窗口函数小结

样本数据库下载：

http://files.cnblogs.com/files/haseo/TSQL2012.rar

1.窗口聚合函数

除了多个OVER子句其实和分组聚合函数一样。

SELECT  custid ,
        orderid ,
        val ,
        SUM(val) OVER ( PARTITION BY custid ) AS custtotal ,
        SUM(val) OVER ( ) AS grandtotal
FROM    Sales.OrderValues;

 

custid	orderid	val	custtotal	grandtotal
1	10643	814.5	4273	1265793
1	10692	878	4273	1265793
1	10702	330	4273	1265793
1	10835	845.8	4273	1265793
1	10952	471.2	4273	1265793
1	11011	933.5	4273	1265793
2	10926	514.4	1402.95	1265793
2	10759	320	1402.95	1265793
2	10625	479.75	1402.95	1265793
2	10308	88.8	1402.95	1265793
3	10365	403.2	7023.98	1265793
3	10507	749.06	7023.98	1265793

 

SELECT  custid ,
        orderid ,
        val ,
        CAST(100.0 * val / SUM(val) OVER ( PARTITION BY custid ) AS NUMERIC(5,2)) AS pctcust ,
        CAST(100.0 * val / SUM(val) OVER ( ) AS NUMERIC(5, 2)) AS pcttotal
FROM    Sales.OrderValues;

custid	orderid	val	pctcust	pcttotal
1	10643	814.5	19.06	0.06
1	10692	878	20.55	0.07
1	10702	330	7.72	0.03
1	10835	845.8	19.79	0.07
1	10952	471.2	11.03	0.04
1	11011	933.5	21.85	0.07
2	10926	514.4	36.67	0.04
2	10759	320	22.81	0.03
2	10625	479.75	34.2	0.04
2	10308	88.8	6.33	0.01
3	10365	403.2	5.74	0.03
3	10507	749.06	10.66	0.06

框架限定（必须要排序）

UNBOUNDED PRECEDING   统计之前所有记录 

<n> ROWS PRECEDING 统计前面几行

CURRENT ROW  当前行

UNBOUNDED FOLLOWING 本组下一行

SELECT  custid ,
        orderid ,
        orderdate ,
        val ,
        SUM(val) OVER ( PARTITION BY custid ORDER BY orderdate, orderid
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW  --可以直接简写为 ROWS UNBOUNDED PRECEDING
--ROWS  BETWEEN 2 PRECEDING AND CURRENT ROW  
) AS runningtotal
FROM    Sales.OrderValues;

custid	orderid	orderdate	val	runningtotal
1	10643	2007/8/25	814.5	814.5
1	10692	2007/10/3	878	1692.5
1	10702	2007/10/13	330	2022.5
1	10835	2008/1/15	845.8	2868.3
1	10952	2008/3/16	471.2	3339.5
1	11011	2008/4/9	933.5	4273
2	10308	2006/9/18	88.8	88.8
2	10625	2007/8/8	479.75	568.55
2	10759	2007/11/28	320	888.55
2	10926	2008/3/4	514.4	1402.95
3	10365	2006/11/27	403.2	403.2
3	10507	2007/4/15	749.06	1152.26
3	10535	2007/5/13	1940.85	3093.11
3	10573	2007/6/19	2082	5175.11

 2.排名函数

SELECT  custid ,
        orderid ,
        val ,
        ROW_NUMBER() OVER ( ORDER BY val ) AS rownum ,
        RANK() OVER ( ORDER BY val ) AS rnk , 
        DENSE_RANK() OVER ( ORDER BY val ) AS densernk ,
        NTILE(100) OVER ( ORDER BY val ) AS ntile100 --分割结果，列子中分了一百组
FROM    Sales.OrderValues;

custid	orderid	val	rownum	rnk	densernk	ntile100
12	10782	12.5	1	1	1	1
27	10807	18.4	2	2	2	1
66	10586	23.8	3	3	3	1
76	10767	28	4	4	4	1
54	10898	30	5	5	5	1
88	10900	33.75	6	6	6	1
48	10883	36	7	7	7	1
41	11051	36	8	7	7	1
71	10815	40	9	9	8	1
38	10674	45	10	10	9	2
53	11057	45	11	10	9	2
75	10271	48	12	12	10	2

3.偏移函数 

SELECT  custid ,
        orderid ,
        orderdate ,
        val ,
        LAG(val) OVER ( PARTITION BY custid ORDER BY orderdate, orderid ) AS prev_val ,--显示前一行
        LAG(val, 3) OVER ( PARTITION BY custid ORDER BY orderdate, orderid ) AS prev_val1 ,--往前显示第三行
        LAG(val, 3, 0) OVER ( PARTITION BY custid ORDER BY orderdate, orderid ) AS prev_val2 ,--往前显示第三行，如过前面每值，则显示0
        LEAD(val) OVER ( PARTITION BY custid ORDER BY orderdate, orderid ) AS next_val --显示后一行
FROM    Sales.OrderValues;

custid	orderid	orderdate	val	prev_val	prev_val1	prev_val2	next_val
1	10643	00:00.0	814.5	NULL	NULL	0	878
1	10692	00:00.0	878	814.5	NULL	0	330
1	10702	00:00.0	330	878	NULL	0	845.8
1	10835	00:00.0	845.8	330	814.5	814.5	471.2
1	10952	00:00.0	471.2	845.8	878	878	933.5
1	11011	00:00.0	933.5	471.2	330	330	NULL
2	10308	00:00.0	88.8	NULL	NULL	0	479.75
2	10625	00:00.0	479.75	88.8	NULL	0	320
2	10759	00:00.0	320	479.75	NULL	0	514.4
2	10926	00:00.0	514.4	320	88.8	88.8	NULL
3	10365	00:00.0	403.2	NULL	NULL	0	749.06
3	10507	00:00.0	749.06	403.2	NULL	0	1940.85
3	10535	00:00.0	1940.85	749.06	NULL	0	2082
3	10573	00:00.0	2082	1940.85	403.2	403.2	813.37

--FIRST_VALUE 和 LAST_VALUE支持window partition,order, and frame clauses.
SELECT  custid ,
        orderid ,
        orderdate ,
        val ,
        FIRST_VALUE(val) OVER ( PARTITION BY custid ORDER BY orderdate, orderid
ROWS BETWEEN UNBOUNDED PRECEDING
AND CURRENT ROW ) AS first_val ,
        LAST_VALUE(val) OVER ( PARTITION BY custid ORDER BY orderdate, orderid
ROWS BETWEEN CURRENT ROW
AND UNBOUNDED FOLLOWING ) AS last_val
FROM    Sales.OrderValues;

custid	orderid	orderdate	val	first_val	last_val
1	11011	2008/4/9	933.5	814.5	933.5
1	10952	2008/3/16	471.2	814.5	933.5
1	10835	2008/1/15	845.8	814.5	933.5
1	10702	2007/10/13	330	814.5	933.5
1	10692	2007/10/3	878	814.5	933.5
1	10643	2007/8/25	814.5	814.5	933.5
2	10926	2008/3/4	514.4	88.8	514.4
2	10759	2007/11/28	320	88.8	514.4
2	10625	2007/8/8	479.75	88.8	514.4
2	10308	2006/9/18	88.8	88.8	514.4
3	10856	2008/1/28	660	403.2	660
3	10682	2007/9/25	375.5	403.2	660
3	10677	2007/9/22	813.37	403.2	660
3	10573	2007/6/19	2082	403.2	660