Optimising views - Mailing list pgsql-performance

From Bastiaan Olij
Subject Optimising views
Date
Msg-id 522001D7.6080500@basenlily.me
Whole thread Raw
Responses Re: Optimising views
List pgsql-performance
Hi all,

I've started using views in an attempt to try and simplify some of my
more complex reporting and running into some snags and I'm trying to
figure out how Postgres decides to join data with my view.

Just to present a simplified representation of what I am trying to
accomplish the following would match pretty well with what I'm doing.
At the center I have a table called "jobs" that holds information about
jobs clients have requested.
Related to my jobs are a number of different tables that each hold
different types of costs.
There is a simple table called "directcosts" that simply holds single
items that represent a cost on that job
There is a more complex invoice/detail structure for invoices received
by suppliers who did things in relation to the job
And a few more structures like that.

What I currently do when I need to report on the total costs per job,
and which works very well, is the following:
----
select jobid, jobdesc, sum(cost)
from (
  select jobid, jobdesc, dcamount as cost
  from jobs
  join directcosts on jobid = dcjobid
  where <some filter for my jobs>
union all
  select jobid, jobdesc, detamount as cost
   from jobs
   join invoiceheader on jobid = invjobid
  join invoicedetail on detinvid = invid
   where <some filter for my jobs>
) totalcosts
group by jobid, jobdesc
----

Work well enough.. But as I'm using the same data in different reports
and I though a view might be smart. So I created a view:
----
create view v_costs as
  select dcjobid as costjobid, sum(dcamount) as costamount
  from directcosts
  group by dcjobid
union all
  select invjobid as costjobid, sum(detamount) as costamount from
invoiceheader
  join finvoicedetail on detinvid = invid
  group by invjobid
----

And rewrote my report to:
----
select jobid, jobdesc, sum(costamount)
from jobs
join v_costs on costjobid = jobid
where <some filter for my jobs>
group by jobid, jobdesc
----

Now what I was hoping for was that postgres would start at my jobs
table, find the records I'm trying to report on and then index scan on
the related tables and start aggregating the amounts.
What it seems to do is to first execute the view to get totals for all
the jobs in the database and join that result set with my 2 or 3 jobs
that match my filter.

What is it about my view that prevents postgres to effectively use it?
The group bys? the union?

Cheers,

Bastiaan Olij


pgsql-performance by date:

Previous
From: bsreejithin
Date:
Subject: Re: How clustering for scale out works in PostgreSQL
Next
From: Mathieu De Zutter
Date:
Subject: Query plan change with multiple elements in IN clause