关于mysql中explain的那些事儿

 

有两种用法：

• 1.EXPLAIN tbl_name
  
• 2.EXPLAIN [EXTENDED] SELECT select_options

为了更好的说明它，我们需要建两张表，下面的语句用于创建一张测试用的订单表：

CREATE TABLE `t_order` (
 `order_id` int(10) unsigned NOT NULL AUTO_INCREMENT COMMENT '订单ID',
 `express_type` tinyint(1) unsigned NOT NULL COMMENT '快递方式',
 `user_id` int(10) unsigned DEFAULT NULL COMMENT '用户ID',
 `add_time` int(10) NOT NULL COMMENT '下单时间',
 PRIMARY KEY (`order_id`),
 KEY `user_id` (`user_id`),
 KEY `express_type` (`express_type`)
) ENGINE=InnoDB AUTO_INCREMENT=100000 DEFAULT CHARSET=utf8 COMMENT='订单记录表'

然后我往其中插入了100000条记录。

下面是订单的扩展表，仅向其中插入了一条记录，关联某笔订单。

CREATE TABLE `t_order_ext` (
 `order_id` int(10) NOT NULL COMMENT '订单ID',
 `user_type` int(11) NOT NULL DEFAULT '0' COMMENT '用户类型',
 `comment` varchar(255) CHARACTER SET utf8 NOT NULL COMMENT '订单备注',
 KEY `order_id` (`order_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8

先看一下第一种用法：

mysql> explain t_order;
+----------+------------------+------+-----+---------+----------------+
| Field    | Type             | Null | Key | Default | Extra          |
+----------+------------------+------+-----+---------+----------------+
| order_id | int(10) unsigned | NO   | PRI | NULL    | auto_increment |
| add_time | int(10)          | NO   |     | NULL    |                |
+----------+------------------+------+-----+---------+----------------+
2 rows in set (0.04 sec)

这其实和describe是等价的：

mysql> describe t_order;
+----------+------------------+------+-----+---------+----------------+
| Field    | Type             | Null | Key | Default | Extra          |
+----------+------------------+------+-----+---------+----------------+
| order_id | int(10) unsigned | NO   | PRI | NULL    | auto_increment |
| add_time | int(10)          | NO   |     | NULL    |                |
+----------+------------------+------+-----+---------+----------------+
2 rows in set (0.00 sec)

这返回的是表的结构，没有什么可说的。

重点是第二种用法，需要深入的了解。

先看一个例子：

mysql> explain select * from t_order;
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
| id | select_type | table   | type | possible_keys | key  | key_len | ref  | rows   | Extra |
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
|  1 | SIMPLE      | t_order | ALL  | NULL          | NULL | NULL    | NULL | 100453 |       |
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
1 row in set (0.03 sec)

加上extended后之后：

mysql> explain extended select * from t_order;
+----+-------------+---------+------+---------------+------+---------+------+--------+----------+-------+
| id | select_type | table   | type | possible_keys | key  | key_len | ref  | rows   | filtered | Extra |
+----+-------------+---------+------+---------------+------+---------+------+--------+----------+-------+
|  1 | SIMPLE      | t_order | ALL  | NULL          | NULL | NULL    | NULL | 100453 |   100.00 |       |
+----+-------------+---------+------+---------------+------+---------+------+--------+----------+-------+
1 row in set, 1 warning (0.00 sec)

