通过GROUP BY grouping sets提升数据分组统计效率
使用 GROUPING SETS 的 GROUP BY 子句可以生成一个等效于由多个简单 GROUP BY 子句的 UNION ALL 生成的结果集
示例:(sql server 2008 R2)
创建测试数据库及表并插入测试数据
use master
CREATE DATABASE db_sales
go
use db_sales
go
CREATE TABLE [dbo].[tb_sale](
[id] [int] IDENTITY(1,1) NOT NULL,
[server] [nvarchar](50) NULL,
[pname] [nvarchar](50) NULL,
[pinpai] [nvarchar](50) NULL,
[dates] [smalldatetime] NULL,
[cnt] [int] NULL
) ON [PRIMARY]
go
INSERT INTO [db_ControlManager_ft].[dbo].[test]([server],[pname],[pinpai],[dates],[cnt])
VALUES('A','computer','hp','2012-01-01',1),
('A','computer','hp','2012-01-02',3),
('A','computer','hp','2012-01-03',5),
('A','computer','hp','2012-01-04',1),
('A','computer','hp','2012-01-05',3),
('A','computer','hp','2012-01-06',5),
('A','computer','dell','2012-01-01',2),
('A','computer','dell','2012-01-02',4),
('A','computer','dell','2012-01-03',6),
('A','computer','dell','2012-01-04',7),
('A','computer','dell','2012-01-05',2),
('A','computer','dell','2012-01-06',4),
('B','computer','hp','2012-01-01',3),
('B','computer','hp','2012-01-02',3),
('B','computer','hp','2012-01-03',3),
('B','computer','hp','2012-01-04',3),
('B','computer','hp','2012-01-05',3),
('B','computer','hp','2012-01-06',2),
('B','computer','dell','2012-01-01',2),
('B','computer','dell','2012-01-02',2),
('B','computer','dell','2012-01-03',2),
('B','computer','dell','2012-01-04',2),
('B','computer','dell','2012-01-05',1),
('B','computer','dell','2012-01-06',1),
('A','TV','hp','2012-01-01',1),
('A','TV','hp','2012-01-02',3),
('A','TV','hp','2012-01-03',5),
('A','TV','hp','2012-01-04',1),
('A','TV','hp','2012-01-05',3),
('A','TV','hp','2012-01-06',5),
('A','TV','dell','2012-01-01',2),
('A','TV','dell','2012-01-02',4),
('A','TV','dell','2012-01-03',6),
('A','TV','dell','2012-01-04',7),
('A','TV','dell','2012-01-05',2),
('A','TV','dell','2012-01-06',4),
('B','TV','hp','2012-01-01',3),
('B','TV','hp','2012-01-02',3),
('B','TV','hp','2012-01-03',3),
('B','TV','hp','2012-01-04',3),
('B','TV','hp','2012-01-05',3),
('B','TV','hp','2012-01-06',2),
('B','TV','dell','2012-01-01',2),
('B','TV','dell','2012-01-02',2),
('B','TV','dell','2012-01-03',2),
('B','TV','dell','2012-01-04',2),
('B','TV','dell','2012-01-05',1),
('B','TV','dell','2012-01-06',1)
go
现在要求分别计算出每天的销量,总销量,每个销售员的总销量,每个产品的总销量,每个品牌的总销量,及每个销售员按品牌的产品销量
使用普通的UNION ALL语句
select N'总销量',null,null,null,null,SUM(cnt) from tb_sale union all select N'每日销量',null,null,null,dates,SUM(cnt) from tb_sale group by dates union all select N'按人员总销量',[server],null,null,null,SUM(cnt) from tb_sale group by [server] union all select N'按品牌总销量',null,null,pinpai,null,SUM(cnt) from tb_sale group by pinpai union all select N'按产品总销量',null,pname,null,null,SUM(cnt) from tb_sale group by pname union all select N'按产品及品牌总销量',null,pname,pinpai,null,SUM(cnt) from tb_sale group by pname,pinpai union all select N'按人员产品及品牌总销量',[server],pname,pinpai,null,SUM(cnt) from tb_sale group by [server],pname,pinpai
使用grouping sets 和grouping_id实现
select [server] as 销售人员 , pname as 产品名称, pinpai as 品牌 , dates as 销售时间, SUM(cnt) as 销售数量, (case when GROUPING_ID([server],pname,pinpai,dates)=15 then N'总销量' when GROUPING_ID([server],pname,pinpai,dates)=14 then N'每日销量' when GROUPING_ID([server],pname,pinpai,dates)=13 then N'按品牌总销量' when GROUPING_ID([server],pname,pinpai,dates)=11 then N'按产品总销量' when GROUPING_ID([server],pname,pinpai,dates)=9 then N'按产品及品牌总销量' when GROUPING_ID([server],pname,pinpai,dates)=7 then N'按人员总销量' when GROUPING_ID([server],pname,pinpai,dates)=1 then N'按人员产品及品牌总销量' end ) as 项目说明 from tb_sale group by GROUPING sets(([server],pname,pinpai),(pname,pinpai),[server],pname,pinpai,dates,()) order by [server],pname,pinpai,dates
使用union操作会增加IO开销,会减少cpu和内存的开销,使用grouping sets会减少IO开销,会增加cpu和内存的消耗.
GROUPING SETS在遇到多个条件时，聚合是一次性从数据库中取出所有需要操作的数据,在内存中对数据库进行聚合操作并生成结果。而UNION ALL是多次扫描表，将返回的结果进行UNION操作.这也就是为什么GROUPING SETS和UNION操作所返回的数据顺序是不同的.
grouping sets的执行方式在group by后面有多列的时候,grouping sets带来的性能提升非常明显
原文链接： http://blog.csdn.net/vince6799/article/details/7321697