CODEWITHSUNDEEP
sqlsql-server
Tony_Henrich
Tony_Henrich

Reputation: 44085

Delete records which are considered duplicates based on same value on a column and keep the newest

I would like to delete records which are considered duplicates based on them having the same value in a certain column and keep one which is considered the newest based on InsertedDate in my example below. I would like a solution which doesn't use a cursor but is set based. Goal: delete all duplicates and keep the newest.

The ddl below creates some duplicates. The records which need to be deleted are: John1 & John2 because they have the same ID as John3 and John3 is the newest record.

Also record John5 needs to be deleted because there's another record with ID = 3 and is newer (John6).

Create table dbo.TestTable (ID int, InsertedDate DateTime, Name varchar(50))

Insert into dbo.TestTable Select 1, '07/01/2009', 'John1'
Insert into dbo.TestTable Select 1, '07/02/2009', 'John2'
Insert into dbo.TestTable Select 1, '07/03/2009', 'John3'
Insert into dbo.TestTable Select 2, '07/03/2009', 'John4'
Insert into dbo.TestTable Select 3, '07/05/2009', 'John5'
Insert into dbo.TestTable Select 3, '07/06/2009', 'John6'

Upvotes: 2

Views: 333

Answers (2)

Remus Rusanu
Remus Rusanu

Reputation: 294287

Just as an academic exercise:

with cte as (
   select *, row_number() over (partition by ID order by InsertedDate desc) as rn
   from TestTable)
delete from cte
where rn <> 1;

Most of the time the solution proposed by Sam performs much better.

Upvotes: 4

Sam Saffron
Sam Saffron

Reputation: 131112

This works: 

delete t 
from TestTable t
left join 
(
    select id, InsertedDate = max(InsertedDate) from TestTable
    group by id
) as sub on sub.id = t.id and sub.InsertedDate = t.InsertedDate
where sub.id is null

If you have to deal with ties it gets a tiny bit trickier.

Upvotes: 2

Related Questions