fastmd

High-performance markdown parser for Node.js and the browser (Rust/WASM core).

Build

pnpm install
pnpm build:wasm   # build WASM (web + Node)
pnpm build       # compile TS and copy WASM into packages/fastmd/dist

Test

pnpm test        # unit tests; integration tests run when WASM is built

Benchmark

fastmd only:

pnpm build:wasm:node
pnpm bench

Compare vs react-markdown and Bun’s native markdown:

1. Build WASM: pnpm build:wasm:node
2. Optional: install react-markdown for comparison:

   pnpm add -D react-markdown react react-dom

3. Run comparison (Node):

   pnpm bench:compare

4. Or run under Bun (uses Bun’s runtime; Bun’s native markdown is included if your Bun version exposes it):

   bun run crates/fastmd-core/bench/compare.bench.ts

Same fixture and iteration count for all parsers; output is ops/sec so you can compare.

Usage

Node.js (sync, no init):

import { parseToTokens, renderHTML } from "@abhi1264/fastmd"
const tokens = parseToTokens("# Hello\n")
console.log(renderHTML(tokens, "# Hello\n"))  // <h1>Hello</h1>

Browser (async init once):

import initfastmd, { parseToTokens } from "@abhi1264/fastmd"
await initfastmd()
const tokens = parseToTokens("# Hello\n")

See docs/INIT.md and docs/STREAMING.md.

Supported syntax

| Feature | Supported | |--------|-----------| | Headings (#–######) | ✅ | | Bold (** / __) | ✅ | | Italic (* / _) | ✅ | | ~~Strikethrough~~ (~~) | ✅ | | Code inline | ✅ | | Fenced code blocks (```) with optional language | ✅ | | Links | ✅ | | Autolinks <https://…> | ✅ | | Blockquotes (>) | ✅ | | Unordered lists (- / * / +) | ✅ | | Ordered lists (1. …) | ✅ | | Horizontal rule (--- / ***) | ✅ |

Not supported: raw HTML, tables, task lists, definition lists, footnotes, front matter. Use a full CommonMark/GFM parser if you need those.

Render options

renderHTML(tokens, src, options) accepts:

• linkTarget – e.g. "_blank" for links
• headingIdPrefix – prefix for heading id (slug from text)
• codeClassPrefix – prefix for code block class (e.g. "language-" for syntax highlighting)

React components (optional)

For custom React components per element (like react-markdown):

import { parseToTokens, renderToReact } from "@abhi1264/fastmd/react"
const tokens = parseToTokens("# Hi\n")
const nodes = renderToReact(tokens, "# Hi\n", {
  a: (props) => <a {...props} target="_blank" />,
})

Requires react as peer. Import from fastmd/react.

Token format

Parse returns a flat Uint32Array: 4 u32 per token [kind, start, end, aux]. Use tokensFromBuffer(buf) to get Token[], or parseToTokens(md) for the full string → tokens pipeline. start/end are UTF-16 code unit offsets for correct src.slice(start, end) in JS.