mysql-优化案例

优化分页

select title,content from product_comment where status=1 and product_id=1111 limit 10,5;

• 第一步优化在status和product_id上构建联合索引
  • 联合索引的顺序依据选择性决定
  • select count(distinct status)/count(*),count(distinct product_id)/count(*) from product_comment
• 但现在依然存在问题，如果limit的偏移较大的情况下会很耗时
• 第二步优化：在覆盖索引上做分页，减少二次查询的数据量
  • status和product_id上构建联合索引，在comment_id上建立主键

select title,content from
(select comment_id from product_comment status=1 and product_id=1111 limit 10,5) a 
join product_comment b on a.comment_id=b.comment_id;

删除重复数据

• 用链接来代替使用in
• 获取重复的评论数量

select product_id,order_id,count(comment_id) from comment group by product_id,order_id having count(comment_id)>1;

• 获取重复评论的comment_id

select comment_id from (select product_id,order_id from comment group by product_id,order_id having count(comment_id)>1) a join comment b on a.product_id=b.product_id and a.order_id=b.order_id;

• 保留最小的comment_id，删除其他的

delete from comment b join
(select product_id,order_id,min(comment_id) as comment_id from comment group by product_id,order_id having count(comment_id)>1) a
on a.product_id=b.product_id and a.order_id=b.order_id and b.comment_id>a.comment_id;

使用join来代替in

SELECT count(*)
FROM emp
WHERE depid not IN( SELECT depid
FROM dep
WHERE dname ='bLByMr' or dname = 'pyuDKO')

• 会先查询子表，然后外层表会依次去同子表的结果集进行比较。如果外层表很大会消耗很多的io。改为只用inner join

EXPLAIN 
SELECT count(*)
FROM emp join dep on emp.depid=dep.depid
WHERE dname !='bLByMr' and dname != 'pyuDKO'

循环取出外层表的每一条记录与子表进行比较: https://www.jianshu.com/p/b723372ccb57

对表过分范式化

• 将一些信息存在表中，避免过多的表连接影响效率

索引覆盖

• 在一些常用的查询上尽量使用到索引覆盖，减少二次查询的机会