基于条件的 MDX 查询到数据透视表

Posted 2023-03-25

【中文标题】基于条件的 MDX 查询到数据透视表

【英文标题】：MDX query to pivot table based on condition

【发布时间】：2015-06-25 10:32:26

【问题描述】：

我正在尝试为数据透视表编写 MDX 查询。

RDBMS 中的类似查询是这样的：

SELECT  stats_Date
        ,ISNULL(SUM(clicks), 0) AS clicks
        ,ISNULL(SUM(CASE WHEN ad_type IN (1,3) THEN clicks END), 0) AS keyword_clicks
        ,ISNULL(SUM(CASE WHEN ad_type IN (2,3) THEN clicks END), 0) AS direct_clicks
FROM    STATS_TABLE (NOLOCK)
WHERE   stats_Date BETWEEN '2015-06-01' AND '2015-06-30'
GROUP BY stats_Date

我有两个维度 [DIM TIME] 和 [DIM AD TYPE]

我已经尝试过以下 MDX 查询：

WITH    
    MEMBER [Measures].[Clicks Keyword] AS     
        IIF
        (
          [DIM AD TYPE].[Ad Type].CurrentMember IS [DIM AD TYPE].[Ad Type].&[1]
         ,[Measures].[clicks]
         ,0
        )
SELECT  
          [Measures].[Clicks]
         ,[Measures].[Clicks Keyword]
         ON COLUMNS
        , 
          [DIM TIME].[CalendarHierarchy].[Date]*[DIM AD TYPE].[Ad Type].[Ad Type]
          ON ROWS
FROM    [CM_STATS_CUBE]
WHERE   ([DIM TIME].[Month].&[201506]:[DIM TIME].[Month].&[201506])

此 MDX 查询的示例输出如下所示：

                       Clicks       Clicks Keyword
20150501    Invalid    (null)       0             
20150501    unknown    200          0
20150501    Keyword    500          0
20150501    Ads        300          300
20150502    Invalid    (null)       0
20150502    unknown    400          0
20150502    Keyword    600          0
20150502    Ads        500          500

但我只想按 stats_date 分组，预期的输出是：

            Clicks      Clicks Keyword
20150501    1000        300
20150502    1500        500    

在 [Adventure Works] 立方体数据库中进行测试的类似示例：

WITH 
  MEMBER [Measures].[Internet Sales Amount US] AS  
    IIF( [Customer].[Customer Geography].CurrentMember IS [Customer].[Customer Geography].[Country].&[United States]
         ,[Measures].[Internet Sales Amount]
         ,NULL
       )
SELECT 
         [Measures].[Internet Sales Amount]
        ,[Measures].[Internet Sales Amount US]
        ON 0
       ,NON EMPTY[Date].[Calendar].[Date] ON 1
FROM [Adventure Works]
WHERE   [Date].[Date].&[20050701]:[Date].[Date].&[20050702]

【问题讨论】：

那个例子不好。它给出了错误。

【参考方案1】：

您无需为交叉加入而烦恼[DIM TIME].[CalendarHierarchy].[Date]*[DIM AD TYPE].[Ad Type].[Ad Type]

WITH 
  MEMBER [Measures].[Clicks Keyword] AS 
    IIF
    (
      [DIM AD TYPE].[Ad Type].CurrentMember IS [DIM AD TYPE].[Ad Type].&[1]
     ,[Measures].[clicks]
     ,0
    ) 
SELECT 
  
    [Measures].[Clicks]
   ,[Measures].[Clicks Keyword]
   ON COLUMNS
 ,[DIM TIME].[CalendarHierarchy].[Date] ON ROWS
FROM [CM_STATS_CUBE]
WHERE 
  [DIM TIME].[Month].&[201506] : [DIM TIME].[Month].&[201506];

我还建议在您的 IIF 函数中使用 null 而不是 0 - 这应该可以整理结果并加快处理速度：

WITH 
  MEMBER [Measures].[Clicks Keyword] AS 
    IIF
    (
      [DIM AD TYPE].[Ad Type].CurrentMember IS [DIM AD TYPE].[Ad Type].&[1]
     ,[Measures].[clicks]
     ,null   //<<<<<<<<<<<<<<<<< better to use null rather than 0
    ) 
SELECT 
  
