8 种常见 SQL 错误用法

1、LIMIT 语句分页查询是最常用的场景之一,但也通常也是最容易出问题的地方。比如对于下面简单的语句,一般 DBA 想到的办法是在 type, name, create_time 字段上加组合索引。这样条件排序都能有效的利用到索引,性能迅速提升。SELECT * FROM   operation WHERE  type = 'SQLStats'        AND name = 'SlowLog' ORDER  BY create_time LIMIT  1000, 10;好吧,可能90%以上的 DBA 解决该问题就到此为止。但当 LIMIT 子句变成 “LIMIT 1000000,10” 时,程序员仍然会抱怨:我只取10条记录为什么还是慢?要知道数据库也并不知道第1000000条记录从什么地方开始,即使有索引也需要从头计算一次。出现这种性能问题,多数情形下是程序员偷懒了。在前端数据浏览翻页,或者大数据分批导出等场景下,是可以将上一页的最大值当成参数作为查询条件的。SQL 重新设计如下:SELECT   * FROM     operation WHERE    type = 'SQLStats' AND      name = 'SlowLog' AND      create_time > '2017-03-16 14:00:00' ORDER BY create_time limit 10;在新设计下查询时间基本固定,不会随着数据量的增长而发生变化。2、隐式转换SQL语句中查询变量和字段定义类型不匹配是另一个常见的错误。比如下面的语句:mysql> explain extended SELECT *      > FROM   my_balance b      > WHERE  b.bpn = 14000000123      >       AND b.isverified IS NULL ;mysql> show warnings;| Warning | 1739 | Cannot use ref access on index 'bpn' due to type or collation conversion on field 'bpn'其中字段 bpn 的定义为 varchar(20),MySQL 的策略是将字符串转换为数字之后再比较。函数作用于表字段,索引失效。上述情况可能是应用程序框架自动填入的参数,而不是程序员的原意。现在应用框架很多很繁杂,使用方便的同时也小心它可能给自己挖坑。3、关联更新、删除虽然 MySQL5.6 引入了物化特性,但需要特别注意它目前仅仅针对查询语句的优化。对于更新或删除需要手工重写成 JOIN。比如下面 UPDATE 语句,MySQL 实际执行的是循环/嵌套子查询(DEPENDENT SUBQUERY),其执行时间可想而知。UPDATE operation o SET    status = 'applying' WHERE  o.id IN (SELECT id                 FROM   (SELECT o.id,                                o.status                         FROM   operation o                         WHERE  o.group = 123                                AND o.status NOT IN ( 'done' )                         ORDER  BY o.parent,                                   o.id                         LIMIT  1) t);执行计划:+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+| id | select_type        | table | type  | possible_keys | key     | key_len | ref   | rows | Extra                                               |+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+| 1  | PRIMARY            | o     | index |               | PRIMARY | 8       |       | 24   | Using where; Using temporary                        || 2  | DEPENDENT SUBQUERY |       |       |               |         |         |       |      | Impossible WHERE noticed after reading const tables || 3  | DERIVED            | o     | ref   | idx_2,idx_5   | idx_5   | 8       | const | 1    | Using where; Using filesort                         |+----+--------------------+-------+-------+---------------+---------+---------+-------+------+-----------------------------------------------------+重写为 JOIN 之后,子查询的选择模式从 DEPENDENT SUBQUERY 变成 DERIVED,执行速度大大加快,从7秒降低到2毫秒。UPDATE operation o        JOIN  (SELECT o.id,                             o.status                      FROM   operation o                      WHERE  o.group = 123                             AND o.status NOT IN ( 'done' )                      ORDER  BY o.parent,                                o.id                      LIMIT  1) t         ON o.id = t.id SET    status = 'applying'执行计划简化为:+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+| id | select_type | table | type | possible_keys | key   | key_len | ref   | rows | Extra                                               |+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+| 1  | PRIMARY     |       |      |               |       |         |       |      | Impossible WHERE noticed after reading const tables || 2  | DERIVED     | o     | ref  | idx_2,idx_5   | idx_5 | 8       | const | 1    | Using where; Using filesort                         |+----+-------------+-------+------+---------------+-------+---------+-------+------+-----------------------------------------------------+4、混合排序MySQL 不能利用索引进行混合排序。