基于Sphinx的中文全文检索引擎Coreseek的安装

按照Sphinx来进行全文检索，默认只支持按字拆分，如果要实现比较好的中文分词，可以使用基于libmmseg的引擎Coreseek。

yum install g++

yum install gcc

yum install make

yum install mysql mysql-devel php-mysql qt4-mysql

wget http://www.coreseek.cn/uploads/sources/mmseg3_0b3.tar.gz

wget http://www.coreseek.cn/uploads/sources/csft3_0b4.tar.gz

tar -xzvf mmseg3_0b3.tar.gz

tar -xzvf csft3_0b4.tar.gz

cd mmseg.3.0b3/

./configure --prefix=/var/mmseg

make

make install

cd ..

cd csft3_0b4

./configure --prefix=/var/coreseek --with-mysql --with-mmseg-includes=/var/mmseg/include/mmseg --with-mmseg-libs=/var/mmseg/lib/

make

make install

cd /var/coreseek/

mkdir dict

cd /home/hfahe/mmseg.3.0b3/data

/var/mmseg/bin/mmseg -u unigram.txt

cp unigram.txt.uni /var/coreseek/dict/uni.lib

cd /var/coreseek/dict/

vi mmseg.ini

输入

[mmseg] 
merge_number_and_ascii=1; 
number_and_ascii_joint=-; 
compress_space=0; 
seperate_number_ascii=1; 

cd /var/coreseek/etc/

cp sphinx.conf.dist sphinx.conf

mysql -h 192.168.1.xxx -u root -pxxx test < example.sql

vi sphinx.conf

修改配置里的数据库ip、用户名、密码、数据库

/var/coreseek/bin/indexer --config /var/coreseek/etc/sphinx.conf

这时可能会出现关于libmysqlclient的错误，解决办法如下：

locate libmysqlclient.so

ln -s /usr/local/lib/mysql/libmysqlclient.so.16 /lib/libmysqlclient.so.16

/var/coreseek/bin/indexer --config /var/coreseek/etc/sphinx.conf --all

/var/coreseek/bin/search --config /var/coreseek/etc/sphinx.conf doc

displaying matches:
1. document=3, weight=1, group_id=2, date_added=Thu Apr 22 15:15:25 2010
        id=3
        group_id=2
        group_id2=7
        date_added=2010-04-22 15:15:25
        title=another doc
        content=this is another group
2. document=4, weight=1, group_id=2, date_added=Thu Apr 22 15:15:25 2010
        id=4
        group_id=2
        group_id2=8
        date_added=2010-04-22 15:15:25
        title=doc number four
        content=this is to test groups

words:
1. 'doc': 2 documents, 2 hits

要支持中文，需要将配置里的charset_type值修改为zh_cn.utf-8，并添加charset_dictpath = /var/coreseek/dict。

还需要将sql_query_pre = SET NAMES utf8这一行配置打开。

/var/coreseek/bin/indexer --config /var/coreseek/etc/sphinx.conf --all

/var/coreseek/bin/search --config /var/coreseek/etc/sphinx.conf 中文

试试是否已经可以正常检索中文了。

coreseek默认的配置文件是etc下的csft.conf，使用此文件时可以不需要加入config的配置。

此时应该已经可以显示正确结果了。