On Wed, May 13, 2009 at 4:14 PM, Tobias Zahn <tobias-zahn@arcor.de> wrote:
> Hello,
> thank you for posting the paper, it was quite interesting to read. I
> think it would be a good idea to give the two-phase optimization a try.
> As far as I know and understand the current (geqo) optimizer source,
> many important parts are already there. For example, we can calculate
> the costs of a given join order. Therefore, it would only be necessary
> to write an algorithm witch chooses the right input for the cost function.
> I would be interested in your opinion.
I'm very interested in any improvements we can make to planning large
join nests. Unfortunately the paper seems to conclude that it's not
really feasible to use heuristics, as had been my hope, but I'd be
very interested in any other approaches we can come up with. I
probably do not have time to implement anything myself, but I'm happy
to help with ideas and code review.
...Robert