kl3x/mysql-mecab repository overview

⁠MySQL with MeCab Docker Image

简体中文⁠

This repository contains the necessary files to build a Docker image based on the official mysql:8.4 image (Oracle Linux 9) that includes the MeCab⁠ Full-Text Parser Plugin. This allows for Japanese full-text searching capabilities within MySQL.

⁠Prerequisites

• Docker⁠ installed on your system.

⁠Building the Image Locally

1. Clone this repository:

   git clone https://github.com/stupidloud/mysql-with-mecab.git
   cd mysql-with-mecab
   

   Copy
2. Build the Docker image:

   docker build -t mysql-mecab:latest .
   

   Copy
   (Note: The first build might take a while as it needs to download the MySQL source code and compile the plugin.)

⁠Using the Image

You can use this image like any other MySQL image. Here's an example using docker run:

docker run --name some-mysql-mecab -e MYSQL_ROOT_PASSWORD=my-secret-pw -d mysql-mecab:latest

Or in a docker-compose.yml file:

services:
  db:
    image: mysql-mecab:latest # Use the locally built image or replace with your published image (e.g., your-dockerhub-username/mysql-mecab:latest)
    container_name: mysql_with_mecab
    environment:
      MYSQL_ROOT_PASSWORD: your_root_password
      MYSQL_DATABASE: your_database
      MYSQL_USER: your_user
      MYSQL_PASSWORD: your_password
    volumes:
      - ./mysql_data:/var/lib/mysql
    ports:
      - "3306:3306"
    restart: always

When MySQL service starts, the MeCab plugin is automatically loaded (via the plugin-load-add configuration). You can verify the installation by connecting to the MySQL instance and running:

SHOW PLUGINS;
SHOW STATUS LIKE 'mecab_charset';

The default dictionary is the UTF-8 IPA dictionary (mecab-ipadic), so full-text indexed columns must use the utf8mb4 character set:

CREATE TABLE articles (
  id INT PRIMARY KEY AUTO_INCREMENT,
  body TEXT,
  FULLTEXT (body) WITH PARSER mecab
) DEFAULT CHARSET=utf8mb4;

innodb_ft_min_token_size is set to 1 so that short Japanese words are indexed. If you change this value on an existing database, rebuild existing full-text indexes.

⁠Japanese stopwords

When the data directory is initialized for the first time, initdb/01-mecab-stopwords.sql creates the stopword table ft_stopwords.ja (InnoDB's default English stopwords plus Japanese particles and auxiliary verbs such as は, の, です, ます, plus full-width and ASCII punctuation/symbols such as 。, ：, ・, ―, /) and enables it with SET PERSIST innodb_ft_server_stopword_table = 'ft_stopwords/ja'. Stopwords are ignored both when indexing and when searching, so documents no longer match just because they share a particle.

• You can add or remove words in ft_stopwords.ja; changes only apply to full-text indexes created or rebuilt afterwards.
• For an existing data directory (initialized before this script existed), run the script manually, then rebuild your full-text indexes.

⁠Automated Builds (GitHub Actions)

On every push to main (or when triggered manually), the Docker Image CI⁠ workflow builds the image, pushes it to Docker Hub as kl3x/mysql-mecab:latest, and syncs this README to the Docker Hub description. It requires the DOCKERHUB_USERNAME and DOCKERHUB_TOKEN repository secrets.

⁠Files

• Dockerfile: Defines the multi-stage build process for the image. The plugin is compiled from the MySQL source tag that exactly matches the mysqld version in the base image.
• mecab.cnf: MySQL configuration file that loads the MeCab plugin with plugin-load-add and sets innodb_ft_min_token_size=1.
• initdb/01-mecab-stopwords.sql: Creates and enables the Japanese stopword table on first initialization.