Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On Fri, Sep 02, 2005 at 11:03:24AM -0400, Tom Lane wrote:
>> I once started to make a btree opclass for XID, and stopped when it
>> occurred to me that XID comparison doesn't obey the transitive law.
> Not having it does affect the planner somehow, right?
> Maybe we could have the opclass but somehow dictate that making indexes
> with it is verboten.
The reason it affects the planner is that the planner assumes that
operators found in a btree opclass obey the normal laws of comparison.
Such an opclass would certainly break predtest.c for instance, as it
uses the assumption of transitivity directly.
(In any case Hannu's problem seems to be unrelated to the datatype,
see followups.)
regards, tom lane