Re: POC: GROUP BY optimization

>> I don't see why not to generate all possible orderings (keeping only the
>> cheapest path for each pathkey variant) and let the optimizer to sort it
>> out.
> 
> I'm assuming costing the full N! possible orderings would be
> prohibitively expensive.

That's true, but for the first step we need to improve cost_sort and only then 
try to find the best pathkeys order by optimal way.

> - If the user requested that order, we assume it "somewhat
> subjectively better" (by giving it a slightly reduced cost).
I don't think so. It's not a SQL way - DB should define the optimal way to 
execute query.


-- 
Teodor Sigaev                                   E-mail: teodor@sigaev.ru
                                                    WWW: http://www.sigaev.ru/