On Mon, Jun 21, 2021 at 1:52 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> You're ignoring the fact that the plan shape we generate now is in fact
> *optimal*, and easily proven to be so, in some very common cases.
As I've said I don't reject the idea that there is room for
disagreement on the specifics. For example perhaps it'll turn out that
only a restricted subset of the cases that Robert originally had in
mind will truly turn out to work as well as hoped. But that just seems
like a case of Robert refining a very preliminary proposal. I
absolutely expect there to be some need to iron out the wrinkles.
> I don't
> think the people whose use-cases get worse are going to be mollified by
> the argument that you reduced their risk, when there is provably no risk.
By definition what we're doing here is throwing away slightly cheaper
plans when the potential downside is much higher than the potential
upside of choosing a reasonable alternative. I don't think that the
downside is particularly likely. In fact I believe that it's fairly
unlikely in general. This is an imperfect trade-off, at least in
theory. I fully own that.
> I'm willing to take some flak if there's not an easy proof that the outer
> scan is single-row, but I don't think we should just up and change it
> for cases where there is.
Seems reasonable to me.
--
Peter Geoghegan