Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Simon Riggs wrote:
>>> Oh, why would someone want to set enable_constraint_exclusion to false?
>>
>> The included functionality performs the exclusion at plan time. If a
>> query was prepared for later execution, it *could* return the wrong
>> answer when the plan was executed at a later time since plans are not
>> invalidated when constraints change. So, in general, this should be set
>> to false except for circumstances where the user can guarantee no such
>> mistake would be made.
> Ah, so there is a small additional restriction (changing constraints on
> planned queries) that this would affect.
This is a stopgap until we have automatic plan invalidation.
regards, tom lane