Query to find duplicates, update appropriately, delete one of those duplicate

Question

I have duplicate rows in below given table and need to cleanse it by observing another duplicate record.
I'd like to know various ways to achieve it. (I'm confused I should use UPDATE.. (CASE WHEN.. THEN).. or MERGE or something else)
Please find below DDL/DML.
create table MyTable
PKey int identity(1,1),
CustID int,
FirstName varchar(10),
LastName varchar(10),
Main varchar(10),
Department varchar(10)
Insert into MyTable
select 101, 'aaa','bbb','VM','Marketing' union
select 101, '', '','','' union
select 102, '', 'yyy', 'Main', 'Marketing' union
select 102, 'xxx','','','' union
select 103, 'ppp', 'qqq', '', 'HR' union
select 103, '', '', 'MF', '' union
select 104, 'mmm', 'nnn', 'f', 'dept'
select * from mytable
--PKey CustID FirstName LastName Main Department
--2 101 aaa bbb VM Marketing
--3 102 xxx yyy Main Marketing
--6 103 ppp qqq MF HR
--7 104 mmm nnn f dept
Cheers,
Vaibhav Chaudhari

Answer 1

Hi Vaibhav,
Manu's has copied as a part of the below code.
create table MyTable
PKey int identity(1,1),
CustID int,
FirstName varchar(10),
LastName varchar(10),
Main varchar(10),
Department varchar(10)
Insert into MyTable--(CustID,FirstName,LastName,Main,Department)
select 101, 'aaa','bbb','VM','Marketing' union
select 101, '', '','','' union
select 102, '', 'yyy', 'Main', 'Marketing' union
select 102, 'xxx','','','' union
select 103, 'ppp', 'qqq', '', 'HR' union
select 103, '', '', 'MF', '' union
select 104, 'mmm', 'nnn', 'f', 'dept'
SELECT * FROM MyTable;
;WITH cte AS
SELECT DISTINCT
MAX(PKey) PKey,
CustID,
MAX(FirstName) AS FirstName,
MAX(LastName)AS LastName,
MAX(Main) AS Main,
MAX(Department) AS Department
FROM mytable
GROUP BY CustID
MERGE mytable AS Tar
USING cte AS Src
ON Tar.PKey = Src.PKey
WHEN MATCHED THEN
UPDATE SET Tar.CustID = Src.CustID, Tar.FirstName = Src.FirstName,Tar.LastName = Src.LastName, Tar.Main = Src.Main,Tar.Department = Src.Department
WHEN NOT MATCHED BY SOURCE THEN
DELETE
SELECT * FROM MyTable
DROP TABLE MyTable;
If you do care about the Identity Pkey, as per the expected output, my understanding on your logic is like below.
;WITH cte AS
SELECT PKey, CustID,V1,V2,V3,V4,ROW_NUMBER() OVER(PARTITION BY CustID ORDER BY v1+v2+v3+v4 DESC) AS RN
FROM MyTable
CROSS APPLY(SELECT CASE WHEN FirstName ='' THEN 0 ELSE 1 END AS v1) AS cat1
CROSS APPLY(SELECT CASE WHEN LastName ='' THEN 0 ELSE 1 END AS v2) AS cat2
CROSS APPLY(SELECT CASE WHEN Main ='' THEN 0 ELSE 1 END AS v3) AS cat3
CROSS APPLY(SELECT CASE WHEN Department ='' THEN 0 ELSE 1 END AS v4) AS cat4
,cte2 AS
SELECT DISTINCT
CustID,
MAX(FirstName) AS FirstName,
MAX(LastName)AS LastName,
MAX(Main) AS Main,
MAX(Department) AS Department
FROM mytable
GROUP BY CustID
,cte3 AS
SELECT c2.CustID,c2.FirstName,c2.LastName,c2.Main,c2.Department,c.PKey FROM cte2 c2 JOIN cte c ON c.CustID = c2.CustID WHERE NOT EXISTS(SELECT 1 FROM cte WHERE RN<c.RN)
MERGE mytable AS Tar
USING cte3 AS Src
ON Tar.PKey = Src.PKey
WHEN MATCHED THEN
UPDATE SET Tar.CustID = Src.CustID, Tar.FirstName = Src.FirstName,Tar.LastName = Src.LastName, Tar.Main = Src.Main,Tar.Department = Src.Department
WHEN NOT MATCHED BY SOURCE THEN
DELETE
If you have any question, feel free to let me know.
Eric Zhang
TechNet Community Support