Update markdown parsing library to github.com/gomarkdown/markdown (#944)

36 files changed, 8266 insertions, 0 deletions

diff --git a/vendor/github.com/gomarkdown/markdown/.gitignore b/vendor/github.com/gomarkdown/markdown/.gitignore
new file mode 100644
index 00000000..8f15b677
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/.gitignore
@@ -0,0 +1,13 @@
+*.out
+*.swp
+*.8
+*.6
+_obj
+_test*
+markdown
+tags
+fuzz-workdir/
+markdown-fuzz.zip
+coverage.txt
+testdata/*_got.md
+testdata/*_ast.txt
diff --git a/vendor/github.com/gomarkdown/markdown/.gitpod b/vendor/github.com/gomarkdown/markdown/.gitpod
new file mode 100644
index 00000000..ad5feff6
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/.gitpod
@@ -0,0 +1,7 @@
+checkoutLocation: "src/github.com/gomarkdown/markdown"
+workspaceLocation: "."
+tasks:
+  - command: >
+      cd /workspace/src/github.com/gomarkdown/markdown &&
+      go get -v ./... &&
+      go test -c
diff --git a/vendor/github.com/gomarkdown/markdown/.travis.yml b/vendor/github.com/gomarkdown/markdown/.travis.yml
new file mode 100644
index 00000000..4ec5d7b0
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/.travis.yml
@@ -0,0 +1,17 @@
+dist: bionic
+language: go
+
+go:
+  - "1.12.x"
+
+install:
+  - go build -v ./...
+
+script:
+  - go test -v ./...
+  - go test -run=^$ -bench=BenchmarkReference -benchmem
+  - ./s/test_with_codecoverage.sh
+  - ./s/ci_fuzzit.sh
+
+after_success:
+  - bash <(curl -s https://codecov.io/bash)
diff --git a/vendor/github.com/gomarkdown/markdown/LICENSE.txt b/vendor/github.com/gomarkdown/markdown/LICENSE.txt
new file mode 100644
index 00000000..68804610
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/LICENSE.txt
@@ -0,0 +1,31 @@
+Markdown is distributed under the Simplified BSD License:
+
+Copyright © 2011 Russ Ross
+Copyright © 2018 Krzysztof Kowalczyk
+Copyright © 2018 Authors
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+
+1.  Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+
+2.  Redistributions in binary form must reproduce the above
+    copyright notice, this list of conditions and the following
+    disclaimer in the documentation and/or other materials provided with
+    the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
+ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGE.
diff --git a/vendor/github.com/gomarkdown/markdown/README.md b/vendor/github.com/gomarkdown/markdown/README.md
new file mode 100644
index 00000000..3bca71db
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/README.md
@@ -0,0 +1,325 @@
+# Markdown Parser and HTML Renderer for Go
+
+[![GoDoc](https://godoc.org/github.com/gomarkdown/markdown?status.svg)](https://godoc.org/github.com/gomarkdown/markdown) [![codecov](https://codecov.io/gh/gomarkdown/markdown/branch/master/graph/badge.svg)](https://codecov.io/gh/gomarkdown/markdown)
+
+Package `github.com/gomarkdown/markdown` is a very fast Go library for parsing [Markdown](https://daringfireball.net/projects/markdown/) documents and rendering them to HTML.
+
+It's fast and supports common extensions.
+
+## Installation
+
+    go get -u github.com/gomarkdown/markdown
+
+API Docs:
+
+- https://godoc.org/github.com/gomarkdown/markdown : top level package
+- https://godoc.org/github.com/gomarkdown/markdown/ast : defines abstract syntax tree of parsed markdown document
+- https://godoc.org/github.com/gomarkdown/markdown/parser : parser
+- https://godoc.org/github.com/gomarkdown/markdown/html : html renderer
+
+## Usage
+
+To convert markdown text to HTML using reasonable defaults:
+
+```go
+md := []byte("## markdown document")
+output := markdown.ToHTML(md, nil, nil)
+```
+
+## Customizing markdown parser
+
+Markdown format is loosely specified and there are multiple extensions invented after original specification was created.
+
+The parser supports several [extensions](https://godoc.org/github.com/gomarkdown/markdown/parser#Extensions).
+
+Default parser uses most common `parser.CommonExtensions` but you can easily use parser with custom extension:
+
+```go
+import (
+    "github.com/gomarkdown/markdown"
+    "github.com/gomarkdown/markdown/parser"
+)
+
+extensions := parser.CommonExtensions | parser.AutoHeadingIDs
+parser := parser.NewWithExtensions(extensions)
+
+md := []byte("markdown text")
+html := markdown.ToHTML(md, parser, nil)
+```
+
+## Customizing HTML renderer
+
+Similarly, HTML renderer can be configured with different [options](https://godoc.org/github.com/gomarkdown/markdown/html#RendererOptions)
+
+Here's how to use a custom renderer:
+
+```go
+import (
+    "github.com/gomarkdown/markdown"
+    "github.com/gomarkdown/markdown/html"
+)
+
+htmlFlags := html.CommonFlags | html.HrefTargetBlank
+opts := html.RendererOptions{Flags: htmlFlags}
+renderer := html.NewRenderer(opts)
+
+md := []byte("markdown text")
+html := markdown.ToHTML(md, nil, renderer)
+```
+
+HTML renderer also supports reusing most of the logic and overriding rendering of only specifc nodes.
+
+You can provide [RenderNodeFunc](https://godoc.org/github.com/gomarkdown/markdown/html#RenderNodeFunc) in [RendererOptions](https://godoc.org/github.com/gomarkdown/markdown/html#RendererOptions).
+
+The function is called for each node in AST, you can implement custom rendering logic and tell HTML renderer to skip rendering this node.
+
+Here's the simplest example that drops all code blocks from the output:
+
+````go
+import (
+    "github.com/gomarkdown/markdown"
+    "github.com/gomarkdown/markdown/ast"
+    "github.com/gomarkdown/markdown/html"
+)
+
+// return (ast.GoToNext, true) to tell html renderer to skip rendering this node
+// (because you've rendered it)
+func renderHookDropCodeBlock(w io.Writer, node ast.Node, entering bool) (ast.WalkStatus, bool) {
+    // skip all nodes that are not CodeBlock nodes
+	if _, ok := node.(*ast.CodeBlock); !ok {
+		return ast.GoToNext, false
+    }
+    // custom rendering logic for ast.CodeBlock. By doing nothing it won't be
+    // present in the output
+	return ast.GoToNext, true
+}
+
+opts := html.RendererOptions{
+    Flags: html.CommonFlags,
+    RenderNodeHook: renderHookDropCodeBlock,
+}
+renderer := html.NewRenderer(opts)
+md := "test\n```\nthis code block will be dropped from output\n```\ntext"
+html := markdown.ToHTML([]byte(s), nil, renderer)
+````
+
+## Sanitize untrusted content
+
+We don't protect against malicious content. When dealing with user-provided
+markdown, run renderer HTML through HTML sanitizer such as [Bluemonday](https://github.com/microcosm-cc/bluemonday).
+
+Here's an example of simple usage with Bluemonday:
+
+```go
+import (
+    "github.com/microcosm-cc/bluemonday"
+    "github.com/gomarkdown/markdown"
+)
+
+// ...
+maybeUnsafeHTML := markdown.ToHTML(md, nil, nil)
+html := bluemonday.UGCPolicy().SanitizeBytes(maybeUnsafeHTML)
+```
+
+## mdtohtml command-line tool
+
+https://github.com/gomarkdown/mdtohtml is a command-line markdown to html
+converter built using this library.
+
+You can also use it as an example of how to use the library.
+
+You can install it with:
+
+    go get -u github.com/gomarkdown/mdtohtml
+
+To run: `mdtohtml input-file [output-file]`
+
+## Features
+
+- **Compatibility**. The Markdown v1.0.3 test suite passes with
+  the `--tidy` option. Without `--tidy`, the differences are
+  mostly in whitespace and entity escaping, where this package is
+  more consistent and cleaner.
+
+- **Common extensions**, including table support, fenced code
+  blocks, autolinks, strikethroughs, non-strict emphasis, etc.
+
+- **Safety**. Markdown is paranoid when parsing, making it safe
+  to feed untrusted user input without fear of bad things
+  happening. The test suite stress tests this and there are no
+  known inputs that make it crash. If you find one, please let me
+  know and send me the input that does it.
+
+  NOTE: "safety" in this context means _runtime safety only_. In order to
+  protect yourself against JavaScript injection in untrusted content, see
+  [this example](https://github.com/gomarkdown/markdown#sanitize-untrusted-content).
+
+- **Fast**. It is fast enough to render on-demand in
+  most web applications without having to cache the output.
+
+- **Thread safety**. You can run multiple parsers in different
+  goroutines without ill effect. There is no dependence on global
+  shared state.
+
+- **Minimal dependencies**. Only depends on standard library packages in Go.
+
+- **Standards compliant**. Output successfully validates using the
+  W3C validation tool for HTML 4.01 and XHTML 1.0 Transitional.
+
+## Extensions
+
+In addition to the standard markdown syntax, this package
+implements the following extensions:
+
+- **Intra-word emphasis supression**. The `_` character is
+  commonly used inside words when discussing code, so having
+  markdown interpret it as an emphasis command is usually the
+  wrong thing. We let you treat all emphasis markers as
+  normal characters when they occur inside a word.
+
+- **Tables**. Tables can be created by drawing them in the input
+  using a simple syntax:
+
+  ```
+  Name    | Age
+  --------|------
+  Bob     | 27
+  Alice   | 23
+  ```
+
+  Table footers are supported as well and can be added with equal signs (`=`):
+
+  ```
+  Name    | Age
+  --------|------
+  Bob     | 27
+  Alice   | 23
+  ========|======
+  Total   | 50
+  ```
+
+- **Fenced code blocks**. In addition to the normal 4-space
+  indentation to mark code blocks, you can explicitly mark them
+  and supply a language (to make syntax highlighting simple). Just
+  mark it like this:
+
+      ```go
+      func getTrue() bool {
+          return true
+      }
+      ```
+
+  You can use 3 or more backticks to mark the beginning of the
+  block, and the same number to mark the end of the block.
+
+- **Definition lists**. A simple definition list is made of a single-line
+  term followed by a colon and the definition for that term.
+
+      Cat
+      : Fluffy animal everyone likes
+
+      Internet
+      : Vector of transmission for pictures of cats
+
+  Terms must be separated from the previous definition by a blank line.
+
+- **Footnotes**. A marker in the text that will become a superscript number;
+  a footnote definition that will be placed in a list of footnotes at the
+  end of the document. A footnote looks like this:
+
+      This is a footnote.[^1]
+
+      [^1]: the footnote text.
+
+- **Autolinking**. We can find URLs that have not been
+  explicitly marked as links and turn them into links.
+
+- **Strikethrough**. Use two tildes (`~~`) to mark text that
+  should be crossed out.
+
+- **Hard line breaks**. With this extension enabled newlines in the input
+  translate into line breaks in the output. This extension is off by default.
+
+- **Non blocking space**. With this extension enabled spaces preceeded by an backslash n the input
+  translate non-blocking spaces in the output. This extension is off by default.
+
+- **Smart quotes**. Smartypants-style punctuation substitution is
+  supported, turning normal double- and single-quote marks into
+  curly quotes, etc.
+
+- **LaTeX-style dash parsing** is an additional option, where `--`
+  is translated into `&ndash;`, and `---` is translated into
+  `&mdash;`. This differs from most smartypants processors, which
+  turn a single hyphen into an ndash and a double hyphen into an
+  mdash.
+
+- **Smart fractions**, where anything that looks like a fraction
+  is translated into suitable HTML (instead of just a few special
+  cases like most smartypant processors). For example, `4/5`
+  becomes `<sup>4</sup>&frasl;<sub>5</sub>`, which renders as
+  <sup>4</sup>&frasl;<sub>5</sub>.
+
+- **MathJaX Support** is an additional feature which is supported by
+  many markdown editor. It translate inline math equation quoted by `$`
+  and display math block quoted by `$$` into MathJax compatible format.
+  hyphen `_` won't break LaTeX render within a math element any more.
+
+  ```
+  $$
+  \left[ \begin{array}{a} a^l_1 \\ ⋮ \\ a^l_{d_l} \end{array}\right]
+  = \sigma(
+   \left[ \begin{matrix}
+   	w^l_{1,1} & ⋯  & w^l_{1,d_{l-1}} \\
+   	⋮ & ⋱  & ⋮  \\
+   	w^l_{d_l,1} & ⋯  & w^l_{d_l,d_{l-1}} \\
+   \end{matrix}\right]  ·
+   \left[ \begin{array}{x} a^{l-1}_1 \\ ⋮ \\ ⋮ \\ a^{l-1}_{d_{l-1}} \end{array}\right] +
+   \left[ \begin{array}{b} b^l_1 \\ ⋮ \\ b^l_{d_l} \end{array}\right])
+   $$
+  ```
+
+- **Ordered list start number**. With this extension enabled an ordered list will start with the
+  the number that was used to start it.
+
+- **Super and subscript**. With this extension enabled sequences between ^ will indicate
+  superscript and ~ will become a subscript. For example: H~2~O is a liquid, 2^10^ is 1024.
+
+- **Block level attributes**, allow setting attributes (ID, classes and key/value pairs) on block
+  level elements. The attribute must be enclosed with braces and be put on a line before the
+  element.
+
+  ```
+  {#id3 .myclass fontsize="tiny"}
+  # Header 1
+  ```
+
+  Will convert into `<h1 id="id3" class="myclass" fontsize="tiny">Header 1</h1>`.
+
+- **Mmark support**, see <https://mmark.nl/syntax> for all new syntax elements this adds.
+
+## Todo
+
+- port https://github.com/russross/blackfriday/issues/348
+- port [LaTeX output](https://github.com/Ambrevar/Blackfriday-LaTeX):
+  renders output as LaTeX.
+- port https://github.com/shurcooL/github_flavored_markdown to markdown
+- port [markdownfmt](https://github.com/shurcooL/markdownfmt): like gofmt,
+  but for markdown.
+- More unit testing
+- Improve unicode support. It does not understand all unicode
+  rules (about what constitutes a letter, a punctuation symbol,
+  etc.), so it may fail to detect word boundaries correctly in
+  some instances. It is safe on all utf-8 input.
+
+## History
+
+markdown is a fork of v2 of https://github.com/russross/blackfriday that is:
+
+- actively maintained (sadly in Feb 2018 blackfriday was inactive for 5 months with many bugs and pull requests accumulated)
+- refactored API (split into ast/parser/html sub-packages)
+
+Blackfriday itself was based on C implementation [sundown](https://github.com/vmg/sundown) which in turn was based on [libsoldout](http://fossil.instinctive.eu/libsoldout/home).
+
+## License
+
+[Simplified BSD License](LICENSE.txt)
diff --git a/vendor/github.com/gomarkdown/markdown/ast/attribute.go b/vendor/github.com/gomarkdown/markdown/ast/attribute.go
new file mode 100644
index 00000000..002c6a2e
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/ast/attribute.go
@@ -0,0 +1,10 @@
+package ast
+
+// An attribute can be attached to block elements. They are specified as
+// {#id .classs key="value"} where quotes for values are mandatory, multiple
+// key/value pairs are separated by whitespace.
+type Attribute struct {
+	ID      []byte
+	Classes [][]byte
+	Attrs   map[string][]byte
+}
diff --git a/vendor/github.com/gomarkdown/markdown/ast/doc.go b/vendor/github.com/gomarkdown/markdown/ast/doc.go
new file mode 100644
index 00000000..376dc67c
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/ast/doc.go
@@ -0,0 +1,4 @@
+/*
+Package ast defines tree representation of a parsed markdown document.
+*/
+package ast
diff --git a/vendor/github.com/gomarkdown/markdown/ast/node.go b/vendor/github.com/gomarkdown/markdown/ast/node.go
new file mode 100644
index 00000000..e6fcba9a
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/ast/node.go
@@ -0,0 +1,559 @@
+package ast
+
+// ListType contains bitwise or'ed flags for list and list item objects.
+type ListType int
+
+// These are the possible flag values for the ListItem renderer.
+// Multiple flag values may be ORed together.
+// These are mostly of interest if you are writing a new output format.
+const (
+	ListTypeOrdered ListType = 1 << iota
+	ListTypeDefinition
+	ListTypeTerm
+
+	ListItemContainsBlock
+	ListItemBeginningOfList // TODO: figure out if this is of any use now
+	ListItemEndOfList
+)
+
+// CellAlignFlags holds a type of alignment in a table cell.
+type CellAlignFlags int
+
+// These are the possible flag values for the table cell renderer.
+// Only a single one of these values will be used; they are not ORed together.
+// These are mostly of interest if you are writing a new output format.
+const (
+	TableAlignmentLeft CellAlignFlags = 1 << iota
+	TableAlignmentRight
+	TableAlignmentCenter = (TableAlignmentLeft | TableAlignmentRight)
+)
+
+func (a CellAlignFlags) String() string {
+	switch a {
+	case TableAlignmentLeft:
+		return "left"
+	case TableAlignmentRight:
+		return "right"
+	case TableAlignmentCenter:
+		return "center"
+	default:
+		return ""
+	}
+}
+
+// DocumentMatters holds the type of a {front,main,back}matter in the document
+type DocumentMatters int
+
+// These are all possible Document divisions.
+const (
+	DocumentMatterNone DocumentMatters = iota
+	DocumentMatterFront
+	DocumentMatterMain
+	DocumentMatterBack
+)
+
+// CitationTypes holds the type of a citation, informative, normative or suppressed
+type CitationTypes int
+
+const (
+	CitationTypeNone CitationTypes = iota
+	CitationTypeSuppressed
+	CitationTypeInformative
+	CitationTypeNormative
+)
+
+// Node defines an ast node
+type Node interface {
+	AsContainer() *Container
+	AsLeaf() *Leaf
+	GetParent() Node
+	SetParent(newParent Node)
+	GetChildren() []Node
+	SetChildren(newChildren []Node)
+}
+
+// Container is a type of node that can contain children
+type Container struct {
+	Parent   Node
+	Children []Node
+
+	Literal []byte // Text contents of the leaf nodes
+	Content []byte // Markdown content of the block nodes
+
+	*Attribute // Block level attribute
+}
+
+// AsContainer returns itself as *Container
+func (c *Container) AsContainer() *Container {
+	return c
+}
+
+// AsLeaf returns nil
+func (c *Container) AsLeaf() *Leaf {
+	return nil
+}
+
+// GetParent returns parent node
+func (c *Container) GetParent() Node {
+	return c.Parent
+}
+
+// SetParent sets the parent node
+func (c *Container) SetParent(newParent Node) {
+	c.Parent = newParent
+}
+
+// GetChildren returns children nodes
+func (c *Container) GetChildren() []Node {
+	return c.Children
+}
+
+// SetChildren sets children node
+func (c *Container) SetChildren(newChildren []Node) {
+	c.Children = newChildren
+}
+
+// Leaf is a type of node that cannot have children
+type Leaf struct {
+	Parent Node
+
+	Literal []byte // Text contents of the leaf nodes
+	Content []byte // Markdown content of the block nodes
+
+	*Attribute // Block level attribute
+}
+
+// AsContainer returns nil
+func (l *Leaf) AsContainer() *Container {
+	return nil
+}
+
+// AsLeaf returns itself as *Leaf
+func (l *Leaf) AsLeaf() *Leaf {
+	return l
+}
+
+// GetParent returns parent node
+func (l *Leaf) GetParent() Node {
+	return l.Parent
+}
+
+// SetParent sets the parent nodd
+func (l *Leaf) SetParent(newParent Node) {
+	l.Parent = newParent
+}
+
+// GetChildren returns nil because Leaf cannot have children
+func (l *Leaf) GetChildren() []Node {
+	return nil
+}
+
+// SetChildren will panic becuase Leaf cannot have children
+func (l *Leaf) SetChildren(newChildren []Node) {
+	panic("leaf node cannot have children")
+}
+
+// Document represents markdown document node, a root of ast
+type Document struct {
+	Container
+}
+
+// DocumentMatter represents markdown node that signals a document
+// division: frontmatter, mainmatter or backmatter.
+type DocumentMatter struct {
+	Container
+
+	Matter DocumentMatters
+}
+
+// BlockQuote represents markdown block quote node
+type BlockQuote struct {
+	Container
+}
+
+// Aside represents an markdown aside node.
+type Aside struct {
+	Container
+}
+
+// List represents markdown list node
+type List struct {
+	Container
+
+	ListFlags       ListType
+	Tight           bool   // Skip <p>s around list item data if true
+	BulletChar      byte   // '*', '+' or '-' in bullet lists
+	Delimiter       byte   // '.' or ')' after the number in ordered lists
+	Start           int    // for ordered lists this indicates the starting number if > 0
+	RefLink         []byte // If not nil, turns this list item into a footnote item and triggers different rendering
+	IsFootnotesList bool   // This is a list of footnotes
+}
+
+// ListItem represents markdown list item node
+type ListItem struct {
+	Container
+
+	ListFlags       ListType
+	Tight           bool   // Skip <p>s around list item data if true
+	BulletChar      byte   // '*', '+' or '-' in bullet lists
+	Delimiter       byte   // '.' or ')' after the number in ordered lists
+	RefLink         []byte // If not nil, turns this list item into a footnote item and triggers different rendering
+	IsFootnotesList bool   // This is a list of footnotes
+}
+
+// Paragraph represents markdown paragraph node
+type Paragraph struct {
+	Container
+}
+
+// Math represents markdown MathAjax inline node
+type Math struct {
+	Leaf
+}
+
+// MathBlock represents markdown MathAjax block node
+type MathBlock struct {
+	Container
+}
+
+// Heading represents markdown heading node
+type Heading struct {
+	Container
+
+	Level        int    // This holds the heading level number
+	HeadingID    string // This might hold heading ID, if present
+	IsTitleblock bool   // Specifies whether it's a title block
+	IsSpecial    bool   // We are a special heading (starts with .#)
+}
+
+// HorizontalRule represents markdown horizontal rule node
+type HorizontalRule struct {
+	Leaf
+}
+
+// Emph represents markdown emphasis node
+type Emph struct {
+	Container
+}
+
+// Strong represents markdown strong node
+type Strong struct {
+	Container
+}
+
+// Del represents markdown del node
+type Del struct {
+	Container
+}
+
+// Link represents markdown link node
+type Link struct {
+	Container
+
+	Destination []byte // Destination is what goes into a href
+	Title       []byte // Title is the tooltip thing that goes in a title attribute
+	NoteID      int    // NoteID contains a serial number of a footnote, zero if it's not a footnote
+	Footnote    Node   // If it's a footnote, this is a direct link to the footnote Node. Otherwise nil.
+	DeferredID  []byte // If a deferred link this holds the original ID.
+}
+
+// CrossReference is a reference node.
+type CrossReference struct {
+	Container
+
+	Destination []byte // Destination is where the reference points to
+}
+
+// Citation is a citation node.
+type Citation struct {
+	Leaf
+
+	Destination [][]byte        // Destination is where the citation points to. Multiple ones are allowed.
+	Type        []CitationTypes // 1:1 mapping of destination and citation type
+	Suffix      [][]byte        // Potential citation suffix, i.e. [@!RFC1035, p. 144]
+}
+
+// Image represents markdown image node
+type Image struct {
+	Container
+
+	Destination []byte // Destination is what goes into a href
+	Title       []byte // Title is the tooltip thing that goes in a title attribute
+}
+
+// Text represents markdown text node
+type Text struct {
+	Leaf
+}
+
+// HTMLBlock represents markdown html node
+type HTMLBlock struct {
+	Leaf
+}
+
+// CodeBlock represents markdown code block node
+type CodeBlock struct {
+	Leaf
+
+	IsFenced    bool   // Specifies whether it's a fenced code block or an indented one
+	Info        []byte // This holds the info string
+	FenceChar   byte
+	FenceLength int
+	FenceOffset int
+}
+
+// Softbreak represents markdown softbreak node
+// Note: not used currently
+type Softbreak struct {
+	Leaf
+}
+
+// Hardbreak represents markdown hard break node
+type Hardbreak struct {
+	Leaf
+}
+
+// NonBlockingSpace represents markdown non-blocking space node
+type NonBlockingSpace struct {
+	Leaf
+}
+
+// Code represents markdown code node
+type Code struct {
+	Leaf
+}
+
+// HTMLSpan represents markdown html span node
+type HTMLSpan struct {
+	Leaf
+}
+
+// Table represents markdown table node
+type Table struct {
+	Container
+}
+
+// TableCell represents markdown table cell node
+type TableCell struct {
+	Container
+
+	IsHeader bool           // This tells if it's under the header row
+	Align    CellAlignFlags // This holds the value for align attribute
+}
+
+// TableHeader represents markdown table head node
+type TableHeader struct {
+	Container
+}
+
+// TableBody represents markdown table body node
+type TableBody struct {
+	Container
+}
+
+// TableRow represents markdown table row node
+type TableRow struct {
+	Container
+}
+
+// TableFooter represents markdown table foot node
+type TableFooter struct {
+	Container
+}
+
+// Caption represents a figure, code or quote caption
+type Caption struct {
+	Container
+}
+
+// CaptionFigure is a node (blockquote or codeblock) that has a caption
+type CaptionFigure struct {
+	Container
+
+	HeadingID string // This might hold heading ID, if present
+}
+
+// Callout is a node that can exist both in text (where it is an actual node) and in a code block.
+type Callout struct {
+	Leaf
+
+	ID []byte // number of this callout
+}
+
+// Index is a node that contains an Index item and an optional, subitem.
+type Index struct {
+	Leaf
+
+	Primary bool
+	Item    []byte
+	Subitem []byte
+	ID      string // ID of the index
+}
+
+// Subscript is a subscript node
+type Subscript struct {
+	Leaf
+}
+
+// Subscript is a superscript node
+type Superscript struct {
+	Leaf
+}
+
+// Footnotes is a node that contains all footnotes
+type Footnotes struct {
+	Container
+}
+
+func removeNodeFromArray(a []Node, node Node) []Node {
+	n := len(a)
+	for i := 0; i < n; i++ {
+		if a[i] == node {
+			return append(a[:i], a[i+1:]...)
+		}
+	}
+	return nil
+}
+
+// AppendChild appends child to children of parent
+// It panics if either node is nil.
+func AppendChild(parent Node, child Node) {
+	RemoveFromTree(child)
+	child.SetParent(parent)
+	newChildren := append(parent.GetChildren(), child)
+	parent.SetChildren(newChildren)
+}
+
+// RemoveFromTree removes this node from tree
+func RemoveFromTree(n Node) {
+	if n.GetParent() == nil {
+		return
+	}
+	// important: don't clear n.Children if n has no parent
+	// we're called from AppendChild and that might happen on a node
+	// that accumulated Children but hasn't been inserted into the tree
+	n.SetChildren(nil)
+	p := n.GetParent()
+	newChildren := removeNodeFromArray(p.GetChildren(), n)
+	if newChildren != nil {
+		p.SetChildren(newChildren)
+	}
+}
+
+// GetLastChild returns last child of node n
+// It's implemented as stand-alone function to keep Node interface small
+func GetLastChild(n Node) Node {
+	a := n.GetChildren()
+	if len(a) > 0 {
+		return a[len(a)-1]
+	}
+	return nil
+}
+
+// GetFirstChild returns first child of node n
+// It's implemented as stand-alone function to keep Node interface small
+func GetFirstChild(n Node) Node {
+	a := n.GetChildren()
+	if len(a) > 0 {
+		return a[0]
+	}
+	return nil
+}
+
+// GetNextNode returns next sibling of node n (node after n)
+// We can't make it part of Container or Leaf because we loose Node identity
+func GetNextNode(n Node) Node {
+	parent := n.GetParent()
+	if parent == nil {
+		return nil
+	}
+	a := parent.GetChildren()
+	len := len(a) - 1
+	for i := 0; i < len; i++ {
+		if a[i] == n {
+			return a[i+1]
+		}
+	}
+	return nil
+}
+
+// GetPrevNode returns previous sibling of node n (node before n)
+// We can't make it part of Container or Leaf because we loose Node identity
+func GetPrevNode(n Node) Node {
+	parent := n.GetParent()
+	if parent == nil {
+		return nil
+	}
+	a := parent.GetChildren()
+	len := len(a)
+	for i := 1; i < len; i++ {
+		if a[i] == n {
+			return a[i-1]
+		}
+	}
+	return nil
+}
+
+// WalkStatus allows NodeVisitor to have some control over the tree traversal.
+// It is returned from NodeVisitor and different values allow Node.Walk to
+// decide which node to go to next.
+type WalkStatus int
+
+const (
+	// GoToNext is the default traversal of every node.
+	GoToNext WalkStatus = iota
+	// SkipChildren tells walker to skip all children of current node.
+	SkipChildren
+	// Terminate tells walker to terminate the traversal.
+	Terminate
+)
+
+// NodeVisitor is a callback to be called when traversing the syntax tree.
+// Called twice for every node: once with entering=true when the branch is
+// first visited, then with entering=false after all the children are done.
+type NodeVisitor interface {
+	Visit(node Node, entering bool) WalkStatus
+}
+
+// NodeVisitorFunc casts a function to match NodeVisitor interface
+type NodeVisitorFunc func(node Node, entering bool) WalkStatus
+
+// Walk traverses tree recursively
+func Walk(n Node, visitor NodeVisitor) WalkStatus {
+	isContainer := n.AsContainer() != nil
+	status := visitor.Visit(n, true) // entering
+	if status == Terminate {
+		// even if terminating, close container node
+		if isContainer {
+			visitor.Visit(n, false)
+		}
+		return status
+	}
+	if isContainer && status != SkipChildren {
+		children := n.GetChildren()
+		for _, n := range children {
+			status = Walk(n, visitor)
+			if status == Terminate {
+				return status
+			}
+		}
+	}
+	if isContainer {
+		status = visitor.Visit(n, false) // exiting
+		if status == Terminate {
+			return status
+		}
+	}
+	return GoToNext
+}
+
+// Visit calls visitor function
+func (f NodeVisitorFunc) Visit(node Node, entering bool) WalkStatus {
+	return f(node, entering)
+}
+
+// WalkFunc is like Walk but accepts just a callback function
+func WalkFunc(n Node, f NodeVisitorFunc) {
+	visitor := NodeVisitorFunc(f)
+	Walk(n, visitor)
+}
diff --git a/vendor/github.com/gomarkdown/markdown/ast/print.go b/vendor/github.com/gomarkdown/markdown/ast/print.go
new file mode 100644
index 00000000..75daf911
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/ast/print.go
@@ -0,0 +1,165 @@
+package ast
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"strings"
+)
+
+// Print is for debugging. It prints a string representation of parsed
+// markdown doc (result of parser.Parse()) to dst.
+//
+// To make output readable, it shortens text output.
+func Print(dst io.Writer, doc Node) {
+	PrintWithPrefix(dst, doc, "  ")
+}
+
+// PrintWithPrefix is like Print but allows customizing prefix used for
+// indentation. By default it's 2 spaces. You can change it to e.g. tab
+// by passing "\t"
+func PrintWithPrefix(w io.Writer, doc Node, prefix string) {
+	// for more compact output, don't print outer Document
+	if _, ok := doc.(*Document); ok {
+		for _, c := range doc.GetChildren() {
+			printRecur(w, c, prefix, 0)
+		}
+	} else {
+		printRecur(w, doc, prefix, 0)
+	}
+}
+
+// ToString is like Dump but returns result as a string
+func ToString(doc Node) string {
+	var buf bytes.Buffer
+	Print(&buf, doc)
+	return buf.String()
+}
+
+func contentToString(d1 []byte, d2 []byte) string {
+	if d1 != nil {
+		return string(d1)
+	}
+	if d2 != nil {
+		return string(d2)
+	}
+	return ""
+}
+
+func getContent(node Node) string {
+	if c := node.AsContainer(); c != nil {
+		return contentToString(c.Literal, c.Content)
+	}
+	leaf := node.AsLeaf()
+	return contentToString(leaf.Literal, leaf.Content)
+}
+
+func shortenString(s string, maxLen int) string {
+	// for cleaner, one-line ouput, replace some white-space chars
+	// with their escaped version
+	s = strings.Replace(s, "\n", `\n`, -1)
+	s = strings.Replace(s, "\r", `\r`, -1)
+	s = strings.Replace(s, "\t", `\t`, -1)
+	if maxLen < 0 {
+		return s
+	}
+	if len(s) < maxLen {
+		return s
+	}
+	// add "..." to indicate truncation
+	return s[:maxLen-3] + "..."
