Remove/Keep record based on group

Question

Post reply

Remove/Keep record based on group

scottcabral

Default port

Points: 1495
More actions
March 28, 2016 at 1:29 pm

#326128

Hi,
I have the following data:
declare @Policy table
(
Download varchar(20),
AccountNum varchar(20),
RootPolicyNum varchar(20),
PolicyNum varchar(20),
CurrentStatus varchar(20)
)
insert into @Policy (Download, AccountNum, RootPolicyNum, PolicyNum, CurrentStatus)
select 'EQAutralia', '57126', '1006784', '1006784', 'Renewed'
UNION ALL
select 'EQAutralia', '57126', '1006784', '1000194', 'Approved'
UNION ALL
select 'EQAutralia', '57126', '1006784', '00XH09000', 'Approved'
UNION ALL
select 'EQAutralia', '57126', 'AU120736', 'AU120736', 'Approved'
UNION ALL
select 'EQCalifornia', '57126', '1006784', '00XG96301', 'Renewed'
select * from @Policy
For every group of Download, AccountNum, and RootPolicyNum, I only want to keep the 'Renewed' CurrentStatus records if they are the only row in a group. So in the example data above, I want to remove the first row (EQAustralia, 57126, 1006784, 1006784, Renewed) since it is part of a group that containins 'Approved' records, but I want to keep the last row (EQCalifornia, 57126, 1006784, 00XG96301, Renewed) because it is the only record for that group and it is CurrentStatus is 'Renewed'
thanks
Scott

Viewing 10 posts - 1 through 9 (of 9 total)

You must be logged in to reply to this topic. Login to reply

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 1

I came up with the following:

if object_id('tempdb..#Policy') is not null

set noexec on;

create table #Policy

(

Download varchar(20),

AccountNum varchar(20),

RootPolicyNum varchar(20),

PolicyNum varchar(20),

CurrentStatus varchar(20)

);

set noexec off;

truncate table #Policy;

insert into #Policy (Download, AccountNum, RootPolicyNum, PolicyNum, CurrentStatus)

select 'EQAutralia', '57126', '1006784', '1006784', 'Renewed'

UNION ALL

select 'EQAutralia', '57126', '1006784', '1000194', 'Approved'

UNION ALL

select 'EQAutralia', '57126', '1006784', '00XH09000', 'Approved'

UNION ALL

select 'EQAutralia', '57126', 'AU120736', 'AU120736', 'Approved'

UNION ALL

select 'EQCalifornia', '57126', '1006784', '00XG96301', 'Renewed';

select * from #Policy;

with Renewals as (

select

Download,

AccountNum,

RootPolicyNum,

PolicyNum,

CurrentStatus,

cnt = count(*) over (partition by Download, AccountNum, RootPolicyNum)

from

#Policy

)

--select * from Renewals where cnt > 1 and CurrentStatus = 'Renewed';

delete from Renewals where cnt > 1 and CurrentStatus = 'Renewed';

select * from #Policy;

There is a problem with this code, however. Can there be a chance where you may have two or more records with a CurrentStatus of 'Renewed' for a given RootPolicyNum.

Recurs1on SSCommitted Points: 1974 More actions · Answer 2

Use the GROUP BY and HAVING operators. Try something like this:

DELETE P

FROM @Policy P

JOIN (

SELECT Download, AccountNum, RootPolicyNum

FROM @Policy

GROUP BY Download, AccountNum, RootPolicyNum

HAVING COUNT(*) > 1

) AS X

ON X.Download = P.Download

AND X.AccountNum = P.AccountNum

AND X.RootPolicyNum = P.RootPolicyNum

WHERE P.CurrentStatus = 'Renewed'

Recurs1on SSCommitted Points: 1974 More actions · Answer 3

I started replying before Lynn's answer posted. Lynn's answer give's you a better estimated execution plan.

scottcabral Default port Points: 1495 More actions · Answer 4

scottcabral

Default port

Points: 1495

March 28, 2016 at 1:52 pm

#1868062

works, thanks!

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 5

scottcabral (3/28/2016)
works, thanks!

It works as long as there aren't two rows of data for the same Download, AccountNum, RootPolicyNum where CurrentStatus = 'Renewed' for both.

drew.allen SSC Guru Points: 76985 More actions · Answer 6

Lynn Pettis (3/28/2016)
scottcabral (3/28/2016)
works, thanks!
It works as long as there aren't two rows of data for the same Download, AccountNum, RootPolicyNum where CurrentStatus = 'Renewed' for both.

I think that can be fixed by replacing the COUNT(*) with a DENSE_RANK().

WITH Renewals as (

select

Download,

AccountNum,

RootPolicyNum,

PolicyNum,

CurrentStatus,

dr = DENSE_RANK() OVER (partition by Download, AccountNum, RootPolicyNum ORDER BY CurrentStatus)

from

#Policy

)

select * from Renewals where dr > 1 and CurrentStatus = 'Renewed';

Drew

J. Drew Allen
Business Intelligence Analyst
Philadelphia, PA

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 7

drew.allen (3/28/2016)
Lynn Pettis (3/28/2016)
scottcabral (3/28/2016)
works, thanks!
It works as long as there aren't two rows of data for the same Download, AccountNum, RootPolicyNum where CurrentStatus = 'Renewed' for both.
I think that can be fixed by replacing the COUNT(*) with a DENSE_RANK().
WITH Renewals as (
select
Download,
AccountNum,
RootPolicyNum,
PolicyNum,
CurrentStatus,
dr = DENSE_RANK() OVER (partition by Download, AccountNum, RootPolicyNum ORDER BY CurrentStatus)
from
#Policy
)
select * from Renewals where dr > 1 and CurrentStatus = 'Renewed';
Drew

What if you wanted to delete one of the 'Renewed' rows if there were two? Don't think this will work.

scottcabral Default port Points: 1495 More actions · Answer 8

I think the Dense_Rank will work.

we should never have a scenerio where we have 2 renewals and have to remove just one.

OceanDrop SSC-Addicted Points: 487 More actions · Answer 9

This could work too !

Delete P

from @Policy P

left join

(

select Download, AccountNum, RootPolicyNum

from @Policy

Group by Download, AccountNum,RootPolicyNum

Having count(distinct CurrentStatus)=1

and max(CurrentStatus)='Renewed'

) Q

on P.Download=Q.Download

and P.AccountNum=Q.AccountNum

and P.RootPolicyNum=Q.RootPolicyNum

where Q.AccountNum is null

select *

from @Policy