本文转载自：http://blog.diginfos.com/index.php?r=article/view&id=126

此处以某消费记录表(consume_record)为例，SQL语句如下：

DELETE consume_record
FROM
    consume_record, 
    (
        SELECT
            min(id) id,
            user_id,
            monetary,
            consume_time
        FROM
            consume_record
        GROUP BY
            user_id,
            monetary,
            consume_time
        HAVING
            count(*) > 1
    ) t2
WHERE
    consume_record.user_id = t2.user_id 
    and consume_record.monetary = t2.monetary
    and consume_record.consume_time  = t2.consume_time
AND consume_record.id > t2.id;

SQL语句分析：

1、查询出重复记录形成一个集合（临时表t2），集合里是每种重复记录的最小ID

(SELECT min(id) id, user_id, monetary, consume_time FROM consume_record GROUP BY user_id, monetary, consume_time HAVING count(*) > 1 ) t2

2、关联<判断重复基准的字段

consume_record.user_id = t2.user_id and consume_record.monetary = t2.monetary and consume_record.consume_time = t2.consume_time

3、根据条件，删除原表中id大于t2中id的记录

DELETE consume_record FROM ... WHERE ... AND consume_record.id > t2.id;

测试效果：
图一为删除前总记录数45541，图二为删除操作、从45541条记录中删除2800条重复记录用时0.09秒，图三为删除后总记录数。贴上测试表，如有需要的小伙伴，下载导入即可进行测试。consume_record.sql

如下语句，用于SQL server对AccountEmail账号信息去重：

DELETE [FSDBtemp].[dbo].[CusUsers]
FROM [FSDBtemp].[dbo].[CusUsers], 
    (
SELECT
            min(cuid) cuid,
            [AccountEmail]
        FROM
            [FSDBtemp].[dbo].[CusUsers]
        GROUP BY
                        [AccountEmail]
        HAVING
            count(*) > 1
    ) t2
WHERE
    [FSDBtemp].[dbo].[CusUsers].AccountEmail = t2.AccountEmail 
AND [FSDBtemp].[dbo].[CusUsers].cuid > t2.cuid