I see that
SELECT (time '00:00', interval '1 hour') OVERLAPS (time '01:30', interval '1 day') AS "False";
(a case in the regression tests) returns false (currently labeled as
returning true, but that is another issue ;).
istm that this *should* return true, since times do wrap around the day
boundary. That is, this should be evaluated as though the second time
period is really a full day long, rather than evaluate to a time period
of zero length.
SQL99 *seems* to ask for the current behavior, rather than a (to me)
more intuitive wrapping behavior. Could someone check their
interpretation of the standard to see if there could be support for the
"wrapping interpretation"? Does anyone else think the first (current)
interpretation is a bit screwy??
- Thomas