Re: xsd:float and xsd:int

From: "Gary Ng" <Gary.Ng@networkinference.com>
Subject: xsd:float and xsd:int
Date: Wed, 3 Sep 2003 18:32:19 +0100

> 
> Hi all, I have a question regarding the desired treatment of
> floats and ints within an "OWL-DL" reasoner implementation.
> 
> The question is: do their value spaces overlap? (or interleaved rather)
> 
> XMLSchema says both float and decimal are different primitive types, but
> their semantics are defined mathematically. 

Agreed.  

> In addition, integer/int is
> a derived
> type of decimal limiting only decimal's pattern space. All their order
> relations
> are also identical.
> 
> Since both value spaces are defined mathematically, for example,
> 1.0^^decimal and 10.0e-1^^float and 1^^int
> should be considered identical, and 3.5^^float is considered greater
> than 3^^int.

Agreed.  OWL, however, does not use the order relationship.

> This has implication towards their classification in DL, as well as
> querying.
> Classes could be inconsistent or satisfiable depending on the treatment
> of float and int.
> And querying could be restrictive if one needs to know which type was
> used in the ontology in order to get the correct (or any) results.

Agreed.

> OWL specs says an implementation needs to support only string and
> integers,
> so if one were to support float as well, which is the desired behaviour?

Well, one should read the XML Schema specifications and see what it says.
The desired behaviour in OWL is dependent on the desired behaviour in RDF
which is, last time I looked, dependent on the XML Schema specifications.

> Cheers
> 
> Gary

Note that the question of whether xsd:integer and xsd:float have
intersecting value spaces is (potentially) different from whether an XML
Schema application is required/permitted to accept an integer where it is
expecting a float.

Peter F. Patel-Schneider

Received on Wednesday, 3 September 2003 13:57:38 UTC