有必要解释一下这个长长的表格里每一列的含义：

id	SELECT识别符。这是SELECT的查询序列号
select_type	SELECT类型,可以为以下任何一种: SIMPLE:简单SELECT(不使用UNION或子查询) PRIMARY:最外面的SELECT UNION:UNION中的第二个或后面的SELECT语句 DEPENDENT UNION:UNION中的第二个或后面的SELECT语句,取决于外面的查询 UNION RESULT:UNION 的结果 SUBQUERY:子查询中的第一个SELECT DEPENDENT SUBQUERY:子查询中的第一个SELECT,取决于外面的查询 DERIVED:导出表的SELECT(FROM子句的子查询)
table	输出的行所引用的表
type	联接类型。下面给出各种联接类型,按照从最佳类型到最坏类型进行排序: system:表仅有一行(=系统表)。这是const联接类型的一个特例。 const:表最多有一个匹配行,它将在查询开始时被读取。因为仅有一行,在这行的列值可被优化器剩余部分认为是常数。const表很快,因为它们只读取一次! eq_ref:对于每个来自于前面的表的行组合,从该表中读取一行。这可能是最好的联接类型,除了const类型。 ref:对于每个来自于前面的表的行组合,所有有匹配索引值的行将从这张表中读取。 ref_or_null:该联接类型如同ref,但是添加了MySQL可以专门搜索包含NULL值的行。 index_merge:该联接类型表示使用了索引合并优化方法。 unique_subquery:该类型替换了下面形式的IN子查询的ref: value IN (SELECT primary_key FROM single_table WHERE some_expr) unique_subquery是一个索引查找函数,可以完全替换子查询,效率更高。 index_subquery:该联接类型类似于unique_subquery。可以替换IN子查询,但只适合下列形式的子查询中的非唯一索引: value IN (SELECT key_column FROM single_table WHERE some_expr) range:只检索给定范围的行,使用一个索引来选择行。 index:该联接类型与ALL相同,除了只有索引树被扫描。这通常比ALL快,因为索引文件通常比数据文件小。 ALL:对于每个来自于先前的表的行组合,进行完整的表扫描。
possible_keys	指出MySQL能使用哪个索引在该表中找到行
key	显示MySQL实际决定使用的键(索引)。如果没有选择索引,键是NULL。
key_len	显示MySQL决定使用的键长度。如果键是NULL,则长度为NULL。
ref	显示使用哪个列或常数与key一起从表中选择行。
rows	显示MySQL认为它执行查询时必须检查的行数。多行之间的数据相乘可以估算要处理的行数。
filtered	显示了通过条件过滤出的行数的百分比估计值。
Extra	该列包含MySQL解决查询的详细信息 Distinct:MySQL发现第1个匹配行后,停止为当前的行组合搜索更多的行。 Not exists:MySQL能够对查询进行LEFT JOIN优化,发现1个匹配LEFT JOIN标准的行后,不再为前面的的行组合在该表内检查更多的行。 range checked for each record (index map: #):MySQL没有发现好的可以使用的索引,但发现如果来自前面的表的列值已知,可能部分索引可以使用。 Using filesort:MySQL需要额外的一次传递,以找出如何按排序顺序检索行。 Using index:从只使用索引树中的信息而不需要进一步搜索读取实际的行来检索表中的列信息。 Using temporary:为了解决查询,MySQL需要创建一个临时表来容纳结果。 Using where:WHERE 子句用于限制哪一个行匹配下一个表或发送到客户。 Using sort_union(...), Using union(...), Using intersect(...):这些函数说明如何为index_merge联接类型合并索引扫描。 Using index for group-by:类似于访问表的Using index方式,Using index for group-by表示MySQL发现了一个索引,可以用来查 询GROUP BY或DISTINCT查询的所有列,而不要额外搜索硬盘访问实际的表。

 

一.select_type的说明

1.UNION:

当通过union来连接多个查询结果时，第二个之后的select其select_type为UNION。

mysql> explain select * from t_order where order_id=100 union select * from t_order where order_id=200;
+----+--------------+------------+-------+---------------+---------+---------+-------+------+-------+
| id | select_type  | table      | type  | possible_keys | key     | key_len | ref   | rows | Extra |
+----+--------------+------------+-------+---------------+---------+---------+-------+------+-------+
|  1 | PRIMARY      | t_order    | const | PRIMARY       | PRIMARY | 4       | const |    1 |       |
|  2 | UNION        | t_order    | const | PRIMARY       | PRIMARY | 4       | const |    1 |       |
| NULL | UNION RESULT | <union1,2> | ALL   | NULL          | NULL    | NULL    | NULL  | NULL |       |
+----+--------------+------------+-------+---------------+---------+---------+-------+------+-------+
3 rows in set (0.34 sec)

2.DEPENDENT UNION与DEPENDENT SUBQUERY:

