On Mon, Oct 31, 2011 at 12:50 PM, Richard Fateman <fateman at eecs.berkeley.edu
> wrote:
> On 10/31/2011 8:46 AM, Raymond Toy wrote:
>
>>
>>
>> Do you want symbolic intervals too?
>>
>> Ray
>>
>> Symbolic endpoints are probably a bad idea. The complexity of the
> endpoint expressions grows very fast.
>
That's what I was thinking. It's useful for some quick tests, but anything
other than that is difficult.
I have written a interval package before. It's not particularly hard to
get the basics right, but the corner cases are a pain. Plus, I didn't do
directed rounding; that would have added more complexity. But it did get
the fact that x*x is non-negative.
Ray