On Thu, Nov 13, 2008 at 5:16 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Brendan Jurd" <direvus@gmail.com> writes:
>> I guess my question is, what's the real benefit of going to all this
>> trouble trying to prove that clauses are false?
>
> Not having to scan gigabytes of data in an excluded partition, for
> instance.
[after RTFMing ...]
The docs also say:
"When this parameter is on, the planner compares query conditions with
table CHECK constraints, and omits scanning tables for which the
conditions contradict the constraints."
I would normally interpret the above to mean that the planner *only*
performs these checks where a table CHECK constraint is relevant. I
dug up the original test case posted by Sergey, and his test table
didn't have any CHECK constraint on it at all (unless you count the
NOT NULL implied by PRIMARY KEY).
Cheers,
BJ