Micromd -- a ~3kB Markdown parser

See it parsing this document here: https://md5crypt.github.io/micromd

• minimal footprint
• 100% linear
• single pass
• AST generation with separate rendering backends
  • DOM backend available
  • string html backend available
  • super easy to implement a custom one!
• ok-ish compatibility (see below)

Rational

This parser offers a fine balance between footprint size and common mark compatibility. It is by no means perfect but it is tiny (~3kB minified) and fast.

Syntax support

• emphasis & inline code

  • bold (**, __)
  • italics (*, _)
  • strikethrough (~~)
  • bold italic (***)
  • code spans (`, ``, ```)
    • escaping in code spans does not work
    • inline code blocks don't stop on line breaks
  • simplified fencing rules
    • utf8 punctuation and whitespace characters like em-dash are not handled correctly

• links and images

  • title and ref not supported
  • href/src cannot contain ( or )
  • alt cannot contain [ or ]
  • escapes are not processed in alt, href, and src

• autolinks

  • supported only with <...> and only for http / https
  • bare URLs are not auto-linked

• hard line breaks

  • more then two whitespaces at end of line are treated as hard break

• headers

  • headers using the --- syntax (setext) are not supported

• horizontal lines

  • only *** and --- is supported, must start at column 0 (no indentation)

• lists

  • lists can't contain other block elements (header, blockquote, code fence)
  • slightly different indentation rule then commonmark, should never be visible in normal use
  • changing list type mid list has no effect
  • setting numeration start works only for first list element

• blockquotes

  • blockquotes can't contain other block elements (header, list, code fence)
  • blockquotes can not be nested

• code fences

  • only triple backticks ``` are supported
  • closing fence must be the only thing on its line
  • a newline is required before the closing fence
  • escaping inside fenced code is not supported

Usage

HTML generation

HTML strings can be generated with this function:

/**
 * Generate a html string output
 * 
 * @param input - the input string
 * @param options - parser options, see {@link MdParseOptions}
 *
 * @returns the complete html string
 * - if {@link MdParseOptions.stream} is set returns "" (empty string)
 */
export function mdRenderHtml(input: string, options: MdRenderHtmlOptions = {})

It accepts the following options:

export interface MdRenderHtmlOptions {
	/**
	 * when true parser will ignore all block elements (lists, quotes, headers, horizontal lines)
	 * and not wrap output in a paragraph
	 */
	inline?: boolean

	/**
	 * this function will be called after each root-level element finishes parsing
	 * useful to keep memory footprint small by consuming elements as the come
	 * 
	 * mdParse always returns undefined when set
	 * 
	 * @note ignored if inline is set
	 */
	stream?: (node: string) => void

	/**
	 * called for each code block to apply syntax coloring
	 *
	 * expects valid html code as response to put inside the
	 * <pre><code></code></pre> block
	 */
	codeStyler?: (code: string, lang?: string) => string
}

DOM generation

If used to display content on screen, why generate html, AST can be used to generate DOM elements directly:

/**
 * Generate DOM elements directly
 * 
 * @param input - the input string
 * @param options - parser options, see {@link MdRenderDomOptions}
 *
 * @returns a DocumentFragment
 */
export function mdRenderDom(input: string, options: MdRenderDomOptions = {})

It accepts the following options:

export interface MdRenderDomOptions {
	/**
	 * when true parser will ignore all block elements (lists, quotes, headers, horizontal lines)
	 * and not wrap output in a paragraph
	 */
	inline?: boolean

	/**
	 * called for each code block to apply syntax coloring
	 *
	 * expects a DocumentFragment or single Element to put into
	 * <pre><code></code></pre> block
	 */
	codeStyler?: (code: string, lang?: string) => Node
}

AST generation

The main parser function looks as follows

/**
 * Run the parser and build the AST
 * 
 * @param input - the input string
 * @param options - parser options, see {@link MdParseOptions}
 *
 * @returns
 * -  by default the first block-level node of the completed AST tree
 * - if {@link MdParseOptions.stream} is set returns undefined
 * - if {@link MdParseOptions.inline} is set returns first line-level node
 */
export function mdParse(input: string, options: MdParseOptions = {})

With options being:

export interface MdParseOptions {
    /**
     * when true parser will ignore all block elements (lists, quotes, headers, horizontal lines)
     * and not wrap output in a paragraph
     */
    inline?: boolean

    /**
     * this function will be called after each root-level element finishes parsing
     * useful to keep memory footprint small by consuming elements as the come
     * 
     * mdParse always returns undefined when set
     * 
     * @note ignored if inline is set
     */
    stream?: (node: MdNode) => void
}

A single AST node looks like this:

export interface MdBaseNode {
	/** node type enum */
	type: MdNodeType

	/** next element in the node list */
	next?: MdNode

	/** first child node */
	tree?: MdNode
}

Each special node has it's own interface, for example:

export interface MdPreNode extends MdBaseNode {
	type: MdNodeType.CODE_BLOCK
	value: string
	lang?: string
}

Adding a custom AST renderer

See src/domRenderer.ts and src/htmlRenderer.ts for examples

Usage example

import { mdRenderHtml } from "micromd"
import { readFileSync } from "fs"
mdRenderHtml(readFileSync("README.md", "utf8"), {stream: console.log})

Playground

Text Formatting

bold and bold
italic and italic
bold italic
this thing
code code code
mixing stuff bold italic

escaping *things* `like this` and a slash \

line line line line line line line line
line break

next paragraph

Links

auto link: http://google.com/__strong__
normal link strong

Images

image:
image link:

Headers

#hash in header

italic in header

link in header

Block quotes

block quote!
hey! look at me

Lists

• unorderd list

• minus also works
  • nesting
    • woah!
  • ok, going back
    • not again!
      • help!
• ok I'm back

1. ordered list
   1. my hands are typing
      • an unordered list!
   2. and back
2. the end

Code blocks

without language specification (class language-none):

**bold** and __bold__  
*italic* and _italic_  
***bold italic***  
~~this thing~~  
`code` ``code`` ```code```  
*mixing `stuff` __bold__ italic*

with language specification (class language-lisp):

((((((((((((((((((((((((((((((((((((((((((([...and 1000 more]
)))))))))))))))))))))))))))))))))))

Lines

---

---

---

---