@diglot/code

Code-block comment extraction for diglot. Finds the prose comments of a code block with a per-language tokenizer, turns each comment line into a code-comment segment, and replaces only the comment text when applying a translation.

pnpm add @diglot/code @diglot/core

Usage

import { applyCode, extractCode } from '@diglot/code'

const { segments } = extractCode(code, 'ts')
const { code: translated, report } = applyCode(code, 'ts', translations)

For adapters:

• findComments(code, language) returns { line, start, end, text } for each prose comment line, where [start, end) are offsets of the text in the code
• commentDraft(comment, language, context) converts a comment to a segment draft; inline code and URLs become placeholders
• replaceComments(code, replacements) substitutes comment texts

Supported languages: JavaScript and TypeScript (js, ts, jsx, tsx, mjs, cjs), Vue and HTML (including <script> and <style> blocks), CSS, SCSS and Less, shell (bash, sh, zsh), dotenv and ini, YAML, TOML, JSON5 and JSONC.

Not extracted: twoslash directives (// ^?, // @errors, // ---cut---), Shiki notations (// [!code ++]), tool pragmas (eslint-disable, @ts-expect-error, #region), commented-out code, and comments with fewer than two words.