gukhanmun-core

Format-neutral core of the Gukhanmun hanja-to-hangul pipeline. The crate is no_std + alloc, so it is suitable for embedded and WASM targets. Format adapters, dictionary backends, and runtime I/O live in separate crates.

What this crate provides

HanjaDictionary is the trait that all dictionary backends implement. The engine works over any value that satisfies it. ChainDictionary composes two backends so that a user-supplied dictionary is consulted before a larger pre-built one.

The lattice segmenter covers all possible dictionary matches for a hanja span and picks the best split using dynamic programming. 行事場所 segments as 行事 + 場所 rather than 行事場 + 所.

Fallback phonetization kicks in for characters not found in any dictionary. It uses a table derived from the Unicode Unihan database (kHangul field) and applies the initial sound law (頭音法則) when the active preset requires it.

The rendering step turns engine output into text. RenderMode controls the shape: hangul-only, hangul(hanja) parentheses, hanja(hangul) parentheses, ruby markup, or the original mixed-script form with selective glosses.

Installation

[dependencies]
gukhanmun-core = "0.1"

Usage

use gukhanmun_core::{MapDictionary, RenderMode, convert_plain_text};

let mut dict = MapDictionary::new();
dict.insert("漢字", "한자");
dict.insert("北京", "베이징");

let output = convert_plain_text("漢字 北京", &dict, RenderMode::HangulOnly);
assert_eq!(output, "한자 베이징");

no_std note

The crate declares #![no_std] and uses extern crate alloc. Callers on std targets get alloc for free; embedded targets need a global allocator. The dictionary trait itself is synchronous and allocation-free on the hot path.

License

GPL-3.0-only. See LICENSE at the repository root.