Re: Planner creating ineffective plans on LEFT OUTER joins

On Thu, 2008-06-26 at 12:57 -0400, Tom Lane wrote:
> Simon Riggs <simon@2ndquadrant.com> writes:
> > On Thu, 2008-06-26 at 12:36 -0400, Robert Haas wrote:
> >> It's my understanding that the philosophy of the PGDG in the past has
> >> been to avoid putting any kind of hints into the system, focusing
> >> rather an improving the planning of queries.
> 
> > It's not a specific hint, its a general goal setting.
> 
> Right.  There are definitely places where we've made engineering
> judgements to not attempt a particular type of optimization because it'd
> be too expensive compared to the typical payoff.  Simon's idea has some
> merit for providing a framework to deal with that type of situation.
> However, just adding a GUC variable isn't going to make anything happen
> --- we'd need some concrete plans about what we'd do with it.

Well, I'm convinced the egg came first. 

So I figure to put the framework in place and then start reviewing
things to see if they can be categorised. Plus I want new optimizer
features to be considered in the light of the new framework. This also
allows us a way of handling optimizer performance bugs. We just
reclassify certain cases as being costs-more solutions, rather than
stripping the code out entirely.

-- Simon Riggs           www.2ndQuadrant.comPostgreSQL Training, Services and Support