但在某些场景,还是有机会使用特殊方法提升性能的。SELECT * FROM   my_order o        INNER JOIN my_appraise a ON a.orderid = o.id ORDER  BY a.is_reply ASC,           a.appraise_time DESC LIMIT  0, 20执行计划显示为全表扫描:+----+-------------+-------+--------+-------------+---------+---------+---------------+---------+-+| id | select_type | table | type   | possible_keys     | key     | key_len | ref      | rows    | Extra    +----+-------------+-------+--------+-------------+---------+---------+---------------+---------+-+|  1 | SIMPLE      | a     | ALL    | idx_orderid | NULL    | NULL    | NULL    | 1967647 | Using filesort ||  1 | SIMPLE      | o     | eq_ref | PRIMARY     | PRIMARY | 122     | a.orderid |       1 | NULL           |+----+-------------+-------+--------+---------+---------+---------+-----------------+---------+-+由于 is_reply 只有0和1两种状态,我们按照下面的方法重写后,执行时间从1.58秒降低到2毫秒。SELECT * FROM   ((SELECT *         FROM   my_order o                 INNER JOIN my_appraise a                         ON a.orderid = o.id                            AND is_reply = 0          ORDER  BY appraise_time DESC          LIMIT  0, 20)         UNION ALL         (SELECT *         FROM   my_order o                 INNER JOIN my_appraise a                         ON a.orderid = o.id                            AND is_reply = 1          ORDER  BY appraise_time DESC          LIMIT  0, 20)) t ORDER  BY  is_reply ASC,           appraisetime DESC LIMIT  20;5、EXISTS语句MySQL 对待 EXISTS 子句时,仍然采用嵌套子查询的执行方式。如下面的 SQL 语句:SELECT *FROM   my_neighbor n        LEFT JOIN my_neighbor_apply sra               ON n.id = sra.neighbor_id                  AND sra.user_id = 'xxx' WHERE  n.topic_status < 4        AND EXISTS(SELECT 1                   FROM   message_info m                   WHERE  n.id = m.neighbor_id                          AND m.inuser = 'xxx')        AND n.topic_type <> 5执行计划为:+----+--------------------+-------+------+-----+------------------------------------------+---------+-------+---------+ -----+| id | select_type        | table | type | possible_keys     | key   | key_len | ref   | rows    | Extra   |+----+--------------------+-------+------+ -----+------------------------------------------+---------+-------+---------+ -----+|  1 | PRIMARY            | n     | ALL  |  | NULL     | NULL    | NULL  | 1086041 | Using where                   ||  1 | PRIMARY            | sra   | ref  |  | idx_user_id | 123     | const |       1 | Using where          ||  2 | DEPENDENT SUBQUERY | m     | ref  |  | idx_message_info   | 122     | const |       1 | Using index condition; Using where |+----+--------------------+-------+------+ -----+------------------------------------------+---------+-------+---------+ -----+去掉 exists 更改为 join,能够避免嵌套子查询,将执行时间从1.93秒降低为1毫秒。SELECT *FROM   my_neighbor n        INNER JOIN message_info m                ON n.id = m.neighbor_id                   AND m.inuser = 'xxx'        LEFT JOIN my_neighbor_apply sra               ON n.id = sra.neighbor_id                  AND sra.user_id = 'xxx' WHERE  n.topic_status < 4        AND n.topic_type <> 5新的执行计划:+----+-------------+-------+--------+ -----+------------------------------------------+---------+ -----+------+ -----+| id | select_type | table | type   | possible_keys     | key       | key_len | ref   | rows | Extra                 |+----+-------------+-------+--------+ -----+------------------------------------------+---------+ -----+------+ -----+|  1 | SIMPLE      | m     | ref    | | idx_message_info   | 122     | const    |    1 | Using index condition ||  1 | SIMPLE      | n     | eq_ref | | PRIMARY   | 122     | ighbor_id |    1 | Using where      ||  1 | SIMPLE      | sra   | ref    | | idx_user_id | 123     | const     |    1 | Using where           |+----+-------------+-------+--------+ -----+------------------------------------------+---------+ -----+------+ -----+6、条件下推外部查询条件不能够下推到复杂的视图或子查询的情况有:聚合子查询;含有 LIMIT 的子查询;UNION 或 UNION ALL 子查询;输出字段中的子查询;如下面的语句,从执行计划可以看出其条件作用于聚合子查询之后:SELECT * FROM   (SELECT target,                Count(*)         FROM   operation         GROUP  BY target) t WHERE  target = 'rm-xxxx'+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+| id | select_type | table      | type  | possible_keys | key         | key_len | ref   | rows | Extra       |+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+|  1 | PRIMARY     || ref   ||| 514     | const |    2 | Using where ||  2 | DERIVED     | operation  | index | idx_4         | idx_4       | 519     | NULL  |   20 | Using index |+----+-------------+------------+-------+---------------+-------------+---------+-------+------+-------------+确定从语义上查询条件可以直接下推后,重写如下:SELECT target,        Count(*) FROM   operation WHERE  target = 'rm-xxxx' GROUP  BY target执行计划变为:+----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |+----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+| 1 | SIMPLE | operation | ref | idx_4 | idx_4 | 514 | const | 1 | Using where; Using index |+----+-------------+-----------+------+---------------+-------+---------+-------+------+--------------------+关于 MySQL 外部条件不能下推的详细解释说明请参考文章:http://mysql.taobao.org/monthly/2016/07/087、提前缩小范围先上初始 SQL 语句:SELECT * FROM   my_order o        LEFT JOIN my_userinfo u               ON o.uid = u.uid       LEFT JOIN my_productinfo p               ON o.pid = p.pid WHERE  ( o.display = 0 )        AND ( o.ostaus = 1 ) ORDER  BY o.selltime DESC LIMIT  0, 15该SQL语句原意是:先做一系列的左连接,然后排序取前15条记录。从执行计划也可以看出,最后一步估算排序记录数为90万,时间消耗为12秒。+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+| id | select_type | table | type   | possible_keys | key     | key_len | ref             | rows   | Extra                                              |+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+|  1 | SIMPLE      | o     | ALL    | NULL          | NULL    | NULL    | NULL            | 909119 | Using where; Using temporary; Using filesort       ||  1 | SIMPLE      | u     | eq_ref | PRIMARY       | PRIMARY | 4       | o.uid |      1 | NULL                                               ||  1 | SIMPLE      | p     | ALL    | PRIMARY       | NULL    | NULL    | NULL            |      6 | Using where; Using join buffer (Block Nested Loop) |+----+-------------+-------+--------+---------------+---------+---------+-----------------+--------+----------------------------------------------------+由于最后 WHERE 条件以及排序均针对最左主表,因此可以先对 my_order 排序提前缩小数据量再做左连接。SQL 重写后如下,执行时间缩小为1毫秒左右。SELECT * FROM (SELECT * FROM   my_order o WHERE  ( o.display = 0 )        AND ( o.ostaus = 1 ) ORDER  BY o.selltime DESC LIMIT  0, 15) o      LEFT JOIN my_userinfo u               ON o.uid = u.uid      LEFT JOIN my_productinfo p               ON o.pid = p.pid ORDER BY  o.selltime DESClimit 0, 15再检查执行计划:子查询物化后(select_type=DERIVED)参与 JOIN。