Re: constraints in query plans - Mailing list pgsql-hackers

From Jeremy Drake
Subject Re: constraints in query plans
Date
Msg-id Pine.BSO.4.64.0610152033390.18508@resin.csoft.net
Whole thread Raw
In response to Re: constraints in query plans  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: constraints in query plans  (Jeff Davis <pgsql@j-davis.com>)
List pgsql-hackers
On Sun, 15 Oct 2006, Tom Lane wrote:

> Jeremy Drake <pgsql@jdrake.com> writes:
> > CREATE TABLE test_domain (
> >  fkey integer not null,
> >  k    integer not null,
> >  x1   integer not null,
> >  x2   integer,
> >  mp   m_or_p not null
> > );
>
> > CREATE INDEX test_domain_k_x1_x2_m ON test_domain (k, x1, x2) WHERE mp = 'm';
> > CREATE INDEX test_domain_k_x1_x2_p ON test_domain (k, x1, x2) WHERE mp = 'p';
>
> Perhaps you need a non-partial index.

I just tried that,
CREATE INDEX test_domain_k_x1_x2_mp ON test_domain (k, x1, x2, mp);

and dropped the others.  That actually works properly.
jeremyd=# explain analyze select * from test_domain where k = 1255 and mp;
            QUERY PLAN
 

----------------------------------------------------------------------------------------------------------------------------------Bitmap
HeapScan on test_domain  (cost=5.37..237.21 rows=66 width=17) (actual time=0.115..0.707 rows=132 loops=1)  Recheck
Cond:(k = 1255)  Filter: mp  ->  Bitmap Index Scan on test_domain_k_x1_x2_mp  (cost=0.00..5.37 rows=66 width=0) (actual
time=0.081..0.081rows=132 loops=1)        Index Cond: ((k = 1255) AND (mp = true))Total runtime: 1.137 ms
 
(6 rows)

I thought I had to refer to all of the columns in order for this to work,
that I could not skip some in the middle, but it seems to work.


-- 
If you can survive death, you can probably survive anything.


pgsql-hackers by date:

Previous
From: "Jim C. Nasby"
Date:
Subject: Re: Threaded python on FreeBSD
Next
From: Bruce Momjian
Date:
Subject: Re: Threaded python on FreeBSD