+}
+
+// get a short name of the type of v which excludes package name
+// and strips "()" from the end
+func getNodeType(node Node) string {
+	s := fmt.Sprintf("%T", node)
+	s = strings.TrimSuffix(s, "()")
+	if idx := strings.Index(s, "."); idx != -1 {
+		return s[idx+1:]
+	}
+	return s
+}
+
+func printDefault(w io.Writer, indent string, typeName string, content string) {
+	content = strings.TrimSpace(content)
+	if len(content) > 0 {
+		fmt.Fprintf(w, "%s%s '%s'\n", indent, typeName, content)
+	} else {
+		fmt.Fprintf(w, "%s%s\n", indent, typeName)
+	}
+}
+
+func getListFlags(f ListType) string {
+	var s string
+	if f&ListTypeOrdered != 0 {
+		s += "ordered "
+	}
+	if f&ListTypeDefinition != 0 {
+		s += "definition "
+	}
+	if f&ListTypeTerm != 0 {
+		s += "term "
+	}
+	if f&ListItemContainsBlock != 0 {
+		s += "has_block "
+	}
+	if f&ListItemBeginningOfList != 0 {
+		s += "start "
+	}
+	if f&ListItemEndOfList != 0 {
+		s += "end "
+	}
+	s = strings.TrimSpace(s)
+	return s
+}
+
+func printRecur(w io.Writer, node Node, prefix string, depth int) {
+	if node == nil {
+		return
+	}
+	indent := strings.Repeat(prefix, depth)
+
+	content := shortenString(getContent(node), 40)
+	typeName := getNodeType(node)
+	switch v := node.(type) {
+	case *Link:
+		content := "url=" + string(v.Destination)
+		printDefault(w, indent, typeName, content)
+	case *Image:
+		content := "url=" + string(v.Destination)
+		printDefault(w, indent, typeName, content)
+	case *List:
+		if v.Start > 1 {
+			content += fmt.Sprintf("start=%d ", v.Start)
+		}
+		if v.Tight {
+			content += "tight "
+		}
+		if v.IsFootnotesList {
+			content += "footnotes "
+		}
+		flags := getListFlags(v.ListFlags)
+		if len(flags) > 0 {
+			content += "flags=" + flags + " "
+		}
+		printDefault(w, indent, typeName, content)
+	case *ListItem:
+		if v.Tight {
+			content += "tight "
+		}
+		if v.IsFootnotesList {
+			content += "footnotes "
+		}
+		flags := getListFlags(v.ListFlags)
+		if len(flags) > 0 {
+			content += "flags=" + flags + " "
+		}
+		printDefault(w, indent, typeName, content)
+	default:
+		printDefault(w, indent, typeName, content)
+	}
+	for _, child := range node.GetChildren() {
+		printRecur(w, child, prefix, depth+1)
+	}
+}
diff --git a/vendor/github.com/gomarkdown/markdown/changes-from-blackfriday.md b/vendor/github.com/gomarkdown/markdown/changes-from-blackfriday.md
new file mode 100644
index 00000000..b618dfef
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/changes-from-blackfriday.md
@@ -0,0 +1,27 @@
+## Changes from blackfriday
+
+This library is derived from blackfriday library. Here's a list of changes.
+
+**Redesigned API**
+
+- split into 3 separate packages: ast, parser and html (for html renderer). This makes the API more manageable. It also separates e.g. parser option from renderer options
+- changed how AST node is represented from union-like representation (manually keeping track of the type of the node) to using interface{} (which is a Go way to combine an arbitrary value with its type)
+
+**Allow re-using most of html renderer logic**
+
+You can implement your own renderer by implementing `Renderer` interface.
+
+Implementing a full renderer is a lot of work and often you just want to tweak html rendering of few node typs.
+
+I've added a way to hook `Renderer.Render` function in html renderer with a custom function that can take over rendering of specific nodes.
+
+I use it myself to do syntax-highlighting of code snippets.
+
+**Speed up go test**
+
+Running `go test` was really slow (17 secs) because it did a poor man's version of fuzzing by feeding the parser all subsets of test strings in order to find panics
+due to incorrect parsing logic.
+
+I've moved that logic to `cmd/crashtest`, so that it can be run on CI but not slow down regular development.
+
+Now `go test` is blazing fast.
diff --git a/vendor/github.com/gomarkdown/markdown/codecov.yml b/vendor/github.com/gomarkdown/markdown/codecov.yml
new file mode 100644
index 00000000..f681ff11
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/codecov.yml
@@ -0,0 +1,8 @@
+coverage:
+  status:
+    project:
+      default:
+        # basic
+        target: 60%
+        threshold: 2%
+        base: auto
diff --git a/vendor/github.com/gomarkdown/markdown/doc.go b/vendor/github.com/gomarkdown/markdown/doc.go
new file mode 100644
index 00000000..9fb77e02
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/doc.go
@@ -0,0 +1,35 @@
+/*
+Package markdown implements markdown parser and HTML renderer.
+
+It parses markdown into AST format which can be serialized to HTML
+(using html.Renderer) or possibly other formats (using alternate renderers).
+
+Convert markdown to HTML
+
+The simplest way to convert markdown document to HTML
+
+  md := []byte("## markdown document")
+  html := markdown.ToHTML(md, nil, nil)
+
+Customizing parsing and HTML rendering
+
+You can customize parser and HTML renderer:
+
+	import (
+		"github.com/gomarkdown/markdown/parser"
+		"github.com/gomarkdown/markdown/renderer"
+		"github.com/gomarkdown/markdown"
+	)
+	extensions := parser.CommonExtensions | parser.AutoHeadingIDs
+	p := parser.NewWithExensions(extensions)
+
+	htmlFlags := html.CommonFlags | html.HrefTargetBlank
+	opts := html.RendererOptions{Flags: htmlFlags}
+	renderer := html.NewRenderer(opts)
+
+	md := []byte("markdown text")
+	html := markdown.ToHTML(md, p, renderer)
+
+For a cmd-line tool see https://github.com/gomarkdown/mdtohtml
+*/
+package markdown
diff --git a/vendor/github.com/gomarkdown/markdown/fuzz.go b/vendor/github.com/gomarkdown/markdown/fuzz.go
new file mode 100644
index 00000000..704182b8
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/fuzz.go
@@ -0,0 +1,9 @@
+// +build gofuzz
+
+package markdown
+
+// Fuzz is to be used by https://github.com/dvyukov/go-fuzz
+func Fuzz(data []byte) int {
+	Parse(data, nil)
+	return 0
+}
diff --git a/vendor/github.com/gomarkdown/markdown/go.mod b/vendor/github.com/gomarkdown/markdown/go.mod
new file mode 100644
index 00000000..899e3237
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/go.mod
@@ -0,0 +1,5 @@
+module github.com/gomarkdown/markdown
+
+go 1.12
+
+require golang.org/dl v0.0.0-20190829154251-82a15e2f2ead // indirect
diff --git a/vendor/github.com/gomarkdown/markdown/html/callouts.go b/vendor/github.com/gomarkdown/markdown/html/callouts.go
new file mode 100644
index 00000000..e377af22
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/html/callouts.go
@@ -0,0 +1,42 @@
+package html
+
+import (
+	"bytes"
+	"io"
+
+	"github.com/gomarkdown/markdown/ast"
+	"github.com/gomarkdown/markdown/parser"
+)
+
+// EscapeHTMLCallouts writes html-escaped d to w. It escapes &, <, > and " characters, *but*
+// expands callouts <<N>> with the callout HTML, i.e. by calling r.callout() with a newly created
+// ast.Callout node.
+func (r *Renderer) EscapeHTMLCallouts(w io.Writer, d []byte) {
+	ld := len(d)
+Parse:
+	for i := 0; i < ld; i++ {
+		for _, comment := range r.opts.Comments {
+			if !bytes.HasPrefix(d[i:], comment) {
+				break
+			}
+
+			lc := len(comment)
+			if i+lc < ld {
+				if id, consumed := parser.IsCallout(d[i+lc:]); consumed > 0 {
+					// We have seen a callout
+					callout := &ast.Callout{ID: id}
+					r.callout(w, callout)
+					i += consumed + lc - 1
+					continue Parse
+				}
+			}
+		}
+
+		escSeq := Escaper[d[i]]
+		if escSeq != nil {
+			w.Write(escSeq)
+		} else {
+			w.Write([]byte{d[i]})
+		}
+	}
+}
diff --git a/vendor/github.com/gomarkdown/markdown/html/doc.go b/vendor/github.com/gomarkdown/markdown/html/doc.go
new file mode 100644
index 00000000..f837c63d
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/html/doc.go
@@ -0,0 +1,43 @@
+/*
+Package html implements HTML renderer of parsed markdown document.
+
+Configuring and customizing a renderer
+
+A renderer can be configured with multiple options:
+
+	import "github.com/gomarkdown/markdown/html"
+
+	flags := html.CommonFlags | html.CompletePage | html.HrefTargetBlank
+	opts := html.RenderOptions{
+		TItle: "A custom title",
+		Flags: flags,
+	}
+	renderer := html.NewRenderer(opts)
+
+You can also re-use most of the logic and customize rendering of selected nodes
+by providing node render hook.
+This is most useful for rendering nodes that allow for design choices, like
+links or code blocks.
+
+	import (
+		"github.com/gomarkdown/markdown/html"
+		"github.com/gomarkdown/markdown/ast"
+	)
+
+	// a very dummy render hook that will output "code_replacements" instead of
+	// <code>${content}</code> emitted by html.Renderer
+	func renderHookCodeBlock(w io.Writer, node *ast.Node, entering bool) (ast.WalkStatus, bool) {
+		_, ok := node.Data.(*ast.CodeBlockData)
+		if !ok {
+			return ast.GoToNext, false
+		}
+		io.WriteString(w, "code_replacement")
+		return ast.GoToNext, true
+	}
+
+	opts := html.RendererOptions{
+		RenderNodeHook: renderHookCodeBlock,
+	}
+	renderer := html.NewRenderer(opts)
+*/
+package html
diff --git a/vendor/github.com/gomarkdown/markdown/html/esc.go b/vendor/github.com/gomarkdown/markdown/html/esc.go
new file mode 100644
index 00000000..89ec9a27
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/html/esc.go
@@ -0,0 +1,50 @@
+package html
+
+import (
+	"html"
+	"io"
+)
+
+var Escaper = [256][]byte{
+	'&': []byte("&amp;"),
+	'<': []byte("&lt;"),
+	'>': []byte("&gt;"),
+	'"': []byte("&quot;"),
+}
+
+// EscapeHTML writes html-escaped d to w. It escapes &, <, > and " characters.
+func EscapeHTML(w io.Writer, d []byte) {
+	var start, end int
+	n := len(d)
+	for end < n {
+		escSeq := Escaper[d[end]]
+		if escSeq != nil {
+			w.Write(d[start:end])
+			w.Write(escSeq)
+			start = end + 1
+		}
+		end++
+	}
+	if start < n && end <= n {
+		w.Write(d[start:end])
+	}
+}
+
+func escLink(w io.Writer, text []byte) {
+	unesc := html.UnescapeString(string(text))
+	EscapeHTML(w, []byte(unesc))
+}
+
+// Escape writes the text to w, but skips the escape character.
+func Escape(w io.Writer, text []byte) {
+	esc := false
+	for i := 0; i < len(text); i++ {
+		if text[i] == '\\' {
+			esc = !esc
+		}
+		if esc && text[i] == '\\' {
+			continue
+		}
+		w.Write([]byte{text[i]})
+	}
+}
diff --git a/vendor/github.com/gomarkdown/markdown/html/renderer.go b/vendor/github.com/gomarkdown/markdown/html/renderer.go
new file mode 100644
index 00000000..367f7dfa
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/html/renderer.go
@@ -0,0 +1,1318 @@
+package html
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// Flags control optional behavior of HTML renderer.
+type Flags int
+
+// IDTag is the tag used for tag identification, it defaults to "id", some renderers
+// may wish to override this and use e.g. "anchor".
+var IDTag = "id"
+
+// HTML renderer configuration options.
+const (
+	FlagsNone               Flags = 0
+	SkipHTML                Flags = 1 << iota // Skip preformatted HTML blocks
+	SkipImages                                // Skip embedded images
+	SkipLinks                                 // Skip all links
+	Safelink                                  // Only link to trusted protocols
+	NofollowLinks                             // Only link with rel="nofollow"
+	NoreferrerLinks                           // Only link with rel="noreferrer"
+	HrefTargetBlank                           // Add a blank target
+	CompletePage                              // Generate a complete HTML page
+	UseXHTML                                  // Generate XHTML output instead of HTML
+	FootnoteReturnLinks                       // Generate a link at the end of a footnote to return to the source
+	FootnoteNoHRTag                           // Do not output an HR after starting a footnote list.
+	Smartypants                               // Enable smart punctuation substitutions
+	SmartypantsFractions                      // Enable smart fractions (with Smartypants)
+	SmartypantsDashes                         // Enable smart dashes (with Smartypants)
+	SmartypantsLatexDashes                    // Enable LaTeX-style dashes (with Smartypants)
+	SmartypantsAngledQuotes                   // Enable angled double quotes (with Smartypants) for double quotes rendering
+	SmartypantsQuotesNBSP                     // Enable « French guillemets » (with Smartypants)
+	TOC                                       // Generate a table of contents
+
+	CommonFlags Flags = Smartypants | SmartypantsFractions | SmartypantsDashes | SmartypantsLatexDashes
+)
+
+var (
+	htmlTagRe = regexp.MustCompile("(?i)^" + htmlTag)
+)
+
+const (
+	htmlTag = "(?:" + openTag + "|" + closeTag + "|" + htmlComment + "|" +
+		processingInstruction + "|" + declaration + "|" + cdata + ")"
+	closeTag              = "</" + tagName + "\\s*[>]"
+	openTag               = "<" + tagName + attribute + "*" + "\\s*/?>"
+	attribute             = "(?:" + "\\s+" + attributeName + attributeValueSpec + "?)"
+	attributeValue        = "(?:" + unquotedValue + "|" + singleQuotedValue + "|" + doubleQuotedValue + ")"
+	attributeValueSpec    = "(?:" + "\\s*=" + "\\s*" + attributeValue + ")"
+	attributeName         = "[a-zA-Z_:][a-zA-Z0-9:._-]*"
+	cdata                 = "<!\\[CDATA\\[[\\s\\S]*?\\]\\]>"
+	declaration           = "<![A-Z]+" + "\\s+[^>]*>"
+	doubleQuotedValue     = "\"[^\"]*\""
+	htmlComment           = "<!---->|<!--(?:-?[^>-])(?:-?[^-])*-->"
+	processingInstruction = "[<][?].*?[?][>]"
+	singleQuotedValue     = "'[^']*'"
+	tagName               = "[A-Za-z][A-Za-z0-9-]*"
+	unquotedValue         = "[^\"'=<>`\\x00-\\x20]+"
+)
+
+// RenderNodeFunc allows reusing most of Renderer logic and replacing
+// rendering of some nodes. If it returns false, Renderer.RenderNode
+// will execute its logic. If it returns true, Renderer.RenderNode will
+// skip rendering this node and will return WalkStatus
+type RenderNodeFunc func(w io.Writer, node ast.Node, entering bool) (ast.WalkStatus, bool)
+
+// RendererOptions is a collection of supplementary parameters tweaking
+// the behavior of various parts of HTML renderer.
+type RendererOptions struct {
+	// Prepend this text to each relative URL.
+	AbsolutePrefix string
+	// Add this text to each footnote anchor, to ensure uniqueness.
+	FootnoteAnchorPrefix string
+	// Show this text inside the <a> tag for a footnote return link, if the
+	// FootnoteReturnLinks flag is enabled. If blank, the string
+	// <sup>[return]</sup> is used.
+	FootnoteReturnLinkContents string
+	// CitationFormatString defines how a citation is rendered. If blnck, the string
+	// <sup>[%s]</sup> is used. Where %s will be substituted with the citation target.
+	CitationFormatString string
+	// If set, add this text to the front of each Heading ID, to ensure uniqueness.
+	HeadingIDPrefix string
+	// If set, add this text to the back of each Heading ID, to ensure uniqueness.
+	HeadingIDSuffix string
+
+	Title string // Document title (used if CompletePage is set)
+	CSS   string // Optional CSS file URL (used if CompletePage is set)
+	Icon  string // Optional icon file URL (used if CompletePage is set)
+	Head  []byte // Optional head data injected in the <head> section (used if CompletePage is set)
+
+	Flags Flags // Flags allow customizing this renderer's behavior
+
+	// if set, called at the start of RenderNode(). Allows replacing
+	// rendering of some nodes
+	RenderNodeHook RenderNodeFunc
+
+	// Comments is a list of comments the renderer should detect when
+	// parsing code blocks and detecting callouts.
+	Comments [][]byte
+
+	// Generator is a meta tag that is inserted in the generated HTML so show what rendered it. It should not include the closing tag.
+	// Defaults (note content quote is not closed) to `  <meta name="GENERATOR" content="github.com/gomarkdown/markdown markdown processor for Go`
+	Generator string
+}
+
+// Renderer implements Renderer interface for HTML output.
+//
+// Do not create this directly, instead use the NewRenderer function.
+type Renderer struct {
+	opts RendererOptions
+
+	closeTag string // how to end singleton tags: either " />" or ">"
+
+	// Track heading IDs to prevent ID collision in a single generation.
+	headingIDs map[string]int
+
+	lastOutputLen int
+	disableTags   int
+
+	sr *SPRenderer
+
+	documentMatter ast.DocumentMatters // keep track of front/main/back matter.
+}
+
+// NewRenderer creates and configures an Renderer object, which
+// satisfies the Renderer interface.
+func NewRenderer(opts RendererOptions) *Renderer {
+	// configure the rendering engine
+	closeTag := ">"
+	if opts.Flags&UseXHTML != 0 {
+		closeTag = " />"
+	}
+
+	if opts.FootnoteReturnLinkContents == "" {
+		opts.FootnoteReturnLinkContents = `<sup>[return]</sup>`
+	}
+	if opts.CitationFormatString == "" {
+		opts.CitationFormatString = `<sup>[%s]</sup>`
+	}
+	if opts.Generator == "" {
+		opts.Generator = `  <meta name="GENERATOR" content="github.com/gomarkdown/markdown markdown processor for Go`
+	}
+
+	return &Renderer{
+		opts: opts,
+
+		closeTag:   closeTag,
+		headingIDs: make(map[string]int),
+
+		sr: NewSmartypantsRenderer(opts.Flags),
+	}
+}
+
+func isHTMLTag(tag []byte, tagname string) bool {
+	found, _ := findHTMLTagPos(tag, tagname)
+	return found
+}
+
+// Look for a character, but ignore it when it's in any kind of quotes, it
+// might be JavaScript
+func skipUntilCharIgnoreQuotes(html []byte, start int, char byte) int {
+	inSingleQuote := false
+	inDoubleQuote := false
+	inGraveQuote := false
+	i := start
+	for i < len(html) {
+		switch {
+		case html[i] == char && !inSingleQuote && !inDoubleQuote && !inGraveQuote:
+			return i
+		case html[i] == '\'':
+			inSingleQuote = !inSingleQuote
+		case html[i] == '"':
+			inDoubleQuote = !inDoubleQuote
+		case html[i] == '`':
+			inGraveQuote = !inGraveQuote
+		}
+		i++
+	}
+	return start
+}
+
+func findHTMLTagPos(tag []byte, tagname string) (bool, int) {
+	i := 0
+	if i < len(tag) && tag[0] != '<' {
+		return false, -1
+	}
+	i++
+	i = skipSpace(tag, i)
+
+	if i < len(tag) && tag[i] == '/' {
+		i++
+	}
+
+	i = skipSpace(tag, i)
+	j := 0
+	for ; i < len(tag); i, j = i+1, j+1 {
+		if j >= len(tagname) {
+			break
+		}
+
+		if strings.ToLower(string(tag[i]))[0] != tagname[j] {
+			return false, -1
+		}
+	}
+
+	if i == len(tag) {
+		return false, -1
+	}
+
+	rightAngle := skipUntilCharIgnoreQuotes(tag, i, '>')
+	if rightAngle >= i {
+		return true, rightAngle
+	}
+
+	return false, -1
+}
+
+func isRelativeLink(link []byte) (yes bool) {
+	// a tag begin with '#'
+	if link[0] == '#' {
+		return true
+	}
+
+	// link begin with '/' but not '//', the second maybe a protocol relative link
+	if len(link) >= 2 && link[0] == '/' && link[1] != '/' {
+		return true
+	}
+
+	// only the root '/'
+	if len(link) == 1 && link[0] == '/' {
+		return true
+	}
+
+	// current directory : begin with "./"
+	if bytes.HasPrefix(link, []byte("./")) {
+		return true
+	}
+
+	// parent directory : begin with "../"
+	if bytes.HasPrefix(link, []byte("../")) {
+		return true
+	}
+
+	return false
+}
+
+func (r *Renderer) ensureUniqueHeadingID(id string) string {
+	for count, found := r.headingIDs[id]; found; count, found = r.headingIDs[id] {
+		tmp := fmt.Sprintf("%s-%d", id, count+1)
+
+		if _, tmpFound := r.headingIDs[tmp]; !tmpFound {
+			r.headingIDs[id] = count + 1
+			id = tmp
+		} else {
+			id = id + "-1"
+		}
+	}
+
+	if _, found := r.headingIDs[id]; !found {
+		r.headingIDs[id] = 0
+	}
+
+	return id
+}
+
+func (r *Renderer) addAbsPrefix(link []byte) []byte {
+	if r.opts.AbsolutePrefix != "" && isRelativeLink(link) && link[0] != '.' {
+		newDest := r.opts.AbsolutePrefix
+		if link[0] != '/' {
+			newDest += "/"
+		}
+		newDest += string(link)
+		return []byte(newDest)
+	}
+	return link
+}
+
+func appendLinkAttrs(attrs []string, flags Flags, link []byte) []string {
+	if isRelativeLink(link) {
+		return attrs
+	}
+	var val []string
+	if flags&NofollowLinks != 0 {
+		val = append(val, "nofollow")
+	}
+	if flags&NoreferrerLinks != 0 {
+		val = append(val, "noreferrer")
+	}
+	if flags&HrefTargetBlank != 0 {
+		attrs = append(attrs, `target="_blank"`)
+	}
+	if len(val) == 0 {
+		return attrs
+	}
+	attr := fmt.Sprintf("rel=%q", strings.Join(val, " "))
+	return append(attrs, attr)
+}
+
+func isMailto(link []byte) bool {
+	return bytes.HasPrefix(link, []byte("mailto:"))
+}
+
+func needSkipLink(flags Flags, dest []byte) bool {
+	if flags&SkipLinks != 0 {
+		return true
+	}
+	return flags&Safelink != 0 && !isSafeLink(dest) && !isMailto(dest)
+}
+
+func isSmartypantable(node ast.Node) bool {
+	switch node.GetParent().(type) {
+	case *ast.Link, *ast.CodeBlock, *ast.Code:
+		return false
+	}
+	return true
+}
+
+func appendLanguageAttr(attrs []string, info []byte) []string {
+	if len(info) == 0 {
+		return attrs
+	}
+	endOfLang := bytes.IndexAny(info, "\t ")
+	if endOfLang < 0 {
+		endOfLang = len(info)
+	}
+	s := `class="language-` + string(info[:endOfLang]) + `"`
+	return append(attrs, s)
+}
+
+func (r *Renderer) outTag(w io.Writer, name string, attrs []string) {
+	s := name
+	if len(attrs) > 0 {
+		s += " " + strings.Join(attrs, " ")
+	}
+	io.WriteString(w, s+">")
+	r.lastOutputLen = 1
+}
+
+func footnoteRef(prefix string, node *ast.Link) string {
+	urlFrag := prefix + string(slugify(node.Destination))
+	nStr := strconv.Itoa(node.NoteID)
+	anchor := `<a href="#fn:` + urlFrag + `">` + nStr + `</a>`
+	return `<sup class="footnote-ref" id="fnref:` + urlFrag + `">` + anchor + `</sup>`
+}
+
+func footnoteItem(prefix string, slug []byte) string {
+	return `<li id="fn:` + prefix + string(slug) + `">`
+}
+
+func footnoteReturnLink(prefix, returnLink string, slug []byte) string {
+	return ` <a class="footnote-return" href="#fnref:` + prefix + string(slug) + `">` + returnLink + `</a>`
+}
+
+func listItemOpenCR(listItem *ast.ListItem) bool {
+	if ast.GetPrevNode(listItem) == nil {
+		return false
+	}
+	ld := listItem.Parent.(*ast.List)
+	return !ld.Tight && ld.ListFlags&ast.ListTypeDefinition == 0
+}
+
+func skipParagraphTags(para *ast.Paragraph) bool {
+	parent := para.Parent
+	grandparent := parent.GetParent()
+	if grandparent == nil || !isList(grandparent) {
+		return false
+	}
+	isParentTerm := isListItemTerm(parent)
+	grandparentListData := grandparent.(*ast.List)
+	tightOrTerm := grandparentListData.Tight || isParentTerm
+	return tightOrTerm
+}
+
+func (r *Renderer) out(w io.Writer, d []byte) {
+	r.lastOutputLen = len(d)
+	if r.disableTags > 0 {
+		d = htmlTagRe.ReplaceAll(d, []byte{})
+	}
+	w.Write(d)
+}
+
+func (r *Renderer) outs(w io.Writer, s string) {
+	r.lastOutputLen = len(s)
+	if r.disableTags > 0 {
+		s = htmlTagRe.ReplaceAllString(s, "")
+	}
+	io.WriteString(w, s)
+}
+
+func (r *Renderer) cr(w io.Writer) {
+	if r.lastOutputLen > 0 {
+		r.outs(w, "\n")
+	}
+}
+
+var (
+	openHTags  = []string{"<h1", "<h2", "<h3", "<h4", "<h5"}
+	closeHTags = []string{"</h1>", "</h2>", "</h3>", "</h4>", "</h5>"}
+)
+
+func headingOpenTagFromLevel(level int) string {
+	if level < 1 || level > 5 {
+		return "<h6"
+	}
+	return openHTags[level-1]
+}
+
+func headingCloseTagFromLevel(level int) string {
+	if level < 1 || level > 5 {
+		return "</h6>"
+	}
+	return closeHTags[level-1]
+}
+
+func (r *Renderer) outHRTag(w io.Writer, attrs []string) {
+	hr := tagWithAttributes("<hr", attrs)
+	r.outOneOf(w, r.opts.Flags&UseXHTML == 0, hr, "<hr />")
+}
+
+func (r *Renderer) text(w io.Writer, text *ast.Text) {
+	if r.opts.Flags&Smartypants != 0 {
+		var tmp bytes.Buffer
+		EscapeHTML(&tmp, text.Literal)
+		r.sr.Process(w, tmp.Bytes())
+	} else {
+		_, parentIsLink := text.Parent.(*ast.Link)
+		if parentIsLink {
+			escLink(w, text.Literal)
+		} else {
+			EscapeHTML(w, text.Literal)
+		}
+	}
+}
+
+func (r *Renderer) hardBreak(w io.Writer, node *ast.Hardbreak) {
+	r.outOneOf(w, r.opts.Flags&UseXHTML == 0, "<br>", "<br />")
+	r.cr(w)
+}
+
+func (r *Renderer) nonBlockingSpace(w io.Writer, node *ast.NonBlockingSpace) {
+	r.outs(w, "&nbsp;")
+}
+
+func (r *Renderer) outOneOf(w io.Writer, outFirst bool, first string, second string) {
+	if outFirst {
+		r.outs(w, first)
+	} else {
+		r.outs(w, second)
+	}
+}
+
+func (r *Renderer) outOneOfCr(w io.Writer, outFirst bool, first string, second string) {
+	if outFirst {
+		r.cr(w)
+		r.outs(w, first)
+	} else {
+		r.outs(w, second)
+		r.cr(w)
+	}
+}
+
+func (r *Renderer) htmlSpan(w io.Writer, span *ast.HTMLSpan) {
+	if r.opts.Flags&SkipHTML == 0 {
+		r.out(w, span.Literal)
+	}
+}
+
+func (r *Renderer) linkEnter(w io.Writer, link *ast.Link) {
+	var attrs []string
+	dest := link.Destination
+	dest = r.addAbsPrefix(dest)
+	var hrefBuf bytes.Buffer
+	hrefBuf.WriteString("href=\"")
+	escLink(&hrefBuf, dest)
+	hrefBuf.WriteByte('"')
+	attrs = append(attrs, hrefBuf.String())
+	if link.NoteID != 0 {
+		r.outs(w, footnoteRef(r.opts.FootnoteAnchorPrefix, link))
+		return
+	}
+
+	attrs = appendLinkAttrs(attrs, r.opts.Flags, dest)
+	if len(link.Title) > 0 {
+		var titleBuff bytes.Buffer
+		titleBuff.WriteString("title=\"")
+		EscapeHTML(&titleBuff, link.Title)
+		titleBuff.WriteByte('"')
+		attrs = append(attrs, titleBuff.String())
+	}
+	r.outTag(w, "<a", attrs)
+}
+
+func (r *Renderer) linkExit(w io.Writer, link *ast.Link) {
+	if link.NoteID == 0 {
+		r.outs(w, "</a>")
+	}
+}
+
+func (r *Renderer) link(w io.Writer, link *ast.Link, entering bool) {
+	// mark it but don't link it if it is not a safe link: no smartypants
+	if needSkipLink(r.opts.Flags, link.Destination) {
+		r.outOneOf(w, entering, "<tt>", "</tt>")
+		return
+	}
+
+	if entering {
+		r.linkEnter(w, link)
+	} else {
+		r.linkExit(w, link)
+	}
+}
+
+func (r *Renderer) imageEnter(w io.Writer, image *ast.Image) {
+	dest := image.Destination
+	dest = r.addAbsPrefix(dest)
+	if r.disableTags == 0 {
+		//if options.safe && potentiallyUnsafe(dest) {
+		//out(w, `<img src="" alt="`)
+		//} else {
+		r.outs(w, `<img src="`)
+		escLink(w, dest)
+		r.outs(w, `" alt="`)
+		//}
+	}
+	r.disableTags++
+}
+
+func (r *Renderer) imageExit(w io.Writer, image *ast.Image) {
+	r.disableTags--
+	if r.disableTags == 0 {
+		if image.Title != nil {
+			r.outs(w, `" title="`)
+			EscapeHTML(w, image.Title)
+		}
+		r.outs(w, `" />`)
+	}
+}
+
+func (r *Renderer) paragraphEnter(w io.Writer, para *ast.Paragraph) {
+	// TODO: untangle this clusterfuck about when the newlines need
+	// to be added and when not.
+	prev := ast.GetPrevNode(para)
+	if prev != nil {
+		switch prev.(type) {
+		case *ast.HTMLBlock, *ast.List, *ast.Paragraph, *ast.Heading, *ast.CaptionFigure, *ast.CodeBlock, *ast.BlockQuote, *ast.Aside, *ast.HorizontalRule:
+			r.cr(w)
+		}
+	}
+
+	if prev == nil {
+		_, isParentBlockQuote := para.Parent.(*ast.BlockQuote)
+		if isParentBlockQuote {
+			r.cr(w)
+		}
+		_, isParentAside := para.Parent.(*ast.Aside)
+		if isParentAside {
+			r.cr(w)
+		}
+	}
+
+	tag := tagWithAttributes("<p", BlockAttrs(para))
+	r.outs(w, tag)
+}
+
+func (r *Renderer) paragraphExit(w io.Writer, para *ast.Paragraph) {
+	r.outs(w, "</p>")
+	if !(isListItem(para.Parent) && ast.GetNextNode(para) == nil) {
+		r.cr(w)
+	}
+}
+
+func (r *Renderer) paragraph(w io.Writer, para *ast.Paragraph, entering bool) {
+	if skipParagraphTags(para) {
+		return
+	}
+	if entering {
+		r.paragraphEnter(w, para)
+	} else {
+		r.paragraphExit(w, para)
+	}
+}
+func (r *Renderer) image(w io.Writer, node *ast.Image, entering bool) {
+	if entering {
+		r.imageEnter(w, node)
+	} else {
+		r.imageExit(w, node)
+	}
+}
+
+func (r *Renderer) code(w io.Writer, node *ast.Code) {
+	r.outs(w, "<code>")
+	EscapeHTML(w, node.Literal)
+	r.outs(w, "</code>")
+}
+
+func (r *Renderer) htmlBlock(w io.Writer, node *ast.HTMLBlock) {
+	if r.opts.Flags&SkipHTML != 0 {
+		return
+	}
+	r.cr(w)
+	r.out(w, node.Literal)
+	r.cr(w)
+}
+
+func (r *Renderer) headingEnter(w io.Writer, nodeData *ast.Heading) {
+	var attrs []string
+	var class string
+	// TODO(miek): add helper functions for coalescing these classes.
