在数据库设计的时候会遇到很多只需要0、1、2这种固定几个值的状态字段，基本上都建议设置为只占一字节的tinyint类型，有些觉得char(1)是一样，毕竟char(1)存储数字和字母时一个字符也只是占一个字节。
mysql是用c++写的，而在c++中字符类型是存放对应ascii码的二进制到存储空间，而整型数字是直接存数字的二进制，虽然最终都是二进制存储，但是环节上有少许不同，同样在msyql查找时也会有所不同，下图摘自小白版c++教程《c++ primer plus》:
今天对tinyint和char(1)做了个简单测试，分表建两个表t1、t2，结构如下：
mysql> show create table t1\G
*************************** 1. row ***************************
Table: t1
Create Table: CREATE TABLE `t1` (
`_id` int(11) NOT NULL AUTO_INCREMENT,
`id` tinyint(4) DEFAULT NULL,
`title` text,
PRIMARY KEY (`_id`),
KEY `id` (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=2400096 DEFAULT CHARSET=utf8
1 row in set (0.00 sec)
mysql> show create table t2\G
*************************** 1. row ***************************
Table: t2
Create Table: CREATE TABLE `t2` (
`_id` int(11) NOT NULL AUTO_INCREMENT,
`id` char(1) DEFAULT NULL,
`title` text,
PRIMARY KEY (`_id`),
KEY `id` (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=2400096 DEFAULT CHARSET=utf8
1 row in set (0.00 sec)
两个表唯一不同为id字段类型，总数据量都为2400096，id只有0、1、2三个，为了两个表的数据一样且磁盘上分布也一样，降低IO对测试的影响，分别加载的数据如下：
mysql> select id,count(*) from t1 group by id;
+------+----------+
| id | count(*) |
+------+----------+
| 0 | 1199998 |
| 1 | 1199998 |
| 2 | 99 |
+------+----------+
3 rows in set (0.55 sec)
mysql> select id,count(*) from t2 group by id;
+------+----------+
| id | count(*) |
+------+----------+
| 0 | 1199998 |
| 1 | 1199998 |
| 2 | 99 |
+------+----------+
3 rows in set (0.77 sec)
查看执行计划：
mysql> explain select _id from test.t2 where id='1';
+----+-------------+-------+------+---------------+------+---------+-------+---------+--------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+------+---------------+------+---------+-------+---------+--------------------------+
| 1 | SIMPLE | t2 | ref | id | id | 4 | const | 1170900 | Using where; Using index |
+----+-------------+-------+------+---------------+------+---------+-------+---------+--------------------------+
1 row in set (0.00 sec)
mysql> explain select _id from test.t1 where id=1;
+----+-------------+-------+------+---------------+------+---------+-------+---------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+------+---------------+------+---------+-------+---------+-------------+
| 1 | SIMPLE | t1 | ref | id | id | 2 | const | 1170601 | Using index |
+----+-------------+-------+------+---------------+------+---------+-------+---------+-------------+
1 row in set (0.00 sec)
两个表都使用了id索引，再看看information_schema.tables的信息是否和之前理解的存储字节大小是否有出入：
mysql> select DATA_LENGTH/1024/1024,INDEX_LENGTH/1024/1024,data_free from tables where table_name in ('t1','t2');
+-----------------------+------------------------+-----------+
| DATA_LENGTH/1024/1024 | INDEX_LENGTH/1024/1024 | data_free |
+-----------------------+------------------------+-----------+
| 310.81250000 | 27.56250000 | 0 |
| 313.81250000 | 29.56250000 | 0 |
+-----------------------+------------------------+-----------+
2 rows in set (0.00 sec)
两个表大小相差不多，确认char(1)和tinyint占字节数相同，现在直接看执行时间：
mysql> show profiles;
+----------+------------+---------------------------------------------------------------+
| Query_ID | Duration | Query |
+----------+------------+---------------------------------------------------------------+
| 1 | 0.60804275 | select count(*) from (select _id from test.t1 where id=1) a |
| 2 | 0.59277575 | select count(*) from (select _id from test.t1 where id=1) a |
| 3 | 0.60398000 | select count(*) from (select _id from test.t1 where id=1) a |
| 4 | 0.69068025 | select count(*) from (select _id from test.t2 where id='1') a |
| 5 | 0.69654200 | select count(*) from (select _id from test.t2 where id='1') a |
| 6 | 0.67788800 | select count(*) from (select _id from test.t2 where id='1') a |
+----------+------------+---------------------------------------------------------------+
这样就很明显可以看出为char(1)字段的t2表查询时消耗时间偏多，如果几条几百条的情况根本看不出char(1)和tinyint的差别，毕竟现在CPU的效率是非常高的，这里测试的利用了id=1的数据，有1199998条，这样就可以看出点差别了！！虽然效率差别不是很大，为了生产环境统一以及提升QPS还是使用短小的整型更好.
转载自：http://blog.51cto.com/xiaozhong991/1892569
