Overview

lindera-sqlite is a C ABI library which exposes a FTS5 tokenizer function.

When used as a custom FTS5 tokenizer this enables application to support Chinese, Japanese and Korean in full-text search.

Build extension

% cargo build --features=embed-cjk

Each embed-* feature embeds a different set of dictionaries into the built extension:

Feature	Embedded dictionaries
embed-ipadic	Japanese (IPADIC)
embed-ipadic-neologd	Japanese (IPADIC NEologd)
embed-unidic	Japanese (UniDic)
embed-ko-dic	Korean (ko-dic)
embed-cc-cedict	Chinese (CC-CEDICT)
embed-jieba	Chinese (Jieba)
embed-cjk	Japanese (IPADIC) + Korean (ko-dic) + Chinese (Jieba)

Set enviromment variable for Lindera configuration

% export LINDERA_CONFIG_PATH=./resources/lindera.yml

Then start SQLite

% sqlite3 example.db

Load extension

sqlite> .load ./target/debug/liblindera_sqlite lindera_fts5_tokenizer_init

Create table using FTS5 with Lindera tokenizer

sqlite> CREATE VIRTUAL TABLE example USING fts5(content, tokenize='lindera_tokenizer');

Insert data

sqlite> INSERT INTO example(content) VALUES ("Ｌｉｎｄｅｒａは形態素解析ｴﾝｼﾞﾝです。ユーザー辞書も利用可能です。");

Search data

sqlite> SELECT * FROM example WHERE content MATCH "Lindera" ORDER BY bm25(example) LIMIT 10;