From: | Andreas Wernitznig <andreas(at)insilico(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | pgsql-bugs(at)postgresql(dot)org |
Subject: | Re: Bug #474: Index using problem |
Date: | 2001-10-05 22:37:10 |
Message-ID: | 20011006003710.10febe1a.andreas@insilico.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
I don't know how the primary key checking and the index scan are related in postgresql (but I am keen to learn more about it).
>From Sybase ASA (SQL Anywhere) I know, that these two functions (pk checking, index search) are decoupled.
(that means even a primary key without an index is possible. This makes only sense in tables with a few rows).
The pg-planner takes the current number of datasets to calculate a query plan to reach the best performance.
If the number of datasets changes (and the user/db-programmer usually knows if it changes) the planner sometimes make wrong decisions.
Then I have to execute a "vacuum analyze" or reconnect in case of foreign key checking.
I would like to tune postgresql to use an index in any case if it is available.
On Fri, 05 Oct 2001 18:01:08 -0400
Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Andreas Wernitznig <andreas(at)insilico(dot)com> writes:
> > Why don't you skip the automatic index creation for primary keys
>
> And how then would we enforce uniqueness of the primary key?
>
> > and let the user decide to create an index, that should be used in any
> > case, regardless what the query planner recommends ?
>
> What? AFAIR the planner does not even *know* whether an index is
> associated with a primary key. It certainly does not give such an
> index any special preference.
>
> regards, tom lane
>
From | Date | Subject | |
---|---|---|---|
Next Message | John Summerfield | 2001-10-05 23:56:38 | Re: Long options for psql in 7.1.3 |
Previous Message | Tom Lane | 2001-10-05 22:01:08 | Re: Bug #474: Index using problem |