重复记录只取其一

重复记录只取其一

左直拳

有时侯会遇到需要过滤重复记录的情况。

所谓的重复记录，不一定就是完全相同的记录，而更多是指一部分字段内容相同的记录。

比如说，有一个表记录访问日志（LOG），结构如下：

要求统计页面的访问IP数量，相同的IP，日期相同，算一个IP；日期不同，就算不同的IP。

无论是用group by date,ip 还是 group by ip,date，都不正确。

最后是这样写：

SELECT COUNT（1）AS IpNum FROM LOG AS a

WHERE NOT EXISTS(

SELECT 1 FROM LOG WHERE IP=a.IP AND Date=a.Date AND Id<a.Id

)

关键之处在于 Id<a.Id，就是说IP相同、日期相同的记录中，我只拿第一条，就是ID最小的那一条。

有三千人溺水，我只挑其中一个来嫖。

与此类似的是删除重复记录。

如果所谓的重复记录，是指除了流水号不同，其他字段都相同，则好办。如上表，

DELETE FROM log AS a WHERE EXISTS(

SELECT 1 FROM log WHERE  IP=a.IP AND Date=a.Date AND Id<a.Id)

如果全部字段都相同，怎么办？

貌似应该用游标：

use [test]
go

IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[a_dist]') AND type in (N'U'))
DROP TABLE [dbo].[a_dist]
GO

create table a_dist(id int,name varchar(20))  

insert into a_dist values(1,'abc') 
insert into a_dist values(1,'abc') 
insert into a_dist values(1,'abc') 
insert into a_dist values(1,'abc')

insert into a_dist values(2,'abc') 
insert into a_dist values(2,'abc') 
insert into a_dist values(2,'abc') 
insert into a_dist values(2,'abc')

insert into a_dist values(2,'abcd') 
insert into a_dist values(2,'abcd') 
insert into a_dist values(2,'abcd') 
insert into a_dist values(2,'abcd')

SELECT  * FROM a_dist;

DECLARE curT CURSOR FOR  SELECT id,name,COUNT(*) AS num FROM a_dist GROUP BY id,name HAVING COUNT(*)>1;
DECLARE @Id INT,@name VARCHAR(20),@top INT;
OPEN curT;
FETCH NEXT FROM curT INTO @Id,@name,@top;
WHILE @@FETCH_STATUS = 0
BEGIN
    SET @top = @top - 1;
    DELETE TOP(@top) FROM [a_dist] WHERE [id]=@Id AND [name]=@name;
    FETCH NEXT FROM curT INTO @Id,@name,@top;
END
CLOSE curT;
DEALLOCATE curT;

SELECT  * FROM a_dist;