pg mysql 联表查询_postgresql多表查询语句优化

今天接到客户反馈，网站这两天经常出现无法访问的情况，查看日志发现是一个页面里的sql查询太慢，需要30多秒导致超时并且因为访问量多堵塞住了，导致其他页面也无法正常访问，所以对这个sql进行优化，下面介绍下优化过程。

这个查询是通过3个表来查询的，一个产品表(prod)，两个别名表(synonym, synonym_cn)，原来的SQL语句是这样的:

select id from prod where prod.id > 0 and (prod.cas = '72-44-6' or (exists (select 1 from synonym where prod.id = synonym.prod and lower(synonym.name) = '72-44-6')) or (exists (select 1 from synonym_cn where prod.id = synonym_cn.prod and lower(synonym_cn.name) = '72-44-6')))

其中prod.cas建了唯一索引，synonym和synonym_cn的name建了lower函数索引。prod的数量是64万，synonym和synonym_cn都是100多万。

分析后发现prod.cas并没有走索引，所以先去掉cas的查询条件进行测试，发现还是很慢，而name的搜索还会结合prod.id，所以重建了synonym和synonym_cn的name索引为联合索引，即prod和lower(name)，但测试后效果依然不理想，所以讲索引改回，并将语句分成三个语句测试：

select id from prod where prod.cas = '72-44-6'

select prod from synonym where lower(synonym.name) = '72-44-6'

select prod from synonym_cn where lower(synonym_cn.name) = '72-44-6'

发现搜索速度都很快，所以讲整体语句改成

select id from prod where prod.id in (select b.id from prod b where b.cas = '72-44-6' union select prod from synonym where lower(synonym.name) = '72-44-6' union select prod from synonym_cn where lower(synonym_cn.name) = '72-44-6')

至此解决问题。