Re: [HACKERS] Fwd: Joins and links - Mailing list pgsql-hackers

From Bob Devine
Subject Re: [HACKERS] Fwd: Joins and links
Date
Msg-id 3783A893.5DC77215@cs.utah.edu
Whole thread Raw
Responses Re: [HACKERS] Fwd: Joins and links
List pgsql-hackers
Leon <leon@udmnet.ru> wrote:
> Earlier I proposed that links should be of type similar to tid,
> so inserts should be fed with values of tid. But this requires
> intermediate step, so there can be a function which takes primary
> key and returns tid, or as you say a function 
> last_touched('other_table_name') - this seems the best choice.

Beware of adding special purpose hard-links as a way to
skip the run-time value comparisons.  A link looks attractive
but it really only works for one-to-one relationships
(any multi-way relationships would require a list of links
to follow) and a link has all of the overhead that a
foreign key requires.

As somone who has developed several commercial dbms systems,
I would discourage doing a special "link" type.  There are
other ways to gain performance -- de-normalize your tables
if you are doing mainly reads; carefully check your storage
layout; and, of course, buy more RAM ;-)

--
Bob Devine  devine@cs.utah.edu


pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: [HACKERS] 6.5 cvs: views doesn't survives after pg_dump (fwd)
Next
From: Lamar Owen
Date:
Subject: Re: [HACKERS] Re: Postgresql 6.5-1 rpms on RedHat 6.0