MSSQL 通过 "updating" 这些值对具有不同值的行进行分组

Question

假设以下数据：

需要对这些数据进行分组才能得到这个结果：

如您所见，id 只是与 STRING_AGG(id, ' ') WITHIN GROUP (ORDER BY id ASC) 聚合。 accountNums 有点棘手。只有当 refNum 有 2 个 accountNums 10040 和 10041 时，才应该用值 2 替换它们。如果任何 refNum 只有其中之一 accountNums，则不应替换它们。任何其他 accountNum.

也是如此

我尝试了以下 SQL 查询，但无法正确分组：

SELECT STRING_AGG(f.id, ' ') WITHIN GROUP ( 
ORDER BY f.id ASC),
f.refNum,
CASE 
    WHEN EXISTS (   SELECT 1 
                    FROM Foo f2 
                    WHERE (f2.accountNum = 10040 
                        OR f2.accountNum = 10041) 
                        AND f2.id != f.id) 
    AND (f.accountNum = 10040 
    OR f.accountNum = 10041) 
    THEN 2 
    ELSE f.accountNum 
END AS accountNumUpdated,
SUM(f.value) AS val 
FROM Foo f 
GROUP BY f.refNum,
    accountNumUpdated

此查询的问题是 accountNumUpdated 在 GROUP BY 子句中未知。此外，我无法将 CASE WHEN 移动到 GROUP BY，因为那里也不允许使用子查询。

由于我找不到解决这些问题的方法，问题是，我需要如何更改查询才能获得我想要的结果？

确切的 SQL 服务器版本是：

Microsoft SQL Server 2017 (RTM) - 14.0.1000.169 (X64) 
    Aug 22 2017 17:04:49 
    Copyright (C) 2017 Microsoft Corporation
    Standard Edition (64-bit) on Windows Server 2016 Standard 10.0 <X64> (Build 14393: ) (Hypervisor)

_{有什么不明白的，尽管问}

Answer 1

尝试对您的结果使用 cte。由于您已经计算出您的价值

; WITH CTE AS (
SELECT f.id,
f.refNum,
CASE 
    WHEN EXISTS (   SELECT 1 
                    FROM Foo f2 
                    WHERE (f2.accountNum = 10040 
                        OR f2.accountNum = 10041) 
                        AND f2.id != f1.id
                        AND f2.refNum = f1.refNum)   --- add this since you want to combine for one refno 
    AND (f.accountNum = 10040 
    OR f.accountNum = 10041) 
    THEN 2 
    ELSE f.accountNum 
END) AS accountNumUpdated,
f.value AS val 
FROM Foo f group by f.id, f.refnum, f.value
)
SELECT STRING_AGG(id, ' ') WITHIN GROUP ( ORDER BY id ASC) , 
max(refNum) as refNum,  max(accountNumUpdated) as accountNumUpdated , sum(val) as val 
FROM CTE GROUP BY accountNumUpdated

MSSQL 通过 "updating" 这些值对具有不同值的行进行分组

MSSQL group rows with different values by "updating" those values

sql

sql-server

group-by

case

sql-server-2017