On 22 April 2014 17:00, Stephen Frost <sfrost@snowman.net> wrote:
> Simon,
>
> This all looks good, and at the risk of being slightly off-topic for
> this thread, I just wanted to mention..
>
> * Simon Riggs (simon@2ndQuadrant.com) wrote:
>> Current proposal ends there, but there is a further optimization that
>> allows us to remove the join altogether if
>> * There is a FK between Fact and Dim
>
> It'd be great if we could start by looking for the above requirement and
> then doing join-removal when it exists and no columns from Dim are
> referenced (outside of the FK reference, which must result in 'true' or
> we've screwed something up).
>
> I had been looking at this about a month ago and just ran out of time to
> play with it, but it doesn't seem like it'd be terribly difficult to
> do..
Yeh, it was on my list. I had a few optimizer changes I've been
looking to make for a while.
Hopefully 9.5 is the year I get the chance at a running start on that
since they're never as easy as you'd hope.
-- Simon Riggs http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services