+	if nodeData.IsTitleblock {
+		class = "title"
+	}
+	if nodeData.IsSpecial {
+		if class != "" {
+			class += " special"
+		} else {
+			class = "special"
+		}
+	}
+	if class != "" {
+		attrs = []string{`class="` + class + `"`}
+	}
+	if nodeData.HeadingID != "" {
+		id := r.ensureUniqueHeadingID(nodeData.HeadingID)
+		if r.opts.HeadingIDPrefix != "" {
+			id = r.opts.HeadingIDPrefix + id
+		}
+		if r.opts.HeadingIDSuffix != "" {
+			id = id + r.opts.HeadingIDSuffix
+		}
+		attrID := `id="` + id + `"`
+		attrs = append(attrs, attrID)
+	}
+	attrs = append(attrs, BlockAttrs(nodeData)...)
+	r.cr(w)
+	r.outTag(w, headingOpenTagFromLevel(nodeData.Level), attrs)
+}
+
+func (r *Renderer) headingExit(w io.Writer, heading *ast.Heading) {
+	r.outs(w, headingCloseTagFromLevel(heading.Level))
+	if !(isListItem(heading.Parent) && ast.GetNextNode(heading) == nil) {
+		r.cr(w)
+	}
+}
+
+func (r *Renderer) heading(w io.Writer, node *ast.Heading, entering bool) {
+	if entering {
+		r.headingEnter(w, node)
+	} else {
+		r.headingExit(w, node)
+	}
+}
+
+func (r *Renderer) horizontalRule(w io.Writer, node *ast.HorizontalRule) {
+	r.cr(w)
+	r.outHRTag(w, BlockAttrs(node))
+	r.cr(w)
+}
+
+func (r *Renderer) listEnter(w io.Writer, nodeData *ast.List) {
+	// TODO: attrs don't seem to be set
+	var attrs []string
+
+	if nodeData.IsFootnotesList {
+		r.outs(w, "\n<div class=\"footnotes\">\n\n")
+		if r.opts.Flags&FootnoteNoHRTag == 0 {
+			r.outHRTag(w, nil)
+			r.cr(w)
+		}
+	}
+	r.cr(w)
+	if isListItem(nodeData.Parent) {
+		grand := nodeData.Parent.GetParent()
+		if isListTight(grand) {
+			r.cr(w)
+		}
+	}
+
+	openTag := "<ul"
+	if nodeData.ListFlags&ast.ListTypeOrdered != 0 {
+		if nodeData.Start > 0 {
+			attrs = append(attrs, fmt.Sprintf(`start="%d"`, nodeData.Start))
+		}
+		openTag = "<ol"
+	}
+	if nodeData.ListFlags&ast.ListTypeDefinition != 0 {
+		openTag = "<dl"
+	}
+	attrs = append(attrs, BlockAttrs(nodeData)...)
+	r.outTag(w, openTag, attrs)
+	r.cr(w)
+}
+
+func (r *Renderer) listExit(w io.Writer, list *ast.List) {
+	closeTag := "</ul>"
+	if list.ListFlags&ast.ListTypeOrdered != 0 {
+		closeTag = "</ol>"
+	}
+	if list.ListFlags&ast.ListTypeDefinition != 0 {
+		closeTag = "</dl>"
+	}
+	r.outs(w, closeTag)
+
+	//cr(w)
+	//if node.parent.Type != Item {
+	//	cr(w)
+	//}
+	parent := list.Parent
+	switch parent.(type) {
+	case *ast.ListItem:
+		if ast.GetNextNode(list) != nil {
+			r.cr(w)
+		}
+	case *ast.Document, *ast.BlockQuote, *ast.Aside:
+		r.cr(w)
+	}
+
+	if list.IsFootnotesList {
+		r.outs(w, "\n</div>\n")
+	}
+}
+
+func (r *Renderer) list(w io.Writer, list *ast.List, entering bool) {
+	if entering {
+		r.listEnter(w, list)
+	} else {
+		r.listExit(w, list)
+	}
+}
+
+func (r *Renderer) listItemEnter(w io.Writer, listItem *ast.ListItem) {
+	if listItemOpenCR(listItem) {
+		r.cr(w)
+	}
+	if listItem.RefLink != nil {
+		slug := slugify(listItem.RefLink)
+		r.outs(w, footnoteItem(r.opts.FootnoteAnchorPrefix, slug))
+		return
+	}
+
+	openTag := "<li>"
+	if listItem.ListFlags&ast.ListTypeDefinition != 0 {
+		openTag = "<dd>"
+	}
+	if listItem.ListFlags&ast.ListTypeTerm != 0 {
+		openTag = "<dt>"
+	}
+	r.outs(w, openTag)
+}
+
+func (r *Renderer) listItemExit(w io.Writer, listItem *ast.ListItem) {
+	if listItem.RefLink != nil && r.opts.Flags&FootnoteReturnLinks != 0 {
+		slug := slugify(listItem.RefLink)
+		prefix := r.opts.FootnoteAnchorPrefix
+		link := r.opts.FootnoteReturnLinkContents
+		s := footnoteReturnLink(prefix, link, slug)
+		r.outs(w, s)
+	}
+
+	closeTag := "</li>"
+	if listItem.ListFlags&ast.ListTypeDefinition != 0 {
+		closeTag = "</dd>"
+	}
+	if listItem.ListFlags&ast.ListTypeTerm != 0 {
+		closeTag = "</dt>"
+	}
+	r.outs(w, closeTag)
+	r.cr(w)
+}
+
+func (r *Renderer) listItem(w io.Writer, listItem *ast.ListItem, entering bool) {
+	if entering {
+		r.listItemEnter(w, listItem)
+	} else {
+		r.listItemExit(w, listItem)
+	}
+}
+
+func (r *Renderer) codeBlock(w io.Writer, codeBlock *ast.CodeBlock) {
+	var attrs []string
+	// TODO(miek): this can add multiple class= attribute, they should be coalesced into one.
+	// This is probably true for some other elements as well
+	attrs = appendLanguageAttr(attrs, codeBlock.Info)
+	attrs = append(attrs, BlockAttrs(codeBlock)...)
+	r.cr(w)
+
+	r.outs(w, "<pre>")
+	code := tagWithAttributes("<code", attrs)
+	r.outs(w, code)
+	if r.opts.Comments != nil {
+		r.EscapeHTMLCallouts(w, codeBlock.Literal)
+	} else {
+		EscapeHTML(w, codeBlock.Literal)
+	}
+	r.outs(w, "</code>")
+	r.outs(w, "</pre>")
+	if !isListItem(codeBlock.Parent) {
+		r.cr(w)
+	}
+}
+
+func (r *Renderer) caption(w io.Writer, caption *ast.Caption, entering bool) {
+	if entering {
+		r.outs(w, "<figcaption>")
+		return
+	}
+	r.outs(w, "</figcaption>")
+}
+
+func (r *Renderer) captionFigure(w io.Writer, figure *ast.CaptionFigure, entering bool) {
+	// TODO(miek): copy more generic ways of mmark over to here.
+	fig := "<figure"
+	if figure.HeadingID != "" {
+		fig += ` id="` + figure.HeadingID + `">`
+	} else {
+		fig += ">"
+	}
+	r.outOneOf(w, entering, fig, "\n</figure>\n")
+}
+
+func (r *Renderer) tableCell(w io.Writer, tableCell *ast.TableCell, entering bool) {
+	if !entering {
+		r.outOneOf(w, tableCell.IsHeader, "</th>", "</td>")
+		r.cr(w)
+		return
+	}
+
+	// entering
+	var attrs []string
+	openTag := "<td"
+	if tableCell.IsHeader {
+		openTag = "<th"
+	}
+	align := tableCell.Align.String()
+	if align != "" {
+		attrs = append(attrs, fmt.Sprintf(`align="%s"`, align))
+	}
+	if ast.GetPrevNode(tableCell) == nil {
+		r.cr(w)
+	}
+	r.outTag(w, openTag, attrs)
+}
+
+func (r *Renderer) tableBody(w io.Writer, node *ast.TableBody, entering bool) {
+	if entering {
+		r.cr(w)
+		r.outs(w, "<tbody>")
+		// XXX: this is to adhere to a rather silly test. Should fix test.
+		if ast.GetFirstChild(node) == nil {
+			r.cr(w)
+		}
+	} else {
+		r.outs(w, "</tbody>")
+		r.cr(w)
+	}
+}
+
+func (r *Renderer) matter(w io.Writer, node *ast.DocumentMatter, entering bool) {
+	if !entering {
+		return
+	}
+	if r.documentMatter != ast.DocumentMatterNone {
+		r.outs(w, "</section>\n")
+	}
+	switch node.Matter {
+	case ast.DocumentMatterFront:
+		r.outs(w, `<section data-matter="front">`)
+	case ast.DocumentMatterMain:
+		r.outs(w, `<section data-matter="main">`)
+	case ast.DocumentMatterBack:
+		r.outs(w, `<section data-matter="back">`)
+	}
+	r.documentMatter = node.Matter
+}
+
+func (r *Renderer) citation(w io.Writer, node *ast.Citation) {
+	for i, c := range node.Destination {
+		attr := []string{`class="none"`}
+		switch node.Type[i] {
+		case ast.CitationTypeNormative:
+			attr[0] = `class="normative"`
+		case ast.CitationTypeInformative:
+			attr[0] = `class="informative"`
+		case ast.CitationTypeSuppressed:
+			attr[0] = `class="suppressed"`
+		}
+		r.outTag(w, "<cite", attr)
+		r.outs(w, fmt.Sprintf(`<a href="#%s">`+r.opts.CitationFormatString+`</a>`, c, c))
+		r.outs(w, "</cite>")
+	}
+}
+
+func (r *Renderer) callout(w io.Writer, node *ast.Callout) {
+	attr := []string{`class="callout"`}
+	r.outTag(w, "<span", attr)
+	r.out(w, node.ID)
+	r.outs(w, "</span>")
+}
+
+func (r *Renderer) index(w io.Writer, node *ast.Index) {
+	// there is no in-text representation.
+	attr := []string{`class="index"`, fmt.Sprintf(`id="%s"`, node.ID)}
+	r.outTag(w, "<span", attr)
+	r.outs(w, "</span>")
+}
+
+// RenderNode renders a markdown node to HTML
+func (r *Renderer) RenderNode(w io.Writer, node ast.Node, entering bool) ast.WalkStatus {
+	if r.opts.RenderNodeHook != nil {
+		status, didHandle := r.opts.RenderNodeHook(w, node, entering)
+		if didHandle {
+			return status
+		}
+	}
+	switch node := node.(type) {
+	case *ast.Text:
+		r.text(w, node)
+	case *ast.Softbreak:
+		r.cr(w)
+		// TODO: make it configurable via out(renderer.softbreak)
+	case *ast.Hardbreak:
+		r.hardBreak(w, node)
+	case *ast.NonBlockingSpace:
+		r.nonBlockingSpace(w, node)
+	case *ast.Emph:
+		r.outOneOf(w, entering, "<em>", "</em>")
+	case *ast.Strong:
+		r.outOneOf(w, entering, "<strong>", "</strong>")
+	case *ast.Del:
+		r.outOneOf(w, entering, "<del>", "</del>")
+	case *ast.BlockQuote:
+		tag := tagWithAttributes("<blockquote", BlockAttrs(node))
+		r.outOneOfCr(w, entering, tag, "</blockquote>")
+	case *ast.Aside:
+		tag := tagWithAttributes("<aside", BlockAttrs(node))
+		r.outOneOfCr(w, entering, tag, "</aside>")
+	case *ast.Link:
+		r.link(w, node, entering)
+	case *ast.CrossReference:
+		link := &ast.Link{Destination: append([]byte("#"), node.Destination...)}
+		r.link(w, link, entering)
+	case *ast.Citation:
+		r.citation(w, node)
+	case *ast.Image:
+		if r.opts.Flags&SkipImages != 0 {
+			return ast.SkipChildren
+		}
+		r.image(w, node, entering)
+	case *ast.Code:
+		r.code(w, node)
+	case *ast.CodeBlock:
+		r.codeBlock(w, node)
+	case *ast.Caption:
+		r.caption(w, node, entering)
+	case *ast.CaptionFigure:
+		r.captionFigure(w, node, entering)
+	case *ast.Document:
+		// do nothing
+	case *ast.Paragraph:
+		r.paragraph(w, node, entering)
+	case *ast.HTMLSpan:
+		r.htmlSpan(w, node)
+	case *ast.HTMLBlock:
+		r.htmlBlock(w, node)
+	case *ast.Heading:
+		r.heading(w, node, entering)
+	case *ast.HorizontalRule:
+		r.horizontalRule(w, node)
+	case *ast.List:
+		r.list(w, node, entering)
+	case *ast.ListItem:
+		r.listItem(w, node, entering)
+	case *ast.Table:
+		tag := tagWithAttributes("<table", BlockAttrs(node))
+		r.outOneOfCr(w, entering, tag, "</table>")
+	case *ast.TableCell:
+		r.tableCell(w, node, entering)
+	case *ast.TableHeader:
+		r.outOneOfCr(w, entering, "<thead>", "</thead>")
+	case *ast.TableBody:
+		r.tableBody(w, node, entering)
+	case *ast.TableRow:
+		r.outOneOfCr(w, entering, "<tr>", "</tr>")
+	case *ast.TableFooter:
+		r.outOneOfCr(w, entering, "<tfoot>", "</tfoot>")
+	case *ast.Math:
+		r.outOneOf(w, true, `<span class="math inline">\(`, `\)</span>`)
+		EscapeHTML(w, node.Literal)
+		r.outOneOf(w, false, `<span class="math inline">\(`, `\)</span>`)
+	case *ast.MathBlock:
+		r.outOneOf(w, entering, `<p><span class="math display">\[`, `\]</span></p>`)
+		if entering {
+			EscapeHTML(w, node.Literal)
+		}
+	case *ast.DocumentMatter:
+		r.matter(w, node, entering)
+	case *ast.Callout:
+		r.callout(w, node)
+	case *ast.Index:
+		r.index(w, node)
+	case *ast.Subscript:
+		r.outOneOf(w, true, "<sub>", "</sub>")
+		if entering {
+			Escape(w, node.Literal)
+		}
+		r.outOneOf(w, false, "<sub>", "</sub>")
+	case *ast.Superscript:
+		r.outOneOf(w, true, "<sup>", "</sup>")
+		if entering {
+			Escape(w, node.Literal)
+		}
+		r.outOneOf(w, false, "<sup>", "</sup>")
+	case *ast.Footnotes:
+		// nothing by default; just output the list.
+	default:
+		panic(fmt.Sprintf("Unknown node %T", node))
+	}
+	return ast.GoToNext
+}
+
+// RenderHeader writes HTML document preamble and TOC if requested.
+func (r *Renderer) RenderHeader(w io.Writer, ast ast.Node) {
+	r.writeDocumentHeader(w)
+	if r.opts.Flags&TOC != 0 {
+		r.writeTOC(w, ast)
+	}
+}
+
+// RenderFooter writes HTML document footer.
+func (r *Renderer) RenderFooter(w io.Writer, _ ast.Node) {
+	if r.documentMatter != ast.DocumentMatterNone {
+		r.outs(w, "</section>\n")
+	}
+
+	if r.opts.Flags&CompletePage == 0 {
+		return
+	}
+	io.WriteString(w, "\n</body>\n</html>\n")
+}
+
+func (r *Renderer) writeDocumentHeader(w io.Writer) {
+	if r.opts.Flags&CompletePage == 0 {
+		return
+	}
+	ending := ""
+	if r.opts.Flags&UseXHTML != 0 {
+		io.WriteString(w, "<!DOCTYPE html PUBLIC \"-//W3C//DTD XHTML 1.0 Transitional//EN\" ")
+		io.WriteString(w, "\"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd\">\n")
+		io.WriteString(w, "<html xmlns=\"http://www.w3.org/1999/xhtml\">\n")
+		ending = " /"
+	} else {
+		io.WriteString(w, "<!DOCTYPE html>\n")
+		io.WriteString(w, "<html>\n")
+	}
+	io.WriteString(w, "<head>\n")
+	io.WriteString(w, "  <title>")
+	if r.opts.Flags&Smartypants != 0 {
+		r.sr.Process(w, []byte(r.opts.Title))
+	} else {
+		EscapeHTML(w, []byte(r.opts.Title))
+	}
+	io.WriteString(w, "</title>\n")
+	io.WriteString(w, r.opts.Generator)
+	io.WriteString(w, "\"")
+	io.WriteString(w, ending)
+	io.WriteString(w, ">\n")
+	io.WriteString(w, "  <meta charset=\"utf-8\"")
+	io.WriteString(w, ending)
+	io.WriteString(w, ">\n")
+	if r.opts.CSS != "" {
+		io.WriteString(w, "  <link rel=\"stylesheet\" type=\"text/css\" href=\"")
+		EscapeHTML(w, []byte(r.opts.CSS))
+		io.WriteString(w, "\"")
+		io.WriteString(w, ending)
+		io.WriteString(w, ">\n")
+	}
+	if r.opts.Icon != "" {
+		io.WriteString(w, "  <link rel=\"icon\" type=\"image/x-icon\" href=\"")
+		EscapeHTML(w, []byte(r.opts.Icon))
+		io.WriteString(w, "\"")
+		io.WriteString(w, ending)
+		io.WriteString(w, ">\n")
+	}
+	if r.opts.Head != nil {
+		w.Write(r.opts.Head)
+	}
+	io.WriteString(w, "</head>\n")
+	io.WriteString(w, "<body>\n\n")
+}
+
+func (r *Renderer) writeTOC(w io.Writer, doc ast.Node) {
+	buf := bytes.Buffer{}
+
+	inHeading := false
+	tocLevel := 0
+	headingCount := 0
+
+	ast.WalkFunc(doc, func(node ast.Node, entering bool) ast.WalkStatus {
+		if nodeData, ok := node.(*ast.Heading); ok && !nodeData.IsTitleblock {
+			inHeading = entering
+			if !entering {
+				buf.WriteString("</a>")
+				return ast.GoToNext
+			}
+			nodeData.HeadingID = fmt.Sprintf("toc_%d", headingCount)
+			if nodeData.Level == tocLevel {
+				buf.WriteString("</li>\n\n<li>")
+			} else if nodeData.Level < tocLevel {
+				for nodeData.Level < tocLevel {
+					tocLevel--
+					buf.WriteString("</li>\n</ul>")
+				}
+				buf.WriteString("</li>\n\n<li>")
+			} else {
+				for nodeData.Level > tocLevel {
+					tocLevel++
+					buf.WriteString("\n<ul>\n<li>")
+				}
+			}
+
+			fmt.Fprintf(&buf, `<a href="#toc_%d">`, headingCount)
+			headingCount++
+			return ast.GoToNext
+		}
+
+		if inHeading {
+			return r.RenderNode(&buf, node, entering)
+		}
+
+		return ast.GoToNext
+	})
+
+	for ; tocLevel > 0; tocLevel-- {
+		buf.WriteString("</li>\n</ul>")
+	}
+
+	if buf.Len() > 0 {
+		io.WriteString(w, "<nav>\n")
+		w.Write(buf.Bytes())
+		io.WriteString(w, "\n\n</nav>\n")
+	}
+	r.lastOutputLen = buf.Len()
+}
+
+func isList(node ast.Node) bool {
+	_, ok := node.(*ast.List)
+	return ok
+}
+
+func isListTight(node ast.Node) bool {
+	if list, ok := node.(*ast.List); ok {
+		return list.Tight
+	}
+	return false
+}
+
+func isListItem(node ast.Node) bool {
+	_, ok := node.(*ast.ListItem)
+	return ok
+}
+
+func isListItemTerm(node ast.Node) bool {
+	data, ok := node.(*ast.ListItem)
+	return ok && data.ListFlags&ast.ListTypeTerm != 0
+}
+
+// TODO: move to internal package
+func skipSpace(data []byte, i int) int {
+	n := len(data)
+	for i < n && isSpace(data[i]) {
+		i++
+	}
+	return i
+}
+
+// TODO: move to internal package
+var validUris = [][]byte{[]byte("http://"), []byte("https://"), []byte("ftp://"), []byte("mailto://")}
+var validPaths = [][]byte{[]byte("/"), []byte("./"), []byte("../")}
+
+func isSafeLink(link []byte) bool {
+	for _, path := range validPaths {
+		if len(link) >= len(path) && bytes.Equal(link[:len(path)], path) {
+			if len(link) == len(path) {
+				return true
+			} else if isAlnum(link[len(path)]) {
+				return true
+			}
+		}
+	}
+
+	for _, prefix := range validUris {
+		// TODO: handle unicode here
+		// case-insensitive prefix test
+		if len(link) > len(prefix) && bytes.Equal(bytes.ToLower(link[:len(prefix)]), prefix) && isAlnum(link[len(prefix)]) {
+			return true
+		}
+	}
+
+	return false
+}
+
+// TODO: move to internal package
+// Create a url-safe slug for fragments
+func slugify(in []byte) []byte {
+	if len(in) == 0 {
+		return in
+	}
+	out := make([]byte, 0, len(in))
+	sym := false
+
+	for _, ch := range in {
+		if isAlnum(ch) {
+			sym = false
+			out = append(out, ch)
+		} else if sym {
+			continue
+		} else {
+			out = append(out, '-')
+			sym = true
+		}
+	}
+	var a, b int
+	var ch byte
+	for a, ch = range out {
+		if ch != '-' {
+			break
+		}
+	}
+	for b = len(out) - 1; b > 0; b-- {
+		if out[b] != '-' {
+			break
+		}
+	}
+	return out[a : b+1]
+}
+
+// TODO: move to internal package
+// isAlnum returns true if c is a digit or letter
+// TODO: check when this is looking for ASCII alnum and when it should use unicode
+func isAlnum(c byte) bool {
+	return (c >= '0' && c <= '9') || isLetter(c)
+}
+
+// isSpace returns true if c is a white-space charactr
+func isSpace(c byte) bool {
+	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
+}
+
+// isLetter returns true if c is ascii letter
+func isLetter(c byte) bool {
+	return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
+}
+
+// isPunctuation returns true if c is a punctuation symbol.
+func isPunctuation(c byte) bool {
+	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
+		if c == r {
+			return true
+		}
+	}
+	return false
+}
+
+// BlockAttrs takes a node and checks if it has block level attributes set. If so it
+// will return a slice each containing a "key=value(s)" string.
+func BlockAttrs(node ast.Node) []string {
+	var attr *ast.Attribute
+	if c := node.AsContainer(); c != nil && c.Attribute != nil {
+		attr = c.Attribute
+	}
+	if l := node.AsLeaf(); l != nil && l.Attribute != nil {
+		attr = l.Attribute
+	}
+	if attr == nil {
+		return nil
+	}
+
+	var s []string
+	if attr.ID != nil {
+		s = append(s, fmt.Sprintf(`%s="%s"`, IDTag, attr.ID))
+	}
+
+	classes := ""
+	for _, c := range attr.Classes {
+		classes += " " + string(c)
+	}
+	if classes != "" {
+		s = append(s, fmt.Sprintf(`class="%s"`, classes[1:])) // skip space we added.
+	}
+
+	// sort the attributes so it remain stable between runs
+	var keys = []string{}
+	for k, _ := range attr.Attrs {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	for _, k := range keys {
+		s = append(s, fmt.Sprintf(`%s="%s"`, k, attr.Attrs[k]))
+	}
+
+	return s
+}
+
+func tagWithAttributes(name string, attrs []string) string {
+	s := name
+	if len(attrs) > 0 {
+		s += " " + strings.Join(attrs, " ")
+	}
+	return s + ">"
+}
diff --git a/vendor/github.com/gomarkdown/markdown/html/smartypants.go b/vendor/github.com/gomarkdown/markdown/html/smartypants.go
new file mode 100644
index 00000000..a09866b0
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/html/smartypants.go
@@ -0,0 +1,444 @@
+package html
+
+import (
+	"bytes"
+	"io"
+)
+
+// SmartyPants rendering
+
+// SPRenderer is a struct containing state of a Smartypants renderer.
+type SPRenderer struct {
+	inSingleQuote bool
+	inDoubleQuote bool
+	callbacks     [256]smartCallback
+}
+
+func wordBoundary(c byte) bool {
+	return c == 0 || isSpace(c) || isPunctuation(c)
+}
+
+func tolower(c byte) byte {
+	if c >= 'A' && c <= 'Z' {
+		return c - 'A' + 'a'
+	}
+	return c
+}
+
+func isdigit(c byte) bool {
+	return c >= '0' && c <= '9'
+}
+
+func smartQuoteHelper(out *bytes.Buffer, previousChar byte, nextChar byte, quote byte, isOpen *bool, addNBSP bool) bool {
+	// edge of the buffer is likely to be a tag that we don't get to see,
+	// so we treat it like text sometimes
+
+	// enumerate all sixteen possibilities for (previousChar, nextChar)
+	// each can be one of {0, space, punct, other}
+	switch {
+	case previousChar == 0 && nextChar == 0:
+		// context is not any help here, so toggle
+		*isOpen = !*isOpen
+	case isSpace(previousChar) && nextChar == 0:
+		// [ "] might be [ "<code>foo...]
+		*isOpen = true
+	case isPunctuation(previousChar) && nextChar == 0:
+		// [!"] hmm... could be [Run!"] or [("<code>...]
+		*isOpen = false
+	case /* isnormal(previousChar) && */ nextChar == 0:
+		// [a"] is probably a close
+		*isOpen = false
+	case previousChar == 0 && isSpace(nextChar):
+		// [" ] might be [...foo</code>" ]
+		*isOpen = false
+	case isSpace(previousChar) && isSpace(nextChar):
+		// [ " ] context is not any help here, so toggle
+		*isOpen = !*isOpen
+	case isPunctuation(previousChar) && isSpace(nextChar):
+		// [!" ] is probably a close
+		*isOpen = false
+	case /* isnormal(previousChar) && */ isSpace(nextChar):
+		// [a" ] this is one of the easy cases
+		*isOpen = false
+	case previousChar == 0 && isPunctuation(nextChar):
+		// ["!] hmm... could be ["$1.95] or [</code>"!...]
+		*isOpen = false
+	case isSpace(previousChar) && isPunctuation(nextChar):
+		// [ "!] looks more like [ "$1.95]
+		*isOpen = true
+	case isPunctuation(previousChar) && isPunctuation(nextChar):
+		// [!"!] context is not any help here, so toggle
+		*isOpen = !*isOpen
+	case /* isnormal(previousChar) && */ isPunctuation(nextChar):
+		// [a"!] is probably a close
+		*isOpen = false
+	case previousChar == 0 /* && isnormal(nextChar) */ :
+		// ["a] is probably an open
+		*isOpen = true
+	case isSpace(previousChar) /* && isnormal(nextChar) */ :
+		// [ "a] this is one of the easy cases
+		*isOpen = true
+	case isPunctuation(previousChar) /* && isnormal(nextChar) */ :
+		// [!"a] is probably an open
+		*isOpen = true
+	default:
+		// [a'b] maybe a contraction?
+		*isOpen = false
+	}
+
+	// Note that with the limited lookahead, this non-breaking
+	// space will also be appended to single double quotes.
+	if addNBSP && !*isOpen {
+		out.WriteString("&nbsp;")
+	}
+
+	out.WriteByte('&')
+	if *isOpen {
+		out.WriteByte('l')
+	} else {
+		out.WriteByte('r')
+	}
+	out.WriteByte(quote)
+	out.WriteString("quo;")
+
+	if addNBSP && *isOpen {
+		out.WriteString("&nbsp;")
+	}
+
+	return true
+}
+
+func (r *SPRenderer) smartSingleQuote(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 2 {
+		t1 := tolower(text[1])
+
+		if t1 == '\'' {
+			nextChar := byte(0)
+			if len(text) >= 3 {
+				nextChar = text[2]
+			}
+			if smartQuoteHelper(out, previousChar, nextChar, 'd', &r.inDoubleQuote, false) {
+				return 1
+			}
+		}
+
+		if (t1 == 's' || t1 == 't' || t1 == 'm' || t1 == 'd') && (len(text) < 3 || wordBoundary(text[2])) {
+			out.WriteString("&rsquo;")
+			return 0
+		}
+
+		if len(text) >= 3 {
+			t2 := tolower(text[2])
+
+			if ((t1 == 'r' && t2 == 'e') || (t1 == 'l' && t2 == 'l') || (t1 == 'v' && t2 == 'e')) &&
+				(len(text) < 4 || wordBoundary(text[3])) {
+				out.WriteString("&rsquo;")
+				return 0
+			}
+		}
+	}
+
+	nextChar := byte(0)
+	if len(text) > 1 {
+		nextChar = text[1]
+	}
+	if smartQuoteHelper(out, previousChar, nextChar, 's', &r.inSingleQuote, false) {
+		return 0
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartParens(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 3 {
+		t1 := tolower(text[1])
+		t2 := tolower(text[2])
+
+		if t1 == 'c' && t2 == ')' {
+			out.WriteString("&copy;")
+			return 2
+		}
+
+		if t1 == 'r' && t2 == ')' {
+			out.WriteString("&reg;")
+			return 2
+		}
+
+		if len(text) >= 4 && t1 == 't' && t2 == 'm' && text[3] == ')' {
+			out.WriteString("&trade;")
+			return 3
+		}
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartDash(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 2 {
+		if text[1] == '-' {
+			out.WriteString("&mdash;")
+			return 1
+		}
+
+		if wordBoundary(previousChar) && wordBoundary(text[1]) {
+			out.WriteString("&ndash;")
+			return 0
+		}
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartDashLatex(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 3 && text[1] == '-' && text[2] == '-' {
+		out.WriteString("&mdash;")
+		return 2
+	}
+	if len(text) >= 2 && text[1] == '-' {
+		out.WriteString("&ndash;")
+		return 1
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartAmpVariant(out *bytes.Buffer, previousChar byte, text []byte, quote byte, addNBSP bool) int {
+	if bytes.HasPrefix(text, []byte("&quot;")) {
+		nextChar := byte(0)
+		if len(text) >= 7 {
+			nextChar = text[6]
+		}
+		if smartQuoteHelper(out, previousChar, nextChar, quote, &r.inDoubleQuote, addNBSP) {
+			return 5
+		}
+	}
+
+	if bytes.HasPrefix(text, []byte("&#0;")) {
+		return 3
+	}
+
+	out.WriteByte('&')
+	return 0
+}
+
+func (r *SPRenderer) smartAmp(angledQuotes, addNBSP bool) func(*bytes.Buffer, byte, []byte) int {
+	var quote byte = 'd'
+	if angledQuotes {
+		quote = 'a'
+	}
+
+	return func(out *bytes.Buffer, previousChar byte, text []byte) int {
+		return r.smartAmpVariant(out, previousChar, text, quote, addNBSP)
+	}
+}
+
+func (r *SPRenderer) smartPeriod(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 3 && text[1] == '.' && text[2] == '.' {
+		out.WriteString("&hellip;")
+		return 2
+	}
+
+	if len(text) >= 5 && text[1] == ' ' && text[2] == '.' && text[3] == ' ' && text[4] == '.' {
+		out.WriteString("&hellip;")
+		return 4
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartBacktick(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if len(text) >= 2 && text[1] == '`' {
+		nextChar := byte(0)
+		if len(text) >= 3 {
+			nextChar = text[2]
+		}
+		if smartQuoteHelper(out, previousChar, nextChar, 'd', &r.inDoubleQuote, false) {
+			return 1
+		}
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartNumberGeneric(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if wordBoundary(previousChar) && previousChar != '/' && len(text) >= 3 {
+		// is it of the form digits/digits(word boundary)?, i.e., \d+/\d+\b
+		// note: check for regular slash (/) or fraction slash (⁄, 0x2044, or 0xe2 81 84 in utf-8)
+		//       and avoid changing dates like 1/23/2005 into fractions.
