Re: Should total_pages be calculated after partition pruning and constraint exclusion?

David Rowley <david(dot)rowley(at)2ndquadrant(dot)com> writes:
> On 16 May 2018 at 11:04, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> No, it should go under "planner improvement". If this were a bug fix,
>> it'd be a candidate for back-patch, which IMO it's not --- if only
>> because of the risk of plan destabilization.

> I'm not going to make a fuss over it, but I guess we must differ in
> opinion as I would count tracking relation sizes of relations we're
> actually not going to scan as a bug.

Dunno, the way in which total_pages is used is so squishy/heuristic
that it's really hard to say that changing the way we calculate it
is necessarily going to lead to better plans. I agree that this is
*probably* an improvement, but I don't think it's absolutely clear.

If there were some way to paint this as connected to other stuff
already done for v11, I'd be happier about pushing it in now --- but
it doesn't seem to be.

regards, tom lane