@lezer/markdown

What is @lezer/markdown?

@lezer/markdown is a parser for Markdown that is part of the Lezer parser system. It is designed to be fast, modular, and customizable, making it suitable for use in editors and other applications that need to process Markdown text.

What are @lezer/markdown's main functionalities?

Parsing Markdown

This feature allows you to parse a string of Markdown text into a syntax tree. The code sample demonstrates how to parse a simple Markdown string and output the resulting syntax tree.

const {parser} = require('@lezer/markdown');
const input = '# Hello World\nThis is a paragraph.';
const tree = parser.parse(input);
console.log(tree.toString());

Customizing the Parser

This feature allows you to customize the Markdown parser by defining custom nodes and inline styles. The code sample shows how to create a custom parser and parse a string with custom nodes.

const {parser} = require('@lezer/markdown');
const {MarkdownParser} = require('@lezer/markdown');
const customParser = new MarkdownParser({
  defineNodes: [{name: 'CustomNode', style: 'custom-style'}],
  parseInline: [{name: 'CustomInline', style: 'custom-inline-style'}]
});
const input = 'This is a custom node.';
const tree = customParser.parse(input);
console.log(tree.toString());

Syntax Highlighting

This feature allows you to apply syntax highlighting to the parsed Markdown text. The code sample demonstrates how to highlight headings and paragraphs in the parsed syntax tree.

const {parser} = require('@lezer/markdown');
const {highlightTree} = require('@lezer/highlight');
const {styleTags, tags} = require('@lezer/highlight');
const input = '# Hello World\nThis is a paragraph.';
const tree = parser.parse(input);
const highlight = highlightTree(tree, styleTags({Heading: tags.heading, Paragraph: tags.content}));
console.log(highlight);

Other packages similar to @lezer/markdown

markdown-it

markdown-it is a fast and flexible Markdown parser that supports plugins and custom rules. It is similar to @lezer/markdown in terms of speed and flexibility but offers a more extensive plugin ecosystem.

remark

remark is a Markdown processor powered by plugins. It is highly customizable and can be used to parse, transform, and compile Markdown. Compared to @lezer/markdown, remark offers a broader range of plugins and transformations.

marked

marked is a low-level Markdown compiler that allows for fast parsing and rendering of Markdown. It is similar to @lezer/markdown in terms of performance but focuses more on rendering HTML from Markdown.

README

@lezer/markdown

This is an incremental Markdown (CommonMark with support for extension) parser that integrates well with the Lezer parser system. It does not in fact use the Lezer runtime (that runs LR parsers, and Markdown can't really be parsed that way), but it produces Lezer-style compact syntax trees and consumes fragments of such trees for its incremental parsing.

Note that this only parses the document, producing a data structure that represents its syntactic form, and doesn't help with outputting HTML. Also, in order to be single-pass and incremental, it doesn't do some things that a conforming CommonMark parser is expected to do—specifically, it doesn't validate link references, so it'll parse [a][b] and similar as a link, even if no [b] reference is declared.

The @codemirror/lang-markdown package integrates this parser with CodeMirror to provide Markdown editor support.

The code is licensed under an MIT license.

Interface

parser: MarkdownParser

The default CommonMark parser.

class MarkdownParser extends Parser

A Markdown parser configuration.

nodeSet: NodeSet

The parser's syntax node types.

configure(spec: MarkdownExtension) → MarkdownParser

Reconfigure the parser.

parseInline(text: string, offset: number) → Element[]

Parse the given piece of inline text at the given offset, returning an array of Element objects representing the inline content.

interface MarkdownConfig

Objects of this type are used to configure the Markdown parser.

props⁠?: readonly NodePropSource[]

Node props to add to the parser's node set.

defineNodes⁠?: readonly (string | NodeSpec)[]

Define new node types for use in parser extensions.

parseBlock⁠?: readonly BlockParser[]

Define additional block parsing logic.

parseInline⁠?: readonly InlineParser[]

Define new inline parsing logic.

remove⁠?: readonly string[]

Remove the named parsers from the configuration.

wrap⁠?: ParseWrapper

Add a parse wrapper (such as a mixed-language parser) to this parser.

type MarkdownExtension = MarkdownConfig | readonly MarkdownExtension[]

To make it possible to group extensions together into bigger extensions (such as the Github-flavored Markdown extension), reconfiguration accepts nested arrays of config objects.

parseCode(config: Object) → MarkdownExtension

Create a Markdown extension to enable nested parsing on code blocks and/or embedded HTML.

config

codeParser⁠?: fn(info: string) → Parser | null

When provided, this will be used to parse the content of code blocks. info is the string after the opening ``` marker, or the empty string if there is no such info or this is an indented code block. If there is a parser available for the code, it should return a function that can construct the parse.

htmlParser⁠?: Parser

The parser used to parse HTML tags (both block and inline).