+		numEnd := 0
+		for len(text) > numEnd && isdigit(text[numEnd]) {
+			numEnd++
+		}
+		if numEnd == 0 {
+			out.WriteByte(text[0])
+			return 0
+		}
+		denStart := numEnd + 1
+		if len(text) > numEnd+3 && text[numEnd] == 0xe2 && text[numEnd+1] == 0x81 && text[numEnd+2] == 0x84 {
+			denStart = numEnd + 3
+		} else if len(text) < numEnd+2 || text[numEnd] != '/' {
+			out.WriteByte(text[0])
+			return 0
+		}
+		denEnd := denStart
+		for len(text) > denEnd && isdigit(text[denEnd]) {
+			denEnd++
+		}
+		if denEnd == denStart {
+			out.WriteByte(text[0])
+			return 0
+		}
+		if len(text) == denEnd || wordBoundary(text[denEnd]) && text[denEnd] != '/' {
+			out.WriteString("<sup>")
+			out.Write(text[:numEnd])
+			out.WriteString("</sup>&frasl;<sub>")
+			out.Write(text[denStart:denEnd])
+			out.WriteString("</sub>")
+			return denEnd - 1
+		}
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartNumber(out *bytes.Buffer, previousChar byte, text []byte) int {
+	if wordBoundary(previousChar) && previousChar != '/' && len(text) >= 3 {
+		if text[0] == '1' && text[1] == '/' && text[2] == '2' {
+			if len(text) < 4 || wordBoundary(text[3]) && text[3] != '/' {
+				out.WriteString("&frac12;")
+				return 2
+			}
+		}
+
+		if text[0] == '1' && text[1] == '/' && text[2] == '4' {
+			if len(text) < 4 || wordBoundary(text[3]) && text[3] != '/' || (len(text) >= 5 && tolower(text[3]) == 't' && tolower(text[4]) == 'h') {
+				out.WriteString("&frac14;")
+				return 2
+			}
+		}
+
+		if text[0] == '3' && text[1] == '/' && text[2] == '4' {
+			if len(text) < 4 || wordBoundary(text[3]) && text[3] != '/' || (len(text) >= 6 && tolower(text[3]) == 't' && tolower(text[4]) == 'h' && tolower(text[5]) == 's') {
+				out.WriteString("&frac34;")
+				return 2
+			}
+		}
+	}
+
+	out.WriteByte(text[0])
+	return 0
+}
+
+func (r *SPRenderer) smartDoubleQuoteVariant(out *bytes.Buffer, previousChar byte, text []byte, quote byte) int {
+	nextChar := byte(0)
+	if len(text) > 1 {
+		nextChar = text[1]
+	}
+	if !smartQuoteHelper(out, previousChar, nextChar, quote, &r.inDoubleQuote, false) {
+		out.WriteString("&quot;")
+	}
+
+	return 0
+}
+
+func (r *SPRenderer) smartDoubleQuote(out *bytes.Buffer, previousChar byte, text []byte) int {
+	return r.smartDoubleQuoteVariant(out, previousChar, text, 'd')
+}
+
+func (r *SPRenderer) smartAngledDoubleQuote(out *bytes.Buffer, previousChar byte, text []byte) int {
+	return r.smartDoubleQuoteVariant(out, previousChar, text, 'a')
+}
+
+func (r *SPRenderer) smartLeftAngle(out *bytes.Buffer, previousChar byte, text []byte) int {
+	i := 0
+
+	for i < len(text) && text[i] != '>' {
+		i++
+	}
+
+	out.Write(text[:i+1])
+	return i
+}
+
+type smartCallback func(out *bytes.Buffer, previousChar byte, text []byte) int
+
+// NewSmartypantsRenderer constructs a Smartypants renderer object.
+func NewSmartypantsRenderer(flags Flags) *SPRenderer {
+	var (
+		r SPRenderer
+
+		smartAmpAngled      = r.smartAmp(true, false)
+		smartAmpAngledNBSP  = r.smartAmp(true, true)
+		smartAmpRegular     = r.smartAmp(false, false)
+		smartAmpRegularNBSP = r.smartAmp(false, true)
+
+		addNBSP = flags&SmartypantsQuotesNBSP != 0
+	)
+
+	if flags&SmartypantsAngledQuotes == 0 {
+		r.callbacks['"'] = r.smartDoubleQuote
+		if !addNBSP {
+			r.callbacks['&'] = smartAmpRegular
+		} else {
+			r.callbacks['&'] = smartAmpRegularNBSP
+		}
+	} else {
+		r.callbacks['"'] = r.smartAngledDoubleQuote
+		if !addNBSP {
+			r.callbacks['&'] = smartAmpAngled
+		} else {
+			r.callbacks['&'] = smartAmpAngledNBSP
+		}
+	}
+	r.callbacks['\''] = r.smartSingleQuote
+	r.callbacks['('] = r.smartParens
+	if flags&SmartypantsDashes != 0 {
+		if flags&SmartypantsLatexDashes == 0 {
+			r.callbacks['-'] = r.smartDash
+		} else {
+			r.callbacks['-'] = r.smartDashLatex
+		}
+	}
+	r.callbacks['.'] = r.smartPeriod
+	if flags&SmartypantsFractions == 0 {
+		r.callbacks['1'] = r.smartNumber
+		r.callbacks['3'] = r.smartNumber
+	} else {
+		for ch := '1'; ch <= '9'; ch++ {
+			r.callbacks[ch] = r.smartNumberGeneric
+		}
+	}
+	r.callbacks['<'] = r.smartLeftAngle
+	r.callbacks['`'] = r.smartBacktick
+	return &r
+}
+
+// Process is the entry point of the Smartypants renderer.
+func (r *SPRenderer) Process(w io.Writer, text []byte) {
+	mark := 0
+	for i := 0; i < len(text); i++ {
+		if action := r.callbacks[text[i]]; action != nil {
+			if i > mark {
+				w.Write(text[mark:i])
+			}
+			previousChar := byte(0)
+			if i > 0 {
+				previousChar = text[i-1]
+			}
+			var tmp bytes.Buffer
+			i += action(&tmp, previousChar, text[i:])
+			w.Write(tmp.Bytes())
+			mark = i + 1
+		}
+	}
+	if mark < len(text) {
+		w.Write(text[mark:])
+	}
+}
diff --git a/vendor/github.com/gomarkdown/markdown/markdown.go b/vendor/github.com/gomarkdown/markdown/markdown.go
new file mode 100644
index 00000000..fd5c1cfb
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/markdown.go
@@ -0,0 +1,85 @@
+package markdown
+
+import (
+	"bytes"
+	"io"
+
+	"github.com/gomarkdown/markdown/ast"
+	"github.com/gomarkdown/markdown/html"
+	"github.com/gomarkdown/markdown/parser"
+)
+
+// Renderer is an interface for implementing custom renderers.
+type Renderer interface {
+	// RenderNode renders markdown node to w.
+	// It's called once for a leaf node.
+	// It's called twice for non-leaf nodes:
+	// * first with entering=true
+	// * then with entering=false
+	//
+	// Return value is a way to tell the calling walker to adjust its walk
+	// pattern: e.g. it can terminate the traversal by returning Terminate. Or it
+	// can ask the walker to skip a subtree of this node by returning SkipChildren.
+	// The typical behavior is to return GoToNext, which asks for the usual
+	// traversal to the next node.
+	RenderNode(w io.Writer, node ast.Node, entering bool) ast.WalkStatus
+
+	// RenderHeader is a method that allows the renderer to produce some
+	// content preceding the main body of the output document. The header is
+	// understood in the broad sense here. For example, the default HTML
+	// renderer will write not only the HTML document preamble, but also the
+	// table of contents if it was requested.
+	//
+	// The method will be passed an entire document tree, in case a particular
+	// implementation needs to inspect it to produce output.
+	//
+	// The output should be written to the supplied writer w. If your
+	// implementation has no header to write, supply an empty implementation.
+	RenderHeader(w io.Writer, ast ast.Node)
+
+	// RenderFooter is a symmetric counterpart of RenderHeader.
+	RenderFooter(w io.Writer, ast ast.Node)
+}
+
+// Parse parsers a markdown document using provided parser. If parser is nil,
+// we use parser configured with parser.CommonExtensions.
+//
+// It returns AST (abstract syntax tree) that can be converted to another
+// format using Render function.
+func Parse(markdown []byte, p *parser.Parser) ast.Node {
+	if p == nil {
+		p = parser.New()
+	}
+	return p.Parse(markdown)
+}
+
+// Render uses renderer to convert parsed markdown document into a different format.
+//
+// To convert to HTML, pass html.Renderer
+func Render(doc ast.Node, renderer Renderer) []byte {
+	var buf bytes.Buffer
+	renderer.RenderHeader(&buf, doc)
+	ast.WalkFunc(doc, func(node ast.Node, entering bool) ast.WalkStatus {
+		return renderer.RenderNode(&buf, node, entering)
+	})
+	renderer.RenderFooter(&buf, doc)
+	return buf.Bytes()
+}
+
+// ToHTML converts markdownDoc to HTML.
+//
+// You can optionally pass a parser and renderer. This allows to customize
+// a parser, use a customized html render or use completely custom renderer.
+//
+// If you pass nil for both, we use parser configured with parser.CommonExtensions
+// and html.Renderer configured with html.CommonFlags.
+func ToHTML(markdown []byte, p *parser.Parser, renderer Renderer) []byte {
+	doc := Parse(markdown, p)
+	if renderer == nil {
+		opts := html.RendererOptions{
+			Flags: html.CommonFlags,
+		}
+		renderer = html.NewRenderer(opts)
+	}
+	return Render(doc, renderer)
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/aside.go b/vendor/github.com/gomarkdown/markdown/parser/aside.go
new file mode 100644
index 00000000..96e25fe0
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/aside.go
@@ -0,0 +1,73 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// returns aisde prefix length
+func (p *Parser) asidePrefix(data []byte) int {
+	i := 0
+	n := len(data)
+	for i < 3 && i < n && data[i] == ' ' {
+		i++
+	}
+	if i+1 < n && data[i] == 'A' && data[i+1] == '>' {
+		if i+2 < n && data[i+2] == ' ' {
+			return i + 3
+		}
+		return i + 2
+	}
+	return 0
+}
+
+// aside ends with at least one blank line
+// followed by something without a aside prefix
+func (p *Parser) terminateAside(data []byte, beg, end int) bool {
+	if p.isEmpty(data[beg:]) <= 0 {
+		return false
+	}
+	if end >= len(data) {
+		return true
+	}
+	return p.asidePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0
+}
+
+// parse a aside fragment
+func (p *Parser) aside(data []byte) int {
+	var raw bytes.Buffer
+	beg, end := 0, 0
+	// identical to quote
+	for beg < len(data) {
+		end = beg
+		// Step over whole lines, collecting them. While doing that, check for
+		// fenced code and if one's found, incorporate it altogether,
+		// irregardless of any contents inside it
+		for end < len(data) && data[end] != '\n' {
+			if p.extensions&FencedCode != 0 {
+				if i := p.fencedCodeBlock(data[end:], false); i > 0 {
+					// -1 to compensate for the extra end++ after the loop:
+					end += i - 1
+					break
+				}
+			}
+			end++
+		}
+		end = skipCharN(data, end, '\n', 1)
+		if pre := p.asidePrefix(data[beg:]); pre > 0 {
+			// skip the prefix
+			beg += pre
+		} else if p.terminateAside(data, beg, end) {
+			break
+		}
+		// this line is part of the aside
+		raw.Write(data[beg:end])
+		beg = end
+	}
+
+	block := p.addBlock(&ast.Aside{})
+	p.block(raw.Bytes())
+	p.finalize(block)
+	return end
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/attribute.go b/vendor/github.com/gomarkdown/markdown/parser/attribute.go
new file mode 100644
index 00000000..5fdb0709
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/attribute.go
@@ -0,0 +1,116 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// attribute parses a (potential) block attribute and adds it to p.
+func (p *Parser) attribute(data []byte) []byte {
+	if len(data) < 3 {
+		return data
+	}
+	i := 0
+	if data[i] != '{' {
+		return data
+	}
+	i++
+
+	// last character must be a } otherwise it's not an attribute
+	end := skipUntilChar(data, i, '\n')
+	if data[end-1] != '}' {
+		return data
+	}
+
+	i = skipSpace(data, i)
+	b := &ast.Attribute{Attrs: make(map[string][]byte)}
+
+	esc := false
+	quote := false
+	trail := 0
+Loop:
+	for ; i < len(data); i++ {
+		switch data[i] {
+		case ' ', '\t', '\f', '\v':
+			if quote {
+				continue
+			}
+			chunk := data[trail+1 : i]
+			if len(chunk) == 0 {
+				trail = i
+				continue
+			}
+			switch {
+			case chunk[0] == '.':
+				b.Classes = append(b.Classes, chunk[1:])
+			case chunk[0] == '#':
+				b.ID = chunk[1:]
+			default:
+				k, v := keyValue(chunk)
+				if k != nil && v != nil {
+					b.Attrs[string(k)] = v
+				} else {
+					// this is illegal in an attribute
+					return data
+				}
+			}
+			trail = i
+		case '"':
+			if esc {
+				esc = !esc
+				continue
+			}
+			quote = !quote
+		case '\\':
+			esc = !esc
+		case '}':
+			if esc {
+				esc = !esc
+				continue
+			}
+			chunk := data[trail+1 : i]
+			if len(chunk) == 0 {
+				return data
+			}
+			switch {
+			case chunk[0] == '.':
+				b.Classes = append(b.Classes, chunk[1:])
+			case chunk[0] == '#':
+				b.ID = chunk[1:]
+			default:
+				k, v := keyValue(chunk)
+				if k != nil && v != nil {
+					b.Attrs[string(k)] = v
+				} else {
+					return data
+				}
+			}
+			i++
+			break Loop
+		default:
+			esc = false
+		}
+	}
+
+	p.attr = b
+	return data[i:]
+}
+
+// key="value" quotes are mandatory.
+func keyValue(data []byte) ([]byte, []byte) {
+	chunk := bytes.SplitN(data, []byte{'='}, 2)
+	if len(chunk) != 2 {
+		return nil, nil
+	}
+	key := chunk[0]
+	value := chunk[1]
+
+	if len(value) < 3 || len(key) == 0 {
+		return nil, nil
+	}
+	if value[0] != '"' || value[len(value)-1] != '"' {
+		return key, nil
+	}
+	return key, value[1 : len(value)-1]
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/block.go b/vendor/github.com/gomarkdown/markdown/parser/block.go
new file mode 100644
index 00000000..18a2dd89
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/block.go
@@ -0,0 +1,1978 @@
+package parser
+
+import (
+	"bytes"
+	"html"
+	"regexp"
+	"strconv"
+	"unicode"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// Parsing block-level elements.
+
+const (
+	charEntity = "&(?:#x[a-f0-9]{1,8}|#[0-9]{1,8}|[a-z][a-z0-9]{1,31});"
+	escapable  = "[!\"#$%&'()*+,./:;<=>?@[\\\\\\]^_`{|}~-]"
+)
+
+var (
+	reBackslashOrAmp      = regexp.MustCompile("[\\&]")
+	reEntityOrEscapedChar = regexp.MustCompile("(?i)\\\\" + escapable + "|" + charEntity)
+
+	// blockTags is a set of tags that are recognized as HTML block tags.
+	// Any of these can be included in markdown text without special escaping.
+	blockTags = map[string]struct{}{
+		"blockquote": struct{}{},
+		"del":        struct{}{},
+		"div":        struct{}{},
+		"dl":         struct{}{},
+		"fieldset":   struct{}{},
+		"form":       struct{}{},
+		"h1":         struct{}{},
+		"h2":         struct{}{},
+		"h3":         struct{}{},
+		"h4":         struct{}{},
+		"h5":         struct{}{},
+		"h6":         struct{}{},
+		"iframe":     struct{}{},
+		"ins":        struct{}{},
+		"math":       struct{}{},
+		"noscript":   struct{}{},
+		"ol":         struct{}{},
+		"pre":        struct{}{},
+		"p":          struct{}{},
+		"script":     struct{}{},
+		"style":      struct{}{},
+		"table":      struct{}{},
+		"ul":         struct{}{},
+
+		// HTML5
+		"address":    struct{}{},
+		"article":    struct{}{},
+		"aside":      struct{}{},
+		"canvas":     struct{}{},
+		"figcaption": struct{}{},
+		"figure":     struct{}{},
+		"footer":     struct{}{},
+		"header":     struct{}{},
+		"hgroup":     struct{}{},
+		"main":       struct{}{},
+		"nav":        struct{}{},
+		"output":     struct{}{},
+		"progress":   struct{}{},
+		"section":    struct{}{},
+		"video":      struct{}{},
+	}
+)
+
+// sanitizeAnchorName returns a sanitized anchor name for the given text.
+// Taken from https://github.com/shurcooL/sanitized_anchor_name/blob/master/main.go#L14:1
+func sanitizeAnchorName(text string) string {
+	var anchorName []rune
+	var futureDash = false
+	for _, r := range text {
+		switch {
+		case unicode.IsLetter(r) || unicode.IsNumber(r):
+			if futureDash && len(anchorName) > 0 {
+				anchorName = append(anchorName, '-')
+			}
+			futureDash = false
+			anchorName = append(anchorName, unicode.ToLower(r))
+		default:
+			futureDash = true
+		}
+	}
+	return string(anchorName)
+}
+
+// Parse block-level data.
+// Note: this function and many that it calls assume that
+// the input buffer ends with a newline.
+func (p *Parser) block(data []byte) {
+	// this is called recursively: enforce a maximum depth
+	if p.nesting >= p.maxNesting {
+		return
+	}
+	p.nesting++
+
+	// parse out one block-level construct at a time
+	for len(data) > 0 {
+		// attributes that can be specific before a block element:
+		//
+		// {#id .class1 .class2 key="value"}
+		if p.extensions&Attributes != 0 {
+			data = p.attribute(data)
+		}
+
+		if p.extensions&Includes != 0 {
+			f := p.readInclude
+			path, address, consumed := p.isInclude(data)
+			if consumed == 0 {
+				path, address, consumed = p.isCodeInclude(data)
+				f = p.readCodeInclude
+			}
+			if consumed > 0 {
+				included := f(p.includeStack.Last(), path, address)
+				p.includeStack.Push(path)
+				p.block(included)
+				p.includeStack.Pop()
+				data = data[consumed:]
+				continue
+			}
+		}
+
+		// user supplied parser function
+		if p.Opts.ParserHook != nil {
+			node, blockdata, consumed := p.Opts.ParserHook(data)
+			if consumed > 0 {
+				data = data[consumed:]
+
+				if node != nil {
+					p.addBlock(node)
+					if blockdata != nil {
+						p.block(blockdata)
+						p.finalize(node)
+					}
+				}
+				continue
+			}
+		}
+
+		// prefixed heading:
+		//
+		// # Heading 1
+		// ## Heading 2
+		// ...
+		// ###### Heading 6
+		if p.isPrefixHeading(data) {
+			data = data[p.prefixHeading(data):]
+			continue
+		}
+
+		// prefixed special heading:
+		// (there are no levels.)
+		//
+		// .# Abstract
+		if p.isPrefixSpecialHeading(data) {
+			data = data[p.prefixSpecialHeading(data):]
+			continue
+		}
+
+		// block of preformatted HTML:
+		//
+		// <div>
+		//     ...
+		// </div>
+		if data[0] == '<' {
+			if i := p.html(data, true); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// title block
+		//
+		// % stuff
+		// % more stuff
+		// % even more stuff
+		if p.extensions&Titleblock != 0 {
+			if data[0] == '%' {
+				if i := p.titleBlock(data, true); i > 0 {
+					data = data[i:]
+					continue
+				}
+			}
+		}
+
+		// blank lines.  note: returns the # of bytes to skip
+		if i := p.isEmpty(data); i > 0 {
+			data = data[i:]
+			continue
+		}
+
+		// indented code block:
+		//
+		//     func max(a, b int) int {
+		//         if a > b {
+		//             return a
+		//         }
+		//         return b
+		//      }
+		if p.codePrefix(data) > 0 {
+			data = data[p.code(data):]
+			continue
+		}
+
+		// fenced code block:
+		//
+		// ``` go
+		// func fact(n int) int {
+		//     if n <= 1 {
+		//         return n
+		//     }
+		//     return n * fact(n-1)
+		// }
+		// ```
+		if p.extensions&FencedCode != 0 {
+			if i := p.fencedCodeBlock(data, true); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// horizontal rule:
+		//
+		// ------
+		// or
+		// ******
+		// or
+		// ______
+		if p.isHRule(data) {
+			p.addBlock(&ast.HorizontalRule{})
+			i := skipUntilChar(data, 0, '\n')
+			data = data[i:]
+			continue
+		}
+
+		// block quote:
+		//
+		// > A big quote I found somewhere
+		// > on the web
+		if p.quotePrefix(data) > 0 {
+			data = data[p.quote(data):]
+			continue
+		}
+
+		// aside:
+		//
+		// A> The proof is too large to fit
+		// A> in the margin.
+		if p.extensions&Mmark != 0 {
+			if p.asidePrefix(data) > 0 {
+				data = data[p.aside(data):]
+				continue
+			}
+		}
+
+		// figure block:
+		//
+		// !---
+		// ![Alt Text](img.jpg "This is an image")
+		// ![Alt Text](img2.jpg "This is a second image")
+		// !---
+		if p.extensions&Mmark != 0 {
+			if i := p.figureBlock(data, true); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// table:
+		//
+		// Name  | Age | Phone
+		// ------|-----|---------
+		// Bob   | 31  | 555-1234
+		// Alice | 27  | 555-4321
+		if p.extensions&Tables != 0 {
+			if i := p.table(data); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// an itemized/unordered list:
+		//
+		// * Item 1
+		// * Item 2
+		//
+		// also works with + or -
+		if p.uliPrefix(data) > 0 {
+			data = data[p.list(data, 0, 0):]
+			continue
+		}
+
+		// a numbered/ordered list:
+		//
+		// 1. Item 1
+		// 2. Item 2
+		if i := p.oliPrefix(data); i > 0 {
+			start := 0
+			if i > 2 && p.extensions&OrderedListStart != 0 {
+				s := string(data[:i-2])
+				start, _ = strconv.Atoi(s)
+				if start == 1 {
+					start = 0
+				}
+			}
+			data = data[p.list(data, ast.ListTypeOrdered, start):]
+			continue
+		}
+
+		// definition lists:
+		//
+		// Term 1
+		// :   Definition a
+		// :   Definition b
+		//
+		// Term 2
+		// :   Definition c
+		if p.extensions&DefinitionLists != 0 {
+			if p.dliPrefix(data) > 0 {
+				data = data[p.list(data, ast.ListTypeDefinition, 0):]
+				continue
+			}
+		}
+
+		if p.extensions&MathJax != 0 {
+			if i := p.blockMath(data); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// document matters:
+		//
+		// {frontmatter}/{mainmatter}/{backmatter}
+		if p.extensions&Mmark != 0 {
+			if i := p.documentMatter(data); i > 0 {
+				data = data[i:]
+				continue
+			}
+		}
+
+		// anything else must look like a normal paragraph
+		// note: this finds underlined headings, too
+		idx := p.paragraph(data)
+		data = data[idx:]
+	}
+
+	p.nesting--
+}
+
+func (p *Parser) addBlock(n ast.Node) ast.Node {
+	p.closeUnmatchedBlocks()
+
+	if p.attr != nil {
+		if c := n.AsContainer(); c != nil {
+			c.Attribute = p.attr
+		}
+		if l := n.AsLeaf(); l != nil {
+			l.Attribute = p.attr
+		}
+		p.attr = nil
+	}
+	return p.addChild(n)
+}
+
+func (p *Parser) isPrefixHeading(data []byte) bool {
+	if data[0] != '#' {
+		return false
+	}
+
+	if p.extensions&SpaceHeadings != 0 {
+		level := skipCharN(data, 0, '#', 6)
+		if level == len(data) || data[level] != ' ' {
+			return false
+		}
+	}
+	return true
+}
+
+func (p *Parser) prefixHeading(data []byte) int {
+	level := skipCharN(data, 0, '#', 6)
+	i := skipChar(data, level, ' ')
+	end := skipUntilChar(data, i, '\n')
+	skip := end
+	id := ""
+	if p.extensions&HeadingIDs != 0 {
+		j, k := 0, 0
+		// find start/end of heading id
+		for j = i; j < end-1 && (data[j] != '{' || data[j+1] != '#'); j++ {
+		}
+		for k = j + 1; k < end && data[k] != '}'; k++ {
+		}
+		// extract heading id iff found
+		if j < end && k < end {
+			id = string(data[j+2 : k])
+			end = j
+			skip = k + 1
+			for end > 0 && data[end-1] == ' ' {
+				end--
+			}
+		}
+	}
+	for end > 0 && data[end-1] == '#' {
+		if isBackslashEscaped(data, end-1) {
+			break
+		}
+		end--
+	}
+	for end > 0 && data[end-1] == ' ' {
+		end--
+	}
+	if end > i {
+		if id == "" && p.extensions&AutoHeadingIDs != 0 {
+			id = sanitizeAnchorName(string(data[i:end]))
+		}
+		block := &ast.Heading{
+			HeadingID: id,
+			Level:     level,
+		}
+		block.Content = data[i:end]
+		p.addBlock(block)
+	}
+	return skip
+}
+
+func (p *Parser) isPrefixSpecialHeading(data []byte) bool {
+	if p.extensions|Mmark == 0 {
+		return false
+	}
+	if len(data) < 4 {
+		return false
+	}
+	if data[0] != '.' {
+		return false
+	}
+	if data[1] != '#' {
+		return false
+	}
+	if data[2] == '#' { // we don't support level, so nack this.
+		return false
+	}
+
+	if p.extensions&SpaceHeadings != 0 {
+		if data[2] != ' ' {
+			return false
+		}
+	}
+	return true
+}
+
+func (p *Parser) prefixSpecialHeading(data []byte) int {
+	i := skipChar(data, 2, ' ') // ".#" skipped
+	end := skipUntilChar(data, i, '\n')
+	skip := end
+	id := ""
+	if p.extensions&HeadingIDs != 0 {
+		j, k := 0, 0
+		// find start/end of heading id
+		for j = i; j < end-1 && (data[j] != '{' || data[j+1] != '#'); j++ {
+		}
+		for k = j + 1; k < end && data[k] != '}'; k++ {
+		}
+		// extract heading id iff found
+		if j < end && k < end {
+			id = string(data[j+2 : k])
+			end = j
+			skip = k + 1
+			for end > 0 && data[end-1] == ' ' {
+				end--
+			}
+		}
+	}
+	for end > 0 && data[end-1] == '#' {
+		if isBackslashEscaped(data, end-1) {
+			break
+		}
+		end--
+	}
+	for end > 0 && data[end-1] == ' ' {
+		end--
+	}
+	if end > i {
+		if id == "" && p.extensions&AutoHeadingIDs != 0 {
+			id = sanitizeAnchorName(string(data[i:end]))
+		}
+		block := &ast.Heading{
+			HeadingID: id,
+			IsSpecial: true,
+			Level:     1, // always level 1.
+		}
+		block.Literal = data[i:end]
+		block.Content = data[i:end]
+		p.addBlock(block)
+	}
+	return skip
+}
+
+func (p *Parser) isUnderlinedHeading(data []byte) int {
+	// test of level 1 heading
+	if data[0] == '=' {
+		i := skipChar(data, 1, '=')
+		i = skipChar(data, i, ' ')
+		if i < len(data) && data[i] == '\n' {
+			return 1
+		}
+		return 0
+	}
+
+	// test of level 2 heading
+	if data[0] == '-' {
+		i := skipChar(data, 1, '-')
+		i = skipChar(data, i, ' ')
+		if i < len(data) && data[i] == '\n' {
+			return 2
+		}
+		return 0
+	}
+
+	return 0
+}
+
+func (p *Parser) titleBlock(data []byte, doRender bool) int {
+	if data[0] != '%' {
+		return 0
+	}
+	splitData := bytes.Split(data, []byte("\n"))
+	var i int
+	for idx, b := range splitData {
+		if !bytes.HasPrefix(b, []byte("%")) {
+			i = idx // - 1
+			break
+		}
+	}
+
+	data = bytes.Join(splitData[0:i], []byte("\n"))
+	consumed := len(data)
+	data = bytes.TrimPrefix(data, []byte("% "))
+	data = bytes.Replace(data, []byte("\n% "), []byte("\n"), -1)
+	block := &ast.Heading{
+		Level:        1,
+		IsTitleblock: true,
+	}
+	block.Content = data
+	p.addBlock(block)
+
+	return consumed
+}
+
+func (p *Parser) html(data []byte, doRender bool) int {
+	var i, j int
+
+	// identify the opening tag
+	if data[0] != '<' {
+		return 0
+	}
+	curtag, tagfound := p.htmlFindTag(data[1:])
+
+	// handle special cases
+	if !tagfound {
+		// check for an HTML comment
+		if size := p.htmlComment(data, doRender); size > 0 {
+			return size
+		}
+
+		// check for an <hr> tag
+		if size := p.htmlHr(data, doRender); size > 0 {
+			return size
+		}
+
+		// no special case recognized
+		return 0
+	}
+
+	// look for an unindented matching closing tag
+	// followed by a blank line
+	found := false
+	/*
+		closetag := []byte("\n</" + curtag + ">")
+		j = len(curtag) + 1
+		for !found {
+			// scan for a closing tag at the beginning of a line
+			if skip := bytes.Index(data[j:], closetag); skip >= 0 {
+				j += skip + len(closetag)
+			} else {
+				break
+			}
+
+			// see if it is the only thing on the line
+			if skip := p.isEmpty(data[j:]); skip > 0 {
+				// see if it is followed by a blank line/eof
+				j += skip
+				if j >= len(data) {
+					found = true
+					i = j
+				} else {
+					if skip := p.isEmpty(data[j:]); skip > 0 {
+						j += skip
+						found = true
+						i = j
+					}
+				}
+			}
+		}
+	*/
+
+	// if not found, try a second pass looking for indented match
+	// but not if tag is "ins" or "del" (following original Markdown.pl)
+	if !found && curtag != "ins" && curtag != "del" {
+		i = 1
+		for i < len(data) {
+			i++
+			for i < len(data) && !(data[i-1] == '<' && data[i] == '/') {
+				i++
+			}
+
+			if i+2+len(curtag) >= len(data) {
+				break
+			}
+
+			j = p.htmlFindEnd(curtag, data[i-1:])
+
+			if j > 0 {
+				i += j - 1
+				found = true
+				break
+			}
+		}
+	}
+
+	if !found {
+		return 0
+	}
+
+	// the end of the block has been found
+	if doRender {
+		// trim newlines
+		end := backChar(data, i, '\n')
+		htmlBLock := &ast.HTMLBlock{ast.Leaf{Content: data[:end]}}
+		p.addBlock(htmlBLock)
+		finalizeHTMLBlock(htmlBLock)
+	}
+
+	return i
+}
+
+func finalizeHTMLBlock(block *ast.HTMLBlock) {
+	block.Literal = block.Content
+	block.Content = nil
+}
+
+// HTML comment, lax form
+func (p *Parser) htmlComment(data []byte, doRender bool) int {
+	i := p.inlineHTMLComment(data)
+	// needs to end with a blank line
+	if j := p.isEmpty(data[i:]); j > 0 {
+		size := i + j
+		if doRender {
+			// trim trailing newlines
+			end := backChar(data, size, '\n')
+			htmlBLock := &ast.HTMLBlock{ast.Leaf{Content: data[:end]}}
+			p.addBlock(htmlBLock)
+			finalizeHTMLBlock(htmlBLock)
+		}
+		return size
+	}
+	return 0
+}
+
+// HR, which is the only self-closing block tag considered
+func (p *Parser) htmlHr(data []byte, doRender bool) int {
+	if len(data) < 4 {
+		return 0
+	}
+	if data[0] != '<' || (data[1] != 'h' && data[1] != 'H') || (data[2] != 'r' && data[2] != 'R') {
+		return 0
+	}
+	if data[3] != ' ' && data[3] != '/' && data[3] != '>' {
+		// not an <hr> tag after all; at least not a valid one
+		return 0
+	}
+	i := 3
+	for i < len(data) && data[i] != '>' && data[i] != '\n' {
+		i++
+	}
+	if i < len(data) && data[i] == '>' {
+		i++
+		if j := p.isEmpty(data[i:]); j > 0 {
+			size := i + j
+			if doRender {
+				// trim newlines
+				end := backChar(data, size, '\n')
+				htmlBlock := &ast.HTMLBlock{ast.Leaf{Content: data[:end]}}
+				p.addBlock(htmlBlock)
+				finalizeHTMLBlock(htmlBlock)
+			}
+			return size
+		}
+	}
+	return 0
+}
+
+func (p *Parser) htmlFindTag(data []byte) (string, bool) {
+	i := skipAlnum(data, 0)
+	key := string(data[:i])
+	if _, ok := blockTags[key]; ok {
+		return key, true
+	}
+	return "", false
+}
+
+func (p *Parser) htmlFindEnd(tag string, data []byte) int {
+	// assume data[0] == '<' && data[1] == '/' already tested
+	if tag == "hr" {
+		return 2
+	}
+	// check if tag is a match
+	closetag := []byte("</" + tag + ">")
+	if !bytes.HasPrefix(data, closetag) {
+		return 0
+	}
+	i := len(closetag)
+
+	// check that the rest of the line is blank
+	skip := 0
+	if skip = p.isEmpty(data[i:]); skip == 0 {
+		return 0
+	}
+	i += skip
+	skip = 0
+
+	if i >= len(data) {
+		return i
+	}
+
+	if p.extensions&LaxHTMLBlocks != 0 {
+		return i
+	}
+	if skip = p.isEmpty(data[i:]); skip == 0 {
+		// following line must be blank
+		return 0
+	}
+
+	return i + skip
+}
+
+func (*Parser) isEmpty(data []byte) int {
+	// it is okay to call isEmpty on an empty buffer
+	if len(data) == 0 {
+		return 0
+	}
+
+	var i int
+	for i = 0; i < len(data) && data[i] != '\n'; i++ {
+		if data[i] != ' ' && data[i] != '\t' {
+			return 0
+		}
+	}
+	i = skipCharN(data, i, '\n', 1)
+	return i
+}
+
+func (*Parser) isHRule(data []byte) bool {
+	i := 0
+
+	// skip up to three spaces
+	for i < 3 && data[i] == ' ' {
+		i++
+	}
+
+	// look at the hrule char
+	if data[i] != '*' && data[i] != '-' && data[i] != '_' {
+		return false
+	}
+	c := data[i]
+
+	// the whole line must be the char or whitespace
+	n := 0
+	for i < len(data) && data[i] != '\n' {
+		switch {
+		case data[i] == c:
+			n++
+		case data[i] != ' ':
+			return false
+		}
+		i++
+	}
+
+	return n >= 3
+}
+
+// isFenceLine checks if there's a fence line (e.g., ``` or ``` go) at the beginning of data,
+// and returns the end index if so, or 0 otherwise. It also returns the marker found.
