MS-SQL 以两列为轴答案

【问题标题】：MS-SQL Pivot by two columnsMS-SQL 以两列为轴
【发布时间】：2018-09-28 09:29:19
【问题描述】：

我在 MS-SQL 数据库中得到以下数据

Manufacturer|Model|Location|Inventory|Sold
------------|-----|--------|---------|-----
Alpha       |One  |USA     |3000     |123
Alpha       |One  |UK      |2300     |53
Beta        |Two  |USA     |1300     |45
Beta        |Two  |UK      |620      |12
Gamma       |Three|USA     |520      |155
Gamma       |Three|UK      |250      |19

我想获得

Manufacturer|Model|Inventory UK|Sold UK|Inventory USA|Sold USA
------------|-----|------------|-------|-------------|--------
Alpha       |One  |2300        |53     |3000         |123
Beta        |Two  |620         |12     |1300         |45
Gamma       |Three|250         |19     |520          |155

如果我理解正确，我应该旋转两次，但目前我被困在如何获得这个

SELECT *
FROM Data
PIVOT (max([Inventory]) FOR Location IN ([UK],[USA])) Piv1
PIVOT (max([Sold]) FOR Location IN ([UK],[USA])) Piv2
group by Manufacturer,Model,..?

当然，我还远远没有理解我应该如何进行。有人可以帮帮我吗？非常感谢

【问题讨论】：

标签： sql-server pivot

【解决方案1】：

这是直接使用 SQL Server 的 PIVOT 运算符执行此操作的一种方法：

SELECT
    Manufacturer,
    Model,
    MAX(UK) AS [Inventory UK],
    MAX(UK1) AS [Sold UK],
    MAX(USA) AS [Inventory USA],
    MAX(USA1) AS [Sold USA]
FROM
(
    SELECT Manufacturer, Model, Location, Location + '1' AS Location1, Inventory, Sold
    FROM Data
) AS P
PIVOT (MAX([Inventory]) FOR Location IN ([UK],[USA])) Piv1
PIVOT (MAX([Sold]) FOR Location1 IN ([UK1],[USA1])) Piv2
GROUP BY Manufacturer, Model
ORDER BY Manufacturer;

Demo

【讨论】：

我最终也遇到了两个枢轴之间的名称冲突的问题。您介意解释一下“Location1”吗？
@Aldo 信不信由你，我通过拼凑查询来凭经验回答。我能做的最多解释就是在不避免名称问题的情况下重复错误消息。为了充分解释，我必须查找PIVOT 的实际实现。

【解决方案2】：

您可以使用条件聚合来透视数据：

declare @tmp table(Manufacturer varchar(50), 
                   Model varchar(50), Location varchar(50), Inventory int, Sold int)

insert into @tmp values
 ('Alpha','One'  ,'USA',3000  ,123)
,('Alpha','One'  ,'UK' ,2300  ,53 )
,('Beta ','Two'  ,'USA',1300  ,45 )
,('Beta ','Two'  ,'UK' ,620   ,12 )
,('Gamma','Three','USA',520   ,155)
,('Gamma','Three','UK' ,250   ,19 )

select Manufacturer, Model
,sum (case when Location ='UK' then Inventory else 0 end) as Inventory_UK
,sum (case when Location ='UK' then Sold else 0 end) as Sold_UK
,sum (case when Location ='USA' then Inventory else 0 end) as Inventory_USA
,sum (case when Location ='USA' then Sold else 0 end) as Sold_USA

from @tmp 
group by Manufacturer, Model
order by Manufacturer, Model

结果：

【讨论】：

效果很好，谢谢！但只是出于好奇......我试图通过两列枢轴来解决它是错误的还是这是唯一的方法？
@Aldo 你也可以使用双轴，但我通常使用 group by 因为我有更多的控制权（在我看来）。
感谢安德里亚的澄清！

【解决方案3】：

您可以通过执行动态 sql 查询来做到这一点。

查询

declare @sql as varchar(max);

select @sql = 'select [Manufacturer], [Model], ' + stuff((
        select distinct ',sum(case [Location] when ' + char(39) + [Location] + char(39) 
        + ' then [Inventory] end) as [Inventory_' + [Location] + ']'
        + ',sum(case [Location] when ' + char(39) + [Location] + char(39) 
        + ' then [Sold] end) as [Sold_' + [Location] + ']'
        from [your_table_name]
        for xml path('')
    )
    , 1, 1, ''
);

select @sql += 'from [your_table_name] group by [Manufacturer], [Model];';

exec(@sql);

`Find a demo here`

【讨论】：