关于Grouping, Rollup,cube,

--准备数据源
select a.StateCode, a.DepCode,a.SendMoney
  into #tmp 
  from 
		(Select '100001' as StateCode ,'310001' as DepCode ,1000 as SendMoney 
		 UNION ALL 
		 Select '100001' as StateCode,'310001' as DepCode,2000  as SendMoney
		 UNION ALL 
		 Select '100001' as StateCode,'310002' as DepCode,1500  as SendMoney
		 UNION ALL 
		 Select '100002' as StateCode,'320001' as DepCode,3000  as SendMoney
		 UNION ALL 
		 Select '100002' as StateCode,'320001' as DepCode,1200  as SendMoney
		 UNION ALL 
		 Select '100003' as StateCode,'330001' as DepCode,1800  as SendMoney
		 UNION ALL 
		 Select '100003' as StateCode,'330002' as DepCode,2100  as SendMoney
		 UNION ALL 
		 Select '100004' as StateCode,'340001' as DepCode,2500  as SendMoney ) a

-- grouping-rollup 运行结果
Select CASE WHEN GROUPING(StateCode)=1 THEN 'Total:'       ELSE StateCode END as StateCode 
      ,CASE WHEN GROUPING(DepCode)=1   THEN 'State Total:' ELSE DepCode   END as DepCode 
      ,Sum(SendMoney) AS SendMoney 
  From #tmp 
 GROUP BY ROLLUP(StateCode,DepCode) 

 

with cube,with rollup和grouping

http://www.jb51.net/article/35103.htm

关于with cube ,with rollup 和 grouping

通过查看sql 2005的帮助文档找到了CUBE 和 ROLLUP 之间的具体区别

CUBE 生成的结果集显示了所选列中值的所有组合的聚合。ROLLUP 生成的结果集显示了所选列中值的某一层次结构的聚合。

再看看对grouping的解释:

当行由 CUBE 或 ROLLUP 运算符添加时,该函数将导致附加列的输出值为 1;当行不由 CUBE 或 ROLLUP 运算符添加时,该函数将导致附加列的输出值为 0。

仅在与包含 CUBE 或 ROLLUP 运算符的 GROUP BY 子句相关联的选择列表中才允许分组。

当看到以上的解释肯定非常的模糊,不知所云和不知道该怎样用,下面通过实例操作来体验一下:

先建表(dbo.PeopleInfo):

复制代码代码如下:

CREATE TABLE [dbo].[PeopleInfo]( 
[id] [int] IDENTITY(1,1) NOT NULL, 
[name] [nchar](10) COLLATE Chinese_PRC_CI_AS NULL, 
[numb] [nchar](10) COLLATE Chinese_PRC_CI_AS NOT NULL, 
[phone] [nchar](10) COLLATE Chinese_PRC_CI_AS NULL, 
[FenShu] [int] NULL 
) ON [PRIMARY]


向表插入数据:

复制代码代码如下:

insert into peopleinfo([name],numb,phone,fenshu) values ('李欢','3223','1365255',80) 
insert into peopleinfo([name],numb,phone,fenshu) values ('李欢','322123','1',90) 
insert into peopleinfo([name],numb,phone,fenshu) values ('李名','3213112352','13152',56) 
insert into peopleinfo([name],numb,phone,fenshu) values ('李名','32132312','13342563',60) 
insert into peopleinfo([name],numb,phone,fenshu) values ('王华','3223','1365255',80)


查询出插入的全部数据:

复制代码代码如下:

select * from dbo.PeopleInfo


结果如图:

 

操作一:先试试:1, 查询所有数据;2,用group by 查询所有数据;3,用with cube。这三种情况的比较

SQL语句如下:

复制代码代码如下:

select * from dbo.PeopleInfo --1, 查询所有数据; 
select [name],numb,sum(fenshu) from dbo.PeopleInfo group by [name],numb --2,用group by 查询所有数据; 
select [name],numb,sum(fenshu) from dbo.PeopleInfo group by [name],numb with cube --3,用with cube。这三种情况的比较


结果如图:

 

结果分析:

  用第三种(用with cube)为什么会多出来有null的字段值呢?通过分析图上的值得组合会发现是怎么回事儿了,以第三条数据(李欢,null,170)为例:它只是把姓名是【李欢】的分为了一组,而没有考虑【numb】,所以有多出来了第三条数据,也说明了170是怎么来的。其他的也是这样。再回顾一下帮助文档的解释:CUBE 生成的结果集显示了所选列中值的所有组合的聚合, 发现明了了许多。

