Somebody wrote to me a few days ago that the BRIN cost estimation is
rather poor. One immediately obvious issue which I think is easily
fixed is the correlation estimate, which is currently hardcoded to 1.
Since a BRIN scan always entails a scan of the relation in physical
order, it's simple to produce a better estimate for that, per the
attached patch. (Note: trying to run this on expression indexes will
cause a crash. I need to complete that part ...)
There are other improvements we've been discussing, but I'm unclear that
I will have time to study them to get a useful patch quickly enough. If
somebody else wants to mess with this, please get in touch.
Thoughts?
--
Álvaro Herrera 33.5S 70.5W
