Re: *_collapse_limit, geqo_threshold

Tom Lane wrote:
> "Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
>> You do, but it's been pretty rare in my experience, and we're
>> considering alternatives which give a lot less flexibility that this.
> 
> I dunno about "considering".  We've already wasted vastly more time on
> this than it's worth.  AFAIR there has never been one single user
> request for the ability to partially constrain join order.  I think we
> should do an enable_join_ordering boolean and quit wasting brainpower on
> the issue.

I think I've found it useful in the past[1], but I also think we
already have a way to give postgres such hints using subselects
and "offset 0".

Instead of SAP-DB's
> select * from (t1 join t2 on <whatever>) join t3 on <whatever>;
ISTM we can already do
> select * from (select t1 join t2 on <whatever> offset 0) as a join t3 on <whatever>;
which seems like a reasonably way of hinting which parenthesis
can be reordered and which can't.


Would these new proposals give (guc's or syntax hacks) anything that
I can't already do?



[1] http://archives.postgresql.org/pgsql-performance/2007-12/msg00088.php