Re: sumskipnan(nan) = 0 ?

[Søren Hauberg]

address@hidden skrev:
> On Sun Sep  9 21:15 , Søren Hauberg  sent:
> > > sum(nan) = nan.
> > I think you are misunderstanding the meaning of NaN. "NaN" means "Not a 
> > Number", so the addition of "a number" and "not a number" is really well 
> > defined. That is NaN+1 can't be a number, and hence must be NaN. Your 
> > arguments could to some extend work for "NA" (that is "Not Available"), 
> > but even then I wouldn't agree with your logic.
>
> The starting point was sumskipnan. If NaN+1=NaN by defintion, that function is
> useless.
>
> What do you mean/refer to when you say that NaN+1 is "well defined"?
Sorry -- I missed a "not" in  my reply. I meant to say, "not well defined".