使用 Group By、Rank Row_Number 删除重复项

Question

我有两个 table。一个是 CustomerOrders，另一个是 OrderCustomerRef - lookup table。

两个 table 都具有一对多关系 - 一个客户可能与多个订单相关联。

CustomerOrders table 有重复的客户（相同的 LName、FName、Email）。但是他们有不同的Cust_IDs.

我需要合并基本客户 table 中的所有重复联系人（一对一）。（此 table 未在此处显示）。

第 1 步：

需要找出哪个 Cust_ID 应该合并到哪个对应的重复客户（相同的 LName、FName、Email）中。具有最新 Order_Date 的联系人应该赢得其对应的重复对方（客户）。 VIP 客户除外 - 无论 Order_Date.

，他们应该始终是获胜者

第 2 步： 更新了 OrderCustomerRef table：用获胜的 Cust_IDs 替换所有失败的副本 Cust_IDs。

第 3 步： 从基本客户 table 中删除所有丢失的联系人（当前范围内没有。我会自己做）。

IF OBJECT_ID('tempdb..#table') IS NOT NULL
DROP TABLE #table;

IF OBJECT_ID('tempdb..#CustomerOrders') IS NOT NULL
DROP TABLE #CustomerOrders;

IF OBJECT_ID('tempdb..#OrderCustomerRef') IS NOT NULL
DROP TABLE #OrderCustomerRef;

CREATE TABLE #CustomerOrders 
(
[PK_ID] INT NOT NULL PRIMARY KEY IDENTITY(1,1),
Cust_ID INT NOT NULL, 
LName VARCHAR(100) NULL, 
FName VARCHAR(100) NULL, 
[Customer_E-mail] VARCHAR(100) NULL,
Order_Date DATETIME NULL,
Customer_Source VARCHAR(100) NULL,
CustomerType VARCHAR(100) NULL
)

INSERT INTO #CustomerOrders (Cust_ID, LName, FName, [Customer_E-mail], Order_Date, Customer_Source, CustomerType)
VALUES 
(1, 'John', 'Smith', 'JSmith@email.com', '2018-11-10 01:40:55.150', 'XYZ Company', 'Regular'),
(2, 'John', 'Smith', 'JSmith@email.com', '2018-10-10 05:05:55.150', 'Internet', 'VIP'),
(3, 'Adam', 'Burns', 'ABurns@email.com', '2017-05-05 00:00:00.000', 'XYZ Company','Regular'),
(3, 'Adam', 'Burns', 'ABurns@email.com', '2017-05-05 00:00:00.000', 'XYZ Company','VIP'),
(4, 'Adam', 'Burns', 'ABurns@email.com', '2017-05-05 00:00:00.000', 'Internet','Regular'),
(5, 'Adam', 'Burns', 'ABurns@email.com', '2017-05-05 00:00:00.000', 'Internet','VIP'),
(6, 'James', 'Snatcher', 'JSnatcher@email.com', '2019-07-07 00:00:00.000', 'XYZ Company', 'Regular'),
(7, 'James', 'Snatcher', 'JSnatcher@email.com', '2019-07-07 00:00:00.000', 'Internet','Regular'),
(9, 'Thomas', 'Johnson', 'TJohnson@email.com', '2016-05-01 00:00:00.000', 'Internet','Regular'),
(9, 'Thomas', 'Johnson', 'TJohnson@email.com', '2015-04-01 00:00:00.000', 'Internet','Regular'),
(10, 'Thomas', 'Johnson', 'TJohnson@email.com', '2014-03-01 00:00:00.000', 'Internet','Regular'),
(11, 'Thomas', 'Johnson', 'TJohnson@email.com', '2013-02-01 00:00:00.000', 'XYZ Company','Regular'),
(12, 'Peter', 'McDonald', 'PMcDonald@email.com', '2013-02-01 00:00:00.000', 'XYZ Company','Regular'),
(13, 'Jose', 'Mainster', 'JMainster@email.com', '2013-02-01 00:00:00.000', 'Internet','Regular'),
(14, 'Kevin', 'Digginton', 'KDigginton@email.com', '2013-02-01 00:00:00.000', 'Internet','Regular'),
(14, 'Kevin', 'Digginton', 'KDigginton@email.com', '2015-09-03 00:00:00.000', 'Internet','Regular')

CREATE TABLE #OrderCustomerRef
(
    Raw_PK INT NOT NULL PRIMARY KEY IDENTITY(1,1),
    OrderID INT NOT NULL, 
    Cust_ID INT NULL, 
    OrderType VARCHAR(100) NULL
)

    INSERT INTO #OrderCustomerRef (OrderID, Cust_ID, OrderType)
    VALUES 
    (1,1,'Online'),
    (2,2,'Online'),
    (3,3,'Online'),
    (4,3,'Online'),
    (5,4,'In Store'),
    (6,5,'Online'),
    (7,6,'Online'),
    (8,7,'In Store'),
    (9,9,'Online'),
    (10,9,'Online'),
    (11,10,'In Store'),
    (12,11,'Online'),
    (13,12,'Online'),
    (14,13,'Online'),
    (15,14,'Online'),
    (16,14,'In Store')

    -- SELECT * FROM #OrderCustomerRef

    SELECT *,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail], Customer_Source ORDER BY Order_Date DESC) AS Rank_1,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail], Customer_Source ORDER BY Order_Date, CustomerType DESC ) AS Rank_CustType,
    RANK() OVER (PARTITION BY Cust_ID, FName, LName, [Customer_E-mail], Customer_Source ORDER BY Order_Date, CustomerType DESC ) AS Rank_CustID,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail] ORDER BY Order_Date DESC) AS Rank_2,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail] ORDER BY Cust_ID) AS Rank_3
    FROM #CustomerOrders