当union作为子查询时，其中第二个union的select_type就是DEPENDENT UNION。
第一个子查询的select_type则是DEPENDENT SUBQUERY。

mysql> explain select * from t_order where order_id in (select order_id from t_order where order_id=100 union select order_id from t_order where order_id=200);
+----+--------------------+------------+-------+---------------+---------+---------+-------+--------+-------------+
| id | select_type        | table      | type  | possible_keys | key     | key_len | ref   | rows   | Extra       |
+----+--------------------+------------+-------+---------------+---------+---------+-------+--------+-------------+
|  1 | PRIMARY            | t_order    | ALL   | NULL          | NULL    | NULL    | NULL  | 100453 | Using where |
|  2 | DEPENDENT SUBQUERY | t_order    | const | PRIMARY       | PRIMARY | 4       | const |      1 | Using index |
|  3 | DEPENDENT UNION    | t_order    | const | PRIMARY       | PRIMARY | 4       | const |      1 | Using index |
| NULL | UNION RESULT       | <union2,3> | ALL   | NULL          | NULL    | NULL    | NULL  |   NULL |             |
+----+--------------------+------------+-------+---------------+---------+---------+-------+--------+-------------+
4 rows in set (0.03 sec)

3.SUBQUERY:

子查询中的第一个select其select_type为SUBQUERY。

mysql> explain select * from t_order where order_id=(select order_id from t_order where order_id=100);
+----+-------------+---------+-------+---------------+---------+---------+-------+------+-------------+
| id | select_type | table   | type  | possible_keys | key     | key_len | ref   | rows | Extra       |
+----+-------------+---------+-------+---------------+---------+---------+-------+------+-------------+
|  1 | PRIMARY     | t_order | const | PRIMARY       | PRIMARY | 4       | const |    1 |             |
|  2 | SUBQUERY    | t_order | const | PRIMARY       | PRIMARY | 4       |       |    1 | Using index |
+----+-------------+---------+-------+---------------+---------+---------+-------+------+-------------+
2 rows in set (0.03 sec)

4.DERIVED:

当子查询是from子句时，其select_type为DERIVED。

mysql> explain select * from (select order_id from t_order where order_id=100) a;
+----+-------------+------------+--------+---------------+---------+---------+------+------+-------------+
| id | select_type | table      | type   | possible_keys | key     | key_len | ref  | rows | Extra       |
+----+-------------+------------+--------+---------------+---------+---------+------+------+-------------+
|  1 | PRIMARY     | <derived2> | system | NULL          | NULL    | NULL    | NULL |    1 |             |
|  2 | DERIVED     | t_order    | const  | PRIMARY       | PRIMARY | 4       |      |    1 | Using index |
+----+-------------+------------+--------+---------------+---------+---------+------+------+-------------+
2 rows in set (0.03 sec)

二.type的说明

1.system，const

见上面4.DERIVED的例子。其中第一行的type就是为system，第二行是const，这两种联接类型是最快的。

2.eq_ref

在t_order表中的order_id是主键，t_order_ext表中的order_id也是主键，该表可以认为是订单表的补充信息表，他们的关系是1对1，在下面的例子中可以看到b表的连接类型是eq_ref，这是极快的联接类型。

mysql> explain select * from t_order a,t_order_ext b where a.order_id=b.order_id;
+----+-------------+-------+--------+---------------+---------+---------+-----------------+------+-------------+
| id | select_type | table | type   | possible_keys | key     | key_len | ref             | rows | Extra       |
+----+-------------+-------+--------+---------------+---------+---------+-----------------+------+-------------+
|  1 | SIMPLE      | b     | ALL    | order_id      | NULL    | NULL    | NULL            |    1 |             |
|  1 | SIMPLE      | a     | eq_ref | PRIMARY       | PRIMARY | 4       | test.b.order_id |    1 | Using where |
+----+-------------+-------+--------+---------------+---------+---------+-----------------+------+-------------+
2 rows in set (0.00 sec)