+// If syntax is not nil, it gets set to the syntax specified in the fence line.
+func isFenceLine(data []byte, syntax *string, oldmarker string) (end int, marker string) {
+	i, size := 0, 0
+
+	n := len(data)
+	// skip up to three spaces
+	for i < n && i < 3 && data[i] == ' ' {
+		i++
+	}
+
+	// check for the marker characters: ~ or `
+	if i >= n {
+		return 0, ""
+	}
+	if data[i] != '~' && data[i] != '`' {
+		return 0, ""
+	}
+
+	c := data[i]
+
+	// the whole line must be the same char or whitespace
+	for i < n && data[i] == c {
+		size++
+		i++
+	}
+
+	// the marker char must occur at least 3 times
+	if size < 3 {
+		return 0, ""
+	}
+	marker = string(data[i-size : i])
+
+	// if this is the end marker, it must match the beginning marker
+	if oldmarker != "" && marker != oldmarker {
+		return 0, ""
+	}
+
+	// TODO(shurcooL): It's probably a good idea to simplify the 2 code paths here
+	// into one, always get the syntax, and discard it if the caller doesn't care.
+	if syntax != nil {
+		syn := 0
+		i = skipChar(data, i, ' ')
+
+		if i >= n {
+			if i == n {
+				return i, marker
+			}
+			return 0, ""
+		}
+
+		syntaxStart := i
+
+		if data[i] == '{' {
+			i++
+			syntaxStart++
+
+			for i < n && data[i] != '}' && data[i] != '\n' {
+				syn++
+				i++
+			}
+
+			if i >= n || data[i] != '}' {
+				return 0, ""
+			}
+
+			// strip all whitespace at the beginning and the end
+			// of the {} block
+			for syn > 0 && isSpace(data[syntaxStart]) {
+				syntaxStart++
+				syn--
+			}
+
+			for syn > 0 && isSpace(data[syntaxStart+syn-1]) {
+				syn--
+			}
+
+			i++
+		} else {
+			for i < n && !isSpace(data[i]) {
+				syn++
+				i++
+			}
+		}
+
+		*syntax = string(data[syntaxStart : syntaxStart+syn])
+	}
+
+	i = skipChar(data, i, ' ')
+	if i >= n || data[i] != '\n' {
+		if i == n {
+			return i, marker
+		}
+		return 0, ""
+	}
+	return i + 1, marker // Take newline into account.
+}
+
+// fencedCodeBlock returns the end index if data contains a fenced code block at the beginning,
+// or 0 otherwise. It writes to out if doRender is true, otherwise it has no side effects.
+// If doRender is true, a final newline is mandatory to recognize the fenced code block.
+func (p *Parser) fencedCodeBlock(data []byte, doRender bool) int {
+	var syntax string
+	beg, marker := isFenceLine(data, &syntax, "")
+	if beg == 0 || beg >= len(data) {
+		return 0
+	}
+
+	var work bytes.Buffer
+	work.WriteString(syntax)
+	work.WriteByte('\n')
+
+	for {
+		// safe to assume beg < len(data)
+
+		// check for the end of the code block
+		fenceEnd, _ := isFenceLine(data[beg:], nil, marker)
+		if fenceEnd != 0 {
+			beg += fenceEnd
+			break
+		}
+
+		// copy the current line
+		end := skipUntilChar(data, beg, '\n') + 1
+
+		// did we reach the end of the buffer without a closing marker?
+		if end >= len(data) {
+			return 0
+		}
+
+		// verbatim copy to the working buffer
+		if doRender {
+			work.Write(data[beg:end])
+		}
+		beg = end
+	}
+
+	if doRender {
+		codeBlock := &ast.CodeBlock{
+			IsFenced: true,
+		}
+		codeBlock.Content = work.Bytes() // TODO: get rid of temp buffer
+
+		if p.extensions&Mmark == 0 {
+			p.addBlock(codeBlock)
+			finalizeCodeBlock(codeBlock)
+			return beg
+		}
+
+		// Check for caption and if found make it a figure.
+		if captionContent, id, consumed := p.caption(data[beg:], []byte("Figure: ")); consumed > 0 {
+			figure := &ast.CaptionFigure{}
+			caption := &ast.Caption{}
+			figure.HeadingID = id
+			p.Inline(caption, captionContent)
+
+			p.addBlock(figure)
+			codeBlock.AsLeaf().Attribute = figure.AsContainer().Attribute
+			p.addChild(codeBlock)
+			finalizeCodeBlock(codeBlock)
+			p.addChild(caption)
+			p.finalize(figure)
+
+			beg += consumed
+
+			return beg
+		}
+
+		// Still here, normal block
+		p.addBlock(codeBlock)
+		finalizeCodeBlock(codeBlock)
+	}
+
+	return beg
+}
+
+func unescapeChar(str []byte) []byte {
+	if str[0] == '\\' {
+		return []byte{str[1]}
+	}
+	return []byte(html.UnescapeString(string(str)))
+}
+
+func unescapeString(str []byte) []byte {
+	if reBackslashOrAmp.Match(str) {
+		return reEntityOrEscapedChar.ReplaceAllFunc(str, unescapeChar)
+	}
+	return str
+}
+
+func finalizeCodeBlock(code *ast.CodeBlock) {
+	c := code.Content
+	if code.IsFenced {
+		newlinePos := bytes.IndexByte(c, '\n')
+		firstLine := c[:newlinePos]
+		rest := c[newlinePos+1:]
+		code.Info = unescapeString(bytes.Trim(firstLine, "\n"))
+		code.Literal = rest
+	} else {
+		code.Literal = c
+	}
+	code.Content = nil
+}
+
+func (p *Parser) table(data []byte) int {
+	i, columns, table := p.tableHeader(data)
+	if i == 0 {
+		return 0
+	}
+
+	p.addBlock(&ast.TableBody{})
+
+	for i < len(data) {
+		pipes, rowStart := 0, i
+		for ; i < len(data) && data[i] != '\n'; i++ {
+			if data[i] == '|' {
+				pipes++
+			}
+		}
+
+		if pipes == 0 {
+			i = rowStart
+			break
+		}
+
+		// include the newline in data sent to tableRow
+		i = skipCharN(data, i, '\n', 1)
+
+		if p.tableFooter(data[rowStart:i]) {
+			continue
+		}
+
+		p.tableRow(data[rowStart:i], columns, false)
+	}
+	if captionContent, id, consumed := p.caption(data[i:], []byte("Table: ")); consumed > 0 {
+		caption := &ast.Caption{}
+		p.Inline(caption, captionContent)
+
+		// Some switcheroo to re-insert the parsed table as a child of the captionfigure.
+		figure := &ast.CaptionFigure{}
+		figure.HeadingID = id
+		table2 := &ast.Table{}
+		// Retain any block level attributes.
+		table2.AsContainer().Attribute = table.AsContainer().Attribute
+		children := table.GetChildren()
+		ast.RemoveFromTree(table)
+
+		table2.SetChildren(children)
+		ast.AppendChild(figure, table2)
+		ast.AppendChild(figure, caption)
+
+		p.addChild(figure)
+		p.finalize(figure)
+
+		i += consumed
+	}
+
+	return i
+}
+
+// check if the specified position is preceded by an odd number of backslashes
+func isBackslashEscaped(data []byte, i int) bool {
+	backslashes := 0
+	for i-backslashes-1 >= 0 && data[i-backslashes-1] == '\\' {
+		backslashes++
+	}
+	return backslashes&1 == 1
+}
+
+// tableHeaders parses the header. If recognized it will also add a table.
+func (p *Parser) tableHeader(data []byte) (size int, columns []ast.CellAlignFlags, table ast.Node) {
+	i := 0
+	colCount := 1
+	for i = 0; i < len(data) && data[i] != '\n'; i++ {
+		if data[i] == '|' && !isBackslashEscaped(data, i) {
+			colCount++
+		}
+	}
+
+	// doesn't look like a table header
+	if colCount == 1 {
+		return
+	}
+
+	// include the newline in the data sent to tableRow
+	j := skipCharN(data, i, '\n', 1)
+	header := data[:j]
+
+	// column count ignores pipes at beginning or end of line
+	if data[0] == '|' {
+		colCount--
+	}
+	if i > 2 && data[i-1] == '|' && !isBackslashEscaped(data, i-1) {
+		colCount--
+	}
+
+	columns = make([]ast.CellAlignFlags, colCount)
+
+	// move on to the header underline
+	i++
+	if i >= len(data) {
+		return
+	}
+
+	if data[i] == '|' && !isBackslashEscaped(data, i) {
+		i++
+	}
+	i = skipChar(data, i, ' ')
+
+	// each column header is of form: / *:?-+:? *|/ with # dashes + # colons >= 3
+	// and trailing | optional on last column
+	col := 0
+	n := len(data)
+	for i < n && data[i] != '\n' {
+		dashes := 0
+
+		if data[i] == ':' {
+			i++
+			columns[col] |= ast.TableAlignmentLeft
+			dashes++
+		}
+		for i < n && data[i] == '-' {
+			i++
+			dashes++
+		}
+		if i < n && data[i] == ':' {
+			i++
+			columns[col] |= ast.TableAlignmentRight
+			dashes++
+		}
+		for i < n && data[i] == ' ' {
+			i++
+		}
+		if i == n {
+			return
+		}
+		// end of column test is messy
+		switch {
+		case dashes < 3:
+			// not a valid column
+			return
+
+		case data[i] == '|' && !isBackslashEscaped(data, i):
+			// marker found, now skip past trailing whitespace
+			col++
+			i++
+			for i < n && data[i] == ' ' {
+				i++
+			}
+
+			// trailing junk found after last column
+			if col >= colCount && i < len(data) && data[i] != '\n' {
+				return
+			}
+
+		case (data[i] != '|' || isBackslashEscaped(data, i)) && col+1 < colCount:
+			// something else found where marker was required
+			return
+
+		case data[i] == '\n':
+			// marker is optional for the last column
+			col++
+
+		default:
+			// trailing junk found after last column
+			return
+		}
+	}
+	if col != colCount {
+		return
+	}
+
+	table = &ast.Table{}
+	p.addBlock(table)
+	p.addBlock(&ast.TableHeader{})
+	p.tableRow(header, columns, true)
+	size = skipCharN(data, i, '\n', 1)
+	return
+}
+
+func (p *Parser) tableRow(data []byte, columns []ast.CellAlignFlags, header bool) {
+	p.addBlock(&ast.TableRow{})
+	i, col := 0, 0
+
+	if data[i] == '|' && !isBackslashEscaped(data, i) {
+		i++
+	}
+
+	n := len(data)
+	for col = 0; col < len(columns) && i < n; col++ {
+		for i < n && data[i] == ' ' {
+			i++
+		}
+
+		cellStart := i
+
+		for i < n && (data[i] != '|' || isBackslashEscaped(data, i)) && data[i] != '\n' {
+			i++
+		}
+
+		cellEnd := i
+
+		// skip the end-of-cell marker, possibly taking us past end of buffer
+		i++
+
+		for cellEnd > cellStart && cellEnd-1 < n && data[cellEnd-1] == ' ' {
+			cellEnd--
+		}
+
+		block := &ast.TableCell{
+			IsHeader: header,
+			Align:    columns[col],
+		}
+		block.Content = data[cellStart:cellEnd]
+		p.addBlock(block)
+	}
+
+	// pad it out with empty columns to get the right number
+	for ; col < len(columns); col++ {
+		block := &ast.TableCell{
+			IsHeader: header,
+			Align:    columns[col],
+		}
+		p.addBlock(block)
+	}
+
+	// silently ignore rows with too many cells
+}
+
+// tableFooter parses the (optional) table footer.
+func (p *Parser) tableFooter(data []byte) bool {
+	colCount := 1
+	for i := 0; i < len(data) && data[i] != '\n'; i++ {
+		if data[i] == '|' && !isBackslashEscaped(data, i) {
+			colCount++
+			continue
+		}
+		// remaining data must be the = character
+		if data[i] != '=' {
+			return false
+		}
+	}
+
+	// doesn't look like a table footer
+	if colCount == 1 {
+		return false
+	}
+
+	p.addBlock(&ast.TableFooter{})
+
+	return true
+}
+
+// returns blockquote prefix length
+func (p *Parser) quotePrefix(data []byte) int {
+	i := 0
+	n := len(data)
+	for i < 3 && i < n && data[i] == ' ' {
+		i++
+	}
+	if i < n && data[i] == '>' {
+		if i+1 < n && data[i+1] == ' ' {
+			return i + 2
+		}
+		return i + 1
+	}
+	return 0
+}
+
+// blockquote ends with at least one blank line
+// followed by something without a blockquote prefix
+func (p *Parser) terminateBlockquote(data []byte, beg, end int) bool {
+	if p.isEmpty(data[beg:]) <= 0 {
+		return false
+	}
+	if end >= len(data) {
+		return true
+	}
+	return p.quotePrefix(data[end:]) == 0 && p.isEmpty(data[end:]) == 0
+}
+
+// parse a blockquote fragment
+func (p *Parser) quote(data []byte) int {
+	var raw bytes.Buffer
+	beg, end := 0, 0
+	for beg < len(data) {
+		end = beg
+		// Step over whole lines, collecting them. While doing that, check for
+		// fenced code and if one's found, incorporate it altogether,
+		// irregardless of any contents inside it
+		for end < len(data) && data[end] != '\n' {
+			if p.extensions&FencedCode != 0 {
+				if i := p.fencedCodeBlock(data[end:], false); i > 0 {
+					// -1 to compensate for the extra end++ after the loop:
+					end += i - 1
+					break
+				}
+			}
+			end++
+		}
+		end = skipCharN(data, end, '\n', 1)
+		if pre := p.quotePrefix(data[beg:]); pre > 0 {
+			// skip the prefix
+			beg += pre
+		} else if p.terminateBlockquote(data, beg, end) {
+			break
+		}
+		// this line is part of the blockquote
+		raw.Write(data[beg:end])
+		beg = end
+	}
+
+	if p.extensions&Mmark == 0 {
+		block := p.addBlock(&ast.BlockQuote{})
+		p.block(raw.Bytes())
+		p.finalize(block)
+		return end
+	}
+
+	if captionContent, id, consumed := p.caption(data[end:], []byte("Quote: ")); consumed > 0 {
+		figure := &ast.CaptionFigure{}
+		caption := &ast.Caption{}
+		figure.HeadingID = id
+		p.Inline(caption, captionContent)
+
+		p.addBlock(figure) // this discard any attributes
+		block := &ast.BlockQuote{}
+		block.AsContainer().Attribute = figure.AsContainer().Attribute
+		p.addChild(block)
+		p.block(raw.Bytes())
+		p.finalize(block)
+
+		p.addChild(caption)
+		p.finalize(figure)
+
+		end += consumed
+
+		return end
+	}
+
+	block := p.addBlock(&ast.BlockQuote{})
+	p.block(raw.Bytes())
+	p.finalize(block)
+
+	return end
+}
+
+// returns prefix length for block code
+func (p *Parser) codePrefix(data []byte) int {
+	n := len(data)
+	if n >= 1 && data[0] == '\t' {
+		return 1
+	}
+	if n >= 4 && data[3] == ' ' && data[2] == ' ' && data[1] == ' ' && data[0] == ' ' {
+		return 4
+	}
+	return 0
+}
+
+func (p *Parser) code(data []byte) int {
+	var work bytes.Buffer
+
+	i := 0
+	for i < len(data) {
+		beg := i
+
+		i = skipUntilChar(data, i, '\n')
+		i = skipCharN(data, i, '\n', 1)
+
+		blankline := p.isEmpty(data[beg:i]) > 0
+		if pre := p.codePrefix(data[beg:i]); pre > 0 {
+			beg += pre
+		} else if !blankline {
+			// non-empty, non-prefixed line breaks the pre
+			i = beg
+			break
+		}
+
+		// verbatim copy to the working buffer
+		if blankline {
+			work.WriteByte('\n')
+		} else {
+			work.Write(data[beg:i])
+		}
+	}
+
+	// trim all the \n off the end of work
+	workbytes := work.Bytes()
+
+	eol := backChar(workbytes, len(workbytes), '\n')
+
+	if eol != len(workbytes) {
+		work.Truncate(eol)
+	}
+
+	work.WriteByte('\n')
+
+	codeBlock := &ast.CodeBlock{
+		IsFenced: false,
+	}
+	// TODO: get rid of temp buffer
+	codeBlock.Content = work.Bytes()
+	p.addBlock(codeBlock)
+	finalizeCodeBlock(codeBlock)
+
+	return i
+}
+
+// returns unordered list item prefix
+func (p *Parser) uliPrefix(data []byte) int {
+	// start with up to 3 spaces
+	i := skipCharN(data, 0, ' ', 3)
+
+	if i >= len(data)-1 {
+		return 0
+	}
+	// need one of {'*', '+', '-'} followed by a space or a tab
+	if (data[i] != '*' && data[i] != '+' && data[i] != '-') ||
+		(data[i+1] != ' ' && data[i+1] != '\t') {
+		return 0
+	}
+	return i + 2
+}
+
+// returns ordered list item prefix
+func (p *Parser) oliPrefix(data []byte) int {
+	// start with up to 3 spaces
+	i := skipCharN(data, 0, ' ', 3)
+
+	// count the digits
+	start := i
+	for i < len(data) && data[i] >= '0' && data[i] <= '9' {
+		i++
+	}
+	if start == i || i >= len(data)-1 {
+		return 0
+	}
+
+	// we need >= 1 digits followed by a dot and a space or a tab
+	if data[i] != '.' || !(data[i+1] == ' ' || data[i+1] == '\t') {
+		return 0
+	}
+	return i + 2
+}
+
+// returns definition list item prefix
+func (p *Parser) dliPrefix(data []byte) int {
+	if len(data) < 2 {
+		return 0
+	}
+	// need a ':' followed by a space or a tab
+	if data[0] != ':' || !(data[1] == ' ' || data[1] == '\t') {
+		return 0
+	}
+	i := skipChar(data, 0, ' ')
+	return i + 2
+}
+
+// parse ordered or unordered list block
+func (p *Parser) list(data []byte, flags ast.ListType, start int) int {
+	i := 0
+	flags |= ast.ListItemBeginningOfList
+	list := &ast.List{
+		ListFlags: flags,
+		Tight:     true,
+		Start:     start,
+	}
+	block := p.addBlock(list)
+
+	for i < len(data) {
+		skip := p.listItem(data[i:], &flags)
+		if flags&ast.ListItemContainsBlock != 0 {
+			list.Tight = false
+		}
+		i += skip
+		if skip == 0 || flags&ast.ListItemEndOfList != 0 {
+			break
+		}
+		flags &= ^ast.ListItemBeginningOfList
+	}
+
+	above := block.GetParent()
+	finalizeList(list)
+	p.tip = above
+	return i
+}
+
+// Returns true if the list item is not the same type as its parent list
+func (p *Parser) listTypeChanged(data []byte, flags *ast.ListType) bool {
+	if p.dliPrefix(data) > 0 && *flags&ast.ListTypeDefinition == 0 {
+		return true
+	} else if p.oliPrefix(data) > 0 && *flags&ast.ListTypeOrdered == 0 {
+		return true
+	} else if p.uliPrefix(data) > 0 && (*flags&ast.ListTypeOrdered != 0 || *flags&ast.ListTypeDefinition != 0) {
+		return true
+	}
+	return false
+}
+
+// Returns true if block ends with a blank line, descending if needed
+// into lists and sublists.
+func endsWithBlankLine(block ast.Node) bool {
+	// TODO: figure this out. Always false now.
+	for block != nil {
+		//if block.lastLineBlank {
+		//return true
+		//}
+		switch block.(type) {
+		case *ast.List, *ast.ListItem:
+			block = ast.GetLastChild(block)
+		default:
+			return false
+		}
+	}
+	return false
+}
+
+func finalizeList(list *ast.List) {
+	items := list.Parent.GetChildren()
+	lastItemIdx := len(items) - 1
+	for i, item := range items {
+		isLastItem := i == lastItemIdx
+		// check for non-final list item ending with blank line:
+		if !isLastItem && endsWithBlankLine(item) {
+			list.Tight = false
+			break
+		}
+		// recurse into children of list item, to see if there are spaces
+		// between any of them:
+		subItems := item.GetParent().GetChildren()
+		lastSubItemIdx := len(subItems) - 1
+		for j, subItem := range subItems {
+			isLastSubItem := j == lastSubItemIdx
+			if (!isLastItem || !isLastSubItem) && endsWithBlankLine(subItem) {
+				list.Tight = false
+				break
+			}
+		}
+	}
+}
+
+// Parse a single list item.
+// Assumes initial prefix is already removed if this is a sublist.
+func (p *Parser) listItem(data []byte, flags *ast.ListType) int {
+	// keep track of the indentation of the first line
+	itemIndent := 0
+	if data[0] == '\t' {
+		itemIndent += 4
+	} else {
+		for itemIndent < 3 && data[itemIndent] == ' ' {
+			itemIndent++
+		}
+	}
+
+	var bulletChar byte = '*'
+	i := p.uliPrefix(data)
+	if i == 0 {
+		i = p.oliPrefix(data)
+	} else {
+		bulletChar = data[i-2]
+	}
+	if i == 0 {
+		i = p.dliPrefix(data)
+		// reset definition term flag
+		if i > 0 {
+			*flags &= ^ast.ListTypeTerm
+		}
+	}
+	if i == 0 {
+		// if in definition list, set term flag and continue
+		if *flags&ast.ListTypeDefinition != 0 {
+			*flags |= ast.ListTypeTerm
+		} else {
+			return 0
+		}
+	}
+
+	// skip leading whitespace on first line
+	i = skipChar(data, i, ' ')
+
+	// find the end of the line
+	line := i
+	for i > 0 && i < len(data) && data[i-1] != '\n' {
+		i++
+	}
+
+	// get working buffer
+	var raw bytes.Buffer
+
+	// put the first line into the working buffer
+	raw.Write(data[line:i])
+	line = i
+
+	// process the following lines
+	containsBlankLine := false
+	sublist := 0
+
+gatherlines:
+	for line < len(data) {
+		i++
+
+		// find the end of this line
+		for i < len(data) && data[i-1] != '\n' {
+			i++
+		}
+
+		// if it is an empty line, guess that it is part of this item
+		// and move on to the next line
+		if p.isEmpty(data[line:i]) > 0 {
+			containsBlankLine = true
+			line = i
+			continue
+		}
+
+		// calculate the indentation
+		indent := 0
+		indentIndex := 0
+		if data[line] == '\t' {
+			indentIndex++
+			indent += 4
+		} else {
+			for indent < 4 && line+indent < i && data[line+indent] == ' ' {
+				indent++
+				indentIndex++
+			}
+		}
+
+		chunk := data[line+indentIndex : i]
+
+		// evaluate how this line fits in
+		switch {
+		// is this a nested list item?
+		case (p.uliPrefix(chunk) > 0 && !p.isHRule(chunk)) || p.oliPrefix(chunk) > 0 || p.dliPrefix(chunk) > 0:
+
+			// to be a nested list, it must be indented more
+			// if not, it is either a different kind of list
+			// or the next item in the same list
+			if indent <= itemIndent {
+				if p.listTypeChanged(chunk, flags) {
+					*flags |= ast.ListItemEndOfList
+				} else if containsBlankLine {
+					*flags |= ast.ListItemContainsBlock
+				}
+
+				break gatherlines
+			}
+
+			if containsBlankLine {
+				*flags |= ast.ListItemContainsBlock
+			}
+
+			// is this the first item in the nested list?
+			if sublist == 0 {
+				sublist = raw.Len()
+				// in the case of dliPrefix we are too late and need to search back for the definition item, which
+				// should be on the previous line, we then adjust sublist to start there.
+				if p.dliPrefix(chunk) > 0 {
+					sublist = backUntilChar(raw.Bytes(), raw.Len()-1, '\n')
+				}
+			}
+
+			// is this a nested prefix heading?
+		case p.isPrefixHeading(chunk), p.isPrefixSpecialHeading(chunk):
+			// if the heading is not indented, it is not nested in the list
+			// and thus ends the list
+			if containsBlankLine && indent < 4 {
+				*flags |= ast.ListItemEndOfList
+				break gatherlines
+			}
+			*flags |= ast.ListItemContainsBlock
+
+		// anything following an empty line is only part
+		// of this item if it is indented 4 spaces
+		// (regardless of the indentation of the beginning of the item)
+		case containsBlankLine && indent < 4:
+			if *flags&ast.ListTypeDefinition != 0 && i < len(data)-1 {
+				// is the next item still a part of this list?
+				next := i
+				for next < len(data) && data[next] != '\n' {
+					next++
+				}
+				for next < len(data)-1 && data[next] == '\n' {
+					next++
+				}
+				if i < len(data)-1 && data[i] != ':' && next < len(data)-1 && data[next] != ':' {
+					*flags |= ast.ListItemEndOfList
+				}
+			} else {
+				*flags |= ast.ListItemEndOfList
+			}
+			break gatherlines
+
+		// a blank line means this should be parsed as a block
+		case containsBlankLine:
+			raw.WriteByte('\n')
+			*flags |= ast.ListItemContainsBlock
+		}
+
+		// if this line was preceded by one or more blanks,
+		// re-introduce the blank into the buffer
+		if containsBlankLine {
+			containsBlankLine = false
+			raw.WriteByte('\n')
+		}
+
+		// add the line into the working buffer without prefix
+		raw.Write(data[line+indentIndex : i])
+
+		line = i
+	}
+
+	rawBytes := raw.Bytes()
+
+	listItem := &ast.ListItem{
+		ListFlags:  *flags,
+		Tight:      false,
+		BulletChar: bulletChar,
+		Delimiter:  '.', // Only '.' is possible in Markdown, but ')' will also be possible in CommonMark
+	}
+	p.addBlock(listItem)
+
+	// render the contents of the list item
+	if *flags&ast.ListItemContainsBlock != 0 && *flags&ast.ListTypeTerm == 0 {
+		// intermediate render of block item, except for definition term
+		if sublist > 0 {
+			p.block(rawBytes[:sublist])
+			p.block(rawBytes[sublist:])
+		} else {
+			p.block(rawBytes)
+		}
+	} else {
+		// intermediate render of inline item
+		para := &ast.Paragraph{}
+		if sublist > 0 {
+			para.Content = rawBytes[:sublist]
+		} else {
+			para.Content = rawBytes
+		}
+		p.addChild(para)
+		if sublist > 0 {
+			p.block(rawBytes[sublist:])
+		}
+	}
+	return line
+}
+
+// render a single paragraph that has already been parsed out
+func (p *Parser) renderParagraph(data []byte) {
+	if len(data) == 0 {
+		return
+	}
+
+	// trim leading spaces
+	beg := skipChar(data, 0, ' ')
+
+	end := len(data)
+	// trim trailing newline
+	if data[len(data)-1] == '\n' {
+		end--
+	}
+
+	// trim trailing spaces
+	for end > beg && data[end-1] == ' ' {
+		end--
+	}
+	para := &ast.Paragraph{}
+	para.Content = data[beg:end]
+	p.addBlock(para)
+}
+
+// blockMath handle block surround with $$
+func (p *Parser) blockMath(data []byte) int {
+	if len(data) <= 4 || data[0] != '$' || data[1] != '$' || data[2] == '$' {
+		return 0
+	}
+
+	// find next $$
+	var end int
+	for end = 2; end+1 < len(data) && (data[end] != '$' || data[end+1] != '$'); end++ {
+	}
+
+	// $$ not match
+	if end+1 == len(data) {
+		return 0
+	}
+
+	// render the display math
+	mathBlock := &ast.MathBlock{}
+	mathBlock.Literal = data[2:end]
+	p.addBlock(mathBlock)
+
+	return end + 2
+}
+
+func (p *Parser) paragraph(data []byte) int {
+	// prev: index of 1st char of previous line
+	// line: index of 1st char of current line
+	// i: index of cursor/end of current line
+	var prev, line, i int
+	tabSize := tabSizeDefault
+	if p.extensions&TabSizeEight != 0 {
+		tabSize = tabSizeDouble
+	}
+	// keep going until we find something to mark the end of the paragraph
+	for i < len(data) {
+		// mark the beginning of the current line
+		prev = line
+		current := data[i:]
+		line = i
+
+		// did we find a reference or a footnote? If so, end a paragraph
+		// preceding it and report that we have consumed up to the end of that
+		// reference:
+		if refEnd := isReference(p, current, tabSize); refEnd > 0 {
+			p.renderParagraph(data[:i])
+			return i + refEnd
+		}
+
+		// did we find a blank line marking the end of the paragraph?
+		if n := p.isEmpty(current); n > 0 {
+			// did this blank line followed by a definition list item?
