Question

I have a big table that has duplicates as such:

• number (primary key),
• group (primary key),
• Division (primary key),
• dateChange.

Example:

1,2,3,20121015

1,2,3,20120101

1,2,3,20110101

2,2,2,20121010

2,2,2,20120101

result should be:

1,2,3,20121015

2,2,2,20121010

I have tried many combinations including group by the primary key with minimum “changeDate”
but nothing seems to work perfectly.

I want to have something like this:

delete from table where (number,group.devision,changeDate) not in 
(select from table(number,group,devision,Max(changeDate)) 
group by (number,group.devision)

But I dont think it is a valid MS-SQL syntax.

Your help will be very appreciated!!

Answer 1

To delete all rows except for the latest for a number, group, Division combination.

;WITH cte
     AS (SELECT ROW_NUMBER() OVER (PARTITION BY number, group, Division 
                                       ORDER BY dateChange DESC) RN
         FROM   YourTable)
DELETE FROM cte
WHERE  RN > 1