3.ref

下面的例子在上面的例子上略作了修改，加上了条件。此时b表的联接类型变成了ref。因为所有与a表中order_id=100的匹配记录都将会从b表获取。这是比较常见的联接类型。

mysql> explain select * from t_order a,t_order_ext b where a.order_id=b.order_id and a.order_id=100;
+----+-------------+-------+-------+---------------+----------+---------+-------+------+-------+
| id | select_type | table | type  | possible_keys | key      | key_len | ref   | rows | Extra |
+----+-------------+-------+-------+---------------+----------+---------+-------+------+-------+
|  1 | SIMPLE      | a     | const | PRIMARY       | PRIMARY  | 4       | const |    1 |       |
|  1 | SIMPLE      | b     | ref   | order_id      | order_id | 4       | const |    1 |       |
+----+-------------+-------+-------+---------------+----------+---------+-------+------+-------+
2 rows in set (0.00 sec)

4.ref_or_null

user_id字段是一个可以为空的字段，并对该字段创建了一个索引。在下面的查询中可以看到联接类型为ref_or_null，这是mysql为含有null的字段专门做的处理。在我们的表设计中应当尽量避免索引字段为NULL，因为这会额外的耗费mysql的处理时间来做优化。

mysql> explain select * from t_order where user_id=100 or user_id is null;
+----+-------------+---------+-------------+---------------+---------+---------+-------+-------+-------------+
| id | select_type | table   | type        | possible_keys | key     | key_len | ref   | rows  | Extra       |
+----+-------------+---------+-------------+---------------+---------+---------+-------+-------+-------------+
|  1 | SIMPLE      | t_order | ref_or_null | user_id       | user_id | 5       | const | 50325 | Using where |
+----+-------------+---------+-------------+---------------+---------+---------+-------+-------+-------------+
1 row in set (0.00 sec)

5.index_merge

经常出现在使用一张表中的多个索引时。mysql会将多个索引合并在一起，如下例:

mysql> explain select * from t_order where order_id=100 or user_id=10;
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+-------------------------------------------+
| id | select_type | table   | type        | possible_keys   | key             | key_len | ref  | rows | Extra                                     |
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+-------------------------------------------+
|  1 | SIMPLE      | t_order | index_merge | PRIMARY,user_id | PRIMARY,user_id | 4,5     | NULL |    2 | Using union(PRIMARY,user_id); Using where |
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+-------------------------------------------+
1 row in set (0.09 sec)

6.unique_subquery

该联接类型用于替换value IN (SELECT primary_key FROM single_table WHERE some_expr)这样的子查询的ref。注意ref列，其中第二行显示的是func，表明unique_subquery是一个函数，而不是一个普通的ref。

mysql> explain select * from t_order where order_id in (select order_id from t_order where user_id=10);
+----+--------------------+---------+-----------------+-----------------+---------+---------+------+--------+-------------+
| id | select_type        | table   | type            | possible_keys   | key     | key_len | ref  | rows   | Extra       |
+----+--------------------+---------+-----------------+-----------------+---------+---------+------+--------+-------------+
|  1 | PRIMARY            | t_order | ALL             | NULL            | NULL    | NULL    | NULL | 100649 | Using where |
|  2 | DEPENDENT SUBQUERY | t_order | unique_subquery | PRIMARY,user_id | PRIMARY | 4       | func |      1 | Using where |
+----+--------------------+---------+-----------------+-----------------+---------+---------+------+--------+-------------+
2 rows in set (0.00 sec)

7.index_subquery

该联接类型与上面的太像了，唯一的差别就是子查询查的不是主键而是非唯一索引。

mysql> explain select * from t_order where user_id in (select user_id from t_order where order_id>10);
+----+--------------------+---------+----------------+-----------------+---------+---------+------+--------+--------------------------+
| id | select_type        | table   | type           | possible_keys   | key     | key_len | ref  | rows   | Extra                    |
+----+--------------------+---------+----------------+-----------------+---------+---------+------+--------+--------------------------+
|  1 | PRIMARY            | t_order | ALL            | NULL            | NULL    | NULL    | NULL | 100649 | Using where              |
|  2 | DEPENDENT SUBQUERY | t_order | index_subquery | PRIMARY,user_id | user_id | 5       | func |  50324 | Using index; Using where |
+----+--------------------+---------+----------------+-----------------+---------+---------+------+--------+--------------------------+
2 rows in set (0.00 sec)

