mysql中的count(primary_key)、count(1)、count(*)的区别

表结构如下：

mysql> show create table user\G;
*************************** 1. row ***************************
       Table: user
Create Table: CREATE TABLE `user` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `name` varchar(50) NOT NULL,
  `pwd` varchar(50) NOT NULL,
  `email` varchar(100) NOT NULL,
  `phone` varchar(30) NOT NULL,
  `sex` enum('F','M','N') NOT NULL DEFAULT 'N',
  `addres` varchar(100) NOT NULL,
  `tag` varchar(100) NOT NULL,
  PRIMARY KEY (`id`),
  KEY `name` (`name`)
) ENGINE=InnoDB AUTO_INCREMENT=5000003 DEFAULT CHARSET=utf8 COMMENT='用户表'
1 row in set (0.00 sec)

下面做一下explain： 1、count(id)

mysql> select count(id) from user;
+-----------+
| count(id) |
+-----------+
|   5000002 |
+-----------+
1 row in set (1.93 sec)

mysql> explain select count(id) from user;
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+

| id | select_type | table | type  | possible_keys | key  | key_len | ref  | rows    | Extra       |

+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | index | NULL          | name | 152     | NULL | 4998401 | Using index |

+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+

1 row in set (0.05 sec)

2、count(1)

mysql> select count(1) from user;
+----------+
| count(1) |
+----------+
|  5000002 |
+----------+
1 row in set (0.90 sec)

mysql> explain select count(1) from user;
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
| id | select_type | table | type  | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | index | NULL          | name | 152     | NULL | 4998401 | Using index |
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

3、count(*)

mysql> select count(*) from user;
+----------+
| count(*) |
+----------+
|  5000002 |
+----------+
1 row in set (0.87 sec)

mysql> explain select count(*) from user;
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
| id | select_type | table | type  | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | index | NULL          | name | 152     | NULL | 4998401 | Using index |
+----+-------------+-------+-------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

比较三个查询，explain的结果一模一样，这说明这三个的效率是一样的吗？

再看看下面三个操作，带上where条件 sex='F'，以下三个操作中间均会重启mysql服务。

1、count(id)

mysql> select count(id) from user where sex='F';
+-----------+
| count(id) |
+-----------+
|   1681259 |
+-----------+
1 row in set (18.87 sec)
mysql> explain select count(id) from user where sex='F';
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | ALL  | NULL          | NULL | NULL    | NULL | 4998401 | Using where |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

2、count(1)

mysql> select count(1) from user where sex='F';
+----------+
| count(1) |
+----------+
|  1681259 |
+----------+
1 row in set (4.81 sec)
mysql> explain select count(1) from user where sex='F';
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | ALL  | NULL          | NULL | NULL    | NULL | 4998401 | Using where |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

3、count(*)

mysql> select count(*) from user where sex='F';
+----------+
| count(*) |
+----------+
|  1681259 |
+----------+
1 row in set (4.69 sec)
mysql> explain select count(*) from user where sex='F';
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | user  | ALL  | NULL          | NULL | NULL    | NULL | 4998401 | Using where |
+----+-------------+-------+------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

以上三种查询有一些差别，其中count(id)用时最长，count(*)比count(1)速度要稍微快一点。

两组查询，带条件的都没有使用到索引，扫描了全表；而没有条件的则使用了索引name。

所以在应用中尽量不使用count(*)和count(1)，杜绝使用count(primary_key)。

网上有很多资料说

没有主键，count(1)比count(*)快；

有主键的话，count(primary_key)最快，但是在上面的测试中发现，count(primary_key)是最慢的，难道是测试不准确？这个有待验证。

如果表只有一个字段，则count(*)是最快的。

说明：

count(1)中的1并不是指第一个column；

count(*)和count(1)一样，包括对值为NULL的统计；

count(column)不包括对值为NULL的统计，这里的column指的不是primary_key；

本文出自 寄凡，http://www.hblpf.com/?post=55