Filter sequence - Mailing list pgsql-general

From André Volpato
Subject Filter sequence
Date
Msg-id 472F6B67.2070306@ecomtecnologia.com.br
Whole thread Raw
Responses Re: Filter sequence  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
Hi all,

I´m experiencing an unexpected behaviour in the planner. I want the
planner to apply a function in the results of a subquery, but its doing
a filter in the hole table.
The results between the mixed filters are the same in the end, but its
taking ages. What I want is the planner to aplly the filters in the
query sequence, eg, filter the subquery, then filter its results again
using a function.

The table:
[ users ]
year | cod | name | age | sex
1997 |  123 | john | 23 | M
1997 |  456 | smith | 68 | M
1998 |  123 | john | 23 | M
1998 |  456 | smith | 68 | M
1999 |  789 | mary | 12 | F
...

The query:

select u2.cod, u2.name
from
(
    select u.cod, u.name
    from users u
    where age between 0 and 44 and sex='F'
    group by u.cod, u.name
) u2
group by u2.cod, u2.name
having
    getSalaryPeriod(1997,1999,u2.cod) > 1000

The function getSalaryPeriod is in pl/pgsql , and basically returns the
accumulated salary in the given period (1997 to 1999).

I want the planner to filter the user age and sex (wich restricts the
results and groups the user cod), and after that, run getSalaryPeriod
ONLY in the results of the subquery u2. Instead, its filtering the age,
sex and the salary in the same point :

Filter: ((age >= 0) AND (age<= 44) AND (sex = 'F'::bpchar) AND
            (getSalaryPeriod(1997 , 1999, (cod)::text, 0) >= 1000))

Right now, I am rewriting the query to filter age/sex in a temp table,
and after that, running getSalary.
This way is EXTREMELY faster, but I think that there must be a better way =)

Any hints ?

-- ACV






pgsql-general by date:

Previous
From: Keaton Adams
Date:
Subject: Is there a way to tell how far along a COPY is in the process?
Next
From: Scott Ribe
Date:
Subject: Re: running postgresql