+			if p.extensions&DefinitionLists != 0 {
+				if i < len(data)-1 && data[i+1] == ':' {
+					listLen := p.list(data[prev:], ast.ListTypeDefinition, 0)
+					return prev + listLen
+				}
+			}
+
+			p.renderParagraph(data[:i])
+			return i + n
+		}
+
+		// an underline under some text marks a heading, so our paragraph ended on prev line
+		if i > 0 {
+			if level := p.isUnderlinedHeading(current); level > 0 {
+				// render the paragraph
+				p.renderParagraph(data[:prev])
+
+				// ignore leading and trailing whitespace
+				eol := i - 1
+				for prev < eol && data[prev] == ' ' {
+					prev++
+				}
+				for eol > prev && data[eol-1] == ' ' {
+					eol--
+				}
+
+				id := ""
+				if p.extensions&AutoHeadingIDs != 0 {
+					id = sanitizeAnchorName(string(data[prev:eol]))
+				}
+
+				block := &ast.Heading{
+					Level:     level,
+					HeadingID: id,
+				}
+				block.Content = data[prev:eol]
+				p.addBlock(block)
+
+				// find the end of the underline
+				return skipUntilChar(data, i, '\n')
+			}
+		}
+
+		// if the next line starts a block of HTML, then the paragraph ends here
+		if p.extensions&LaxHTMLBlocks != 0 {
+			if data[i] == '<' && p.html(current, false) > 0 {
+				// rewind to before the HTML block
+				p.renderParagraph(data[:i])
+				return i
+			}
+		}
+
+		// if there's a prefixed heading or a horizontal rule after this, paragraph is over
+		if p.isPrefixHeading(current) || p.isPrefixSpecialHeading(current) || p.isHRule(current) {
+			p.renderParagraph(data[:i])
+			return i
+		}
+
+		// if there's a fenced code block, paragraph is over
+		if p.extensions&FencedCode != 0 {
+			if p.fencedCodeBlock(current, false) > 0 {
+				p.renderParagraph(data[:i])
+				return i
+			}
+		}
+
+		// if there's a figure block, paragraph is over
+		if p.extensions&Mmark != 0 {
+			if p.figureBlock(current, false) > 0 {
+				p.renderParagraph(data[:i])
+				return i
+			}
+		}
+
+		// if there's a definition list item, prev line is a definition term
+		if p.extensions&DefinitionLists != 0 {
+			if p.dliPrefix(current) != 0 {
+				ret := p.list(data[prev:], ast.ListTypeDefinition, 0)
+				return ret + prev
+			}
+		}
+
+		// if there's a list after this, paragraph is over
+		if p.extensions&NoEmptyLineBeforeBlock != 0 {
+			if p.uliPrefix(current) != 0 ||
+				p.oliPrefix(current) != 0 ||
+				p.quotePrefix(current) != 0 ||
+				p.codePrefix(current) != 0 {
+				p.renderParagraph(data[:i])
+				return i
+			}
+		}
+
+		// otherwise, scan to the beginning of the next line
+		nl := bytes.IndexByte(data[i:], '\n')
+		if nl >= 0 {
+			i += nl + 1
+		} else {
+			i += len(data[i:])
+		}
+	}
+
+	p.renderParagraph(data[:i])
+	return i
+}
+
+// skipChar advances i as long as data[i] == c
+func skipChar(data []byte, i int, c byte) int {
+	n := len(data)
+	for i < n && data[i] == c {
+		i++
+	}
+	return i
+}
+
+// like skipChar but only skips up to max characters
+func skipCharN(data []byte, i int, c byte, max int) int {
+	n := len(data)
+	for i < n && max > 0 && data[i] == c {
+		i++
+		max--
+	}
+	return i
+}
+
+// skipUntilChar advances i as long as data[i] != c
+func skipUntilChar(data []byte, i int, c byte) int {
+	n := len(data)
+	for i < n && data[i] != c {
+		i++
+	}
+	return i
+}
+
+func skipAlnum(data []byte, i int) int {
+	n := len(data)
+	for i < n && isAlnum(data[i]) {
+		i++
+	}
+	return i
+}
+
+func skipSpace(data []byte, i int) int {
+	n := len(data)
+	for i < n && isSpace(data[i]) {
+		i++
+	}
+	return i
+}
+
+func backChar(data []byte, i int, c byte) int {
+	for i > 0 && data[i-1] == c {
+		i--
+	}
+	return i
+}
+
+func backUntilChar(data []byte, i int, c byte) int {
+	for i > 0 && data[i-1] != c {
+		i--
+	}
+	return i
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/callout.go b/vendor/github.com/gomarkdown/markdown/parser/callout.go
new file mode 100644
index 00000000..15858aa9
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/callout.go
@@ -0,0 +1,29 @@
+package parser
+
+import (
+	"bytes"
+	"strconv"
+)
+
+// IsCallout detects a callout in the following format: <<N>> Where N is a integer > 0.
+func IsCallout(data []byte) (id []byte, consumed int) {
+	if !bytes.HasPrefix(data, []byte("<<")) {
+		return nil, 0
+	}
+	start := 2
+	end := bytes.Index(data[start:], []byte(">>"))
+	if end < 0 {
+		return nil, 0
+	}
+
+	b := data[start : start+end]
+	b = bytes.TrimSpace(b)
+	i, err := strconv.Atoi(string(b))
+	if err != nil {
+		return nil, 0
+	}
+	if i <= 0 {
+		return nil, 0
+	}
+	return b, start + end + 2 // 2 for >>
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/caption.go b/vendor/github.com/gomarkdown/markdown/parser/caption.go
new file mode 100644
index 00000000..54d3f741
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/caption.go
@@ -0,0 +1,70 @@
+package parser
+
+import (
+	"bytes"
+)
+
+// caption checks for a caption, it returns the caption data and a potential "headingID".
+func (p *Parser) caption(data, caption []byte) ([]byte, string, int) {
+	if !bytes.HasPrefix(data, caption) {
+		return nil, "", 0
+	}
+	j := len(caption)
+	data = data[j:]
+	end := p.linesUntilEmpty(data)
+
+	data = data[:end]
+
+	id, start := captionID(data)
+	if id != "" {
+		return data[:start], id, end + j
+	}
+
+	return data, "", end + j
+}
+
+// linesUntilEmpty scans lines up to the first empty line.
+func (p *Parser) linesUntilEmpty(data []byte) int {
+	line, i := 0, 0
+
+	for line < len(data) {
+		i++
+
+		// find the end of this line
+		for i < len(data) && data[i-1] != '\n' {
+			i++
+		}
+
+		if p.isEmpty(data[line:i]) == 0 {
+			line = i
+			continue
+		}
+
+		break
+	}
+	return i
+}
+
+// captionID checks if the caption *ends* in {#....}. If so the text after {# is taken to be
+// the ID/anchor of the entire figure block.
+func captionID(data []byte) (string, int) {
+	end := len(data)
+
+	j, k := 0, 0
+	// find start/end of heading id
+	for j = 0; j < end-1 && (data[j] != '{' || data[j+1] != '#'); j++ {
+	}
+	for k = j + 1; k < end && data[k] != '}'; k++ {
+	}
+	// remains must be whitespace.
+	for l := k + 1; l < end; l++ {
+		if !isSpace(data[l]) {
+			return "", 0
+		}
+	}
+
+	if j > 0 && k > 0 && j+2 < k {
+		return string(data[j+2 : k]), j
+	}
+	return "", 0
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/citation.go b/vendor/github.com/gomarkdown/markdown/parser/citation.go
new file mode 100644
index 00000000..8ea1fbee
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/citation.go
@@ -0,0 +1,86 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// citation parses a citation. In its most simple form [@ref], we allow multiple
+// being separated by semicolons and a sub reference inside ala pandoc: [@ref, p. 23].
+// Each citation can have a modifier: !, ? or - wich mean:
+//
+// ! - normative
+// ? - formative
+// - - suppressed
+//
+// The suffix starts after a comma, we strip any whitespace before and after. If the output
+// allows for it, this can be rendered.
+func citation(p *Parser, data []byte, offset int) (int, ast.Node) {
+	// look for the matching closing bracket
+	i := offset + 1
+	for level := 1; level > 0 && i < len(data); i++ {
+		switch {
+		case data[i] == '\n':
+			// no newlines allowed.
+			return 0, nil
+
+		case data[i-1] == '\\':
+			continue
+
+		case data[i] == '[':
+			level++
+
+		case data[i] == ']':
+			level--
+			if level <= 0 {
+				i-- // compensate for extra i++ in for loop
+			}
+		}
+	}
+
+	if i >= len(data) {
+		return 0, nil
+	}
+
+	node := &ast.Citation{}
+
+	citations := bytes.Split(data[1:i], []byte(";"))
+	for _, citation := range citations {
+		var suffix []byte
+		citation = bytes.TrimSpace(citation)
+		j := 0
+		if citation[j] != '@' {
+			// not a citation, drop out entirely.
+			return 0, nil
+		}
+		if c := bytes.Index(citation, []byte(",")); c > 0 {
+			part := citation[:c]
+			suff := citation[c+1:]
+			part = bytes.TrimSpace(part)
+			suff = bytes.TrimSpace(suff)
+
+			citation = part
+			suffix = suff
+		}
+
+		citeType := ast.CitationTypeInformative
+		j = 1
+		switch citation[j] {
+		case '!':
+			citeType = ast.CitationTypeNormative
+			j++
+		case '?':
+			citeType = ast.CitationTypeInformative
+			j++
+		case '-':
+			citeType = ast.CitationTypeSuppressed
+			j++
+		}
+		node.Destination = append(node.Destination, citation[j:])
+		node.Type = append(node.Type, citeType)
+		node.Suffix = append(node.Suffix, suffix)
+	}
+
+	return i + 1, node
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/esc.go b/vendor/github.com/gomarkdown/markdown/parser/esc.go
new file mode 100644
index 00000000..0a79aa35
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/esc.go
@@ -0,0 +1,20 @@
+package parser
+
+// isEscape returns true if byte i is prefixed by an odd number of backslahses.
+func isEscape(data []byte, i int) bool {
+	if i == 0 {
+		return false
+	}
+	if i == 1 {
+		return data[0] == '\\'
+	}
+	j := i - 1
+	for ; j >= 0; j-- {
+		if data[j] != '\\' {
+			break
+		}
+	}
+	j++
+	// odd number of backslahes means escape
+	return (i-j)%2 != 0
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/figures.go b/vendor/github.com/gomarkdown/markdown/parser/figures.go
new file mode 100644
index 00000000..6615449c
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/figures.go
@@ -0,0 +1,119 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// sFigureLine checks if there's a figure line (e.g., !--- ) at the beginning of data,
+// and returns the end index if so, or 0 otherwise.
+func sFigureLine(data []byte, oldmarker string) (end int, marker string) {
+	i, size := 0, 0
+
+	n := len(data)
+	// skip up to three spaces
+	for i < n && i < 3 && data[i] == ' ' {
+		i++
+	}
+
+	// check for the marker characters: !
+	if i+1 >= n {
+		return 0, ""
+	}
+	if data[i] != '!' || data[i+1] != '-' {
+		return 0, ""
+	}
+	i++
+
+	c := data[i] // i.e. the -
+
+	// the whole line must be the same char or whitespace
+	for i < n && data[i] == c {
+		size++
+		i++
+	}
+
+	// the marker char must occur at least 3 times
+	if size < 3 {
+		return 0, ""
+	}
+	marker = string(data[i-size : i])
+
+	// if this is the end marker, it must match the beginning marker
+	if oldmarker != "" && marker != oldmarker {
+		return 0, ""
+	}
+
+	// there is no syntax modifier although it might be an idea to re-use this space for something?
+
+	i = skipChar(data, i, ' ')
+	if i >= n || data[i] != '\n' {
+		if i == n {
+			return i, marker
+		}
+		return 0, ""
+	}
+	return i + 1, marker // Take newline into account.
+}
+
+// figureBlock returns the end index if data contains a figure block at the beginning,
+// or 0 otherwise. It writes to out if doRender is true, otherwise it has no side effects.
+// If doRender is true, a final newline is mandatory to recognize the figure block.
+func (p *Parser) figureBlock(data []byte, doRender bool) int {
+	beg, marker := sFigureLine(data, "")
+	if beg == 0 || beg >= len(data) {
+		return 0
+	}
+
+	var raw bytes.Buffer
+
+	for {
+		// safe to assume beg < len(data)
+
+		// check for the end of the code block
+		figEnd, _ := sFigureLine(data[beg:], marker)
+		if figEnd != 0 {
+			beg += figEnd
+			break
+		}
+
+		// copy the current line
+		end := skipUntilChar(data, beg, '\n') + 1
+
+		// did we reach the end of the buffer without a closing marker?
+		if end >= len(data) {
+			return 0
+		}
+
+		// verbatim copy to the working buffer
+		if doRender {
+			raw.Write(data[beg:end])
+		}
+		beg = end
+	}
+
+	if !doRender {
+		return beg
+	}
+
+	figure := &ast.CaptionFigure{}
+	p.addBlock(figure)
+	p.block(raw.Bytes())
+
+	defer p.finalize(figure)
+
+	if captionContent, id, consumed := p.caption(data[beg:], []byte("Figure: ")); consumed > 0 {
+		caption := &ast.Caption{}
+		p.Inline(caption, captionContent)
+
+		figure.HeadingID = id
+
+		p.addChild(caption)
+
+		beg += consumed
+	}
+
+	p.finalize(figure)
+	return beg
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/include.go b/vendor/github.com/gomarkdown/markdown/parser/include.go
new file mode 100644
index 00000000..2448a685
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/include.go
@@ -0,0 +1,129 @@
+package parser
+
+import (
+	"bytes"
+	"path"
+	"path/filepath"
+)
+
+// isInclude parses {{...}}[...], that contains a path between the {{, the [...] syntax contains
+// an address to select which lines to include. It is treated as an opaque string and just given
+// to readInclude.
+func (p *Parser) isInclude(data []byte) (filename string, address []byte, consumed int) {
+	i := skipCharN(data, 0, ' ', 3) // start with up to 3 spaces
+	if len(data[i:]) < 3 {
+		return "", nil, 0
+	}
+	if data[i] != '{' || data[i+1] != '{' {
+		return "", nil, 0
+	}
+	start := i + 2
+
+	// find the end delimiter
+	i = skipUntilChar(data, i, '}')
+	if i+1 >= len(data) {
+		return "", nil, 0
+	}
+	end := i
+	i++
+	if data[i] != '}' {
+		return "", nil, 0
+	}
+	filename = string(data[start:end])
+
+	if i+1 < len(data) && data[i+1] == '[' { // potential address specification
+		start := i + 2
+
+		end = skipUntilChar(data, start, ']')
+		if end >= len(data) {
+			return "", nil, 0
+		}
+		address = data[start:end]
+		return filename, address, end + 1
+	}
+
+	return filename, address, i + 1
+}
+
+func (p *Parser) readInclude(from, file string, address []byte) []byte {
+	if p.Opts.ReadIncludeFn != nil {
+		return p.Opts.ReadIncludeFn(from, file, address)
+	}
+
+	return nil
+}
+
+// isCodeInclude parses <{{...}} which is similar to isInclude the returned bytes are, however wrapped in a code block.
+func (p *Parser) isCodeInclude(data []byte) (filename string, address []byte, consumed int) {
+	i := skipCharN(data, 0, ' ', 3) // start with up to 3 spaces
+	if len(data[i:]) < 3 {
+		return "", nil, 0
+	}
+	if data[i] != '<' {
+		return "", nil, 0
+	}
+	start := i
+
+	filename, address, consumed = p.isInclude(data[i+1:])
+	if consumed == 0 {
+		return "", nil, 0
+	}
+	return filename, address, start + consumed + 1
+}
+
+// readCodeInclude acts like include except the returned bytes are wrapped in a fenced code block.
+func (p *Parser) readCodeInclude(from, file string, address []byte) []byte {
+	data := p.readInclude(from, file, address)
+	if data == nil {
+		return nil
+	}
+	ext := path.Ext(file)
+	buf := &bytes.Buffer{}
+	buf.Write([]byte("```"))
+	if ext != "" { // starts with a dot
+		buf.WriteString(" " + ext[1:] + "\n")
+	} else {
+		buf.WriteByte('\n')
+	}
+	buf.Write(data)
+	buf.WriteString("```\n")
+	return buf.Bytes()
+}
+
+// incStack hold the current stack of chained includes. Each value is the containing
+// path of the file being parsed.
+type incStack struct {
+	stack []string
+}
+
+func newIncStack() *incStack {
+	return &incStack{stack: []string{}}
+}
+
+// Push updates i with new.
+func (i *incStack) Push(new string) {
+	if path.IsAbs(new) {
+		i.stack = append(i.stack, path.Dir(new))
+		return
+	}
+	last := ""
+	if len(i.stack) > 0 {
+		last = i.stack[len(i.stack)-1]
+	}
+	i.stack = append(i.stack, path.Dir(filepath.Join(last, new)))
+}
+
+// Pop pops the last value.
+func (i *incStack) Pop() {
+	if len(i.stack) == 0 {
+		return
+	}
+	i.stack = i.stack[:len(i.stack)-1]
+}
+
+func (i *incStack) Last() string {
+	if len(i.stack) == 0 {
+		return ""
+	}
+	return i.stack[len(i.stack)-1]
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/inline.go b/vendor/github.com/gomarkdown/markdown/parser/inline.go
new file mode 100644
index 00000000..81766b85
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/inline.go
@@ -0,0 +1,1284 @@
+package parser
+
+import (
+	"bytes"
+	"regexp"
+	"strconv"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// Parsing of inline elements
+
+var (
+	urlRe    = `((https?|ftp):\/\/|\/)[-A-Za-z0-9+&@#\/%?=~_|!:,.;\(\)]+`
+	anchorRe = regexp.MustCompile(`^(<a\shref="` + urlRe + `"(\stitle="[^"<>]+")?\s?>` + urlRe + `<\/a>)`)
+
+	// TODO: improve this regexp to catch all possible entities:
+	htmlEntityRe = regexp.MustCompile(`&[a-z]{2,5};`)
+)
+
+// Inline parses text within a block.
+// Each function returns the number of consumed chars.
+func (p *Parser) Inline(currBlock ast.Node, data []byte) {
+	// handlers might call us recursively: enforce a maximum depth
+	if p.nesting >= p.maxNesting || len(data) == 0 {
+		return
+	}
+	p.nesting++
+	beg, end := 0, 0
+
+	n := len(data)
+	for end < n {
+		handler := p.inlineCallback[data[end]]
+		if handler == nil {
+			end++
+			continue
+		}
+		consumed, node := handler(p, data, end)
+		if consumed == 0 {
+			// no action from the callback
+			end++
+			continue
+		}
+		// copy inactive chars into the output
+		ast.AppendChild(currBlock, newTextNode(data[beg:end]))
+		if node != nil {
+			ast.AppendChild(currBlock, node)
+		}
+		beg = end + consumed
+		end = beg
+	}
+
+	if beg < n {
+		if data[end-1] == '\n' {
+			end--
+		}
+		ast.AppendChild(currBlock, newTextNode(data[beg:end]))
+	}
+	p.nesting--
+}
+
+// single and double emphasis parsing
+func emphasis(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+	c := data[0]
+
+	n := len(data)
+	if n > 2 && data[1] != c {
+		// whitespace cannot follow an opening emphasis;
+		// strikethrough only takes two characters '~~'
+		if isSpace(data[1]) {
+			return 0, nil
+		}
+		if p.extensions&SuperSubscript != 0 && c == '~' {
+			// potential subscript, no spaces, except when escaped, helperEmphasis does
+			// not check that for us, so walk the bytes and check.
+			ret := skipUntilChar(data[1:], 0, c)
+			if ret == 0 {
+				return 0, nil
+			}
+			ret++ // we started with data[1:] above.
+			for i := 1; i < ret; i++ {
+				if isSpace(data[i]) && !isEscape(data, i) {
+					return 0, nil
+				}
+			}
+			sub := &ast.Subscript{}
+			sub.Literal = data[1:ret]
+			return ret + 1, sub
+		}
+		ret, node := helperEmphasis(p, data[1:], c)
+		if ret == 0 {
+			return 0, nil
+		}
+
+		return ret + 1, node
+	}
+
+	if n > 3 && data[1] == c && data[2] != c {
+		if isSpace(data[2]) {
+			return 0, nil
+		}
+		ret, node := helperDoubleEmphasis(p, data[2:], c)
+		if ret == 0 {
+			return 0, nil
+		}
+
+		return ret + 2, node
+	}
+
+	if n > 4 && data[1] == c && data[2] == c && data[3] != c {
+		if c == '~' || isSpace(data[3]) {
+			return 0, nil
+		}
+		ret, node := helperTripleEmphasis(p, data, 3, c)
+		if ret == 0 {
+			return 0, nil
+		}
+
+		return ret + 3, node
+	}
+
+	return 0, nil
+}
+
+func codeSpan(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+
+	// count the number of backticks in the delimiter
+	nb := skipChar(data, 0, '`')
+
+	// find the next delimiter
+	i, end := 0, 0
+	for end = nb; end < len(data) && i < nb; end++ {
+		if data[end] == '`' {
+			i++
+		} else {
+			i = 0
+		}
+	}
+
+	// no matching delimiter?
+	if i < nb && end >= len(data) {
+		return 0, nil
+	}
+
+	// trim outside whitespace
+	fBegin := nb
+	for fBegin < end && data[fBegin] == ' ' {
+		fBegin++
+	}
+
+	fEnd := end - nb
+	for fEnd > fBegin && data[fEnd-1] == ' ' {
+		fEnd--
+	}
+
+	// render the code span
+	if fBegin != fEnd {
+		code := &ast.Code{}
+		code.Literal = data[fBegin:fEnd]
+		return end, code
+	}
+
+	return end, nil
+}
+
+// newline preceded by two spaces becomes <br>
+func maybeLineBreak(p *Parser, data []byte, offset int) (int, ast.Node) {
+	origOffset := offset
+	offset = skipChar(data, offset, ' ')
+
+	if offset < len(data) && data[offset] == '\n' {
+		if offset-origOffset >= 2 {
+			return offset - origOffset + 1, &ast.Hardbreak{}
+		}
+		return offset - origOffset, nil
+	}
+	return 0, nil
+}
+
+// newline without two spaces works when HardLineBreak is enabled
+func lineBreak(p *Parser, data []byte, offset int) (int, ast.Node) {
+	if p.extensions&HardLineBreak != 0 {
+		return 1, &ast.Hardbreak{}
+	}
+	return 0, nil
+}
+
+type linkType int
+
+const (
+	linkNormal linkType = iota
+	linkImg
+	linkDeferredFootnote
+	linkInlineFootnote
+	linkCitation
+)
+
+func isReferenceStyleLink(data []byte, pos int, t linkType) bool {
+	if t == linkDeferredFootnote {
+		return false
+	}
+	return pos < len(data)-1 && data[pos] == '[' && data[pos+1] != '^'
+}
+
+func maybeImage(p *Parser, data []byte, offset int) (int, ast.Node) {
+	if offset < len(data)-1 && data[offset+1] == '[' {
+		return link(p, data, offset)
+	}
+	return 0, nil
+}
+
+func maybeInlineFootnoteOrSuper(p *Parser, data []byte, offset int) (int, ast.Node) {
+	if offset < len(data)-1 && data[offset+1] == '[' {
+		return link(p, data, offset)
+	}
+
+	if p.extensions&SuperSubscript != 0 {
+		ret := skipUntilChar(data[offset:], 1, '^')
+		if ret == 0 {
+			return 0, nil
+		}
+		for i := offset; i < offset+ret; i++ {
+			if isSpace(data[i]) && !isEscape(data, i) {
+				return 0, nil
+			}
+		}
+		sup := &ast.Superscript{}
+		sup.Literal = data[offset+1 : offset+ret]
+		return offset + ret, sup
+	}
+
+	return 0, nil
+}
+
+// '[': parse a link or an image or a footnote or a citation
+func link(p *Parser, data []byte, offset int) (int, ast.Node) {
+	// no links allowed inside regular links, footnote, and deferred footnotes
+	if p.insideLink && (offset > 0 && data[offset-1] == '[' || len(data)-1 > offset && data[offset+1] == '^') {
+		return 0, nil
+	}
+
+	var t linkType
+	switch {
+	// special case: ![^text] == deferred footnote (that follows something with
+	// an exclamation point)
+	case p.extensions&Footnotes != 0 && len(data)-1 > offset && data[offset+1] == '^':
+		t = linkDeferredFootnote
+	// ![alt] == image
+	case offset >= 0 && data[offset] == '!':
+		t = linkImg
+		offset++
+	// [@citation], [@-citation], [@?citation], [@!citation]
+	case p.extensions&Mmark != 0 && len(data)-1 > offset && data[offset+1] == '@':
+		t = linkCitation
+	// [text] == regular link
+	// ^[text] == inline footnote
+	// [^refId] == deferred footnote
+	case p.extensions&Footnotes != 0:
+		if offset >= 0 && data[offset] == '^' {
+			t = linkInlineFootnote
+			offset++
+		} else if len(data)-1 > offset && data[offset+1] == '^' {
+			t = linkDeferredFootnote
+		}
+	default:
+		t = linkNormal
+	}
+
+	data = data[offset:]
+
+	if t == linkCitation {
+		return citation(p, data, 0)
+	}
+
+	var (
+		i                               = 1
+		noteID                          int
+		title, link, linkID, altContent []byte
+		textHasNl                       = false
+	)
+
+	if t == linkDeferredFootnote {
+		i++
+	}
+
+	// look for the matching closing bracket
+	for level := 1; level > 0 && i < len(data); i++ {
+		switch {
+		case data[i] == '\n':
+			textHasNl = true
+
+		case data[i-1] == '\\':
+			continue
+
+		case data[i] == '[':
+			level++
+
+		case data[i] == ']':
+			level--
+			if level <= 0 {
+				i-- // compensate for extra i++ in for loop
+			}
+		}
+	}
+
+	if i >= len(data) {
+		return 0, nil
+	}
+
+	txtE := i
+	i++
+	var footnoteNode ast.Node
+
+	// skip any amount of whitespace or newline
+	// (this is much more lax than original markdown syntax)
+	i = skipSpace(data, i)
+
+	// inline style link
+	switch {
+	case i < len(data) && data[i] == '(':
+		// skip initial whitespace
+		i++
+
+		i = skipSpace(data, i)
+
+		linkB := i
+
+		// look for link end: ' " )
+	findlinkend:
+		for i < len(data) {
+			switch {
+			case data[i] == '\\':
+				i += 2
+
+			case data[i] == ')' || data[i] == '\'' || data[i] == '"':
+				break findlinkend
+
+			default:
+				i++
+			}
+		}
+
+		if i >= len(data) {
+			return 0, nil
+		}
+		linkE := i
+
+		// look for title end if present
+		titleB, titleE := 0, 0
+		if data[i] == '\'' || data[i] == '"' {
+			i++
+			titleB = i
+
+		findtitleend:
+			for i < len(data) {
+				switch {
+				case data[i] == '\\':
+					i += 2
+
+				case data[i] == ')':
+					break findtitleend
+
+				default:
+					i++
+				}
+			}
+
+			if i >= len(data) {
+				return 0, nil
+			}
+
+			// skip whitespace after title
+			titleE = i - 1
+			for titleE > titleB && isSpace(data[titleE]) {
+				titleE--
+			}
+
+			// check for closing quote presence
+			if data[titleE] != '\'' && data[titleE] != '"' {
+				titleB, titleE = 0, 0
+				linkE = i
+			}
+		}
+
+		// remove whitespace at the end of the link
+		for linkE > linkB && isSpace(data[linkE-1]) {
+			linkE--
+		}
+
+		// remove optional angle brackets around the link
+		if data[linkB] == '<' {
+			linkB++
+		}
+		if data[linkE-1] == '>' {
+			linkE--
+		}
+
+		// build escaped link and title
+		if linkE > linkB {
+			link = data[linkB:linkE]
+		}
+
+		if titleE > titleB {
+			title = data[titleB:titleE]
+		}
+
+		i++
+
+	// reference style link
+	case isReferenceStyleLink(data, i, t):
+		var id []byte
+		altContentConsidered := false
+
+		// look for the id
+		i++
+		linkB := i
+		i = skipUntilChar(data, i, ']')
+
+		if i >= len(data) {
+			return 0, nil
+		}
+		linkE := i
+
+		// find the reference
+		if linkB == linkE {
+			if textHasNl {
+				var b bytes.Buffer
+
+				for j := 1; j < txtE; j++ {
+					switch {
+					case data[j] != '\n':
+						b.WriteByte(data[j])
+					case data[j-1] != ' ':
+						b.WriteByte(' ')
+					}
+				}
+
+				id = b.Bytes()
+			} else {
+				id = data[1:txtE]
+				altContentConsidered = true
+			}
+		} else {
+			id = data[linkB:linkE]
+		}
+
+		// find the reference with matching id
+		lr, ok := p.getRef(string(id))
+		if !ok {
+			return 0, nil
+		}
+
+		// keep link and title from reference
+		linkID = id
+		link = lr.link
+		title = lr.title
+		if altContentConsidered {
+			altContent = lr.text
+		}
+		i++
+
+	// shortcut reference style link or reference or inline footnote
+	default:
+		var id []byte
+
+		// craft the id
+		if textHasNl {
+			var b bytes.Buffer
+
+			for j := 1; j < txtE; j++ {
+				switch {
+				case data[j] != '\n':
+					b.WriteByte(data[j])
+				case data[j-1] != ' ':
+					b.WriteByte(' ')
+				}
+			}
+
+			id = b.Bytes()
+		} else {
+			if t == linkDeferredFootnote {
+				id = data[2:txtE] // get rid of the ^
+			} else {
+				id = data[1:txtE]
+			}
+		}
+
+		footnoteNode = &ast.ListItem{}
+		if t == linkInlineFootnote {
+			// create a new reference
+			noteID = len(p.notes) + 1
+
+			var fragment []byte
+			if len(id) > 0 {
+				if len(id) < 16 {
+					fragment = make([]byte, len(id))
+				} else {
+					fragment = make([]byte, 16)
+				}
+				copy(fragment, slugify(id))
+			} else {
+				fragment = append([]byte("footnote-"), []byte(strconv.Itoa(noteID))...)
+			}
+
+			ref := &reference{
+				noteID:   noteID,
+				hasBlock: false,
+				link:     fragment,
+				title:    id,
+				footnote: footnoteNode,
+			}
+
+			p.notes = append(p.notes, ref)
+			p.refsRecord[string(ref.link)] = struct{}{}
+
+			link = ref.link
+			title = ref.title
+		} else {
+			// find the reference with matching id
+			lr, ok := p.getRef(string(id))
+			if !ok {
+				return 0, nil
+			}
+
+			if t == linkDeferredFootnote && !p.isFootnote(lr) {
+				lr.noteID = len(p.notes) + 1
+				lr.footnote = footnoteNode
+				p.notes = append(p.notes, lr)
+				p.refsRecord[string(lr.link)] = struct{}{}
+			}
+
+			// keep link and title from reference
+			link = lr.link
+			// if inline footnote, title == footnote contents
+			title = lr.title
+			noteID = lr.noteID
+		}
+
+		// rewind the whitespace
+		i = txtE + 1
+	}
+
+	var uLink []byte
+	if t == linkNormal || t == linkImg {
+		if len(link) > 0 {
+			var uLinkBuf bytes.Buffer
+			unescapeText(&uLinkBuf, link)
+			uLink = uLinkBuf.Bytes()
+		}
+
+		// links need something to click on and somewhere to go
+		if len(uLink) == 0 || (t == linkNormal && txtE <= 1) {
+			return 0, nil
+		}
+	}
+
+	// call the relevant rendering function
+	switch t {
+	case linkNormal:
+		link := &ast.Link{
+			Destination: normalizeURI(uLink),
+			Title:       title,
+			DeferredID:  linkID,
+		}
+		if len(altContent) > 0 {
+			ast.AppendChild(link, newTextNode(altContent))
+		} else {
+			// links cannot contain other links, so turn off link parsing
+			// temporarily and recurse
+			insideLink := p.insideLink
+			p.insideLink = true
+			p.Inline(link, data[1:txtE])
+			p.insideLink = insideLink
+		}
+		return i, link
+
+	case linkImg:
+		image := &ast.Image{
+			Destination: uLink,
+			Title:       title,
+		}
+		ast.AppendChild(image, newTextNode(data[1:txtE]))
+		return i + 1, image
+
+	case linkInlineFootnote, linkDeferredFootnote:
+		link := &ast.Link{
+			Destination: link,
+			Title:       title,
+			NoteID:      noteID,
+			Footnote:    footnoteNode,
+		}
+		if t == linkDeferredFootnote {
+			link.DeferredID = data[2:txtE]
+		}
+		if t == linkInlineFootnote {
+			i++
+		}
+		return i, link
+
+	default:
+		return 0, nil
+	}
+}
+
+func (p *Parser) inlineHTMLComment(data []byte) int {
+	if len(data) < 5 {
+		return 0
+	}
+	if data[0] != '<' || data[1] != '!' || data[2] != '-' || data[3] != '-' {
+		return 0
+	}
+	i := 5
+	// scan for an end-of-comment marker, across lines if necessary
+	for i < len(data) && !(data[i-2] == '-' && data[i-1] == '-' && data[i] == '>') {
+		i++
+	}
+	// no end-of-comment marker
+	if i >= len(data) {
+		return 0
+	}
+	return i + 1
+}
+
+func stripMailto(link []byte) []byte {
+	if bytes.HasPrefix(link, []byte("mailto://")) {
+		return link[9:]
+	} else if bytes.HasPrefix(link, []byte("mailto:")) {
+		return link[7:]
+	} else {
+		return link
+	}
+}
+
+// autolinkType specifies a kind of autolink that gets detected.
+type autolinkType int
+
+// These are the possible flag values for the autolink renderer.
