Are you suggesting a general rule that
passing zero arguments to a one-argument function is implementation defined?
Or only in the case of these four particular functions?
I guess I'm trying to think of a good
reason why we would not want an error in that case?
-Rob
"David A. Wheeler" <>
wrote on 03/07/2007 02:23:45 PM:
> > > robert_weir:
> > > >
> > > > I don't know if I'd favor having some implementations
give an
> error while others give zero. If someone has a document that
> silently fails (gives zero) and then gives an error when they give
> the spreadsheet to someone else, then this will certainly cause
> some user confusion.
> > >
> > > > Also, this is not an arcane function that will be used
by few
> people. This is not a Bessel function. MIN and MAX are
very basic
> functions that all will use. So the impact of confusion would
be
> multiplied. So I'd recommend having a single, complete, unambiguous
> definition for it.
> > >
> > > Okay. Giving MAX/MIN/etc 0 parameters _is_ a little
arcane, but
> I agree that the functions themselves certainly are NOT arcane. I
> see your point.
> > >
> > > In that case, I think we ought to return "0" when
handed 0
> parameters. Note: This means it should be "*" not
"+" in the arg list.
> > >
> > > Any objections?
> >
> > I was mainly referring the case where, for example, a range is
given but
> > there is no numerical value (in the case of MIN/MAX) in that
range. So,
> > since most apps return 0 then and given the explanation of Andreas
I'm
> > fine with specifying this as "return 0 if no values are
in the set",
> > which could also mean to not care about whether parameters are
handed or
> > not. It seems that currently only Gnumeric allows no parameters.
> >
> > Handing no arguments at all to those functions actually doesn't
make
> > sense. Do we still explicitly want to allow that and say that
apps
> > should return 0 then, or stay with the current {}+ syntax and
remove the
> > "apps should return 0" semantics instead?
>
> Oops, sounds like there's some miscommunication here; there are two
> different meanings of "no relevant values", and looks like
I
> understood something different.
>
> Here are the cases I see:
> * In the case where there _IS_ at least one parameter provided (e.
> g., a reference), but no numbers are found in the reference, I
> believe that we MUST return 0 in these cases. Most apps do so,
and
> people depend on it. Sounds like all are agreeing there.
> * In the case of "0 parameters", many apps (including Excel)
do NOT
> accept this at all. Proposal: Let's keep the {}+ notation, meaning
> that compliant apps are NOT required to accept 0 parameters for them
> (so an Error would be fine to return)... it'd be an implementation-
> defined extension to accept 0 parameters in those cases.
>
> --- David A. Wheeler