虽然估算行扫描仍然为90万,但是利用了索引以及 LIMIT 子句后,实际执行时间变得很小。+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+| id | select_type | table      | type   | possible_keys | key     | key_len | ref   | rows   | Extra                                              |+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+|  1 | PRIMARY     || ALL    | NULL          | NULL    | NULL    | NULL  |     15 | Using temporary; Using filesort                    ||  1 | PRIMARY     | u          | eq_ref | PRIMARY       | PRIMARY | 4       | o.uid |      1 | NULL                                               ||  1 | PRIMARY     | p          | ALL    | PRIMARY       | NULL    | NULL    | NULL  |      6 | Using where; Using join buffer (Block Nested Loop) ||  2 | DERIVED     | o          | index  | NULL          | idx_1   | 5       | NULL  | 909112 | Using where                                        |+----+-------------+------------+--------+---------------+---------+---------+-------+--------+----------------------------------------------------+8、中间结果集下推再来看下面这个已经初步优化过的例子(左连接中的主表优先作用查询条件):SELECT    a.*,           c.allocated FROM      (               SELECT   resourceid               FROM     my_distribute d                    WHERE    isdelete = 0                    AND      cusmanagercode = '1234567'                    ORDER BY salecode limit 20) a LEFT JOIN           (               SELECT   resourcesid, sum(ifnull(allocation, 0) * 12345) allocated               FROM     my_resources                    GROUP BY resourcesid) c ON        a.resourceid = c.resourcesid那么该语句还存在其它问题吗?不难看出子查询 c 是全表聚合查询,在表数量特别大的情况下会导致整个语句的性能下降。其实对于子查询 c,左连接最后结果集只关心能和主表 resourceid 能匹配的数据。因此我们可以重写语句如下,执行时间从原来的2秒下降到2毫秒。SELECT    a.*,           c.allocated FROM      (                    SELECT   resourceid                    FROM     my_distribute d                    WHERE    isdelete = 0                    AND      cusmanagercode = '1234567'                    ORDER BY salecode limit 20) a LEFT JOIN           (                    SELECT   resourcesid, sum(ifnull(allocation, 0) * 12345) allocated                    FROM     my_resources r,                             (                                      SELECT   resourceid                                      FROM     my_distribute d                                      WHERE    isdelete = 0                                      AND      cusmanagercode = '1234567'                                      ORDER BY salecode limit 20) a                    WHERE    r.resourcesid = a.resourcesid                    GROUP BY resourcesid) c ON        a.resourceid = c.resourcesid但是子查询 a 在我们的SQL语句中出现了多次。这种写法不仅存在额外的开销,还使得整个语句显的繁杂。使用 WITH 语句再次重写:WITH a AS (          SELECT   resourceid          FROM     my_distribute d          WHERE    isdelete = 0          AND      cusmanagercode = '1234567'          ORDER BY salecode limit 20)SELECT    a.*,           c.allocated FROM      a LEFT JOIN           (                    SELECT   resourcesid, sum(ifnull(allocation, 0) * 12345) allocated                    FROM     my_resources r,                             a                    WHERE    r.resourcesid = a.resourcesid                    GROUP BY resourcesid) c ON        a.resourceid = c.resourcesid总结数据库编译器产生执行计划,决定着SQL的实际执行方式。但是编译器只是尽力服务,所有数据库的编译器都不是尽善尽美的。上述提到的多数场景,在其它数据库中也存在性能问题。了解数据库编译器的特性,才能避规其短处,写出高性能的SQL语句。程序员在设计数据模型以及编写SQL语句时,要把算法的思想或意识带进来。编写复杂SQL语句要养成使用 WITH 语句的习惯。简洁且思路清晰的SQL语句也能减小数据库的负担 。来源:yq.aliyun.com/articles/72501ENDPHP

返回列表
上一篇:
下一篇: