Simon Riggs <simon@2ndquadrant.com> writes:
> I would like to represent 'infinity' as interval->months = INT_MAX
Per the discussion upthread, what we need when creating an
infinite value is to set
month = INT32_MAX, day = INT32_MAX, time = INT64_MAX (for +Infinity)
month = INT32_MIN, day = INT32_MIN, time = INT64_MIN (for -Infinity)
else we'll need to put special-case logic into comparisons.
However, I think that we only need to restrict ordinary values from
not having month equal to INT32_MIN/MAX, that is the full range
of the other two fields can still be available for normal use.
So when *testing* for an infinity, you would only need to look at the
month field.
> Currently
> postgres=# select 'infinity'::timestamp = 'infinity'::timestamp;
> ?column?
> ----------
> t
Correct, we'd need that here too, if only because btree indexes and
sorting require the reflexive axiom to hold.
> so I was thinking that
> postgres=# select 'infinity'::timestamp - 'infinity'::timestamp;
> would be zero rather than an error, for least surprise.
Wrong. This case needs to be undefined, because "infinity"
isn't a specific value. That's what makes it okay to define, say,
infinity plus any finite value as infinity. There are very
well-defined rules about how to calculate with infinity, and
not following them is not the way to proceed here.
tl;dr: we should model it after the behavior of IEEE float infinities,
except we'll want to throw errors where those produce NaNs.
regards, tom lane