嵌套选择中分区依据返回奇怪的结果

Question

我在SQLServer 2008中遇到了一个我不理解的奇怪行为。我想快速地将唯一的客户与唯一的付款配对。

使用此查询，我得到了期望的结果。每个CustomerId与不同的PaymentId配对。

SELECT CustomerId, PaymentId, RowNumber1, RowNumber2
FROM (
    SELECT
        c.Id as CustomerId,
        p.Id as PaymentId,
        ROW_NUMBER() OVER (PARTITION BY p.Id ORDER BY p.Id) AS RowNumber1,
        ROW_NUMBER() OVER (PARTITION BY c.Id ORDER BY c.Id) AS RowNumber2
    FROM Customer as c
    CROSS JOIN Payment as p
) AS INNERSELECT WHERE RowNumber2 = 1

+------------+-----------+------------+------------+
| CustomerId | PaymentId | RowNumber1 | RowNumber2 |
+------------+-----------+------------+------------+
|          4 |         1 |          1 |          1 |
|          5 |         2 |          2 |          1 |
+------------+-----------+------------+------------+

但是，如果我从外部选择中删除RowNumber1列，结果似乎会改变。现在，即使我没有触摸内部选择语句，PaymentId的每个值都为1。

SELECT CustomerId, PaymentId, RowNumber2
FROM (
    SELECT
        c.Id as CustomerId,
        p.Id as PaymentId,
        ROW_NUMBER() OVER (PARTITION BY p.Id ORDER BY p.Id) AS RowNumber1,
        ROW_NUMBER() OVER (PARTITION BY c.Id ORDER BY c.Id) AS RowNumber2
    FROM Customer as c
    CROSS JOIN Payment as p
) AS INNERSELECT WHERE RowNumber2 = 1

+------------+-----------+------------+
| CustomerId | PaymentId | RowNumber2 |
+------------+-----------+------------+
|          4 |         1 |          1 |
|          5 |         1 |          1 |
+------------+-----------+------------+

谁能向我解释为什么从外部选择中删除一列会导致PaymentId列中的值发生变化？在不需要结果集中的行号的情况下，我还可以使用什么其他方法来实现期望的目标？

Answer 1

这是因为通常未定义子查询中row_number（）的顺序。进行交叉连接时，行可以按任何顺序排列

它可能是：

CUSTOMERID PAYMENTID
         4         1    
         4         2    
         5         2    
         5         1

或者可能是

CUSTOMERID PAYMENTID
         4         1    
         4         2    
         5         1    
         5         2

当您通过CUSTOMERID计算第一个记录集分区中的row_number时，会得到

CUSTOMERID PAYMENTID ROWNUMBER
         4         1         1  
         4         2         2
         5         2         1
         5         1         2

当您通过CUSTOMERID计算第二个记录集分区中的row_number时，您将获得

CUSTOMERID PAYMENTID ROWNUMBER
         4         1         1  
         4         2         2
         5         1         1
         5         2         2

如果您只将随机客户和付款配对，则可以执行此操作

with cte_cust as (
    select id, row_number() over (order by id) as row_num
    from Customer
), cte_pay as (
    select id, row_number() over (order by id) as row_num
    from Payment
)
select
    c.id as CustomerId,
    p.id as PaymentId
from cte_cust as c
    inner join cte_pay as p on p.row_num = c.row_num

请注意，如果您的客户多于付款，则某些客户将不会出现在结果中（反之亦然）。

sql fiddle demo