SQL 优化案例之变更表结构

从慢日志报表中看到一条很长的SQL

select
  id
from
  myinfo
WHERE 1 = 1
  and ((
    SUBSTRING_INDEX(location_axis, '$', 3) like concat('%$', 2334)
    or SUBSTRING_INDEX(location_axis, '$', 3) like concat(concat('%$', 2334), '$%')
  )
  or (
    SUBSTRING_INDEX(location_axis, '$', 3) like concat('%$', 2337)
    or SUBSTRING_INDEX(location_axis, '$', 3) like concat(concat('%$', 2337), '$%')
  )
  or (
    SUBSTRING_INDEX(location_axis, '$', 3) like concat('%$', 2340)
    or SUBSTRING_INDEX(location_axis, '$', 3) like concat(concat('%$', 2340), '$%')
  )
  or (
    SUBSTRING_INDEX(location_axis, '$', 3) like concat('%$', 2353)
    or SUBSTRING_INDEX(location_axis, '$', 3) like concat(concat('%$', 2353), '$%')
  )
  or (
    SUBSTRING_INDEX(location_axis, '$', 3) like concat('%$', 2367)
    or SUBSTRING_INDEX(location_axis, '$', 3) like concat(concat('%$', 2367), '$%')
  )
  or ......

询问开发得知是通过程序拼成的sql

表中存入的location_axis是区域信息，存三个区域，用$分隔，如 $1$2$3，可以看出要实现的是匹配值是否在location_axis中的第一个域或第二个域中

匹配的值是从另一个表中根据条件获取的如：

select it_area_id itinfo where dep_id=''

问题：繁琐、前缀为%无法使用索引，导致比较慢

优化方式：

将第一个地域编码和第二个分别存在单独的列中，然后和表it_area_id进行关联查询

变更表结构

alter table myinfo add first_axis varchar(10),add sec_axis varchar(10);
update myinfo t1 inner join (select id,SUBSTRING_INDEX(SUBSTRING_INDEX(location_axis,'$',2),'$',-1) as c1 from itinfo)t2 using(id) set t1.first_axis=t2.c1;
update myinfo t1 inner join (select id,SUBSTRING_INDEX(SUBSTRING_INDEX(location_axis,'$',3),'$',-1) as c1 from itinfo)t2 using(id) set t1.sec_axis=t2.c1;

查询语句变为

select opr_bo_id from myinfo t1 inner join itinfo t2 on t1.first_axis=t2.it_area_id or t1.sec_axis=t2.it_area_id where t2.dep_id='';

结果达到毫秒级别