Re: [Maxima-discuss] SIGN1 strangeness, was: Problem with bfloat

[Richard F. <fa...@be...>]

A solution to this problem of inf-inf (and a number of others)
is to tag each inf uniquely.  So then inf-inf is zero  if those infs are
identically tagged.

Thus

k :  limit(1/x,x,0);

k-k

returns 0

  On 9/3/2014 9:08 AM, Raymond Toy wrote:
>>>>>> "Robert" == Robert Dodier <rob...@gm...> writes:
>      Robert> On 2014-09-01, David Billinghurst <dbm...@gm...> wrote:
>      >> I:87^611$
>      >> M:91^211$
>      >> sign(sqrt(I)-M);
>
>      Robert> Thanks for the example. Here's a proposed fix. This causes $FLOAT to
>      Robert> always fail (even for Lisps which otherwise allow float infinity)
>      Robert> and therefore SIGN1 takes a different route and sign returns 'pos.
>
>      Robert> The patch assumes that (float inf) minus (float inf) != 0.
>      Robert> I'm pretty sure that's required by IEEE 754 (in fact I think the
>      Robert> result is supposed to be NaN) but I can't quote chapter & verse
>      Robert> at the moment.
>
> Yes, inf - inf is NaN, in IEEE754 arithmetic.  As is inf*0. I would
> add a comment to the code saying so for the next user who isn't so
> familiar with IEEE754.
>
> On the other hand, didn't you also write some kind of float-infinity-p
> function already? If not, we should probably add a float-infinity-p
> and float-nan-p functions just for cases like this.
>
> --
> Ray
>
>
> ------------------------------------------------------------------------------
> Slashdot TV.
> Video for Nerds.  Stuff that matters.
> http://tv.slashdot.org/
> _______________________________________________
> Maxima-discuss mailing list
> Max...@li...
> https://lists.sourceforge.net/lists/listinfo/maxima-discuss