SQL條件透視

Question

是的，這是另一個透視問題......我已經讀過幾乎所有以前的問題，我似乎無法破解我所需要的查詢。

這裏是我的表如下所示：

FirmName Account Balance Pmt Revolving Installment  Mortgage 
Amex 12345 10000 2000 1 0 0 
Discover 54321 20000 4000 1 0 0 
Chase 13579 100000 1500 0 0 1 
Wells Fargo 2468 40000 900 0 1 0 

最後三位列（週轉，分期付款，&按揭）規定的列應如何捲成的類型。每個結果都需要基於類型和行數的三列。結果應該是多列的一行。下面是結果應該是什麼樣子：

Revolving1_Firm Revolving1_Balance Revolving1_Pmt Revolving2_Firm Revolving2_Balance Revolving2_Pmt Realestate1_Firm Realestate1_Balance  Realestate1_Pmt Vehicle1_Firm Vehicle1_Balance Vehicle1_Pmt 

Amex  10000 2000 Discover 20000 4000 Chase 100000 1500 Wells Fargo 40000 900 

你怎麼支點，基於比特字段（週轉，分期付款，&按揭），並保持適當的數量，使得每一列得到計數＃追加到它？

Answer 1

這必須用動態SQL來完成。首先，您需要確定每種類型的最大數量（並且可能還會將3位列轉換爲單個LOAN_TYPE列，因爲這是您的分區），然後使用ROW_NUMBER（）OVER（分區BY LOAN_TYPE ORDER BY FirmName ）在與之相關的標準化數據上，以便將事情放在他們的右列中。

我不得不問，在數據庫中做這件事有多重要 - 因爲架構不固定，很難看到這樣做的效用。

不過，只比大多數動態樞軸稍微複雜一點，所以如果上面的提示沒有讓你在那裏，而你仍然想讓我刺穿它，我會嘗試發佈一些實際的工作代碼後來。

SET NOCOUNT ON 

DECLARE @t AS TABLE 
    (
    FirmName varchar(50) NOT NULL 
    ,Account varchar(50) NOT NULL 
    ,Balance money NOT NULL 
    ,Pmt money NOT NULL 
    ,Revolving bit NOT NULL 
    ,Installment bit NOT NULL 
    ,Mortgage bit NOT NULL 
    ) ; 
INSERT INTO @t 
VALUES ('Amex', '12345', 10000, 2000, 1, 0, 0) ; 
INSERT INTO @t 
VALUES ('Discover', '54321', 20000, 4000, 1, 0, 0) ; 
INSERT INTO @t 
VALUES ('Chase', '13579', 100000, 1500, 0, 0, 1) ; 
INSERT INTO @t 
VALUES ('Wells Fargo', '2468', 40000, 900, 0, 1, 0) ; 

WITH n1 
      AS (
       SELECT FirmName 
         ,Account 
         ,Balance 
         ,Pmt 
         ,LoanType 
         ,LoanTypeFlag 
       FROM  @t UNPIVOT (LoanTypeFlag FOR LoanType IN ([Revolving], [Installment], [Mortgage])) AS unpvt 
      ), 
     n2 
      AS (
       SELECT FirmName 
         ,Balance 
         ,Pmt 
         ,LoanType 
       FROM  n1 
       WHERE  LoanTypeFlag = 1 
      ), 
     n3 
      AS (
       SELECT FirmName 
         ,Balance 
         ,Pmt 
         ,LoanType 
         ,ROW_NUMBER() OVER (PARTITION BY LoanType ORDER BY FirmName) AS SequenceNumber 
       FROM  n2 
      ), 
     n4 
      AS (
       SELECT LoanType + CONVERT(varchar, SequenceNumber) AS Column_Prefix 
         ,FirmName AS Firm 
         ,CONVERT(varchar(50), Balance) AS Balance 
         ,CONVERT(varchar(50), Pmt) AS Pmt 
       FROM  n3 
      ), 
     n5 
      AS (
       SELECT Column_Prefix + '_' + Col AS Col_Nm 
         ,Val 
       FROM  n4 UNPIVOT (Val FOR Col IN ([Firm], [Balance], [Pmt])) AS unpvt 
      ) 
    SELECT * 
    FROM n5 PIVOT (MAX(Val) FOR Col_Nm IN ([Installment1_Firm], [Installment1_Balance], [Installment1_Pmt], 
               [Mortgage1_Firm], [Mortgage1_Balance], [Mortgage1_Pmt], [Revolving1_Firm], 
               [Revolving1_Balance], [Revolving1_Pmt], [Revolving2_Firm], 
               [Revolving2_Balance], [Revolving2_Pmt])) AS pvt 

你的主要剩下的問題是最終的PIVOT列表，因爲一切最終PIVOT之前符合在實體價值的階段VARCHAR（50）（你可以動態生成這是我提到的）和類型安全。

而且，如果該位標誌不是相互排斥的，您會有一些重複...

，如果你使用的是表單生成系統最終PIVOT名單是相對固定的，我想是這樣的，你能不要使用動態SQL來生成該列表，但它會使系統稍微不具有前瞻性。

這將產生pivot_list（可以簡化）：

WITH n1 
      AS (
       SELECT FirmName 
         ,Account 
         ,Balance 
         ,Pmt 
         ,LoanType 
         ,LoanTypeFlag 
       FROM  @t UNPIVOT (LoanTypeFlag FOR LoanType IN ([Revolving], [Installment], [Mortgage])) AS unpvt 
      ), 
     n2 
      AS (
       SELECT FirmName 
         ,Balance 
         ,Pmt 
         ,LoanType 
       FROM  n1 
       WHERE  LoanTypeFlag = 1 
      ), 
     n3 
      AS (
       SELECT FirmName 
         ,Balance 
         ,Pmt 
         ,LoanType 
         ,ROW_NUMBER() OVER (PARTITION BY LoanType ORDER BY FirmName) AS SequenceNumber 
       FROM  n2 
      ), 
     n4 
      AS (
       SELECT LoanType + CONVERT(varchar, SequenceNumber) AS Column_Prefix 
         ,FirmName AS Firm 
         ,CONVERT(varchar(50), Balance) AS Balance 
         ,CONVERT(varchar(50), Pmt) AS Pmt 
       FROM  n3 
      ), 
     n5 
      AS (
       SELECT Column_Prefix + '_' + Col AS Col_Nm 
         ,Val 
       FROM  n4 UNPIVOT (Val FOR Col IN ([Firm], [Balance], [Pmt])) AS unpvt 
      ), 
     pivot_list(pivot_list) 
      AS (
       SELECT ',' + QUOTENAME(Col_Nm) 
       FROM  n5 
       FOR  XML PATH('') 
      ) 
    SELECT STUFF(pivot_list, 1, 1, '') AS pivot_list 
    FROM pivot_list