    [Measures].[Clicks]
   ,[Measures].[Clicks Keyword]
   ON COLUMNS
 , NON EMPTY  //<<<<<<<<<<<<<<<<< now if Clicks and Clicks Keyword are both null the respective row will be excluded
  [DIM TIME].[CalendarHierarchy].[Date] ON ROWS
FROM [CM_STATS_CUBE]
WHERE 
  [DIM TIME].[Month].&[201506] : [DIM TIME].[Month].&[201506];

---

编辑

我没有详细阅读您的脚本 - 抱歉。你可以只聚合一组两个元组：

WITH 
  MEMBER [Measures].[Clicks Keyword] AS 
    Sum
    (
     
      ([DIM AD TYPE].[Ad Type].&[1],[Measures].[clicks])
     ,([DIM AD TYPE].[Ad Type].&[3],[Measures].[clicks])
     
    ) 
SELECT 
  
    [Measures].[Clicks]
   ,[Measures].[Clicks Keyword]
   ON COLUMNS
 , NON EMPTY  //<<<<<<<<<<<<<<<<< now if Clicks and Clicks Keyword are both null the respective row will be excluded
  [DIM TIME].[CalendarHierarchy].[Date] ON ROWS
FROM [CM_STATS_CUBE]
WHERE 
  [DIM TIME].[Month].&[201506] : [DIM TIME].[Month].&[201506];

您发布的AdvWrks 示例只是一个元组：

WITH 
  MEMBER [Measures].[Internet Sales Amount US] AS  
      (
        [Customer].[Customer Geography].[Country].&[United States]
       ,[Measures].[Internet Sales Amount]
      )
SELECT 
         [Measures].[Internet Sales Amount]
        ,[Measures].[Internet Sales Amount US]
        ON 0
       ,NON EMPTY[Date].[Calendar].[Date] ON 1
FROM [Adventure Works]
WHERE   [Date].[Date].&[20050701]:[Date].[Date].&[20050702]

如果您想在加拿大添加，那么似乎有三种可行的选择：

1.

WITH 
  MEMBER [Measures].[Internet Sales Amount US & Canada] AS  
      (
        [Customer].[Customer Geography].[Country].&[United States]
       ,[Measures].[Internet Sales Amount]
      )
    +
      (
        [Customer].[Customer Geography].[Country].&[Canada]
       ,[Measures].[Internet Sales Amount]
      )

2.

 WITH 
  MEMBER [Measures].[Internet Sales Amount US & Canada] AS 
    Aggregate
    (
      
        [Customer].[Customer Geography].[Country].&[United States]
       ,[Customer].[Customer Geography].[Country].&[Canada]
      
     ,[Measures].[Internet Sales Amount]
    ) 

3。 （切换到总和）

WITH 
   MEMBER [Measures].[Internet Sales Amount US & Canada] AS 
    Sum
    (
      
        (
          [Customer].[Customer Geography].[Country].&[Canada]
         ,[Measures].[Internet Sales Amount]
        )
       ,(
          [Customer].[Customer Geography].[Country].&[United States]
         ,[Measures].[Internet Sales Amount]
        )
      
    ) 

【讨论】：

谢谢.. 但我试过了，输出列中全是零或空值。

有人可以尝试在[Adventure Works] 测试数据库上显示类似的查询吗？

@Omesh 请进行我向您的计算成员[Measures].[Clicks Keyword] 建议的更改，即将null 作为分支之一。然后还要注意我已将关键字NON EMPTY 添加到您的ON ROWS 声明中

也试过了，但还是没有成功。我在[Adventure Works] 数据库中编辑了类似示例的问题。您可以在此测试此查询的输出。

[DIM AD TYPE].[Ad Type].CurrentMember 将如何工作？没有设置范围。

【参考方案2】：

试试这个：

WITH 
  MEMBER [Measures].[Clicks Keyword] AS 
  AGGREGATE([DIM AD TYPE].[Ad Type].&[1], [DIM AD TYPE].[Ad Type].&[3], [Measures].[Clicks])

SELECT NON EMPTY
  
    [Measures].[Clicks]
   ,[Measures].[Clicks Keyword]
   ON COLUMNS
 , NON EMPTY  
  [DIM TIME].[CalendarHierarchy].[Date] ON ROWS
FROM [CM_STATS_CUBE]
WHERE 
  ([DIM TIME].[Month].&[201506] : [DIM TIME].[Month].&[201506]);

【讨论】：

我们怎样才能包含这个条件ad_type IN (1,3)

谢谢！也试过了，但似乎没有用。

它有什么问题？