Josh Berkus <josh@agliodbs.com> writes:
> On 1/10/11 7:25 AM, Tom Lane wrote:
>> I'm a bit worried though that there might be other
>> cases where the estimator comes up with 1.0 selectivity but it'd still
>> be worth considering a bitmap scan.
> Well, I think the answer is to apply the other fixes, and test. If
> there are other cases of selectivity=1.0, they'll show up. People are
> pretty fast to complain if indexes aren't used, and we have a good
> production test case available once you implement the other operators.
"Implement the other operators"? I don't think we're on the same page
here. What I'm talking about is a one-line change in indxpath.c to not
short-circuit consideration of a bitmap indexscan.
regards, tom lane