MySQL max() 函数的需扫描where条件过滤后的所有行:
在测试环境中重现:
测试版本:Server version: 5.1.58-log MySQL Community Server (GPL)
testtable表中的索引
mysql> show index from testtable; +-----------+------------+------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | +-----------+------------+------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+ | testtable | 0 | PRIMARY | 1 | id | A | 2 | NULL | NULL | | BTREE | | | testtable | 1 | key_number | 1 | number | A | 2 | NULL | NULL | YES | BTREE | | +-----------+------------+------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
对比的sql为:
select sql_no_cache max(id) from testtable where number=98; select sql_no_cache id from testtable where number=98 order by id desc limit 1;
查看执行计划:
mysql> explain select sql_no_cache max(id) from testtable where number=98; +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ | id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra | +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ | 1 | SIMPLE | testtable | ref | key_number | key_number | 5 | const | 4 | Using where; Using index | +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ 1 row in set (0.00 sec)
mysql> explain select sql_no_cache id from testtable where number=98 order by id desc limit 1; +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ | id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra | +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ | 1 | SIMPLE | testtable | ref | key_number | key_number | 5 | const | 4 | Using where; Using index | +----+-------------+-----------+------+---------------+------------+---------+-------+------+--------------------------+ 1 row in set (0.00 sec) 执行计划显示完全一样。
其中,number为98 对应的记录有4行:
mysql> select count(*) from testtable where number=98;
+----------+ | count(*) | +----------+ | 4 | +----------+ 1 row in set (0.00 sec)
执行前查看innodb_rows_read
#innodb_rows_read 从InnoDB表读取的行数
mysql> show global status like 'innodb_rows_read'; +------------------+-------+ | Variable_name | Value | +------------------+-------+ | Innodb_rows_read | 1022 | +------------------+-------+ 1 row in set (0.00 sec)
执行sql1 mysql> select sql_no_cache max(id) from testtable where number=98; +---------+ | max(id) | +---------+ | 13 | +---------+
1 row in set (0.00 sec)
执行后查看innodb_rows_read,发现innodb_rows_read增加了4,即number为98 对应的记录有4行 mysql> show global status like 'innodb_rows_read'; +------------------+-------+ | Variable_name | Value | +------------------+-------+ | Innodb_rows_read | 1026 | +------------------+-------+ 1 row in set (0.00 sec)
执行sql2 mysql> select sql_no_cache id from testtable where number=98 order by id desc limit 1; +----+ | id | +----+ | 13 | +----+ 1 row in set (0.00 sec)
执行后查看innodb_rows_read,发现innodb_rows_read增加了1 mysql> show global status like 'innodb_rows_read'; +------------------+-------+ | Variable_name | Value | +------------------+-------+ | Innodb_rows_read | 1027 | +------------------+-------+ 1 row in set (0.00 sec)
测试得出:
select sql_no_cache max(id) from testtable where number=98;
需要读取 number=98 的所有行,才能得到最大的id
select sql_no_cache id from testtable where number=98 order by id desc limit 1;
由于id是主键,number是第二索引,只需扫描1行即可得到最大的id
请慎用max()函数,特别是频繁执行的sql,若需用到可转化为测试中的 order by id desc limit 1
因为往往min()或者max()函数往往会造成全表扫描