Re: ISSUE-126 (Revisit Datatypes): A new proposal for the real <-> float <-> double conundrum

On 4 Jul 2008, at 16:09, Alan Ruttenberg wrote:

> On Jul 4, 2008, at 11:07 AM, Bijan Parsia wrote:
>
>> Because the floats are lexicographically ordered that means that  
>> if we increment the representation of a float as an integer then  
>> we move to the next float....
>
> Not if the next one is NaN.

NaN isn't in the sequence. In particular, I believe that all  
comparisons with NaNs are false (including equality). Thus the  
datatype >NaN is *necessarily* empty. As is =NaN, actually. To test  
for NaNness, you need a special predicate.

> Not every bit sequence in the floats corresponds to a number.

The sequences representing numbers do, and they are lexicographically  
ordered. You don't need to consider NaN when computing interval sizes.

Cheers,
Bijan.

Received on Friday, 4 July 2008 16:22:59 UTC