MYSQL的UPDATE子查询，UPDATE时避免使用子查询

大概率

 

MYSQL的UPDATE子查询，UPDATE时避免使用子查询

近日写mysql储存过程的时候，有个SQL执行不动:

UPDATE t_csi_comment
SET is_valid = 0
WHERE
        comment_id IN (
                SELECT
                        comment_id
                FROM
                        (
                                SELECT
                                        *
                                FROM
                                        t_csi_comment
                                WHERE
                                        DATE_FORMAT(comment_time, '%Y%m%d') BETWEEN 20170425
                                AND 20170528
                                ORDER BY
                                        comment_id DESC
                        ) a
                GROUP BY
                        openid,
                        dlr_code
                HAVING
                        count(1) > 2
        );
很奇怪，按道理这条SQL的检索量小于10W，应该怎么慢也不会几分钟不动的地步。

单独执行子查询:

SELECT
        comment_id
FROM
        (
                SELECT
                        *
                FROM
                        t_csi_comment
                WHERE
                        DATE_FORMAT(comment_time, '%Y%m%d') BETWEEN 20170425
                AND 20170528
                ORDER BY
                        comment_id DESC
        ) a
GROUP BY
        openid,
        dlr_code
HAVING
        count(1) > 2;
结果比想象中的快，1秒都不到，EXPLAIN后检索量不到4W行。我就郁闷了。

EXPLANIN第一条update语句:



注意：select_type 里出现了 DEPENDENT SUBQUERY。

这意味着什么？——子查询取决于外面的查询，Mysql 先执行外查询，内查询根据这个查询结果(如执行计划里所述，38196 rows)的每一条记录组成新的查询语句后执行。多重子查询情况下，我已经不想去解析它是如何转换SQL了。 Mysql在这点上并不比人类聪明。

解决办法（子查询转换成联表查询）：

UPDATE t_csi_comment a INNER JOIN
 (
        SELECT
                comment_id
        FROM
                (
                        SELECT
                                *
                        FROM
                                t_csi_comment
                        WHERE
                                DATE_FORMAT(comment_time, '%Y%m%d') BETWEEN 20170425
                        AND 20170528
                        ORDER BY
                                comment_id DESC
                ) a
        GROUP BY
                openid,
                dlr_code
        HAVING
                count(1) > 2
) b ON a.comment_id = b.comment_id;
SET a.is_valid = 0
        


毫秒级别完工。

按理说，越复杂的程序逻辑关系要越明朗，出现复杂SQL的几率要越低。但是总会有一块业务相对复杂多变，无法把控，或者就是整个系统的架构不够明朗，脱离不了复杂SQL。这是在UPDATE时发现的子查询问题，在其它SQL语句中肯定也会有所体现，这是Mysql的查询机制问题，子查询会让Mysql变笨。所以还是慎用子查询，各种复杂SQL下尽量先测试吧。
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。互联网+时代，时刻要保持学习，携手千锋PHP,Dream It Possible。

转载于:https://my.oschina.net/u/3432750/blog/911855