Re: [QUESTIONS] DANGER WILL ROBINSON! - Mailing list pgsql-hackers

From The Hermit Hacker
Subject Re: [QUESTIONS] DANGER WILL ROBINSON!
Date
Msg-id Pine.NEB.3.95.980309130102.23315P-100000@hub.org
Whole thread Raw
Responses Re: [QUESTIONS] DANGER WILL ROBINSON!  (Karl Denninger <karl@mcs.net>)
List pgsql-hackers
Redirected to 'the proper list' - pgsql-hackers@postgresql.org

On Mon, 9 Mar 1998, Bruce Momjian wrote:

> >
> > WARNING!
> >
> > Postgres 6.3 has MAJOR trouble with btree-indexed text fields.
> >
> > Performance levels are *10x* worse than the same indexed fields in "varchar"
> > format!
> >
> > Be EXTREMELY careful - I got bit in the ass by this this morning, and it was
> > very fortunate that I figured out what was going on.
> >
> > The reason I changed this over was that I had dumped the table and it came
> > out of the pg_dump program with a negative size.  So I figured I'd change it
> > to TEXT and that would resolve the problem.  BIG mistake.
> >
> > Be on guard for this folks.
> >
> > Developers, you might want to look into this - there's no good reason for
> > this kind of behavior, is there?
>
> No good reason at all.  As far as I know, text and varchar() behave
> identically in the backend, except for the input functions which limit
> the length of varchar.

    Karl...just curious, but what does an 'explain' show for the two
different situations?  10x worse almost sounds like the indices aren't
even being used, don't they?



pgsql-hackers by date:

Previous
From: Andrew Martin
Date:
Subject: Re: [HACKERS] How to...?
Next
From: Bruce Momjian
Date:
Subject: Re: [HACKERS] How to...?