unifying symbolic and numeric




Robert Dodier wrote:

>
>Eventually I'd like to see a unification of the symbolic and numerical stuff.
>
>  
>
I agree, in some sense.

 Beware though that anything that comes from the numerical/ hardware 
side of the community has constraints that start with  the fixed 
finite-size representation premise. That is, you have some number of 
bits, which could be 32, 64, 80, 128.. to encode each object.  This 
leads to problems like trapping overflows, exceptional operands etc. 
because there are more than 2^32 (etc) different rational numbers. and 
even more real numbers :)

Ideas from the computer algebra community generally are motivated from 
how we can "arithmetize" mathematics so as to reflect in the computer 
all the properties we are used to, starting with integers.

The hardware community viewpoint cannot even express all integers.

A plausible viewpoint: We can treat the non-exceptional hardware floats 
as a special speed-optimized representation for a subset of the rational 
numbers.  We can construct larger subsets... But Inf and NaN are not 
rational numbers.