Question

I have a table that looks like the following

where what should be a unique identifier,drg, is not, so I started on the personal challenge of fixing it. I could do this easily a myriad of ways but I’m trying to do some things in a different fashion to learn some new techniques. What needs to be deleted from the table is any row whose drg is not unique and whose eff_date is **less than ** the max eff_date.

To find what I needed I used

select d1.drg, MAX(d1.eff_date) as maxEffDate,d2.drgdesc
    from drgtable d1
    inner join drgtable d2 on d1.drg=d2.drg 
    group by d1.drg,d2.drgdesc
    having MAX(d1.eff_date) = MAX(d2.eff_date)

Instead of just selecting this into a new table and dropping the old one, I’d like to do it in a delete statement.

My though process on the strange looking query below was to find the rows that weren’t in the table from the above query I used and delete them. I got close, but am stuck. Also, if there’s a way easier way to do this, please let me know. Thanks

–the purpose of this was to find the rows in d3 that were not in d4 and delete them. didn’t quite get that far

delete from 
drgtable
where 
(
select * from drgtable as d3
left join 
(
select d1.drg, MAX(d1.eff_date) as maxEffDate,d2.drgdesc
    from drgtable d1
    inner join drgtable d2 on d1.drg=d2.drg 
    group by d1.drg,d2.drgdesc
    having MAX(d1.eff_date) = MAX(d2.eff_date)

) d4 on d4.drg =d3.drg and d4.maxEffDate = d3.eff_date and d3.drgdesc = d4.drgdesc
) 

Answer 1

This should work:

;WITH cte As
(
    SELECT  *,
            MAX(eff_date) OVER(partition by drg) as MaxEffDate
    FROM    drgtable
)
DELETE 
FROM    cte
WHERE   eff_date <> MaxEffDate

Plus, plenty of new tricks in there for you to learn. 🙂

(note: this does assume that you are on SQL Server 2005 or higher)