Float class with infinity and stuff, interval computations

[Fernando Perez]

Gerald Squelart wrote:

> Hi all,
> 
> Before I start coding away, I'd like to know if there's a module that
> implements a float-like class that includes +/-infinity, NaN and possibly
> +/-overflow.
> I had a quick look at the Vault, but didn't see anything like that.
> If it effectively does not exist, is anybody be interested? (i.e. should I
> publish it?)
> 
> Looking at the bigger picture (wow, sounds serious ;-), I'm working on an
> interval computations class. Same questions as above...

Take a look at numarray, which fixes some of the limitations of Numeric's 
IEEE floating point handling. But it's oriented towards array manipulations, 
so it may not quite fit what you have in mind (you may want to do scalar 
interval-based calculations).

Cheers,

f.