Re: Ordered Partitioned Table Scans

On Sun, Mar 24, 2019 at 11:06 AM David Rowley
<david(dot)rowley(at)2ndquadrant(dot)com> wrote:
>
> On Sat, 23 Mar 2019 at 19:42, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> >
> > David Rowley <david(dot)rowley(at)2ndquadrant(dot)com> writes:
> > > On Sat, 23 Mar 2019 at 05:40, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> > >> BTW, another thing we could possibly do to answer this objection is to
> > >> give the ordered-Append node an artificially pessimistic startup cost,
> > >> such as the sum or the max of its children's startup costs. That's
> > >> pretty ugly and unprincipled, but maybe it's better than not having the
> > >> ability to generate the plan shape at all?
> >
> > > I admit to having thought of that while trying to get to sleep last
> > > night, but I was too scared to even suggest it. It's pretty much how
> > > MergeAppend would cost it anyway. I agree it's not pretty to lie
> > > about the startup cost, but it does kinda seem silly to fall back on a
> > > more expensive MergeAppend when we know fine well Append is cheaper.
> >
> > Yeah. I'm starting to think that this might actually be the way to go,
>
> Here's a version with it done that way.

FTR this patch doesn't apply since single child [Merge]Append
suppression (8edd0e7946) has been pushed.