# External stroke-order source

This directory is a verbatim unpacked copy of `hanzi-writer-data@2.0.1` from
the public npm registry. It was downloaded on 2026-08-02 for the standalone
Hanyu website and was not copied from the Hanyu Flutter app's generated stroke
assets.

## Upstream

- Data package: `hanzi-writer-data@2.0.1`
- Repository: <https://github.com/chanind/hanzi-writer-data>
- Original project: <https://github.com/skishore/makemeahanzi>
- npm tarball: <https://registry.npmjs.org/hanzi-writer-data/-/hanzi-writer-data-2.0.1.tgz>
- npm integrity: `sha512-nbQwM+MaryGoq7pBMIZLCd3lFq03nXuJuwku1+6UbjL58uU+9OULVcMkoNvNuJSoIV7f1bbPRfD4D/LQa5S7qg==`
- Downloaded tarball SHA-256: `72baf3d82b114e60d6e40ea05f24d2262a05cd39d544e2f322ba2fceb7beff15`

The package contains 9,574 per-character JSON files. Each character record
contains ordered vector `strokes` plus matching `medians` used for animation
and handwriting-quiz guidance.

## Website coverage

The seven local HSK 3.0 vocabulary sources contain 11,000 vocabulary records
and 3,088 distinct Han characters. A character-by-character comparison against
this external package found:

- 3,088 / 3,088 unique HSK characters covered
- 11,000 / 11,000 vocabulary records fully covered
- 0 missing HSK characters

This proves complete coverage for the website's current HSK vocabulary. It does
not claim coverage of every CJK Unified Ideograph or of an arbitrary set of
10,000 distinct rare characters.

## License

The data is distributed under the Arphic Public License. The complete upstream
license is preserved in `ARPHICPL.TXT`, with translated copies under `APL/`.

The browser runtime is stored separately under
`assets/vendor/hanzi-writer/`. Hanzi Writer is MIT licensed; its `LICENSE` and
`COPYING.md` files are preserved with the runtime.