Re: ACTION-419: Sync with Birte on Datatypes for canonicalisation

On 29 Mar 2011, at 18:19, Andy Seaborne wrote:

> On 29/03/11 18:04, Bijan Parsia wrote:
>> On 29 Mar 2011, at 17:54, Birte Glimm wrote:
>>> Hm, so "01"^^xsd:integer becomes "1.0"^^xsd:decimal?
>>> That's what we do in HermiT since otherwise you wouldn't be able to
>>> properly implement OWL reasoning with facets, but then I don't
>>> understand the XML Schema Datatypes spec, which seems to suggest that
>>> "1" is the canonical form of "01" for integers (definitely not "1.0").
>>> In other places the spec says, however, that derived types inherit the
>>> canonical forms from the primitive type and for integer the primitive
>>> type is decimal, so from that one could understand "1.0" as decimal to
>>> be the canonical representation of integer "01".
>> Are you using Schema 1.1 (recommended even though not a recommendation!).
> Curious - why?

My and Boris's experience in the OWL WG is that the 1.1 specs are *much* superior in organization, clarity, and detail and well as nailing down various aspects.

> (W3C question: as it's still a WD is it going to progress?)

AFAICT, it's going to progress (checked for evidence just before emailing).

>> "All the following go to xsd:decimal normalized to remove all leading and trailing
>> zeros. Decimal point is not included if fractional part is zero.:"
>> So, "01"^^xsd:integer and "1.0"^^xsd:decimal both go to "1"^^xsd:decimal.
> Oh ... a bit annoying ...

That's a me artefact.

The canonical lexical form obviously doesn't include the datatype. So properly,

"01"^^xsd:integer has a canonical lexical form of "1".
"1.0"^^xsd:decimal has a canonical lexical form of "1". (in 1.1)

Of course, these are the same value, thus should have only one canonical  lexical form.

> requires the decimal point.

That is v. annoying!

But I see why. Integer and decimal share values. Thus, having "1" for the "integer" and "1.0 for "the decimal" isn't really meeting the definition of canonicality.

> As does Turtle in the short form.
> Canonicalization and setting the datatype are different.

Yep, my bad. Sorry.


Received on Tuesday, 29 March 2011 17:47:20 UTC