所需的输出应如下所示：

*异常： - 丢失客户 ID 1、3（应该是赢的，但由于有重复的对应方，它是 VIP，它正在丢失） - 中奖客户 ID 2、5（因为是 VIP，有例外）

例如：##OrderCustomerRef 中所有出现的 Cust_ID of John Smith 和 Cust_ID of 1 都应替换为 John Smith with Cust_ID of 2，所有出现的 Cust_ID of Adam Burns with Cust_ID of 3 should be replaced with Adam Burns with Cust_ID of 5

一般规则： - 丢失客户 ID 7、10、11、4 - 获胜客户 ID 6、9、12、13、14

例如：##OrderCustomerRef 中所有出现的 7 的 Cust_ID 应替换为 6，所有出现的 10 的 Cust_ID 应替换为 9*

最终我应该在 ##OrderCustomerRef table

中只有客户 ID 6、9、12、13、14、2、5

使用 Rank_CustType_1、column_1、column_2 我可以弄清楚第 1 步。但是我在第 2 步中仍然遇到问题 - 更新 OrderCustomerRef table：所有失败的 Cust_IDs 应该替换为相应的重复获胜 Cust_IDs.

我试过了。但这仍然不能取代失去 Cust_ID。

SELECT *,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail] ORDER BY Order_Date, CustomerType DESC) AS Rank_CustType_1,
    RANK() OVER (PARTITION BY FName, LName, [Customer_E-mail] ORDER BY Cust_ID) AS Rank_3
INTO #table
FROM #CustomerOrders

; with cte as (
    select Cust_ID, FName, LName, [Customer_E-mail], max(t.Rank_CustType_1) as Rank_CustType_1
    ,(select distinct Cust_ID from #table a where a.Cust_ID = t.Cust_ID and Rank_3 = 1) column_1
    ,(select distinct Cust_ID from #table a where a.Cust_ID = t.Cust_ID and Rank_3 <> 1) column_2

from #table t
group by Cust_ID, FName, LName, [Customer_E-mail]
    )

    update b
    set Cust_ID = case  
    when b.Cust_ID = cte.Cust_ID and
     b.Cust_ID = ISNULL(cte.column_1,'') and Rank_CustType_1 != 1 then b.Cust_ID 
    when b.Cust_ID = cte.Cust_ID and
     b.Cust_ID = ISNULL(cte.column_2,'') and Rank_CustType_1 != 1 then cte.column_2     
    when b.Cust_ID = cte.Cust_ID and Rank_CustType_1 = 1 and cte.column_1 is null and cte.column_2 is not null then cte.column_2
    when b.Cust_ID = cte.Cust_ID and Rank_CustType_1 = 1 and cte.column_1 is not null and cte.column_2 is null then cte.column_1 
    end  
    from #OrderCustomerRef b
    inner join cte on b.Cust_ID = cte.Cust_ID;

    select * from #OrderCustomerRef;

Answer 1

根据您提供的信息，我使用了以下 CTE 来显示您想要的结果：

WITH DaCTE -- To rank the existing rows
AS  (
        SELECT pk_ID
            , cust_ID
            , fname
            , lname
            , [customer_e-mail]
            , Order_Date
            , Customer_Source
            , customertype
            , ROW_NUMBER() OVER (PARTITION BY fname, lname, [customer_e-mail] ORDER BY customertype DESC, order_date DESC, cust_id) as RankYo -- Orders by the criteria provided but while you suggested 3 should lose to 5, they have the same criteria so either one could win based on ordering
        FROM #customerorders
    )
, NewSource -- To show winning Customer ID next to Original ID
AS  (
        SELECT co.pk_ID
            , DaCTE.cust_ID as NewCustomerID
            , co.cust_ID as OriginalCustomerID
            , co.fname
            , co.lname
            , co.[customer_e-mail]
            , co.Order_Date
            , co.Customer_Source
            , co.customertype
        FROM DaCTE
        INNER JOIN #CustomerOrders as co
            ON co.fname = DaCTE.FName
            AND co.lname = DaCTE.LName
            AND co.[customer_e-mail] = DaCTE.[Customer_E-mail]
        WHERE DaCTE.RankYo = 1 -- filter to show only the winning IDs based on resulting rank from previous CTE
    )
SELECT *
/*UPDATE ocr --commented out so you can see the results before running update
SET ocr.Cust_ID = ns.NewCustomerID*/
FROM #OrderCustomerRef as ocr
INNER JOIN NewSource as ns
    ON ns.OriginalCustomerID = ocr.Cust_ID

Answer 2

您可以使用 CTE（通用 Table 表达式）尝试此操作，如上述答案中 Antoine Hernandez 所解释的那样，您还可以使用 UNION 和 EXCEPT 运算符从 table 中删除重复项。

i.g 使用 EXCEPT 运算符

SELECT * FROM  #customerorders
EXCEPT
SELECT * FROM  #customerorders WHERE 1=0

i.g 使用 UNION 运算符

SELECT * FROM  #customerorders
UNION
SELECT * FROM  #customerorders WHERE 1=0

有关如何使用 CTE 删除重复项的更多信息，请遵循此 link：Remove Duplicates Using CTE

有关如何使用 UNION 和 EXCEPT 运算符删除重复项的更多信息，请遵循此 link： Remove Duplicates Using UNION And EXCEPT Operator

使用 Group By、Rank Row_Number 删除重复项

Duplicates removal using Group By, Rank, Row_Number

sql-server

group-by

row-number

ranking

duplicates