questions about DefinedNamespaces (OWL, RDFS, etc etc)

[jclerman]

I have some questions about the new DefinedNamespace:

1. Can we restore a non-private API to get the URI of the namespace itself? Something like the uri() method on the now-deprecated ClosedNamespace:

rdflib.OWL.uri()   # "http://www.w3.org/2002/07/owl#"

2. I couldn't find a documented way to get the complete set of URIs within a DefinedNamespace. It appears necessary to do something like this:

from rdflib import OWL, URIRef
owl_uris: Set[URIRef] = {OWL[str(x)] for x in OWL.__annotations__}  # this can be generalized to a function

It would be nice to have the above functionality implemented as, say, a method (on DefinedNamespaceMeta?). On ClosedNamespace, it looks like it was possible to get this result via dir(OWL), since __dir__ was defined on that class.

[edmondchuc]

Can we restore a non-private API to get the URI of the namespace itself? Something like the uri() method on the now-deprecated ClosedNamespace

You can get the URI of a DefinedNamespace by calling str() on it.

>>> import rdflib
>>> str(rdflib.OWL)
'http://www.w3.org/2002/07/owl#'

I couldn't find a documented way to get the complete set of URIs within a DefinedNamespace. It appears necessary to do something like this

I've added this as a feature in this PR #1626.

[jclerman]

You can get the URI of a DefinedNamespace by calling str() on it.

Thanks @edmondchuc! I realized that I also seem to have figured that out in another part of my brain, since I used it in the 2nd part of my question.

I've added this [a way to get the complete set of URIs within a DefinedNamespace ] as a feature in this PR #1626.

Nice, thank you. Regarding that PR, upon further thought, overriding __dir__ is perhaps not the best approach - it'd break code for anyone relying on the default behavior of dir(OWL). Is there any reason not to create a method uris() on the DefinedNamespace class instead?

Related but separate: I'm not clear why OWL, RDF, etc aren't designed as instances of the DefinedNamespace class, instead of subclasses of it. I would have thought that making them instances would be more intuitive - is there an advantage to them being classes?

[jclerman]

Regarding that PR, upon further thought, overriding dir is perhaps not the best approach - it'd break code for anyone relying on the default behavior of dir(OWL). Is there any reason not to create a method uris() on the DefinedNamespace class instead?

Update: Apologies, I realize now why a public method is a problem, since it could collide with a member of the namespace.