Re: Performance on inserts

On Sat, Aug 26, 2000 at 05:09:53PM -0400, Tom Lane wrote:
> Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
> > Jules Bean <jules(at)jellybean(dot)co(dot)uk> writes:
> >> Is there any chance you could generate a patch against released 7.0.2
> >> to add just this functionality... It would be the kiss of life for my
> >> code!
>
> > Will look at it.
>
> I looked at it and decided that I don't want to mess with it. The
> BTP_CHAIN logic in 7.0 is so weird and fragile that it's hard to tell
> what will happen if we try to split anything but the last page of a
> chain of duplicates. The idea I'd had of just dropping in the whole
> btree module from current sources doesn't look very practical either;
> a lot of changes that span module boundaries would have to be backed out
> of it.
>
> My recommendation is to try out current sources (from a nightly snapshot
> or CVS). I realize that running a non-released version might make you
> uncomfortable, and rightly so; but I believe that current sources are in
> pretty good shape right now. In any case, working out any bugs lurking
> in current strikes me as a much more profitable activity than trying to
> back-patch the old btree code.

OK. Thanks very much for going through this with me.

I'm actually going to simply do without the index for this release of
my software -- it's an internal project, and the speed problems of
having no index aren't disastrous. However, I'd rather not fight any
instabilities, this project runs long-term and mostly unattended.

I'll certainly keenly upgrade to 7.1 when it comes out, though, with
the new btree logic.

Jules