Skip site navigation (1) Skip section navigation (2)

Re: regression ? 8.4 do not apply One-Time Filter to subquery

From: Robert Haas <robertmhaas(at)gmail(dot)com>
To: Sergey Burladyan <eshkinkot(at)gmail(dot)com>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: regression ? 8.4 do not apply One-Time Filter to subquery
Date: 2009-07-22 18:19:52
Message-ID: 603c8f070907221119v63b9efe5rcb2cd5a251640c8d@mail.gmail.com (view raw or flat)
Thread:
Lists: pgsql-performance
On Wed, Jul 1, 2009 at 8:08 PM, Sergey Burladyan<eshkinkot(at)gmail(dot)com> wrote:
>
> 8.4 from CVS HEAD:
> EXPLAIN ANALYZE select * from (select n, 1 as r from generate_series(1, 100000) as n union all select n, 2 from generate_series(1, 100000) as n) as x where r = 3;
>                                                           QUERY PLAN
> --------------------------------------------------------------------------------------------------------------------------------
>  Result  (cost=0.00..30.00 rows=10 width=36) (actual time=90.723..90.723 rows=0 loops=1)
>   ->  Append  (cost=0.00..30.00 rows=10 width=36) (actual time=90.720..90.720 rows=0 loops=1)
>         ->  Function Scan on generate_series n  (cost=0.00..15.00 rows=5 width=36) (actual time=45.191..45.191 rows=0 loops=1)
>               Filter: (1 = 3)
>         ->  Function Scan on generate_series n  (cost=0.00..15.00 rows=5 width=36) (actual time=45.522..45.522 rows=0 loops=1)
>               Filter: (2 = 3)
>  Total runtime: 118.709 ms
> (7 rows)
>
> 8.3.7:
> EXPLAIN ANALYZE select * from (select n, 1 as r from generate_series(1, 100000) as n union all select n, 2 from generate_series(1, 100000) as n) as x where r = 3;
>                                                QUERY PLAN
> ----------------------------------------------------------------------------------------------------------
>  Result  (cost=0.00..25.02 rows=2 width=8) (actual time=0.005..0.005 rows=0 loops=1)
>   ->  Append  (cost=0.00..25.02 rows=2 width=8) (actual time=0.004..0.004 rows=0 loops=1)
>         ->  Result  (cost=0.00..12.50 rows=1 width=4) (actual time=0.001..0.001 rows=0 loops=1)
>               One-Time Filter: false
>               ->  Function Scan on generate_series n  (cost=0.00..12.50 rows=1 width=4) (never executed)
>         ->  Result  (cost=0.00..12.50 rows=1 width=4) (actual time=0.000..0.000 rows=0 loops=1)
>               One-Time Filter: false
>               ->  Function Scan on generate_series n  (cost=0.00..12.50 rows=1 width=4) (never executed)
>  Total runtime: 0.053 ms
> (9 rows)
>
> Is it right ?

This might be related to this fix by Tom.

http://archives.postgresql.org/message-id/20090711040933.7A08375331E@cvs.postgresql.org

...Robert

In response to

pgsql-performance by date

Next:From: Victor de BuenDate: 2009-07-22 18:28:23
Subject: Re: Atomic access to large arrays
Previous:From: Robert HaasDate: 2009-07-22 18:17:49
Subject: Re: Odd performance / query plan with bitmasked field as opposed to equality

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group