Question

我正在尝试根据最大值或最小值来聚合行，其中max / min由业务规则而不是内在值确定。换句话说，我需要创建自定义最大/最小 - 不使用CLR。我在下面的代码段中有一个工作自包含的示例，但我觉得必须有一个更简单的方法。

在这种特殊情况下，我需要IPA列的业务优先级由下面[IPAMap] CTE中显示的[Seq]值定义。然后，我从[TestData]表/ cte（此示例中只有四行）中获取一组任意行，并根据优先级将它们聚合为两行，并选择正确的IPA值。

--Table [IPAMap] holds all possible values for [IPA], and reflects the relative
--'priority' of each with the [Seq] column.  This table is used as a lookup for
--business rules.
;WITH IPAMap (Seq, IPA) AS ( 
SELECT 1, 'Q' UNION ALL 
SELECT 2, 'S' UNION ALL 
SELECT 3, 'A' UNION ALL 
SELECT 4, 'L'
) 
--Table [TestData] represents the real data. 
, TestData (CustomerID, IPA) AS ( 
SELECT '123', 'A' UNION ALL 
SELECT '123', 'S' UNION ALL 
SELECT '234', 'L' UNION ALL 
SELECT '234', 'Q' 
) 
--Table [PartyIPASeq] finds the lowest Seq rank from [TestData] 
,PartyIPASeq (CustomerID, IPASeq) AS ( 
 SELECT  
   CustomerID,  
   MIN(IP.Seq) 
 FROM TestData [TD] 
  INNER JOIN IPAMap [IP] ON IP.IPA = TD.IPA 
  GROUP BY CustomerID 
) 
--The final selection provides the key and the 'IPA' with the lowest rank 
SELECT TD.CustomerID, TD.IPA 
FROM TestData [TD] 
 INNER JOIN IPAMap [IP] ON IP.IPA = TD.IPA 
 INNER JOIN PartyIPASeq [SEQ] ON SEQ.CustomerID = [TD].CustomerID 
   AND SEQ.IPASeq = IP.Seq

上述查询的结果如下：

CustomerID  IPA
234         S
123         Q

这是正确的，但我希望有人能告诉我更简单的方法来完成同样的工作。

Answer 1

看起来答案是使用带有charindex顺序的窗口函数，如下所示：

WITH TestData (CustomerID, IPA) AS ( 
    SELECT '123', 'A' UNION ALL 
    SELECT '123', 'S' UNION ALL 
    SELECT '234', 'L' UNION ALL 
    SELECT '234', 'Q' 
    ) 
SELECT CustomerID, IPA
FROM (
        SELECT CustomerID, 
        IPA, 
        ROW_NUMBER() OVER (PARTITION BY CustomerID ORDER BY CHARINDEX(IPA, 'QSAL')) rownum
        FROM TestData
    ) A
WHERE rownum = 1

如果业务数据使用多字符值，则CHARINDEX将替换为PATINDEX函数。

查找最大/最小任意优先级

1 个答案: