chinese-characters-decomposition

Typed JSON dataset of Chinese character decompositions (CCD), with a parser for the source TSV.

The data comes from the Wikimedia Commons CCD table (21,169 entries), which lists each character together with its stroke count, structural composition type, sub-components, Cangjie signature, and Kangxi radical.

Install

npm install chinese-characters-decomposition

The package ships an ESM entry point (index.js) with TypeScript declarations (index.d.ts), so it works in any modern Node.js, TypeScript, or bundled environment. If you only need the raw data, import chinese-characters-decomposition/ccd.json to skip the JS wrapper.

Usage

import ccd, { type CcdEntry } from "chinese-characters-decomposition";

const entry: CcdEntry | undefined = ccd.find((e) => e.component === "好");
// {
//   component: "好",
//   strokes: 6,
//   compositionType: "吅",
//   leftComponent: "女",
//   leftStrokes: 3,
//   rightComponent: "子",
//   rightStrokes: 3,
//   signature: "VND",
//   notes: "/",
//   section: "女",
// }

Or import the raw JSON directly:

import dataset from "chinese-characters-decomposition/ccd.json";

Entry shape

| Field | Type | Description | | ----------------- | ---------------- | -------------------------------------------------------------- | | component | string | The character being decomposed. | | strokes | number | Total stroke count. | | compositionType | string | Structure indicator (一, 吕, 回, 吅, 咒, …, or *). | | leftComponent | string \| null | First sub-component (null when atomic). | | leftStrokes | number | Stroke count of the first sub-component. | | rightComponent | string \| null | Second sub-component (null when atomic). | | rightStrokes | number | Stroke count of the second sub-component. | | signature | string | Cangjie-style code (may be empty). | | notes | string | Free-form source notes (/, */, ?/, …). | | section | string \| null | Kangxi radical the character is filed under. |

In the source TSV, missing/atomic sub-components and sections are encoded as *; this parser converts them to null.

Regenerating ccd.json

scripts/parse.ts rebuilds ccd.json from data/ccd.tsv. It runs automatically on npm publish via prepublishOnly.

bun run build   # regenerate ccd.json from data/ccd.tsv

License

MIT. The underlying CCD table is distributed under the terms of the Wikimedia Commons project.