Re: Performance pb vs SQLServer.

Steinar H. Gunderson wrote:

>On Sun, Aug 14, 2005 at 07:27:38PM -0500, John Arbash Meinel wrote:
>
>
>>My guess is that this is part of a larger query. There isn't really much
>>you can do. If you want all 3.2M rows, then you have to wait for them to
>>be pulled in.
>>
>>
>
>To me, it looks like he'll get 88 rows, not 3.2M. Surely we must be able to
>do something better than a full sequential scan in this case?
>
>test=# create table foo ( bar char(4) );
>CREATE TABLE
>test=# insert into foo values ('0000');
>INSERT 24773320 1
>test=# insert into foo values ('0000');
>INSERT 24773321 1
>test=# insert into foo values ('1111');
>INSERT 24773322 1
>test=# select * from foo group by bar;
> bar
>------
> 1111
> 0000
>(2 rows)
>
>I considered doing some odd magic with generate_series() and subqueries with
>LIMIT 1, but it was a bit too weird in the end :-)
>
>/* Steinar */
>
>
I think a plain "GROUP BY" is not smart enough to detect it doesn't need
all rows (since it is generally used because you want to get aggregate
values of other columns).
I think you would want something like SELECT DISTINCT, possibly with an
ORDER BY rather than a GROUP BY (which was my final suggestion).

John
=:->