... btw, it occurs to me that at least as a stopgap,
"set enable_nestloop = off" would be worth trying.
The killer problem with rowcount-1 estimates is that they
encourage the planner to use nestloops when it shouldn't.

                        regards, tom lane


Reply via email to