+const (
+	notAutolink autolinkType = iota
+	normalAutolink
+	emailAutolink
+)
+
+// '<' when tags or autolinks are allowed
+func leftAngle(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+
+	if p.extensions&Mmark != 0 {
+		id, consumed := IsCallout(data)
+		if consumed > 0 {
+			node := &ast.Callout{}
+			node.ID = id
+			return consumed, node
+		}
+	}
+
+	altype, end := tagLength(data)
+	if size := p.inlineHTMLComment(data); size > 0 {
+		end = size
+	}
+	if end <= 2 {
+		return end, nil
+	}
+	if altype == notAutolink {
+		htmlTag := &ast.HTMLSpan{}
+		htmlTag.Literal = data[:end]
+		return end, htmlTag
+	}
+
+	var uLink bytes.Buffer
+	unescapeText(&uLink, data[1:end+1-2])
+	if uLink.Len() <= 0 {
+		return end, nil
+	}
+	link := uLink.Bytes()
+	node := &ast.Link{
+		Destination: link,
+	}
+	if altype == emailAutolink {
+		node.Destination = append([]byte("mailto:"), link...)
+	}
+	ast.AppendChild(node, newTextNode(stripMailto(link)))
+	return end, node
+}
+
+// '\\' backslash escape
+var escapeChars = []byte("\\`*_{}[]()#+-.!:|&<>~")
+
+func escape(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+
+	if len(data) <= 1 {
+		return 2, nil
+	}
+
+	if p.extensions&NonBlockingSpace != 0 && data[1] == ' ' {
+		return 2, &ast.NonBlockingSpace{}
+	}
+
+	if p.extensions&BackslashLineBreak != 0 && data[1] == '\n' {
+		return 2, &ast.Hardbreak{}
+	}
+
+	if bytes.IndexByte(escapeChars, data[1]) < 0 {
+		return 0, nil
+	}
+
+	return 2, newTextNode(data[1:2])
+}
+
+func unescapeText(ob *bytes.Buffer, src []byte) {
+	i := 0
+	for i < len(src) {
+		org := i
+		for i < len(src) && src[i] != '\\' {
+			i++
+		}
+
+		if i > org {
+			ob.Write(src[org:i])
+		}
+
+		if i+1 >= len(src) {
+			break
+		}
+
+		ob.WriteByte(src[i+1])
+		i += 2
+	}
+}
+
+// '&' escaped when it doesn't belong to an entity
+// valid entities are assumed to be anything matching &#?[A-Za-z0-9]+;
+func entity(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+
+	end := skipCharN(data, 1, '#', 1)
+	end = skipAlnum(data, end)
+
+	if end < len(data) && data[end] == ';' {
+		end++ // real entity
+	} else {
+		return 0, nil // lone '&'
+	}
+
+	ent := data[:end]
+	// undo &amp; escaping or it will be converted to &amp;amp; by another
+	// escaper in the renderer
+	if bytes.Equal(ent, []byte("&amp;")) {
+		ent = []byte{'&'}
+	}
+
+	return end, newTextNode(ent)
+}
+
+func linkEndsWithEntity(data []byte, linkEnd int) bool {
+	entityRanges := htmlEntityRe.FindAllIndex(data[:linkEnd], -1)
+	return entityRanges != nil && entityRanges[len(entityRanges)-1][1] == linkEnd
+}
+
+// hasPrefixCaseInsensitive is a custom implementation of
+//     strings.HasPrefix(strings.ToLower(s), prefix)
+// we rolled our own because ToLower pulls in a huge machinery of lowercasing
+// anything from Unicode and that's very slow. Since this func will only be
+// used on ASCII protocol prefixes, we can take shortcuts.
+func hasPrefixCaseInsensitive(s, prefix []byte) bool {
+	if len(s) < len(prefix) {
+		return false
+	}
+	delta := byte('a' - 'A')
+	for i, b := range prefix {
+		if b != s[i] && b != s[i]+delta {
+			return false
+		}
+	}
+	return true
+}
+
+var protocolPrefixes = [][]byte{
+	[]byte("http://"),
+	[]byte("https://"),
+	[]byte("ftp://"),
+	[]byte("file://"),
+	[]byte("mailto:"),
+}
+
+const shortestPrefix = 6 // len("ftp://"), the shortest of the above
+
+func maybeAutoLink(p *Parser, data []byte, offset int) (int, ast.Node) {
+	// quick check to rule out most false hits
+	if p.insideLink || len(data) < offset+shortestPrefix {
+		return 0, nil
+	}
+	for _, prefix := range protocolPrefixes {
+		endOfHead := offset + 8 // 8 is the len() of the longest prefix
+		if endOfHead > len(data) {
+			endOfHead = len(data)
+		}
+		if hasPrefixCaseInsensitive(data[offset:endOfHead], prefix) {
+			return autoLink(p, data, offset)
+		}
+	}
+	return 0, nil
+}
+
+func autoLink(p *Parser, data []byte, offset int) (int, ast.Node) {
+	// Now a more expensive check to see if we're not inside an anchor element
+	anchorStart := offset
+	offsetFromAnchor := 0
+	for anchorStart > 0 && data[anchorStart] != '<' {
+		anchorStart--
+		offsetFromAnchor++
+	}
+
+	anchorStr := anchorRe.Find(data[anchorStart:])
+	if anchorStr != nil {
+		anchorClose := &ast.HTMLSpan{}
+		anchorClose.Literal = anchorStr[offsetFromAnchor:]
+		return len(anchorStr) - offsetFromAnchor, anchorClose
+	}
+
+	// scan backward for a word boundary
+	rewind := 0
+	for offset-rewind > 0 && rewind <= 7 && isLetter(data[offset-rewind-1]) {
+		rewind++
+	}
+	if rewind > 6 { // longest supported protocol is "mailto" which has 6 letters
+		return 0, nil
+	}
+
+	origData := data
+	data = data[offset-rewind:]
+
+	if !isSafeLink(data) {
+		return 0, nil
+	}
+
+	linkEnd := 0
+	for linkEnd < len(data) && !isEndOfLink(data[linkEnd]) {
+		linkEnd++
+	}
+
+	// Skip punctuation at the end of the link
+	if (data[linkEnd-1] == '.' || data[linkEnd-1] == ',') && data[linkEnd-2] != '\\' {
+		linkEnd--
+	}
+
+	// But don't skip semicolon if it's a part of escaped entity:
+	if data[linkEnd-1] == ';' && data[linkEnd-2] != '\\' && !linkEndsWithEntity(data, linkEnd) {
+		linkEnd--
+	}
+
+	// See if the link finishes with a punctuation sign that can be closed.
+	var copen byte
+	switch data[linkEnd-1] {
+	case '"':
+		copen = '"'
+	case '\'':
+		copen = '\''
+	case ')':
+		copen = '('
+	case ']':
+		copen = '['
+	case '}':
+		copen = '{'
+	default:
+		copen = 0
+	}
+
+	if copen != 0 {
+		bufEnd := offset - rewind + linkEnd - 2
+
+		openDelim := 1
+
+		/* Try to close the final punctuation sign in this same line;
+		 * if we managed to close it outside of the URL, that means that it's
+		 * not part of the URL. If it closes inside the URL, that means it
+		 * is part of the URL.
+		 *
+		 * Examples:
+		 *
+		 *      foo http://www.pokemon.com/Pikachu_(Electric) bar
+		 *              => http://www.pokemon.com/Pikachu_(Electric)
+		 *
+		 *      foo (http://www.pokemon.com/Pikachu_(Electric)) bar
+		 *              => http://www.pokemon.com/Pikachu_(Electric)
+		 *
+		 *      foo http://www.pokemon.com/Pikachu_(Electric)) bar
+		 *              => http://www.pokemon.com/Pikachu_(Electric))
+		 *
+		 *      (foo http://www.pokemon.com/Pikachu_(Electric)) bar
+		 *              => foo http://www.pokemon.com/Pikachu_(Electric)
+		 */
+
+		for bufEnd >= 0 && origData[bufEnd] != '\n' && openDelim != 0 {
+			if origData[bufEnd] == data[linkEnd-1] {
+				openDelim++
+			}
+
+			if origData[bufEnd] == copen {
+				openDelim--
+			}
+
+			bufEnd--
+		}
+
+		if openDelim == 0 {
+			linkEnd--
+		}
+	}
+
+	var uLink bytes.Buffer
+	unescapeText(&uLink, data[:linkEnd])
+
+	if uLink.Len() > 0 {
+		node := &ast.Link{
+			Destination: uLink.Bytes(),
+		}
+		ast.AppendChild(node, newTextNode(uLink.Bytes()))
+		return linkEnd, node
+	}
+
+	return linkEnd, nil
+}
+
+func isEndOfLink(char byte) bool {
+	return isSpace(char) || char == '<'
+}
+
+var validUris = [][]byte{[]byte("http://"), []byte("https://"), []byte("ftp://"), []byte("mailto://")}
+var validPaths = [][]byte{[]byte("/"), []byte("./"), []byte("../")}
+
+func isSafeLink(link []byte) bool {
+	nLink := len(link)
+	for _, path := range validPaths {
+		nPath := len(path)
+		linkPrefix := link[:nPath]
+		if nLink >= nPath && bytes.Equal(linkPrefix, path) {
+			if nLink == nPath {
+				return true
+			} else if isAlnum(link[nPath]) {
+				return true
+			}
+		}
+	}
+
+	for _, prefix := range validUris {
+		// TODO: handle unicode here
+		// case-insensitive prefix test
+		nPrefix := len(prefix)
+		if nLink > nPrefix {
+			linkPrefix := bytes.ToLower(link[:nPrefix])
+			if bytes.Equal(linkPrefix, prefix) && isAlnum(link[nPrefix]) {
+				return true
+			}
+		}
+	}
+
+	return false
+}
+
+// return the length of the given tag, or 0 is it's not valid
+func tagLength(data []byte) (autolink autolinkType, end int) {
+	var i, j int
+
+	// a valid tag can't be shorter than 3 chars
+	if len(data) < 3 {
+		return notAutolink, 0
+	}
+
+	// begins with a '<' optionally followed by '/', followed by letter or number
+	if data[0] != '<' {
+		return notAutolink, 0
+	}
+	if data[1] == '/' {
+		i = 2
+	} else {
+		i = 1
+	}
+
+	if !isAlnum(data[i]) {
+		return notAutolink, 0
+	}
+
+	// scheme test
+	autolink = notAutolink
+
+	// try to find the beginning of an URI
+	for i < len(data) && (isAlnum(data[i]) || data[i] == '.' || data[i] == '+' || data[i] == '-') {
+		i++
+	}
+
+	if i > 1 && i < len(data) && data[i] == '@' {
+		if j = isMailtoAutoLink(data[i:]); j != 0 {
+			return emailAutolink, i + j
+		}
+	}
+
+	if i > 2 && i < len(data) && data[i] == ':' {
+		autolink = normalAutolink
+		i++
+	}
+
+	// complete autolink test: no whitespace or ' or "
+	switch {
+	case i >= len(data):
+		autolink = notAutolink
+	case autolink != notAutolink:
+		j = i
+
+		for i < len(data) {
+			if data[i] == '\\' {
+				i += 2
+			} else if data[i] == '>' || data[i] == '\'' || data[i] == '"' || isSpace(data[i]) {
+				break
+			} else {
+				i++
+			}
+
+		}
+
+		if i >= len(data) {
+			return autolink, 0
+		}
+		if i > j && data[i] == '>' {
+			return autolink, i + 1
+		}
+
+		// one of the forbidden chars has been found
+		autolink = notAutolink
+	}
+	i += bytes.IndexByte(data[i:], '>')
+	if i < 0 {
+		return autolink, 0
+	}
+	return autolink, i + 1
+}
+
+// look for the address part of a mail autolink and '>'
+// this is less strict than the original markdown e-mail address matching
+func isMailtoAutoLink(data []byte) int {
+	nb := 0
+
+	// address is assumed to be: [-@._a-zA-Z0-9]+ with exactly one '@'
+	for i, c := range data {
+		if isAlnum(c) {
+			continue
+		}
+
+		switch c {
+		case '@':
+			nb++
+
+		case '-', '.', '_':
+			break
+
+		case '>':
+			if nb == 1 {
+				return i + 1
+			}
+			return 0
+		default:
+			return 0
+		}
+	}
+
+	return 0
+}
+
+// look for the next emph char, skipping other constructs
+func helperFindEmphChar(data []byte, c byte) int {
+	i := 0
+
+	for i < len(data) {
+		for i < len(data) && data[i] != c && data[i] != '`' && data[i] != '[' {
+			i++
+		}
+		if i >= len(data) {
+			return 0
+		}
+		// do not count escaped chars
+		if i != 0 && data[i-1] == '\\' {
+			i++
+			continue
+		}
+		if data[i] == c {
+			return i
+		}
+
+		if data[i] == '`' {
+			// skip a code span
+			tmpI := 0
+			i++
+			for i < len(data) && data[i] != '`' {
+				if tmpI == 0 && data[i] == c {
+					tmpI = i
+				}
+				i++
+			}
+			if i >= len(data) {
+				return tmpI
+			}
+			i++
+		} else if data[i] == '[' {
+			// skip a link
+			tmpI := 0
+			i++
+			for i < len(data) && data[i] != ']' {
+				if tmpI == 0 && data[i] == c {
+					tmpI = i
+				}
+				i++
+			}
+			i++
+			for i < len(data) && (data[i] == ' ' || data[i] == '\n') {
+				i++
+			}
+			if i >= len(data) {
+				return tmpI
+			}
+			if data[i] != '[' && data[i] != '(' { // not a link
+				if tmpI > 0 {
+					return tmpI
+				}
+				continue
+			}
+			cc := data[i]
+			i++
+			for i < len(data) && data[i] != cc {
+				if tmpI == 0 && data[i] == c {
+					return i
+				}
+				i++
+			}
+			if i >= len(data) {
+				return tmpI
+			}
+			i++
+		}
+	}
+	return 0
+}
+
+func helperEmphasis(p *Parser, data []byte, c byte) (int, ast.Node) {
+	i := 0
+
+	// skip one symbol if coming from emph3
+	if len(data) > 1 && data[0] == c && data[1] == c {
+		i = 1
+	}
+
+	for i < len(data) {
+		length := helperFindEmphChar(data[i:], c)
+		if length == 0 {
+			return 0, nil
+		}
+		i += length
+		if i >= len(data) {
+			return 0, nil
+		}
+
+		if i+1 < len(data) && data[i+1] == c {
+			i++
+			continue
+		}
+
+		if data[i] == c && !isSpace(data[i-1]) {
+
+			if p.extensions&NoIntraEmphasis != 0 {
+				if !(i+1 == len(data) || isSpace(data[i+1]) || isPunctuation(data[i+1])) {
+					continue
+				}
+			}
+
+			emph := &ast.Emph{}
+			p.Inline(emph, data[:i])
+			return i + 1, emph
+		}
+	}
+
+	return 0, nil
+}
+
+func helperDoubleEmphasis(p *Parser, data []byte, c byte) (int, ast.Node) {
+	i := 0
+
+	for i < len(data) {
+		length := helperFindEmphChar(data[i:], c)
+		if length == 0 {
+			return 0, nil
+		}
+		i += length
+
+		if i+1 < len(data) && data[i] == c && data[i+1] == c && i > 0 && !isSpace(data[i-1]) {
+			var node ast.Node = &ast.Strong{}
+			if c == '~' {
+				node = &ast.Del{}
+			}
+			p.Inline(node, data[:i])
+			return i + 2, node
+		}
+		i++
+	}
+	return 0, nil
+}
+
+func helperTripleEmphasis(p *Parser, data []byte, offset int, c byte) (int, ast.Node) {
+	i := 0
+	origData := data
+	data = data[offset:]
+
+	for i < len(data) {
+		length := helperFindEmphChar(data[i:], c)
+		if length == 0 {
+			return 0, nil
+		}
+		i += length
+
+		// skip whitespace preceded symbols
+		if data[i] != c || isSpace(data[i-1]) {
+			continue
+		}
+
+		switch {
+		case i+2 < len(data) && data[i+1] == c && data[i+2] == c:
+			// triple symbol found
+			strong := &ast.Strong{}
+			em := &ast.Emph{}
+			ast.AppendChild(strong, em)
+			p.Inline(em, data[:i])
+			return i + 3, strong
+		case i+1 < len(data) && data[i+1] == c:
+			// double symbol found, hand over to emph1
+			length, node := helperEmphasis(p, origData[offset-2:], c)
+			if length == 0 {
+				return 0, nil
+			}
+			return length - 2, node
+		default:
+			// single symbol found, hand over to emph2
+			length, node := helperDoubleEmphasis(p, origData[offset-1:], c)
+			if length == 0 {
+				return 0, nil
+			}
+			return length - 1, node
+		}
+	}
+	return 0, nil
+}
+
+// math handle inline math wrapped with '$'
+func math(p *Parser, data []byte, offset int) (int, ast.Node) {
+	data = data[offset:]
+
+	// too short, or block math
+	if len(data) <= 2 || data[1] == '$' {
+		return 0, nil
+	}
+
+	// find next '$'
+	var end int
+	for end = 1; end < len(data) && data[end] != '$'; end++ {
+	}
+
+	// $ not match
+	if end == len(data) {
+		return 0, nil
+	}
+
+	// create inline math node
+	math := &ast.Math{}
+	math.Literal = data[1:end]
+	return end + 1, math
+}
+
+func newTextNode(d []byte) *ast.Text {
+	return &ast.Text{ast.Leaf{Literal: d}}
+}
+
+func normalizeURI(s []byte) []byte {
+	return s // TODO: implement
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/matter.go b/vendor/github.com/gomarkdown/markdown/parser/matter.go
new file mode 100644
index 00000000..92686357
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/matter.go
@@ -0,0 +1,36 @@
+package parser
+
+import (
+	"bytes"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+func (p *Parser) documentMatter(data []byte) int {
+	if data[0] != '{' {
+		return 0
+	}
+
+	consumed := 0
+	matter := ast.DocumentMatterNone
+	if bytes.HasPrefix(data, []byte("{frontmatter}")) {
+		consumed = len("{frontmatter}")
+		matter = ast.DocumentMatterFront
+	}
+	if bytes.HasPrefix(data, []byte("{mainmatter}")) {
+		consumed = len("{mainmatter}")
+		matter = ast.DocumentMatterMain
+	}
+	if bytes.HasPrefix(data, []byte("{backmatter}")) {
+		consumed = len("{backmatter}")
+		matter = ast.DocumentMatterBack
+	}
+	if consumed == 0 {
+		return 0
+	}
+	node := &ast.DocumentMatter{Matter: matter}
+	p.addBlock(node)
+	p.finalize(node)
+
+	return consumed
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/options.go b/vendor/github.com/gomarkdown/markdown/parser/options.go
new file mode 100644
index 00000000..d3d0c088
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/options.go
@@ -0,0 +1,32 @@
+package parser
+
+import (
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// Flags control optional behavior of parser.
+type Flags int
+
+// Options is a collection of supplementary parameters tweaking the behavior of various parts of the parser.
+type Options struct {
+	ParserHook    BlockFunc
+	ReadIncludeFn ReadIncludeFunc
+
+	Flags Flags // Flags allow customizing parser's behavior
+}
+
+// Parser renderer configuration options.
+const (
+	FlagsNone        Flags = 0
+	SkipFootnoteList Flags = 1 << iota // Skip adding the footnote list (regardless if they are parsed)
+)
+
+// BlockFunc allows to registration of a parser function. If successful it
+// returns an ast.Node, a buffer that should be parsed as a block and the the number of bytes consumed.
+type BlockFunc func(data []byte) (ast.Node, []byte, int)
+
+// ReadIncludeFunc should read the file under path and returns the read bytes,
+// from will be set to the name of the current file being parsed. Initially
+// this will be empty. address is the optional address specifier of which lines
+// of the file to return. If this function is not set no data will be read.
+type ReadIncludeFunc func(from, path string, address []byte) []byte
diff --git a/vendor/github.com/gomarkdown/markdown/parser/parser.go b/vendor/github.com/gomarkdown/markdown/parser/parser.go
new file mode 100644
index 00000000..c7302dfd
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/parser.go
@@ -0,0 +1,812 @@
+/*
+Package parser implements parser for markdown text that generates AST (abstract syntax tree).
+*/
+package parser
+
+import (
+	"bytes"
+	"fmt"
+	"strings"
+	"unicode/utf8"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// Extensions is a bitmask of enabled parser extensions.
+type Extensions int
+
+// Bit flags representing markdown parsing extensions.
+// Use | (or) to specify multiple extensions.
+const (
+	NoExtensions           Extensions = 0
+	NoIntraEmphasis        Extensions = 1 << iota // Ignore emphasis markers inside words
+	Tables                                        // Parse tables
+	FencedCode                                    // Parse fenced code blocks
+	Autolink                                      // Detect embedded URLs that are not explicitly marked
+	Strikethrough                                 // Strikethrough text using ~~test~~
+	LaxHTMLBlocks                                 // Loosen up HTML block parsing rules
+	SpaceHeadings                                 // Be strict about prefix heading rules
+	HardLineBreak                                 // Translate newlines into line breaks
+	NonBlockingSpace                              // Translate backspace spaces into line non-blocking spaces
+	TabSizeEight                                  // Expand tabs to eight spaces instead of four
+	Footnotes                                     // Pandoc-style footnotes
+	NoEmptyLineBeforeBlock                        // No need to insert an empty line to start a (code, quote, ordered list, unordered list) block
+	HeadingIDs                                    // specify heading IDs  with {#id}
+	Titleblock                                    // Titleblock ala pandoc
+	AutoHeadingIDs                                // Create the heading ID from the text
+	BackslashLineBreak                            // Translate trailing backslashes into line breaks
+	DefinitionLists                               // Parse definition lists
+	MathJax                                       // Parse MathJax
+	OrderedListStart                              // Keep track of the first number used when starting an ordered list.
+	Attributes                                    // Block Attributes
+	SuperSubscript                                // Super- and subscript support: 2^10^, H~2~O.
+	EmptyLinesBreakList                           // 2 empty lines break out of list
+	Includes                                      // Support including other files.
+	Mmark                                         // Support Mmark syntax, see https://mmark.nl/syntax
+
+	CommonExtensions Extensions = NoIntraEmphasis | Tables | FencedCode |
+		Autolink | Strikethrough | SpaceHeadings | HeadingIDs |
+		BackslashLineBreak | DefinitionLists | MathJax
+)
+
+// The size of a tab stop.
+const (
+	tabSizeDefault = 4
+	tabSizeDouble  = 8
+)
+
+// for each character that triggers a response when parsing inline data.
+type inlineParser func(p *Parser, data []byte, offset int) (int, ast.Node)
+
+// ReferenceOverrideFunc is expected to be called with a reference string and
+// return either a valid Reference type that the reference string maps to or
+// nil. If overridden is false, the default reference logic will be executed.
+// See the documentation in Options for more details on use-case.
+type ReferenceOverrideFunc func(reference string) (ref *Reference, overridden bool)
+
+// Parser is a type that holds extensions and the runtime state used by
+// Parse, and the renderer. You can not use it directly, construct it with New.
+type Parser struct {
+
+	// ReferenceOverride is an optional function callback that is called every
+	// time a reference is resolved. It can be set before starting parsing.
+	//
+	// In Markdown, the link reference syntax can be made to resolve a link to
+	// a reference instead of an inline URL, in one of the following ways:
+	//
+	//  * [link text][refid]
+	//  * [refid][]
+	//
+	// Usually, the refid is defined at the bottom of the Markdown document. If
+	// this override function is provided, the refid is passed to the override
+	// function first, before consulting the defined refids at the bottom. If
+	// the override function indicates an override did not occur, the refids at
+	// the bottom will be used to fill in the link details.
+	ReferenceOverride ReferenceOverrideFunc
+
+	Opts Options
+
+	// after parsing, this is AST root of parsed markdown text
+	Doc ast.Node
+
+	extensions Extensions
+
+	refs           map[string]*reference
+	refsRecord     map[string]struct{}
+	inlineCallback [256]inlineParser
+	nesting        int
+	maxNesting     int
+	insideLink     bool
+	indexCnt       int // incremented after every index
+
+	// Footnotes need to be ordered as well as available to quickly check for
+	// presence. If a ref is also a footnote, it's stored both in refs and here
+	// in notes. Slice is nil if footnotes not enabled.
+	notes []*reference
+
+	tip                  ast.Node // = doc
+	oldTip               ast.Node
+	lastMatchedContainer ast.Node // = doc
+	allClosed            bool
+
+	// Attributes are attached to block level elements.
+	attr *ast.Attribute
+
+	includeStack *incStack
+}
+
+// New creates a markdown parser with CommonExtensions.
+//
+// You can then call `doc := p.Parse(markdown)` to parse markdown document
+// and `markdown.Render(doc, renderer)` to convert it to another format with
+// a renderer.
+func New() *Parser {
+	return NewWithExtensions(CommonExtensions)
+}
+
+// NewWithExtensions creates a markdown parser with given extensions.
+func NewWithExtensions(extension Extensions) *Parser {
+	p := Parser{
+		refs:         make(map[string]*reference),
+		refsRecord:   make(map[string]struct{}),
+		maxNesting:   16,
+		insideLink:   false,
+		Doc:          &ast.Document{},
+		extensions:   extension,
+		allClosed:    true,
+		includeStack: newIncStack(),
+	}
+	p.tip = p.Doc
+	p.oldTip = p.Doc
+	p.lastMatchedContainer = p.Doc
+
+	p.inlineCallback[' '] = maybeLineBreak
+	p.inlineCallback['*'] = emphasis
+	p.inlineCallback['_'] = emphasis
+	if p.extensions&Strikethrough != 0 {
+		p.inlineCallback['~'] = emphasis
+	}
+	p.inlineCallback['`'] = codeSpan
+	p.inlineCallback['\n'] = lineBreak
+	p.inlineCallback['['] = link
+	p.inlineCallback['<'] = leftAngle
+	p.inlineCallback['\\'] = escape
+	p.inlineCallback['&'] = entity
+	p.inlineCallback['!'] = maybeImage
+	if p.extensions&Mmark != 0 {
+		p.inlineCallback['('] = maybeShortRefOrIndex
+	}
+	p.inlineCallback['^'] = maybeInlineFootnoteOrSuper
+	if p.extensions&Autolink != 0 {
+		p.inlineCallback['h'] = maybeAutoLink
+		p.inlineCallback['m'] = maybeAutoLink
+		p.inlineCallback['f'] = maybeAutoLink
+		p.inlineCallback['H'] = maybeAutoLink
+		p.inlineCallback['M'] = maybeAutoLink
+		p.inlineCallback['F'] = maybeAutoLink
+	}
+	if p.extensions&MathJax != 0 {
+		p.inlineCallback['$'] = math
+	}
+
+	return &p
+}
+
+func (p *Parser) getRef(refid string) (ref *reference, found bool) {
+	if p.ReferenceOverride != nil {
+		r, overridden := p.ReferenceOverride(refid)
+		if overridden {
+			if r == nil {
+				return nil, false
+			}
+			return &reference{
+				link:     []byte(r.Link),
+				title:    []byte(r.Title),
+				noteID:   0,
+				hasBlock: false,
+				text:     []byte(r.Text)}, true
+		}
+	}
+	// refs are case insensitive
+	ref, found = p.refs[strings.ToLower(refid)]
+	return ref, found
+}
+
+func (p *Parser) isFootnote(ref *reference) bool {
+	_, ok := p.refsRecord[string(ref.link)]
+	return ok
+}
+
+func (p *Parser) finalize(block ast.Node) {
+	p.tip = block.GetParent()
+}
+
+func (p *Parser) addChild(node ast.Node) ast.Node {
+	for !canNodeContain(p.tip, node) {
+		p.finalize(p.tip)
+	}
+	ast.AppendChild(p.tip, node)
+	p.tip = node
+	return node
+}
+
+func canNodeContain(n ast.Node, v ast.Node) bool {
+	switch n.(type) {
+	case *ast.List:
+		return isListItem(v)
+	case *ast.Document, *ast.BlockQuote, *ast.Aside, *ast.ListItem, *ast.CaptionFigure:
+		return !isListItem(v)
+	case *ast.Table:
+		switch v.(type) {
+		case *ast.TableHeader, *ast.TableBody, *ast.TableFooter:
+			return true
+		default:
+			return false
+		}
+	case *ast.TableHeader, *ast.TableBody, *ast.TableFooter:
+		_, ok := v.(*ast.TableRow)
+		return ok
+	case *ast.TableRow:
+		_, ok := v.(*ast.TableCell)
+		return ok
+	}
+	return false
+}
+
+func (p *Parser) closeUnmatchedBlocks() {
+	if p.allClosed {
+		return
+	}
+	for p.oldTip != p.lastMatchedContainer {
+		parent := p.oldTip.GetParent()
+		p.finalize(p.oldTip)
+		p.oldTip = parent
+	}
+	p.allClosed = true
+}
+
+// Reference represents the details of a link.
+// See the documentation in Options for more details on use-case.
+type Reference struct {
+	// Link is usually the URL the reference points to.
+	Link string
+	// Title is the alternate text describing the link in more detail.
+	Title string
+	// Text is the optional text to override the ref with if the syntax used was
+	// [refid][]
+	Text string
+}
+
+// Parse generates AST (abstract syntax tree) representing markdown document.
+//
+// The result is a root of the tree whose underlying type is *ast.Document
+//
+// You can then convert AST to html using html.Renderer, to some other format
+// using a custom renderer or transform the tree.
+func (p *Parser) Parse(input []byte) ast.Node {
+	p.block(input)
+	// Walk the tree and finish up some of unfinished blocks
+	for p.tip != nil {
+		p.finalize(p.tip)
+	}
+	// Walk the tree again and process inline markdown in each block
+	ast.WalkFunc(p.Doc, func(node ast.Node, entering bool) ast.WalkStatus {
+		switch node.(type) {
+		case *ast.Paragraph, *ast.Heading, *ast.TableCell:
+			p.Inline(node, node.AsContainer().Content)
+			node.AsContainer().Content = nil
+		}
+		return ast.GoToNext
+	})
+
+	if p.Opts.Flags&SkipFootnoteList == 0 {
+		p.parseRefsToAST()
+	}
+	return p.Doc
+}
+
+func (p *Parser) parseRefsToAST() {
+	if p.extensions&Footnotes == 0 || len(p.notes) == 0 {
+		return
+	}
+	p.tip = p.Doc
+	list := &ast.List{
+		IsFootnotesList: true,
+		ListFlags:       ast.ListTypeOrdered,
+	}
+	p.addBlock(&ast.Footnotes{})
+	block := p.addBlock(list)
+	flags := ast.ListItemBeginningOfList
+	// Note: this loop is intentionally explicit, not range-form. This is
+	// because the body of the loop will append nested footnotes to p.notes and
+	// we need to process those late additions. Range form would only walk over
+	// the fixed initial set.
+	for i := 0; i < len(p.notes); i++ {
+		ref := p.notes[i]
+		p.addChild(ref.footnote)
+		block := ref.footnote
+		listItem := block.(*ast.ListItem)
+		listItem.ListFlags = flags | ast.ListTypeOrdered
+		listItem.RefLink = ref.link
+		if ref.hasBlock {
+			flags |= ast.ListItemContainsBlock
+			p.block(ref.title)
+		} else {
+			p.Inline(block, ref.title)
+		}
+		flags &^= ast.ListItemBeginningOfList | ast.ListItemContainsBlock
+	}
+	above := list.Parent
+	finalizeList(list)
+	p.tip = above
+
+	ast.WalkFunc(block, func(node ast.Node, entering bool) ast.WalkStatus {
+		switch node.(type) {
+		case *ast.Paragraph, *ast.Heading:
+			p.Inline(node, node.AsContainer().Content)
+			node.AsContainer().Content = nil
+		}
+		return ast.GoToNext
+	})
+}
+
+//
+// Link references
+//
+// This section implements support for references that (usually) appear
+// as footnotes in a document, and can be referenced anywhere in the document.
+// The basic format is:
+//
+//    [1]: http://www.google.com/ "Google"
+//    [2]: http://www.github.com/ "Github"
+//
+// Anywhere in the document, the reference can be linked by referring to its
+// label, i.e., 1 and 2 in this example, as in:
+//
+//    This library is hosted on [Github][2], a git hosting site.
+//
+// Actual footnotes as specified in Pandoc and supported by some other Markdown
+// libraries such as php-markdown are also taken care of. They look like this:
+//
+//    This sentence needs a bit of further explanation.[^note]
+//
+//    [^note]: This is the explanation.
+//
+// Footnotes should be placed at the end of the document in an ordered list.
+// Inline footnotes such as:
+//
+//    Inline footnotes^[Not supported.] also exist.
+//
+// are not yet supported.
+
+// reference holds all information necessary for a reference-style links or
+// footnotes.
