kakasi

kakasi is a Rust library to transliterate hiragana, katakana and kanji (Japanese text) into rōmaji (Latin/Roman alphabet).

It was ported from the pykakasi library which itself is a port of the original kakasi library written in C.

Usage

Transliterate:

rust let res = kakasi::convert("こんにちは世界!"); assert_eq!(res.hiragana, "こんにちはせかい!"); assert_eq!(res.romaji, "konnichiha sekai!");

Check if a string contains Japanese characters:

```rust use kakasi::IsJapanese;

asserteq!(kakasi::isjapanese("Abc"), IsJapanese::False); asserteq!(kakasi::isjapanese("日本"), IsJapanese::Maybe); asserteq!(kakasi::isjapanese("ラスト"), IsJapanese::True); ```

CLI

```sh $ cargo install kakasi

Convert to romaji

$ kakasi こんにちは世界! konnichiha sekai!

Convert to hiragana

$ kakasi -k こんにちは世界! こんにちはせかい!

Read from file

$ kakasi -f rust_article.txt

Read from STDIN

$ echo "こんにちは世界!" | kakasi ```

Performance

CPU: AMD Ryzen 7 5700G

| Text | Conversion time | Speed | | -------------------------------- | --------------- | ---------- | | Sentence (161 B) | 7.0911 µs | 22.70 MB/s | | Rust wikipedia article (31705 B) | 1.5055 ms | 21.06 MB/s |

CLI comparison

Time to convert a 100KB test file using the CLI:

| Library | Time | Speed | | ---------------------------------------------------------- | -------- | ---------- | | kakasi (Rust) | 7.4 ms | 13.5 MB/s | | kakasi (C) | 33.5 ms | 2.99 MB/s | | pykakasi (Python) | 810.6 ms | 0.123 MB/s |

Test commands:

CLI performance was measured with hyperfine.

sh hyperfine --warmup 3 'cat 100K.txt | kakasi-rs' hyperfine --warmup 3 'cat 100K.txt | kakasi -i utf-8 -Ka -Ha -Ja -Sa -s' hyperfine --warmup 3 'cat 100K.txt | python bin/kakasi -Ka -Ha -Ja -Sa -s'

License

kakasi is published under the GNU GPL-3.0 license.

The Kakasi dictionaries (Files: codegen/dict/kakasidict.utf8, codegen/dict/itajidict.utf8, codegen/dict/hepburn.utf8) were taken from the pykakasi project, published under the GNU GPL-3.0 license.

pykakasi

Copyright (C) 2010-2021 Hiroshi Miura and contributors(see AUTHORS)

The dictionaries originate from the kakasi project, published under the GNU GPL-2.0 license.

original kakasi

Copyright (C) 1992 1993 1994
Hironobu Takahashi (takahasi@tiny.or.jp),
Masahiko Sato (masahiko@sato.riec.tohoku.ac.jp),
Yukiyoshi Kameyama, Miki Inooka, Akihiko Sasaki, Dai Ando, Junichi Okukawa,
Katsushi Sato and Nobuhiro Yamagishi

For testing I included a copy of the Japanese Rust wikipedia article (tests/rust_article.txt). The article is published under the Creative Commons Attribution-ShareAlike License 3.0.