Subject: sign(exp(2009)) --> floating point overflow
From: Stavros Macrakis
Date: Sun, 18 Oct 2009 10:14:19 -0400
On Sun, Oct 18, 2009 at 9:55 AM, Dieter Kaiser <drdieterkaiser at web.de>wrote:
> ...all expressions with a constant like %e, %pi, ... are converted to a
> float. Therefore, we have the bug for sign(%pi^2009) too.
>
> This code is needed to get answers for expressions which involve the
> numerical constants....
>
Maxima should be using exact techniques as much as possible, especially
since we don't have any way to characterize (e.g. by interval arithmetic)
the accuracy of numeric results. It is not that hard to decide that
%pi^2009 or for that matter %pi^234234 - %e^234234 is positive without
numeric calculation.
-s