数据库环境：SQL SERVER 2005
现有一个产品销售实时表，表数据如下：
字段name是产品名称，字段type是销售类型，1表示售出，2表示退货，字段num是数量，字段ctime是操作时间。
要求：
在一行中统计24小时内所有货物的销售（售出，退货）数据，把日期考虑在内。
分析：
这实际上是行转列的一个应用，在进行行转列之前，需要补全24小时的所有数据。补全数据可以通过系统的数字辅助表
spt_values来实现，进行行转列时，根据type和处理后的ctime分组即可。
1.建表，导入数据
CREATE TABLE snake (name VARCHAR(10 ),type INT,num INT, ctime DATETIME )INSERT INTO snake VALUES(' 方便面', 1,10 ,'2015-08-10 16:20:05')INSERT INTO snake VALUES(' 香烟A ', 2,2 ,'2015-08-10 18:21:10')INSERT INTO snake VALUES(' 香烟A ', 1,5 ,'2015-08-10 20:21:10')INSERT INTO snake VALUES(' 香烟B', 1,6 ,'2015-08-10 20:21:10')INSERT INTO snake VALUES(' 香烟B', 2,9 ,'2015-08-10 20:21:10')INSERT INTO snake VALUES(' 香烟C', 2,9 ,'2015-08-10 20:21:10')
2.补全24小时的数据
/*枚举0-23自然数列*/WITH x0 AS ( SELECT number AS h FROM master..spt_values WHERE type = 'P' AND number >= 0 AND number <= 23 ),/*找出表所有的日期*/ x1 AS ( SELECT DISTINCT CONVERT(VARCHAR(100), ctime, 23) AS d FROM snake ),/*补全所有日期的24小时*/ x2 AS ( SELECT x1.d , x0.h FROM x1 CROSS JOIN x0 ), x3 AS ( SELECT name , type , num , DATEPART(hour, ctime) AS h FROM snake ),/*整理行转列需要用到的数据*/ x4 AS ( SELECT x2.d , x2.h , x3.name , x3.type , x3.num FROM x2 LEFT JOIN x3 ON x3.h = x2.h )
3.行转列
SELECT ISNULL([0], 0) AS [00] , ISNULL([1], 0) AS [01] , ISNULL([2], 0) AS [02] , ISNULL([3], 0) AS [03] , ISNULL([4], 0) AS [04] , ISNULL([5], 0) AS [05] , ISNULL([6], 0) AS [06] , ISNULL([3], 7) AS [07] , ISNULL([8], 0) AS [08] , ISNULL([9], 0) AS [09] , ISNULL([10], 0) AS [10] , ISNULL([3], 11) AS [11] , ISNULL([12], 0) AS [12] , ISNULL([13], 0) AS [13] , ISNULL([14], 0) AS [14] , ISNULL([3], 15) AS [15] , ISNULL([16], 0) AS [16] , ISNULL([17], 0) AS [17] , ISNULL([18], 0) AS [18] , ISNULL([19], 15) AS [19] , ISNULL([20], 0) AS [20] , ISNULL([21], 0) AS [21] , ISNULL([22], 0) AS [22] , ISNULL([23], 15) AS [23] , type , d AS date FROM ( SELECT d , h , type , num FROM x4 ) t PIVOT( SUM(num) FOR h IN ( [0], [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23] ) ) t WHERE type IS NOT NULL
来看一下最终效果，只有1天的数据，可能看起来不是很直观。
本文的技术点有2个：
1.利用数字辅助表补全缺失的记录
2.pivot行转列函数的使用