+//
+// Consider this markdown with reference-style links:
+//
+//     [link][ref]
+//
+//     [ref]: /url/ "tooltip title"
+//
+// It will be ultimately converted to this HTML:
+//
+//     <p><a href=\"/url/\" title=\"title\">link</a></p>
+//
+// And a reference structure will be populated as follows:
+//
+//     p.refs["ref"] = &reference{
+//         link: "/url/",
+//         title: "tooltip title",
+//     }
+//
+// Alternatively, reference can contain information about a footnote. Consider
+// this markdown:
+//
+//     Text needing a footnote.[^a]
+//
+//     [^a]: This is the note
+//
+// A reference structure will be populated as follows:
+//
+//     p.refs["a"] = &reference{
+//         link: "a",
+//         title: "This is the note",
+//         noteID: <some positive int>,
+//     }
+//
+// TODO: As you can see, it begs for splitting into two dedicated structures
+// for refs and for footnotes.
+type reference struct {
+	link     []byte
+	title    []byte
+	noteID   int // 0 if not a footnote ref
+	hasBlock bool
+	footnote ast.Node // a link to the Item node within a list of footnotes
+
+	text []byte // only gets populated by refOverride feature with Reference.Text
+}
+
+func (r *reference) String() string {
+	return fmt.Sprintf("{link: %q, title: %q, text: %q, noteID: %d, hasBlock: %v}",
+		r.link, r.title, r.text, r.noteID, r.hasBlock)
+}
+
+// Check whether or not data starts with a reference link.
+// If so, it is parsed and stored in the list of references
+// (in the render struct).
+// Returns the number of bytes to skip to move past it,
+// or zero if the first line is not a reference.
+func isReference(p *Parser, data []byte, tabSize int) int {
+	// up to 3 optional leading spaces
+	if len(data) < 4 {
+		return 0
+	}
+	i := 0
+	for i < 3 && data[i] == ' ' {
+		i++
+	}
+
+	noteID := 0
+
+	// id part: anything but a newline between brackets
+	if data[i] != '[' {
+		return 0
+	}
+	i++
+	if p.extensions&Footnotes != 0 {
+		if i < len(data) && data[i] == '^' {
+			// we can set it to anything here because the proper noteIds will
+			// be assigned later during the second pass. It just has to be != 0
+			noteID = 1
+			i++
+		}
+	}
+	idOffset := i
+	for i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != ']' {
+		i++
+	}
+	if i >= len(data) || data[i] != ']' {
+		return 0
+	}
+	idEnd := i
+	// footnotes can have empty ID, like this: [^], but a reference can not be
+	// empty like this: []. Break early if it's not a footnote and there's no ID
+	if noteID == 0 && idOffset == idEnd {
+		return 0
+	}
+	// spacer: colon (space | tab)* newline? (space | tab)*
+	i++
+	if i >= len(data) || data[i] != ':' {
+		return 0
+	}
+	i++
+	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
+		i++
+	}
+	if i < len(data) && (data[i] == '\n' || data[i] == '\r') {
+		i++
+		if i < len(data) && data[i] == '\n' && data[i-1] == '\r' {
+			i++
+		}
+	}
+	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
+		i++
+	}
+	if i >= len(data) {
+		return 0
+	}
+
+	var (
+		linkOffset, linkEnd   int
+		titleOffset, titleEnd int
+		lineEnd               int
+		raw                   []byte
+		hasBlock              bool
+	)
+
+	if p.extensions&Footnotes != 0 && noteID != 0 {
+		linkOffset, linkEnd, raw, hasBlock = scanFootnote(p, data, i, tabSize)
+		lineEnd = linkEnd
+	} else {
+		linkOffset, linkEnd, titleOffset, titleEnd, lineEnd = scanLinkRef(p, data, i)
+	}
+	if lineEnd == 0 {
+		return 0
+	}
+
+	// a valid ref has been found
+
+	ref := &reference{
+		noteID:   noteID,
+		hasBlock: hasBlock,
+	}
+
+	if noteID > 0 {
+		// reusing the link field for the id since footnotes don't have links
+		ref.link = data[idOffset:idEnd]
+		// if footnote, it's not really a title, it's the contained text
+		ref.title = raw
+	} else {
+		ref.link = data[linkOffset:linkEnd]
+		ref.title = data[titleOffset:titleEnd]
+	}
+
+	// id matches are case-insensitive
+	id := string(bytes.ToLower(data[idOffset:idEnd]))
+
+	p.refs[id] = ref
+
+	return lineEnd
+}
+
+func scanLinkRef(p *Parser, data []byte, i int) (linkOffset, linkEnd, titleOffset, titleEnd, lineEnd int) {
+	// link: whitespace-free sequence, optionally between angle brackets
+	if data[i] == '<' {
+		i++
+	}
+	linkOffset = i
+	for i < len(data) && data[i] != ' ' && data[i] != '\t' && data[i] != '\n' && data[i] != '\r' {
+		i++
+	}
+	linkEnd = i
+	if linkEnd < len(data) && data[linkOffset] == '<' && data[linkEnd-1] == '>' {
+		linkOffset++
+		linkEnd--
+	}
+
+	// optional spacer: (space | tab)* (newline | '\'' | '"' | '(' )
+	for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
+		i++
+	}
+	if i < len(data) && data[i] != '\n' && data[i] != '\r' && data[i] != '\'' && data[i] != '"' && data[i] != '(' {
+		return
+	}
+
+	// compute end-of-line
+	if i >= len(data) || data[i] == '\r' || data[i] == '\n' {
+		lineEnd = i
+	}
+	if i+1 < len(data) && data[i] == '\r' && data[i+1] == '\n' {
+		lineEnd++
+	}
+
+	// optional (space|tab)* spacer after a newline
+	if lineEnd > 0 {
+		i = lineEnd + 1
+		for i < len(data) && (data[i] == ' ' || data[i] == '\t') {
+			i++
+		}
+	}
+
+	// optional title: any non-newline sequence enclosed in '"() alone on its line
+	if i+1 < len(data) && (data[i] == '\'' || data[i] == '"' || data[i] == '(') {
+		i++
+		titleOffset = i
+
+		// look for EOL
+		for i < len(data) && data[i] != '\n' && data[i] != '\r' {
+			i++
+		}
+		if i+1 < len(data) && data[i] == '\n' && data[i+1] == '\r' {
+			titleEnd = i + 1
+		} else {
+			titleEnd = i
+		}
+
+		// step back
+		i--
+		for i > titleOffset && (data[i] == ' ' || data[i] == '\t') {
+			i--
+		}
+		if i > titleOffset && (data[i] == '\'' || data[i] == '"' || data[i] == ')') {
+			lineEnd = titleEnd
+			titleEnd = i
+		}
+	}
+
+	return
+}
+
+// The first bit of this logic is the same as Parser.listItem, but the rest
+// is much simpler. This function simply finds the entire block and shifts it
+// over by one tab if it is indeed a block (just returns the line if it's not).
+// blockEnd is the end of the section in the input buffer, and contents is the
+// extracted text that was shifted over one tab. It will need to be rendered at
+// the end of the document.
+func scanFootnote(p *Parser, data []byte, i, indentSize int) (blockStart, blockEnd int, contents []byte, hasBlock bool) {
+	if i == 0 || len(data) == 0 {
+		return
+	}
+
+	// skip leading whitespace on first line
+	for i < len(data) && data[i] == ' ' {
+		i++
+	}
+
+	blockStart = i
+
+	// find the end of the line
+	blockEnd = i
+	for i < len(data) && data[i-1] != '\n' {
+		i++
+	}
+
+	// get working buffer
+	var raw bytes.Buffer
+
+	// put the first line into the working buffer
+	raw.Write(data[blockEnd:i])
+	blockEnd = i
+
+	// process the following lines
+	containsBlankLine := false
+
+gatherLines:
+	for blockEnd < len(data) {
+		i++
+
+		// find the end of this line
+		for i < len(data) && data[i-1] != '\n' {
+			i++
+		}
+
+		// if it is an empty line, guess that it is part of this item
+		// and move on to the next line
+		if p.isEmpty(data[blockEnd:i]) > 0 {
+			containsBlankLine = true
+			blockEnd = i
+			continue
+		}
+
+		n := 0
+		if n = isIndented(data[blockEnd:i], indentSize); n == 0 {
+			// this is the end of the block.
+			// we don't want to include this last line in the index.
+			break gatherLines
+		}
+
+		// if there were blank lines before this one, insert a new one now
+		if containsBlankLine {
+			raw.WriteByte('\n')
+			containsBlankLine = false
+		}
+
+		// get rid of that first tab, write to buffer
+		raw.Write(data[blockEnd+n : i])
+		hasBlock = true
+
+		blockEnd = i
+	}
+
+	if data[blockEnd-1] != '\n' {
+		raw.WriteByte('\n')
+	}
+
+	contents = raw.Bytes()
+
+	return
+}
+
+// isPunctuation returns true if c is a punctuation symbol.
+func isPunctuation(c byte) bool {
+	for _, r := range []byte("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~") {
+		if c == r {
+			return true
+		}
+	}
+	return false
+}
+
+// isSpace returns true if c is a white-space charactr
+func isSpace(c byte) bool {
+	return c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '\f' || c == '\v'
+}
+
+// isLetter returns true if c is ascii letter
+func isLetter(c byte) bool {
+	return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')
+}
+
+// isAlnum returns true if c is a digit or letter
+// TODO: check when this is looking for ASCII alnum and when it should use unicode
+func isAlnum(c byte) bool {
+	return (c >= '0' && c <= '9') || isLetter(c)
+}
+
+// TODO: this is not used
+// Replace tab characters with spaces, aligning to the next TAB_SIZE column.
+// always ends output with a newline
+func expandTabs(out *bytes.Buffer, line []byte, tabSize int) {
+	// first, check for common cases: no tabs, or only tabs at beginning of line
+	i, prefix := 0, 0
+	slowcase := false
+	for i = 0; i < len(line); i++ {
+		if line[i] == '\t' {
+			if prefix == i {
+				prefix++
+			} else {
+				slowcase = true
+				break
+			}
+		}
+	}
+
+	// no need to decode runes if all tabs are at the beginning of the line
+	if !slowcase {
+		for i = 0; i < prefix*tabSize; i++ {
+			out.WriteByte(' ')
+		}
+		out.Write(line[prefix:])
+		return
+	}
+
+	// the slow case: we need to count runes to figure out how
+	// many spaces to insert for each tab
+	column := 0
+	i = 0
+	for i < len(line) {
+		start := i
+		for i < len(line) && line[i] != '\t' {
+			_, size := utf8.DecodeRune(line[i:])
+			i += size
+			column++
+		}
+
+		if i > start {
+			out.Write(line[start:i])
+		}
+
+		if i >= len(line) {
+			break
+		}
+
+		for {
+			out.WriteByte(' ')
+			column++
+			if column%tabSize == 0 {
+				break
+			}
+		}
+
+		i++
+	}
+}
+
+// Find if a line counts as indented or not.
+// Returns number of characters the indent is (0 = not indented).
+func isIndented(data []byte, indentSize int) int {
+	if len(data) == 0 {
+		return 0
+	}
+	if data[0] == '\t' {
+		return 1
+	}
+	if len(data) < indentSize {
+		return 0
+	}
+	for i := 0; i < indentSize; i++ {
+		if data[i] != ' ' {
+			return 0
+		}
+	}
+	return indentSize
+}
+
+// Create a url-safe slug for fragments
+func slugify(in []byte) []byte {
+	if len(in) == 0 {
+		return in
+	}
+	out := make([]byte, 0, len(in))
+	sym := false
+
+	for _, ch := range in {
+		if isAlnum(ch) {
+			sym = false
+			out = append(out, ch)
+		} else if sym {
+			continue
+		} else {
+			out = append(out, '-')
+			sym = true
+		}
+	}
+	var a, b int
+	var ch byte
+	for a, ch = range out {
+		if ch != '-' {
+			break
+		}
+	}
+	for b = len(out) - 1; b > 0; b-- {
+		if out[b] != '-' {
+			break
+		}
+	}
+	return out[a : b+1]
+}
+
+func isListItem(d ast.Node) bool {
+	_, ok := d.(*ast.ListItem)
+	return ok
+}
diff --git a/vendor/github.com/gomarkdown/markdown/parser/ref.go b/vendor/github.com/gomarkdown/markdown/parser/ref.go
new file mode 100644
index 00000000..0b59a196
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/parser/ref.go
@@ -0,0 +1,89 @@
+package parser
+
+import (
+	"bytes"
+	"fmt"
+
+	"github.com/gomarkdown/markdown/ast"
+)
+
+// parse '(#r)', where r does not contain spaces. Or.
+// (!item) (!item, subitem), for an index, (!!item) signals primary.
+func maybeShortRefOrIndex(p *Parser, data []byte, offset int) (int, ast.Node) {
+	if len(data[offset:]) < 4 {
+		return 0, nil
+	}
+	// short ref first
+	data = data[offset:]
+	i := 1
+	switch data[i] {
+	case '#': // cross ref
+		i++
+	Loop:
+		for i < len(data) {
+			c := data[i]
+			switch {
+			case c == ')':
+				break Loop
+			case !isAlnum(c):
+				if c == '_' || c == '-' || c == ':' {
+					i++
+					continue
+				}
+				i = 0
+				break Loop
+			}
+			i++
+		}
+		if i >= len(data) {
+			return 0, nil
+		}
+		if data[i] != ')' {
+			return 0, nil
+		}
+
+		id := data[2:i]
+		node := &ast.CrossReference{}
+		node.Destination = id
+
+		return i + 1, node
+
+	case '!': // index
+		i++
+		start := i
+		i = skipUntilChar(data, start, ')')
+
+		// did we reach the end of the buffer without a closing marker?
+		if i >= len(data) {
+			return 0, nil
+		}
+
+		if len(data[start:i]) < 1 {
+			return 0, nil
+		}
+
+		idx := &ast.Index{}
+
+		idx.ID = fmt.Sprintf("idxref:%d", p.indexCnt)
+		p.indexCnt++
+
+		idx.Primary = data[start] == '!'
+		buf := data[start:i]
+
+		if idx.Primary {
+			buf = buf[1:]
+		}
+		items := bytes.Split(buf, []byte(","))
+		switch len(items) {
+		case 1:
+			idx.Item = bytes.TrimSpace(items[0])
+			return i + 1, idx
+		case 2:
+			idx.Item = bytes.TrimSpace(items[0])
+			idx.Subitem = bytes.TrimSpace(items[1])
+			return i + 1, idx
+		}
+	}
+
+	return 0, nil
+}
diff --git a/vendor/github.com/gomarkdown/markdown/todo.md b/vendor/github.com/gomarkdown/markdown/todo.md
new file mode 100644
index 00000000..be8bb55c
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/todo.md
@@ -0,0 +1,7 @@
+# Things to do
+
+[ ] docs: add examples like https://godoc.org/github.com/dgrijalva/jwt-go (put in foo_example_test.go). Or see https://github.com/garyburd/redigo/blob/master/redis/zpop_example_test.go#L5 / https://godoc.org/github.com/garyburd/redigo/redis or https://godoc.org/github.com/go-redis/redis
+
+[ ] figure out expandTabs and parser.TabSizeEight. Are those used?
+
+[ ] SoftbreakData is not used
diff --git a/vendor/github.com/gomarkdown/markdown/tracking-perf.md b/vendor/github.com/gomarkdown/markdown/tracking-perf.md
new file mode 100644
index 00000000..40b95183
--- /dev/null
+++ b/vendor/github.com/gomarkdown/markdown/tracking-perf.md
@@ -0,0 +1,189 @@
+## Tracking perf changes
+
+Initial performance:
+```
+goos: darwin
+goarch: amd64
+pkg: github.com/gomarkdown/markdown
+BenchmarkEscapeHTML-8                           	 2000000	       823 ns/op	       0 B/op	       0 allocs/op
+BenchmarkSmartDoubleQuotes-8                    	  300000	      5033 ns/op	    9872 B/op	      56 allocs/op
+BenchmarkReferenceAmps-8                        	  100000	     19538 ns/op	   26776 B/op	     150 allocs/op
+BenchmarkReferenceAutoLinks-8                   	  100000	     17574 ns/op	   24544 B/op	     132 allocs/op
+BenchmarkReferenceBackslashEscapes-8            	   30000	     50977 ns/op	   76752 B/op	     243 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8   	  200000	      8546 ns/op	   12864 B/op	      65 allocs/op
+BenchmarkReferenceCodeBlocks-8                  	  200000	      9000 ns/op	   14912 B/op	      70 allocs/op
+BenchmarkReferenceCodeSpans-8                   	  200000	      8856 ns/op	   14992 B/op	      69 allocs/op
+BenchmarkReferenceHardWrappedPara-8             	  200000	      6599 ns/op	   11312 B/op	      57 allocs/op
+BenchmarkReferenceHorizontalRules-8             	  100000	     15483 ns/op	   23536 B/op	      98 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8          	  200000	      6839 ns/op	   12150 B/op	      62 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8            	  100000	     19940 ns/op	   28488 B/op	     117 allocs/op
+BenchmarkReferenceInlineHTMLComments-8          	  200000	      7455 ns/op	   13440 B/op	      64 allocs/op
+BenchmarkReferenceLinksInline-8                 	  100000	     16425 ns/op	   23664 B/op	     147 allocs/op
+BenchmarkReferenceLinksReference-8              	   30000	     54895 ns/op	   66464 B/op	     416 allocs/op
+BenchmarkReferenceLinksShortcut-8               	  100000	     17647 ns/op	   23776 B/op	     158 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8         	  200000	      9367 ns/op	   14832 B/op	      95 allocs/op
+BenchmarkReferenceMarkdownBasics-8              	   10000	    129772 ns/op	  130848 B/op	     378 allocs/op
+BenchmarkReferenceMarkdownSyntax-8              	    3000	    502365 ns/op	  461411 B/op	    1411 allocs/op
+BenchmarkReferenceNestedBlockquotes-8           	  200000	      7028 ns/op	   12688 B/op	      64 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8    	   20000	     79686 ns/op	  107520 B/op	     374 allocs/op
+BenchmarkReferenceStrongAndEm-8                 	  200000	     10020 ns/op	   17792 B/op	      78 allocs/op
+BenchmarkReferenceTabs-8                        	  200000	     12025 ns/op	   18224 B/op	      81 allocs/op
+BenchmarkReferenceTidyness-8                    	  200000	      8985 ns/op	   14432 B/op	      71 allocs/op
+PASS
+ok  	github.com/gomarkdown/markdown	45.375s
+```
+
+After switching to using interface{} for Node.Data:
+```
+BenchmarkEscapeHTML-8                           	 2000000	       929 ns/op	       0 B/op	       0 allocs/op
+BenchmarkSmartDoubleQuotes-8                    	  300000	      5126 ns/op	    9248 B/op	      56 allocs/op
+BenchmarkReferenceAmps-8                        	  100000	     19927 ns/op	   17880 B/op	     154 allocs/op
+BenchmarkReferenceAutoLinks-8                   	  100000	     20732 ns/op	   17360 B/op	     141 allocs/op
+BenchmarkReferenceBackslashEscapes-8            	   30000	     50267 ns/op	   38128 B/op	     244 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8   	  200000	      8988 ns/op	   10912 B/op	      67 allocs/op
+BenchmarkReferenceCodeBlocks-8                  	  200000	      8611 ns/op	   12256 B/op	      74 allocs/op
+BenchmarkReferenceCodeSpans-8                   	  200000	      8256 ns/op	   11248 B/op	      69 allocs/op
+BenchmarkReferenceHardWrappedPara-8             	  200000	      6739 ns/op	    9856 B/op	      57 allocs/op
+BenchmarkReferenceHorizontalRules-8             	  100000	     15503 ns/op	   15600 B/op	     104 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8          	  200000	      6874 ns/op	   10278 B/op	      62 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8            	  100000	     22271 ns/op	   18552 B/op	     121 allocs/op
+BenchmarkReferenceInlineHTMLComments-8          	  200000	      8315 ns/op	   10736 B/op	      64 allocs/op
+BenchmarkReferenceLinksInline-8                 	  100000	     16155 ns/op	   16912 B/op	     152 allocs/op
+BenchmarkReferenceLinksReference-8              	   30000	     52387 ns/op	   38192 B/op	     445 allocs/op
+BenchmarkReferenceLinksShortcut-8               	  100000	     17111 ns/op	   16592 B/op	     167 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8         	  200000	      9164 ns/op	   12048 B/op	      97 allocs/op
+BenchmarkReferenceMarkdownBasics-8              	   10000	    129262 ns/op	   87264 B/op	     416 allocs/op
+BenchmarkReferenceMarkdownSyntax-8              	    3000	    496873 ns/op	  293906 B/op	    1559 allocs/op
+BenchmarkReferenceNestedBlockquotes-8           	  200000	      6854 ns/op	   10192 B/op	      64 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8    	   20000	     79633 ns/op	   55024 B/op	     447 allocs/op
+BenchmarkReferenceStrongAndEm-8                 	  200000	      9637 ns/op	   12176 B/op	      78 allocs/op
+BenchmarkReferenceTabs-8                        	  100000	     12164 ns/op	   13776 B/op	      87 allocs/op
+BenchmarkReferenceTidyness-8                    	  200000	      8677 ns/op	   11296 B/op	      75 allocs/op
+```
+
+Not necessarily faster, but uses less bytes per op (but sometimes more allocs).
+
+After tweaking the API:
+```
+$ ./s/run-bench.sh
+
+go test -bench=. -test.benchmem
+goos: darwin
+goarch: amd64
+pkg: github.com/gomarkdown/markdown
+BenchmarkEscapeHTML-8                           	 2000000	       834 ns/op	       0 B/op	       0 allocs/op
+BenchmarkSmartDoubleQuotes-8                    	  300000	      3486 ns/op	    6160 B/op	      27 allocs/op
+BenchmarkReferenceAmps-8                        	  100000	     18158 ns/op	   14792 B/op	     125 allocs/op
+BenchmarkReferenceAutoLinks-8                   	  100000	     16824 ns/op	   14272 B/op	     112 allocs/op
+BenchmarkReferenceBackslashEscapes-8            	   30000	     44066 ns/op	   35040 B/op	     215 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8   	  200000	      6868 ns/op	    7824 B/op	      38 allocs/op
+BenchmarkReferenceCodeBlocks-8                  	  200000	      7157 ns/op	    9168 B/op	      45 allocs/op
+BenchmarkReferenceCodeSpans-8                   	  200000	      6663 ns/op	    8160 B/op	      40 allocs/op
+BenchmarkReferenceHardWrappedPara-8             	  300000	      4821 ns/op	    6768 B/op	      28 allocs/op
+BenchmarkReferenceHorizontalRules-8             	  100000	     13033 ns/op	   12512 B/op	      75 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8          	  300000	      4998 ns/op	    7190 B/op	      33 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8            	  100000	     17696 ns/op	   15464 B/op	      92 allocs/op
+BenchmarkReferenceInlineHTMLComments-8          	  300000	      5506 ns/op	    7648 B/op	      35 allocs/op
+BenchmarkReferenceLinksInline-8                 	  100000	     14450 ns/op	   13824 B/op	     123 allocs/op
+BenchmarkReferenceLinksReference-8              	   30000	     52561 ns/op	   35104 B/op	     416 allocs/op
+BenchmarkReferenceLinksShortcut-8               	  100000	     15616 ns/op	   13504 B/op	     138 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8         	  200000	      7772 ns/op	    8960 B/op	      68 allocs/op
+BenchmarkReferenceMarkdownBasics-8              	   10000	    121436 ns/op	   84176 B/op	     387 allocs/op
+BenchmarkReferenceMarkdownSyntax-8              	    3000	    487404 ns/op	  290818 B/op	    1530 allocs/op
+BenchmarkReferenceNestedBlockquotes-8           	  300000	      5098 ns/op	    7104 B/op	      35 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8    	   20000	     74422 ns/op	   51936 B/op	     418 allocs/op
+BenchmarkReferenceStrongAndEm-8                 	  200000	      7888 ns/op	    9088 B/op	      49 allocs/op
+BenchmarkReferenceTabs-8                        	  200000	     10061 ns/op	   10688 B/op	      58 allocs/op
+BenchmarkReferenceTidyness-8                    	  200000	      7152 ns/op	    8208 B/op	      46 allocs/op
+ok  	github.com/gomarkdown/markdown	40.809s
+```
+
+After refactoring Renderer:
+```
+BenchmarkEscapeHTML-8                                    2000000               883 ns/op               0 B/op          0 allocs/op
+BenchmarkSmartDoubleQuotes-8                              300000              3717 ns/op            6208 B/op         29 allocs/op
+BenchmarkReferenceAmps-8                                  100000             19135 ns/op           14680 B/op        123 allocs/op
+BenchmarkReferenceAutoLinks-8                             100000             17142 ns/op           14176 B/op        110 allocs/op
+BenchmarkReferenceBackslashEscapes-8                       30000             54616 ns/op           35088 B/op        217 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8             200000              7993 ns/op            7872 B/op         40 allocs/op
+BenchmarkReferenceCodeBlocks-8                            200000              8285 ns/op            9216 B/op         47 allocs/op
+BenchmarkReferenceCodeSpans-8                             200000              7684 ns/op            8208 B/op         42 allocs/op
+BenchmarkReferenceHardWrappedPara-8                       200000              5595 ns/op            6816 B/op         30 allocs/op
+BenchmarkReferenceHorizontalRules-8                       100000             16444 ns/op           12560 B/op         77 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8                    200000              5415 ns/op            7238 B/op         35 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8                      100000             19867 ns/op           15512 B/op         94 allocs/op
+BenchmarkReferenceInlineHTMLComments-8                    200000              6026 ns/op            7696 B/op         37 allocs/op
+BenchmarkReferenceLinksInline-8                           100000             14864 ns/op           13664 B/op        120 allocs/op
+BenchmarkReferenceLinksReference-8                         30000             52479 ns/op           34816 B/op        401 allocs/op
+BenchmarkReferenceLinksShortcut-8                         100000             15812 ns/op           13472 B/op        135 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8                   200000              7767 ns/op            8880 B/op         68 allocs/op
+BenchmarkReferenceMarkdownBasics-8                         10000            131065 ns/op           84048 B/op        386 allocs/op
+BenchmarkReferenceMarkdownSyntax-8                          2000            515604 ns/op          289953 B/op       1501 allocs/op
+BenchmarkReferenceNestedBlockquotes-8                     200000              5655 ns/op            7152 B/op         37 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8               20000             84188 ns/op           51984 B/op        420 allocs/op
+BenchmarkReferenceStrongAndEm-8                           200000              8664 ns/op            9136 B/op         51 allocs/op
+BenchmarkReferenceTabs-8                                  100000             11110 ns/op           10736 B/op         60 allocs/op
+BenchmarkReferenceTidyness-8                              200000              7628 ns/op            8256 B/op         48 allocs/op
+ok      github.com/gomarkdown/markdown  40.841s
+```
+
+After Node refactor to have Children array:
+```
+BenchmarkEscapeHTML-8                                    2000000               901 ns/op               0 B/op          0 allocs/op
+BenchmarkSmartDoubleQuotes-8                              300000              3905 ns/op            6224 B/op         31 allocs/op
+BenchmarkReferenceAmps-8                                  100000             22216 ns/op           15560 B/op        157 allocs/op
+BenchmarkReferenceAutoLinks-8                             100000             20335 ns/op           14824 B/op        146 allocs/op
+BenchmarkReferenceBackslashEscapes-8                       20000             69174 ns/op           37392 B/op        316 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8             200000              8443 ns/op            7968 B/op         48 allocs/op
+BenchmarkReferenceCodeBlocks-8                            200000              9250 ns/op            9392 B/op         58 allocs/op
+BenchmarkReferenceCodeSpans-8                             200000              8515 ns/op            8432 B/op         54 allocs/op
+BenchmarkReferenceHardWrappedPara-8                       200000              5738 ns/op            6856 B/op         34 allocs/op
+BenchmarkReferenceHorizontalRules-8                       100000             20864 ns/op           13648 B/op         93 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8                    200000              6187 ns/op            7310 B/op         40 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8                       50000             23793 ns/op           16128 B/op        114 allocs/op
+BenchmarkReferenceInlineHTMLComments-8                    200000              7060 ns/op            7840 B/op         44 allocs/op
+BenchmarkReferenceLinksInline-8                           100000             18432 ns/op           14496 B/op        153 allocs/op
+BenchmarkReferenceLinksReference-8                         20000             67666 ns/op           37136 B/op        502 allocs/op
+BenchmarkReferenceLinksShortcut-8                         100000             19324 ns/op           13984 B/op        162 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8                   200000              8998 ns/op            9320 B/op         83 allocs/op
+BenchmarkReferenceMarkdownBasics-8                         10000            160908 ns/op           88152 B/op        518 allocs/op
+BenchmarkReferenceMarkdownSyntax-8                          2000            707160 ns/op          303801 B/op       2044 allocs/op
+BenchmarkReferenceNestedBlockquotes-8                     200000              6740 ns/op            7248 B/op         45 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8               10000            115808 ns/op           55052 B/op        626 allocs/op
+BenchmarkReferenceStrongAndEm-8                           100000             10540 ns/op            9416 B/op         72 allocs/op
+BenchmarkReferenceTabs-8                                  100000             13171 ns/op           10968 B/op         77 allocs/op
+BenchmarkReferenceTidyness-8                              200000              8903 ns/op            8404 B/op         62 allocs/op
+PASS
+ok      github.com/gomarkdown/markdown  43.477s
+```
+It's slower (but opens up possibilities for further improvements).
+
+After refactoring to make ast.Node a top-level thing.
+```
+BenchmarkEscapeHTML-8                                    2000000               829 ns/op               0 B/op          0 allocs/op
+BenchmarkSmartDoubleQuotes-8                              300000              3998 ns/op            6192 B/op         31 allocs/op
+BenchmarkReferenceAmps-8                                   50000             27389 ns/op           15480 B/op        153 allocs/op
+BenchmarkReferenceAutoLinks-8                              50000             23106 ns/op           14656 B/op        137 allocs/op
+BenchmarkReferenceBackslashEscapes-8                       10000            112435 ns/op           36696 B/op        315 allocs/op
+BenchmarkReferenceBlockquotesWithCodeBlocks-8             200000              9227 ns/op            7856 B/op         46 allocs/op
+BenchmarkReferenceCodeBlocks-8                            200000             10469 ns/op            9248 B/op         54 allocs/op
+BenchmarkReferenceCodeSpans-8                             200000             10522 ns/op            8368 B/op         54 allocs/op
+BenchmarkReferenceHardWrappedPara-8                       200000              6354 ns/op            6784 B/op         34 allocs/op
+BenchmarkReferenceHorizontalRules-8                        50000             32393 ns/op           13952 B/op         87 allocs/op
+BenchmarkReferenceInlineHTMLAdvances-8                    200000              6894 ns/op            7238 B/op         40 allocs/op
+BenchmarkReferenceInlineHTMLSimple-8                       50000             32942 ns/op           15864 B/op        110 allocs/op
+BenchmarkReferenceInlineHTMLComments-8                    200000              8181 ns/op            7776 B/op         44 allocs/op
+BenchmarkReferenceLinksInline-8                           100000             21679 ns/op           14400 B/op        148 allocs/op
+BenchmarkReferenceLinksReference-8                         20000             83928 ns/op           36688 B/op        473 allocs/op
+BenchmarkReferenceLinksShortcut-8                         100000             22053 ns/op           13872 B/op        153 allocs/op
+BenchmarkReferenceLiterQuotesInTitles-8                   100000             10784 ns/op            9296 B/op         81 allocs/op
+BenchmarkReferenceMarkdownBasics-8                          5000            237097 ns/op           87760 B/op        480 allocs/op
+BenchmarkReferenceMarkdownSyntax-8                          1000           1465402 ns/op          300769 B/op       1896 allocs/op
+BenchmarkReferenceNestedBlockquotes-8                     200000              7461 ns/op            7152 B/op         45 allocs/op
+BenchmarkReferenceOrderedAndUnorderedLists-8                5000            212256 ns/op           53724 B/op        553 allocs/op
+BenchmarkReferenceStrongAndEm-8                           100000             13018 ns/op            9264 B/op         72 allocs/op
+BenchmarkReferenceTabs-8                                  100000             15005 ns/op           10752 B/op         71 allocs/op
+BenchmarkReferenceTidyness-8                              200000             10308 ns/op            8292 B/op         58 allocs/op
+PASS
+ok      github.com/gomarkdown/markdown  42.176s
+```