On Mon, 2007-05-03 at 19:02 +0100, Eike Rathke wrote:
> It seems people actually rely on the behavior that MIN/MAX return 0 if
> no value. I just wonder if there is any mathematical justification for
> that.
Well, on the totally ordered set of representable reals, the min of a
finite set is the inf of that set, and the inf of the empty set would be
the largest representable real (no mistake here, I really meant min and
largest).
That would surprise people even more. I don't think there is any real
conflict with mathematical definitions in this case, even if technically
a min (or max) should be a member of the set whose min/max is beinign
calculated.
> If there is, fine. If it contradicts, we should change it. I'd not
> suggest to make this implementation-defined, if all agree that returning
> 0 is the thing to do anyway, which Kspread, Gnumeric and Excel do, then
> OOo may as well follow.
>
I think we should stick with 0 even if personally I would prefer an
ERROR.
Andreas
--
Andreas J. Guelzow, Professor
Dept. of Mathematical & Computing Sciences
Concordia University College of Alberta
