Thomas Lockhart <lockhart@alumni.caltech.edu> writes:
> date: 1999/05/31 19:32:47; author: tgl; state: Exp; lines: +61 -5
> Generate a more specific error message when an operator used
> in an index doesn't have a restriction selectivity estimator.
> Tom, was there anything more here than the new elog error exit itself?
> It used to ignore the missing estimator, or fail farther in to the
> code?
That code useta look something like
fmgr(get_oprrest(operatorOID), ...)
so that if get_oprrest returned 0 you'd get an error message along the
lines of "fmgr: no function cache entry for OID 0". This was pretty
unhelpful, of course, and someone complained about it a few weeks ago;
so I added a test for missing oprrest. That wasn't what broke things
... what broke things was my removal of seemingly bogus oprrest links
from pg_operator, which I think I did on 4/10:
revision 1.56
date: 1999/04/10 23:53:00; author: tgl; state: Exp; lines: +99 -99
Fix another batch of bogosities in pg_operator table.
These were bogus selectivity-estimator links, like a '>' operator
pointing to intltsel when it should use intgtsel.
regards, tom lane