next, ok := parseLinkLabel(raw, 0)
+	if !ok {
+		return nil, nil
+	}
+	position := next
+
+	if position >= len(raw) || raw[position] != ':' {
+		return nil, nil
+	}
+	position++
+
+	destination, next, ok := parseLinkDestination(raw, nextNonWhitespace(raw, position))
+	if !ok {
+		return nil, nil
+	}
+	position = next
+
+	absoluteDestination := relativeToAbsolutePosition(ranges, destination.Position)
+	ret := &ReferenceDefinition{
+		RawDestination: Range{absoluteDestination, absoluteDestination + destination.End - destination.Position},
+		markdown:       markdown,
+		rawLabel:       raw[label.Position:label.End],
+	}
+
+	if position < len(raw) && isWhitespaceByte(raw[position]) {
+		title, next, ok := parseLinkTitle(raw, nextNonWhitespace(raw, position))
+		if !ok {
+			if nextLine, skippedNonWhitespace := nextLine(raw, position); !skippedNonWhitespace {
+				return ret, trimBytesFromRanges(ranges, nextLine)
+			}
+			return nil, nil
+		}
+		if nextLine, skippedNonWhitespace := nextLine(raw, next); !skippedNonWhitespace {
+			ret.rawTitle = raw[title.Position:title.End]
+			return ret, trimBytesFromRanges(ranges, nextLine)
+		}
+	}
+
+	if nextLine, skippedNonWhitespace := nextLine(raw, position); !skippedNonWhitespace {
+		return ret, trimBytesFromRanges(ranges, nextLine)
+	}
+
+	return nil, nil
+}