Re: Use unique index for longer pathkeys.

On Mon, Jul 14, 2014 at 4:02 PM, Kyotaro HORIGUCHI <
horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> At Mon, 14 Jul 2014 11:01:52 +0530, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
wrote in <CAA4eK1+6b6Wjwf51oZMrL+mKFH8xUp9J-pEhQvoR8SE7sWyTWw(at)mail(dot)gmail(dot)com
>
> > On Fri, Jun 13, 2014 at 1:11 PM, Kyotaro HORIGUCHI <
> > horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> > I am bit worried about the extra cycles added by this patch as compare
> > to your previous patch; example
> > During trimming of paths, it will build index paths
(build_index_pathkeys())
> > which it will anyway do again during creation of index paths:
> >
create_index_paths()->get_index_paths()->build_index_paths()->build_index_pathkeys()
>
> I felt the same thing. On stupid measure to reduce cycles would
> be caching created pathkeys in IndexOptInfo. I'll try in the next
> patch.

I suggest to wait for overall review of patch before trying this out,
we can try to see what is the best way to avoid this if required, once
other part of patch is reviewed and proved to be problem free.

Other than this I agree with the other points you mentioned in mail
and may be you can just handle those in next version of your patch.

With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com