Re: Limit query output to a single record based on combination of fields

From: Urs Metzger <urs_at_ursmetzger.de>
Date: Mon, 12 May 2008 17:40:07 +0200
Message-ID: <g09obl$p44$1@online.de>

Sashi schrieb:
> On May 12, 10:55 am, Urs Metzger <u..._at_ursmetzger.de> wrote:

>> Sashi schrieb:
>>
>>> Hi all,
>>> I have a query as follows:
>>> Select building, floor from T1
>>> I would like to limit the output where the building, floor combination
>>> is present only once in the output.
>>> For example, if the output is
>>> blgdA, 2
>>> bldgA, 3
>>> bldgA, 3
>>> bldgB, 1
>>> bldgB, 2
>>> bldgB, 2
>>> I would like to filter it so that the output is only
>>> blgdA, 2
>>> bldgB, 1
>>> I'm not sure about how to do this.
>>> Can someone help out?
>>> Thanks,
>>> Sashi
>> Search the docs for GROUP BY and HAVING.
>>
>> htht,
>> Urs Metzger

>
> Thank you.
> So I modified the query thus:
> select bldg, count(Floor)
> from T
> group by bldg
> having count(Floor) =1
>
> However, when I try to include the floor number
> select lbcode_key, count(floor ), floor
> it fails: ORA-00979: not a GROUP BY expression 
> I want to be able to see the floor number.
> Also there is another field "type" and I need to filter it by this
> field as well, to include the regexp '%Switch'.
>
> So my original query should have been:
> Select building, floor from T1 where type like '%Switch'
>
> However, the where clause is disallowed when using group by and the
> having clause doesn't like this either.
>
> How do I do this?

You need to group by (at least) all columns which you don't aggregate. And you don't need "count(floor)" in the select list - you already know it to be 1.

WHERE is perfectly legal, but only before the GROUP BY clause.

select bldg, floor
from T
where type like '%Switch'
group by bldg, floor
having count(*) = 1

hth,
Urs Metzger Received on Mon May 12 2008 - 10:40:07 CDT