Re: how to avoid repeating expensive computation in select - Mailing list pgsql-general

From Tom Lane
Subject Re: how to avoid repeating expensive computation in select
Date
Msg-id 25774.1296757004@sss.pgh.pa.us
Whole thread Raw
In response to how to avoid repeating expensive computation in select  (Bob Price <rjp_email@yahoo.com>)
Responses Re: how to avoid repeating expensive computation in select  (Pavel Stehule <pavel.stehule@gmail.com>)
Re: how to avoid repeating expensive computation in select  (Nicklas Avén <nicklas.aven@jordogskog.no>)
List pgsql-general
Bob Price <rjp_email@yahoo.com> writes:
> I would like to know if there is a way in PostgreSQL to avoid repeating an expensive computation in a SELECT where
theresult is needed both as a returned value and as an expression in the WHERE clause. 

Use a subselect.  You might need OFFSET 0 to prevent the planner from
"flattening" the subselect, eg

    SELECT whatever FROM
      (SELECT *, expensivefunc(value) AS score FROM mytable OFFSET 0) ss
    WHERE id LIKE '%z%' AND score > 0.5;

Keep in mind that in the above formulation, expensivefunc will be
evaluated at rows that don't pass the LIKE test.  So you probably want
to push down as much as you can into the sub-select's WHERE clause.
The planner will not help you with that if you put in the OFFSET 0
optimization-fence.  It's a good idea to use EXPLAIN (or even better
EXPLAIN VERBOSE, if you're using >= 8.4) to confirm that you're getting
the plan you want.

            regards, tom lane

pgsql-general by date:

Previous
From: "David Johnston"
Date:
Subject: Re: how to avoid repeating expensive computation in select
Next
From: "Wappler, Robert"
Date:
Subject: Re: set theory question