silk-lang

TypeScript utilities for Silk, published on npm as silk-lang. Each tool is a subpath export. So far that is syntax highlighting for web highlighters: Shiki and TanStack Highlight.

bun add silk-lang

Shiki

Shiki reads TextMate grammars, so this package ships the grammar from silk-lang/silk-vscode unchanged. That repository owns it; grammar/silk.tmLanguage.json is a copy that CI checks against silk-vscode's main.

import { createHighlighter } from 'shiki'
import silk from 'silk-lang/shiki'

const highlighter = await createHighlighter({ langs: [silk], themes: ['github-dark'] })
const html = highlighter.codeToHtml(source, { lang: 'silk', theme: 'github-dark' })

The raw grammar is also exported as silk-lang/silk.tmLanguage.json for other TextMate consumers.

TanStack Highlight

TanStack Highlight does not read TextMate or tree-sitter grammars. A language is a tokenizer that returns source ranges, so this package includes a hand-written one that follows the native lexer.

import { createHighlighter } from '@tanstack/highlight/core'
import silk from 'silk-lang/tanstack'

const highlighter = createHighlighter({ languages: [silk] })
const html = highlighter.highlightToHtml(source, { lang: 'silk' })

To keep the two from drifting, CI highlights every .silk file in silk-lang/silk with both and fails if any character gets a different category (comment, string, number, literal, keyword, operator, type or function). Member names after . are a TanStack-only property class.

Develop

bun install
bun run test                # tokenizer tests and TextMate agreement on test/fixtures
bun run compare-silk ../silk  # agreement on a whole silk-lang/silk checkout
bun run sync-grammar        # copy the grammar from silk-vscode main
bun run build               # dist/

When the grammar changes in silk-vscode, run bun run sync-grammar here and update src/tanstack.ts until compare-silk reports no differences.

Release

Releases use Changesets. With a change that should reach npm, run bun changeset, pick the bump and describe it for the changelog, and commit the generated file in .changeset/.

On main, the Release workflow keeps a "Version packages" pull request open that applies the pending changesets to package.json and CHANGELOG.md. Merging it publishes to npm through trusted publishing, so no npm token is needed.