操作二:1,用with cube;2,用with rollup 这两种情况的比较

SQL语句如下:

复制代码代码如下:

select [name],numb,sum(fenshu) from dbo.PeopleInfo group by [name],numb with cube --用with cube。 
select [name],numb,sum(fenshu) from dbo.PeopleInfo group by [name],numb with rollup --用with rollup。


结果如图:

 

结果分析:

  为什么with cube 比 with rollup多出来一部分呢?原来它没有显示,以【numb】分组而不考虑【name】的数据情况。再回顾一下帮助文档的解释:ROLLUP 生成的结果集显示了所选列中值的某一层次结构的聚合,那这个【某一层次】又是以什么为标准的呢?我的猜想是:距离group up最近的字段必须考虑在分组内。

证明猜想实例:

操作:用两个group up 交换字段位置的sql语句和一个在group up 后面增加一个字段的sql语句进行比较:

SQL语句如下:

复制代码代码如下:

select [name],numb from dbo.PeopleInfo group by [name],numb with rollup 
select [name],numb from dbo.PeopleInfo group by numb,[name] with rollup 
select [name],numb,phone from dbo.PeopleInfo group by [name],numb,phone with rollup


结果如图:

 

通过结果图的比较发现猜想是正确的。

---------------------------------------------------grouping-------------------------------------------------

现在来看看grouping的实例:

SQL语句看看与with rollup的结合(与with cube的结合是一样的):

复制代码代码如下:

select [name],numb,grouping(numb) from dbo.PeopleInfo group by [name],numb with rollup


结果如图:

 

结果分析:

结合帮助文档的解释:当行由 CUBE 或 ROLLUP 运算符添加时,该函数将导致附加列的输出值为 1;当行不由 CUBE 或 ROLLUP 运算符添加时,该函数将导致附加列的输出值为 0。  很容易理解再此就不多解释了。

 

 

ROLLUP和CUBE关键字,都是用来为GROUP BY语句返回的结果添加汇总信息,也可以说,它们是对分组结果进行二次分组

SELECT    
    DEPT   AS 部门,  
    SEX    AS 性别,  
    AVG(SALARY) AS 平均工资  
FROM    
(    
    --姓名  性别  部门  工资    
    VALUES    
    ('张三','','市场部',4000),  
    ('赵红','','技术部',2000),  
    ('李四','','市场部',5000),  
    ('李白','','技术部',5000),  
    ('王五','','市场部',3000),  
    ('王蓝','','技术部',4000)  
) AS EMPLOY(NAME,SEX,DEPT,SALARY)  
GROUP BY DEPT,SEX
--GROUP BY ROLLUP(DEPT,SEX)  
ORDER BY 部门,性别  

SELECT    
    DEPT   AS 部门,  
    SEX    AS 性别,  
    AVG(SALARY) AS 平均工资  
FROM    
(    
    --姓名  性别  部门  工资    
    VALUES    
    ('张三','','市场部',4000),  
    ('赵红','','技术部',2000),  
    ('李四','','市场部',5000),  
    ('李白','','技术部',5000),  
    ('王五','','市场部',3000),  
    ('王蓝','','技术部',4000)  
) AS EMPLOY(NAME,SEX,DEPT,SALARY)  
GROUP BY ROLLUP(DEPT,SEX)  
ORDER BY 部门,性别  

 

SELECT    
   DEPT   AS 部门,  
    SEX    AS 性别,      
    AVG(SALARY) AS 平均工资  
FROM    
(    
    --姓名  性别  部门  工资    
    VALUES    
    ('张三','','市场部',4000),  
    ('赵红','','技术部',2000),  
    ('李四','','市场部',5000),  
    ('李白','','技术部',5000),  
    ('王五','','市场部',3000),  
    ('王蓝','','技术部',4000)  
) AS EMPLOY(NAME,SEX,DEPT,SALARY)  
GROUP BY CUBE(DEPT,SEX)  
ORDER BY 部门,性别  

 

 

 http://lxw1234.com/archives/2015/04/193.htm

GROUPING SETS

  1. GROUP BY GROUPING SETS (A,B,C)  等价与  GROUP BY A  
  2.                                         UNION ALL  
  3.                                         GROUP BY B  
  4.                                         UNION ALL  
  5.                                         GROUP BY C  
posted @ 2012-06-07 16:32  streetpasser  阅读(245)  评论(0编辑  收藏  举报