@discord/markdown-wasm

Prerequisites

• wasm-pack

Build

pnpm build

Usage

import { parse } from '@discord/markdown-wasm';

const ast = parse('Hello world!');

parse takes optional flags second, picking the syntax to parse as an OR of Syntax bits, which the package also exports. Every syntax parses by default, and with 0:

import { Syntax, parse } from '@discord/markdown-wasm';

const noMentions = parse('@everyone **hi**', Syntax.ALL & ~Syntax.MENTIONS);

Initialising

The package has no top-level await, so importing it never turns the module graph async — it can be require()d, and it bundles for targets that do not support top-level await.

The cost is that the wasm has to be instantiated before the first parse. Node does that itself on first use, reading and compiling the file synchronously. Nowhere else can: the bytes only arrive over a fetch, and browsers refuse to compile buffers over 4KB synchronously on the main thread anyway. So everywhere else, init once at startup:

import { init, parse } from '@discord/markdown-wasm';

await init();

const ast = parse('Hello world!'); // and everything downstream stays sync

If your bundler or host already hands you the wasm, skip the fetch with initSync. Pass a WebAssembly.Module, not bytes — compiling bytes is what the 4KB cap applies to:

import { initSync, parse } from '@discord/markdown-wasm';

initSync(wasmModule);

Parsing before either call throws.

The lossless tree

parseLossless returns a tree for rich text instead: nodes for what the markdown means, and tokens holding every character of the content in order, down to delimiters, escapes, markers and line endings. Joining the tokens' text gives the content back exactly. Every node and token has a span, in UTF-16 code units, so it indexes the string directly:

import { parseLossless } from '@discord/markdown-wasm';

const content = '**hi** 🦶';
const tree = parseLossless(content);
// tree.children[0] is the paragraph; its first child, the bold, spans `**hi**`
const { start, end } = tree.children[0].children[0].span;
content.slice(start, end); // '**hi**'

It takes flags as parse does. The types are LosslessNode and friends in @discord/markdown-types.