GitHub Flavored Markdown

GFM: MarkdownConfig[]

Extension bundle containing Table, TaskList, Strikethrough, and Autolink.

Table: MarkdownConfig

This extension provides GFM-style tables, using syntax like this:

| head 1 | head 2 |
| ---    | ---    |
| cell 1 | cell 2 |

TaskList: MarkdownConfig

Extension providing GFM-style task list items, where list items can be prefixed with [ ] or [x] to add a checkbox.

Strikethrough: MarkdownConfig

An extension that implements GFM-style Strikethrough syntax using ~~ delimiters.

Autolink: MarkdownConfig

Extension that implements autolinking for www./http:///https:///mailto:/xmpp: URLs and email addresses.

Other extensions

Subscript: MarkdownConfig

Extension providing Pandoc-style subscript using ~ markers.

Superscript: MarkdownConfig

Extension providing Pandoc-style superscript using ^ markers.

Emoji: MarkdownConfig

Extension that parses two colons with only letters, underscores, and numbers between them as Emoji nodes.

Extension

The parser can, to a certain extent, be extended to handle additional syntax.

interface NodeSpec

Used in the configuration to define new syntax node types.

name: string

The node's name.

block⁠?: boolean

Should be set to true if this type represents a block node.

composite⁠?: fn(cx: BlockContext, line: Line, value: number) → boolean

If this is a composite block, this should hold a function that, at the start of a new line where that block is active, checks whether the composite block should continue (return value) and optionally adjusts the line's base position and registers nodes for any markers involved in the block's syntax.

style⁠?: Tag | readonly Tag[] | Object<Tag | readonly Tag[]>

Add highlighting tag information for this node. The value of this property may either by a tag or array of tags to assign directly to this node, or an object in the style of styleTags's argument to assign more complicated rules.

class BlockContext implements PartialParse

Block-level parsing functions get access to this context object.

lineStart: number

The start of the current line.

parser: MarkdownParser

The parser configuration used.

depth: number

The number of parent blocks surrounding the current block.

parentType(depth⁠?: number = this.depth - 1) → NodeType

Get the type of the parent block at the given depth. When no depth is passed, return the type of the innermost parent.

nextLine() → boolean

Move to the next input line. This should only be called by (non-composite) block parsers that consume the line directly, or leaf block parser nextLine methods when they consume the current line (and return true).

prevLineEnd() → number

The end position of the previous line.

startComposite(type: string, start: number, value⁠?: number = 0)

Start a composite block. Should only be called from block parser functions that return null.

addElement(elt: Element)

Add a block element. Can be called by block parsers.

addLeafElement(leaf: LeafBlock, elt: Element)

Add a block element from a leaf parser. This makes sure any extra composite block markup (such as blockquote markers) inside the block are also added to the syntax tree.

elt(type: string, from: number, to: number, children⁠?: readonly Element[]) → Element

elt(tree: Tree, at: number) → Element

Create an Element object to represent some syntax node.

interface BlockParser

Block parsers handle block-level structure. There are three general types of block parsers:

• Composite block parsers, which handle things like lists and blockquotes. These define a parse method that starts a composite block and returns null when it recognizes its syntax.

• Eager leaf block parsers, used for things like code or HTML blocks. These can unambiguously recognize their content from its first line. They define a parse method that, if it recognizes the construct, moves the current line forward to the line beyond the end of the block, add a syntax node for the block, and return true.

• Leaf block parsers that observe a paragraph-like construct as it comes in, and optionally decide to handle it at some point. This is used for "setext" (underlined) headings and link references. These define a leaf method that checks the first line of the block and returns a LeafBlockParser object if it wants to observe that block.

name: string

The name of the parser. Can be used by other block parsers to specify precedence.

parse⁠?: fn(cx: BlockContext, line: Line) → boolean | null

The eager parse function, which can look at the block's first line and return false to do nothing, true if it has parsed (and moved past a block), or null if it has started a composite block.

leaf⁠?: fn(cx: BlockContext, leaf: LeafBlock) → LeafBlockParser | null

A leaf parse function. If no regular parse functions match for a given line, its content will be accumulated for a paragraph-style block. This method can return an object that overrides that style of parsing in some situations.

endLeaf⁠?: fn(cx: BlockContext, line: Line, leaf: LeafBlock) → boolean

Some constructs, such as code blocks or newly started blockquotes, can interrupt paragraphs even without a blank line. If your construct can do this, provide a predicate here that recognizes lines that should end a paragraph (or other non-eager leaf block).

before⁠?: string

When given, this parser will be installed directly before the block parser with the given name. The default configuration defines block parsers with names LinkReference, IndentedCode, FencedCode, Blockquote, HorizontalRule, BulletList, OrderedList, ATXHeading, HTMLBlock, and SetextHeading.

after⁠?: string

When given, the parser will be installed directly after the parser with the given name.

