Shigeru HANADA <shigeru.hanada@gmail.com> writes:
> Just after my post, Fujita-san posted another v7 patch[1], so I merged
> v7 patches into v8 patch.
I've committed a modified version of this, but right after pushing it
I had a better idea about what the AnalyzeForeignTable API should do.
An issue that I'd not previously figured out is how analysis of an
inheritance tree could deal with foreign-table members, because it wants
to estimate the members' sizes before collecting the actual sample rows.
However, given that we've got the work split into a precheck phase and
a sample collection phase, that's not hard to solve: we could insist
that the FDW give back a size estimate in the precheck phase not the
sample collection phase. I'm off to fix that up ...
regards, tom lane