8.range

按指定的范围进行检索，很常见。

mysql> explain select * from t_order where user_id in (100,200,300);
+----+-------------+---------+-------+---------------+---------+---------+------+------+-------------+
| id | select_type | table   | type  | possible_keys | key     | key_len | ref  | rows | Extra       |
+----+-------------+---------+-------+---------------+---------+---------+------+------+-------------+
|  1 | SIMPLE      | t_order | range | user_id       | user_id | 5       | NULL |    3 | Using where |
+----+-------------+---------+-------+---------------+---------+---------+------+------+-------------+
1 row in set (0.00 sec)

9.index

在进行统计时非常常见，此联接类型实际上会扫描索引树，仅比ALL快些。

mysql> explain select count(*) from t_order;
+----+-------------+---------+-------+---------------+---------+---------+------+--------+-------------+
| id | select_type | table   | type  | possible_keys | key     | key_len | ref  | rows   | Extra       |
+----+-------------+---------+-------+---------------+---------+---------+------+--------+-------------+
|  1 | SIMPLE      | t_order | index | NULL          | user_id | 5       | NULL | 100649 | Using index |
+----+-------------+---------+-------+---------------+---------+---------+------+--------+-------------+
1 row in set (0.00 sec)

10.ALL

完整的扫描全表，最慢的联接类型，尽可能的避免。

mysql> explain select * from t_order;
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
| id | select_type | table   | type | possible_keys | key  | key_len | ref  | rows   | Extra |
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
|  1 | SIMPLE      | t_order | ALL  | NULL          | NULL | NULL    | NULL | 100649 |       |
+----+-------------+---------+------+---------------+------+---------+------+--------+-------+
1 row in set (0.00 sec)

三.extra的说明

1.Distinct

MySQL发现第1个匹配行后,停止为当前的行组合搜索更多的行。对于此项没有找到合适的例子，求指点。

2.Not exists

因为b表中的order_id是主键，不可能为NULL，所以mysql在用a表的order_id扫描t_order表，并查找b表的行时，如果在b表发现一个匹配的行就不再继续扫描b了，因为b表中的order_id字段不可能为NULL。这样避免了对b表的多次扫描。

mysql> explain select count(1) from t_order a left join t_order_ext b on a.order_id=b.order_id where b.order_id is null;
+----+-------------+-------+-------+---------------+--------------+---------+-----------------+--------+--------------------------------------+
| id | select_type | table | type  | possible_keys | key          | key_len | ref             | rows   | Extra                                |
+----+-------------+-------+-------+---------------+--------------+---------+-----------------+--------+--------------------------------------+
|  1 | SIMPLE      | a     | index | NULL          | express_type | 1       | NULL            | 100395 | Using index                          |
|  1 | SIMPLE      | b     | ref   | order_id      | order_id     | 4       | test.a.order_id |      1 | Using where; Using index; Not exists |
+----+-------------+-------+-------+---------------+--------------+---------+-----------------+--------+--------------------------------------+
2 rows in set (0.01 sec)

3.Range checked for each record

这种情况是mysql没有发现好的索引可用，速度比没有索引要快得多。

mysql> explain select * from t_order t, t_order_ext s where s.order_id>=t.order_id and s.order_id<=t.order_id and t.express_type>5;
+----+-------------+-------+-------+----------------------+--------------+---------+------+------+------------------------------------------------+
| id | select_type | table | type  | possible_keys        | key          | key_len | ref  | rows | Extra                                          |
+----+-------------+-------+-------+----------------------+--------------+---------+------+------+------------------------------------------------+
|  1 | SIMPLE      | t     | range | PRIMARY,express_type | express_type | 1       | NULL |    1 | Using where                                    |
|  1 | SIMPLE      | s     | ALL   | order_id             | NULL         | NULL    | NULL |    1 | Range checked for each record (index map: 0x1) |
+----+-------------+-------+-------+----------------------+--------------+---------+------+------+------------------------------------------------+
2 rows in set (0.00 sec)