interface LeafBlockParser

Objects that are used to override paragraph-style blocks should conform to this interface.

nextLine(cx: BlockContext, line: Line, leaf: LeafBlock) → boolean

Update the parser's state for the next line, and optionally finish the block. This is not called for the first line (the object is constructed at that line), but for any further lines. When it returns true, the block is finished. It is okay for the function to consume the current line or any subsequent lines when returning true.

finish(cx: BlockContext, leaf: LeafBlock) → boolean

Called when the block is finished by external circumstances (such as a blank line or the start of another construct). If this parser can handle the block up to its current position, it should finish the block and return true.

class Line

Data structure used during block-level per-line parsing.

text: string

The line's full text.

baseIndent: number

The base indent provided by the composite contexts (that have been handled so far).

basePos: number

The string position corresponding to the base indent.

pos: number

The position of the next non-whitespace character beyond any list, blockquote, or other composite block markers.

indent: number

The column of the next non-whitespace character.

next: number

The character code of the character after pos.

skipSpace(from: number) → number

Skip whitespace after the given position, return the position of the next non-space character or the end of the line if there's only space after from.

moveBase(to: number)

Move the line's base position forward to the given position. This should only be called by composite block parsers or markup skipping functions.

moveBaseColumn(indent: number)

Move the line's base position forward to the given column.

addMarker(elt: Element)

Store a composite-block-level marker. Should be called from markup skipping functions when they consume any non-whitespace characters.

countIndent(to: number, from⁠?: number = 0, indent⁠?: number = 0) → number

Find the column position at to, optionally starting at a given position and column.

findColumn(goal: number) → number

Find the position corresponding to the given column.

class LeafBlock

Data structure used to accumulate a block's content during leaf block parsing.

parsers: LeafBlockParser[]

The block parsers active for this block.

start: number

The start position of the block.

content: string

The block's text content.

class InlineContext

Inline parsing functions get access to this context, and use it to read the content and emit syntax nodes.

parser: MarkdownParser

The parser that is being used.

text: string

The text of this inline section.

offset: number

The starting offset of the section in the document.

char(pos: number) → number

Get the character code at the given (document-relative) position.

end: number

The position of the end of this inline section.

slice(from: number, to: number) → string

Get a substring of this inline section. Again uses document-relative positions.

addDelimiter(type: DelimiterType, from: number, to: number, open: boolean, close: boolean) → number

Add a delimiter at this given position. open and close indicate whether this delimiter is opening, closing, or both. Returns the end of the delimiter, for convenient returning from parse functions.

hasOpenLink: boolean

Returns true when there is an unmatched link or image opening token before the current position.

addElement(elt: Element) → number

Add an inline element. Returns the end of the element.

findOpeningDelimiter(type: DelimiterType) → number | null

Find an opening delimiter of the given type. Returns null if no delimiter is found, or an index that can be passed to takeContent otherwise.

takeContent(startIndex: number) → Element[]

Remove all inline elements and delimiters starting from the given index (which you should get from findOpeningDelimiter, resolve delimiters inside of them, and return them as an array of elements.

skipSpace(from: number) → number

Skip space after the given (document) position, returning either the position of the next non-space character or the end of the section.

elt(type: string, from: number, to: number, children⁠?: readonly Element[]) → Element

elt(tree: Tree, at: number) → Element

Create an Element for a syntax node.

interface InlineParser

Inline parsers are called for every character of parts of the document that are parsed as inline content.

name: string

This parser's name, which can be used by other parsers to indicate a relative precedence.

parse(cx: InlineContext, next: number, pos: number) → number

The parse function. Gets the next character and its position as arguments. Should return -1 if it doesn't handle the character, or add some element or delimiter and return the end position of the content it parsed if it can.

before⁠?: string

When given, this parser will be installed directly before the parser with the given name. The default configuration defines inline parsers with names Escape, Entity, InlineCode, HTMLTag, Emphasis, HardBreak, Link, and Image. When no before or after property is given, the parser is added to the end of the list.

after⁠?: string

When given, the parser will be installed directly after the parser with the given name.

interface DelimiterType

Delimiters are used during inline parsing to store the positions of things that might be delimiters, if another matching delimiter is found. They are identified by objects with these properties.

resolve⁠?: string

If this is given, the delimiter should be matched automatically when a piece of inline content is finished. Such delimiters will be matched with delimiters of the same type according to their open and close properties. When a match is found, the content between the delimiters is wrapped in a node whose name is given by the value of this property.

When this isn't given, you need to match the delimiter eagerly using the findOpeningDelimiter and takeContent methods.

mark⁠?: string

If the delimiter itself should, when matched, create a syntax node, set this to the name of the syntax node.

class Element

Elements are used to compose syntax nodes during parsing.

type: number

The node's id.

from: number

The start of the node, as an offset from the start of the document.

to: number

The end of the node.