在实际应用中，很可能会碰到一些需要删除某些字段的重复记录，我现在把我能想到的写下来，望高手们补充。

1、
具体实现如下：
Table         Create Table                                           
------------  --------------------------------------------------------
users_groups  CREATE TABLE `users_groups` (                          
                `id` int(10) unsigned NOT NULL AUTO_INCREMENT,       
                `uid` int(11) NOT NULL,                              
                `gid` int(11) NOT NULL,                              
                PRIMARY KEY (`id`)                                   
              ) ENGINE=InnoDB AUTO_INCREMENT=15 DEFAULT CHARSET=utf8

users_groups.txt内容：
1,11,502
2,107,502
3,100,503
4,110,501
5,112,501
6,104,502
7,100,502
8,100,501
9,102,501
10,104,502
11,100,502
12,100,501
13,102,501
14,110,501
mysql> load data infile 'c://users_groups.txt' into table users_groups fields
rminated by ',' lines terminated by '/n';
Query OK, 14 rows affected (0.05 sec)
Records: 14  Deleted: 0  Skipped: 0  Warnings: 0

mysql> select * from users_groups;

query result(14 records)

14 rows in set (0.00 sec)

mysql> create temporary table tmp_wrap select * from users_groups group by uid having count(1) > 1 union all
select * from users_groups group by uid having count(1) = 1;
Query OK, 7 rows affected (0.11 sec)
Records: 7  Duplicates: 0  Warnings: 0

mysql> truncate table users_groups;
Query OK, 14 rows affected (0.03 sec)

mysql> insert into users_groups select * from tmp_wrap;
Query OK, 7 rows affected (0.03 sec)
Records: 7  Duplicates: 0  Warnings: 0

mysql> select * from users_groups;

query result(7 records)

mysql> drop table tmp_wrap;
Query OK, 0 rows affected (0.05 sec)

2、还有一个很精简的办法。

查找重复的，并且除掉最小的那个。

delete users_groups as a from users_groups as a,
(
select *,min(id) from users_groups group by uid having count(1) > 1
) as b
 where a.uid = b.uid and a.id > b.id;

(7 row(s)affected)
(0 ms taken)

query result(7 records)

3、现在来看一下这两个办法的效率。

运行一下以下SQL 语句

create index f_uid on users_groups(uid);
explain select * from users_groups group by uid having count(1) > 1 union all
select * from users_groups group by uid having count(1) = 1;

explain select * from  users_groups as a,
(
select *,min(id) from users_groups group by uid having count(1) > 1
) as b
 where a.uid = b.uid and a.id > b.id;

query result(3 records)

query result(3 records)

很明显的第二个比第一个扫描的函数要少。