今天在线上发现一个慢sql:
SELECT COUNT(o.order_sn) FROM eload_order_info AS o,eload_users AS u ,eload_wj_link AS wj WHERE o.wj_linkid = wj.id and u.user_id = wj.user_id AND o.pay_time >= ‘1488297600‘ AND o.pay_time <= ‘1488383940‘ ; +-------------------+ | COUNT(o.order_sn) | +-------------------+ | 1432 | +-------------------+ 1 row in set (8.03 sec)
我们来看一下执行计划：
explain SELECT COUNT(o.order_sn) FROM eload_order_info AS o,eload_users AS u ,eload_wj_link AS wj WHERE o.wj_linkid = wj.id and u.user_id = wj.user_id AND o.pay_time >= ‘1488297600‘ AND o.pay_time <= ‘1488383940‘ \G *************************** 1. row *************************** id: 1 select_type: SIMPLE table: wj type: index possible_keys: PRIMARY,user_id key: user_id key_len: 4 ref: NULL rows: 401244 Extra: Using index *************************** 2. row *************************** id: 1 select_type: SIMPLE table: o type: ref possible_keys: wj_linkid,wj_linkid_2 key: wj_linkid key_len: 4 ref: rosewholesale_db.wj.id rows: 14 Extra: Using where *************************** 3. row *************************** id: 1 select_type: SIMPLE table: u type: eq_ref possible_keys: PRIMARY key: PRIMARY key_len: 4 ref: rosewholesale_db.wj.user_id rows: 1 Extra: Using where; Using index 3 rows in set (0.00 sec)
我们来看一下这三个表的行数：
>select count(*) from eload_order_info; +----------+ | count(*) | +----------+ | 3258824 | +----------+ 1 row in set (2.42 sec) >select count(*) from eload_users; +----------+ | count(*) | +----------+ | 6505146 | +----------+ 1 row in set (3.00 sec) >select count(*) from eload_wj_link; +----------+ | count(*) | +----------+ | 400153 | +----------+ 1 row in set (0.09 sec)
从执行计划我们可以看到mysql选择了eload_wj_link做为驱动表，如果我们指定eload_order_info作为驱动表的话可以pay_time的范围来过滤结果集。这时我们可以通过straight_jion来指定表连接的顺序。
对上面的sql进行改写后我们来看一下运行时间：
>SELECT COUNT(o.order_sn) FROM eload_order_info AS o straight_join eload_wj_link AS wj on o.wj_linkid = wj.id, eload_users AS u WHERE u.user_id = wj.user_id AND o.pay_time >= ‘1488297600‘ AND o.pay_time <= ‘1488383940‘ ; +-------------------+ | COUNT(o.order_sn) | +-------------------+ | 1512 | +-------------------+ 1 row in set (2.18 sec)
速度从原来的8.02s提升到2.18s,这个改善还是挺大的，我们再来看一下改造后的语句的执行计划：
>explain SELECT COUNT(o.order_sn) FROM eload_order_info AS o straight_join eload_wj_link AS wj on o.wj_linkid = wj.id, eload_users AS u WHERE u.user_id = wj.user_id AND o.pay_time >= ‘1488297600‘ AND o.pay_time <= ‘1488383940‘ \G *************************** 1. row *************************** id: 1 select_type: SIMPLE table: o type: ALL possible_keys: wj_linkid,wj_linkid_2 key: NULL key_len: NULL ref: NULL rows: 3351498 Extra: Using where *************************** 2. row *************************** id: 1 select_type: SIMPLE table: wj type: eq_ref possible_keys: PRIMARY,user_id key: PRIMARY key_len: 4 ref: rosewholesale_db.o.wj_linkid rows: 1 Extra: *************************** 3. row *************************** id: 1 select_type: SIMPLE table: u type: eq_ref possible_keys: PRIMARY key: PRIMARY key_len: 4 ref: rosewholesale_db.wj.user_id rows: 1 Extra: Using where; Using index 3 rows in set (0.00 sec)
虽然扫描的行数多了，但是实际执行速度快了很多。当涉及到多表连接时如果你认为按自己的顺序比MySQL推荐的顺序进行连接的效率高的话，就可以通过STRAIGHT_JOIN来确定连接顺序。
时间： 2024-08-06 16:05:34