SQL Server 2018如何将信息分组到不同的列中

我有一张桌子的格式非常混乱。这是表中间列的快照，这就是为什么数据中有大洞，其他列中有这些行的内容

+---------------+--------------------+--------------+--------------+
|    upsell1    |      upsell2       |   upsell3    |   upsell4    |
+---------------+--------------------+--------------+--------------+
|               |                    | Car Kit      | Scented Tabs |
|               |                    |              |              |
|               |                    |              |              |
|               | Fiters             |              |              |
| NULL          | NULL               | NULL         | NULL         |
|               |                    |              |              |
|               | Car Kit            | Scented Tabs |              |
|               |                    |              |              |
|               |                    |              |              |

+---------------+--------------------+--------------+--------------+

我希望得到一个输出，对每个产品进行分组和计数，无论它出现在哪个列中

+--------------+---------------+
|   Product    | Product Count |
+--------------+---------------+
| Car Kit      |             2 |
| Scented Tabs |             2 |
| Fiters       |             1 |
+--------------+---------------+

通常情况下，如果它都在一列中，那么使用group by命令将X作为列1，countX作为列2是很容易的，但是我正在寻找一种方法来将这些不同的信息正确地分组在一起

---

注意：如果您无法控制糟糕的模式，我无法对表的结构进行太多更改，这让我很懊恼

with CTE as
(
select upsell1 as product, 1 as counter
from MyTable
where upsell1 is not null

union all
select upsell2, 1
from MyTable
where upsell2 is not null

union all
select upsell3, 1
from MyTable
where upsell3 is not null

union all
select upsell4, 1
from MyTable
where upsell4 is not null
)
select product, sum(counter)
from CTE
group by product

试试这个

    declare @test table (
    upsell1 varchar(100),
    upsell2 varchar(100),       
    upsell3 varchar(100),
    upsell4 varchar(100)

)
insert into @test values ('','','Car Kit','Scented Tabs')
insert into @test values ('','','','')
insert into @test values ('','','','')
insert into @test values ('','Fiters','','')
insert into @test values (null,null,null,null)
insert into @test values ('','','','')
insert into @test values ('','Car Kit','Scented Tabs','')
insert into @test values ('','','','')
insert into @test values ('','','','')

select 
    abc,
    count(*) 
from (
select *
from @test) as src
unpivot (
abc for anyThing in (upsell1,upsell2,upsell3,upsell4)
) as ttt
group by abc

---

由于没有类型假设，我会选择使用UNPIVOT的安全方法：

SELECT * INTO #t FROM (VALUES
    ('','','Car Kit','Scented Tabs'),
    ('','','',''),
    ('','','',''),
    ('','Fiters','',''),
    (null,null,null,null),
    ('','','',''),
    ('','Car Kit','Scented Tabs',''),
    ('','','',''),
    ('','','','')
) T(upsell1,upsell2,upsell3,upsell4)

SELECT Product, COUNT(Product) [Product Count] FROM
(
    SELECT
        CAST(upsell1 AS nvarchar(MAX)) upsell1,
        CAST(upsell2 AS nvarchar(MAX)) upsell2,
        CAST(upsell3 AS nvarchar(MAX)) upsell3,
        CAST(upsell4 AS nvarchar(MAX)) upsell4
    FROM #t
) T
UNPIVOT (Product for X IN (upsell1,upsell2,upsell3,upsell4)) P
WHERE Product != ''
GROUP BY Product

结果

我将使用apply来完成此操作：

你的问题有点不清楚空格是什么。您可能需要：

where v.upsell is not null and v.upsell <> ''

甚至：

where v.upsell is not null and ltrim(rtrim(v.upsell)) <> ''

---

正如您所发现的，由于非规范化设计，您正在与此进行斗争。您必须将这些列转换为行UNPIVOT或UNION，然后进行标准计数。示例中的空格是空字符串还是空字符串？空字符串是空字符串，但也有空字符串出于某些疯狂的原因。请给它一个别名Product count如果在多个upsell中有相同的数据，此计数是否同时计数场？@tomdemaine是的。Union all不会从结果中删除重复项。也就是说，戈登有更好的解决方案

where v.upsell is not null and v.upsell <> ''

where v.upsell is not null and ltrim(rtrim(v.upsell)) <> ''