On 5/17/2012 1:37 PM, Raymond Toy wrote:
>
>
> On Thu, May 17, 2012 at 1:02 PM, Stavros Macrakis
> <macrakis at alum.mit.edu <mailto:macrakis at alum.mit.edu>> wrote:
>
> On Thu, May 17, 2012 at 3:50 PM, Raymond Toy
> <toy.raymond at gmail.com <mailto:toy.raymond at gmail.com>> wrote:
>
>
> I take this to mean we shouldn't have atan(inf) -> %pi/2.
> Which means erf(inf) and friends should just return a noun.
> But 0*inf always returns 0. Shouldn't we return und for that?
>
>
> We should either return und or leave it unsimplified. Similarly
> for inf-inf, inf^0, is(equal(inf,inf)), etc. which currently
> return nonsense results.
>
This topic has been addressed a few times. One (probably incomplete)
solution is to
represent inf as a rational number type, (1/0), und as (0/0) minf
as (-1 / 0) and combine them
as appropriate.
((rat) 0 0) multiplied by <anything> is then 0*anything / 0 or
und. (that is, ((rat) 0 0). )
oddly enough, a lot of the work is done by REMOVING code, namely no
longer checking
for division by zero.
some additional code to display 0/0 as und etc.
I wrote a paper on this...
RJF