4.Using filesort

在有排序子句的情况下很常见的一种情况。此时mysql会根据联接类型浏览所有符合条件的记录，并保存排序关键字和行指针，然后排序关键字并按顺序检索行。

mysql> explain select * from t_order order by express_type;
+----+-------------+---------+------+---------------+------+---------+------+--------+----------------+
| id | select_type | table   | type | possible_keys | key  | key_len | ref  | rows   | Extra          |
+----+-------------+---------+------+---------------+------+---------+------+--------+----------------+
|  1 | SIMPLE      | t_order | ALL  | NULL          | NULL | NULL    | NULL | 100395 | Using filesort |
+----+-------------+---------+------+---------------+------+---------+------+--------+----------------+
1 row in set (0.00 sec)

5.Using index

这是性能很高的一种情况。当查询所需的数据可以直接从索引树中检索到时，就会出现。上面的例子中有很多这样的例子，不再多举例了。

6.Using temporary

发生这种情况一般都是需要进行优化的。mysql需要创建一张临时表用来处理此类查询。

mysql> explain select * from t_order a left join t_order_ext b on a.order_id=b.order_id group by b.order_id;
+----+-------------+-------+------+---------------+----------+---------+-----------------+--------+---------------------------------+
| id | select_type | table | type | possible_keys | key      | key_len | ref             | rows   | Extra                           |
+----+-------------+-------+------+---------------+----------+---------+-----------------+--------+---------------------------------+
|  1 | SIMPLE      | a     | ALL  | NULL          | NULL     | NULL    | NULL            | 100395 | Using temporary; Using filesort |
|  1 | SIMPLE      | b     | ref  | order_id      | order_id | 4       | test.a.order_id |      1 |                                 |
+----+-------------+-------+------+---------------+----------+---------+-----------------+--------+---------------------------------+
2 rows in set (0.00 sec)

7.Using where

当有where子句时，extra都会有说明。

8.Using sort_union(...)/Using union(...)/Using intersect(...)

下面的例子中user_id是一个检索范围，此时mysql会使用sort_union函数来进行索引的合并。而当user_id是一个固定值时，请参看上面type说明5.index_merge的例子，此时会使用union函数进行索引合并。

mysql> explain select * from t_order where order_id=100 or user_id>10;
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+------------------------------------------------+
| id | select_type | table   | type        | possible_keys   | key             | key_len | ref  | rows | Extra                                          |
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+------------------------------------------------+
|  1 | SIMPLE      | t_order | index_merge | PRIMARY,user_id | user_id,PRIMARY | 5,4     | NULL |    2 | Using sort_union(user_id,PRIMARY); Using where |
+----+-------------+---------+-------------+-----------------+-----------------+---------+------+------+------------------------------------------------+
1 row in set (0.00 sec)

对于Using intersect的例子可以参看下例，user_id与express_type发生了索引交叉合并。

mysql> explain select * from t_order where express_type=1 and user_id=100;
+----+-------------+---------+-------------+----------------------+----------------------+---------+------+------+----------------------------------------------------+
| id | select_type | table   | type        | possible_keys        | key                  | key_len | ref  | rows | Extra                                              |
+----+-------------+---------+-------------+----------------------+----------------------+---------+------+------+----------------------------------------------------+
|  1 | SIMPLE      | t_order | index_merge | user_id,express_type | user_id,express_type | 5,1     | NULL |    1 | Using intersect(user_id,express_type); Using where |
+----+-------------+---------+-------------+----------------------+----------------------+---------+------+------+----------------------------------------------------+
1 row in set (0.00 sec)

9.Using index for group-by

表明可以在索引中找到分组所需的所有数据，不需要查询实际的表。

mysql> explain select user_id from t_order group by user_id;
+----+-------------+---------+-------+---------------+---------+---------+------+------+--------------------------+
| id | select_type | table   | type  | possible_keys | key     | key_len | ref  | rows | Extra                    |
+----+-------------+---------+-------+---------------+---------+---------+------+------+--------------------------+
|  1 | SIMPLE      | t_order | range | NULL          | user_id | 5       | NULL |    3 | Using index for group-by |
+----+-------------+---------+-------+---------------+---------+---------+------+------+--------------------------+
1 row in set (0.00 sec)

除了上面的三个说明，还需要注意rows的数值，多行之间的数值是乘积的关系，可以估算大概要处理的行数，如果乘积很大，那就很有优化的必要了。