Re: Float and double in constants, but not descriptions (was Re: A summary of the proposal for a datatype system of OWL 2 (ISSUE-126))

On Jul 23, 2008, at 2:34 PM, Michael Smith wrote:

> If others feel this makes the burden of implementation too high, I  
> think
> a compromise is to not support float and double in class restrictions.

FWIW, this is the way I would lean. The only issue is sanity checking  
the constants, as the float and double lexical productions are the same.

So "2.0e2000"^^xsd:float and "2.0e2000"^^xsd:double are both valid,  
though both map to INF, as best I understand.

Perhaps there is nothing to do about this.

Also, I have a nagging feeling about there being a need for two INFs,  
one for double and one for float, but maybe this is misplaced. The  
issue is that the range of real that INF maps to is different for  
floats and doubles.

So "2.0e200"^^xsd:float maps to INF (from a float point of view) but   
"2.0e200"^^xsd:double maps to a real value.

-Alan

Received on Thursday, 24 July 2008 09:52:04 UTC