add rml:URI to ontology, shapes and spec

[pmaria]

In a recent discussion in kg-construct/rml-questions#28 that it might be useful to support generation of URIs next to IRIs would be to facilitate Linked Data dereferencing. As HTTP only supports URIs, implementing IRI dereferencing inevitably requires IRI to URI mapping below the surface (For example how DBpedia did it). It might therefor be valuable to also be able to generate URIs.

Note by @DylanVanAssche on this:

Maybe we need to have rr:IRI and rr:URI in the new spec as rr:termType?
This way, mapping rules can explicitly being clear about this.

Is this indeed a useful enough feature to add to RML?

[bjdmeest]

I vote yes. generating RDF is thé thing RML does, if we don't support all types of RDF, then what are we doing?

[DylanVanAssche]

There's a proposal and people seems to like it. It is straightforward to support URIs once you have IRIs so implementation-wise this is not a problem. If we ever want to generate a KG that is actually Linked Data on the Web, we desperately need this.

So maintainers, can we fix the documentation?

[dachafra]

As there is a proposal, these are the tasks required to be implemented:

• Spec change: define rml:URI (@pmaria @andimou)
• Ontology change: add new rml:URI (@anaigmo )
• SHACL change: allow rml:URI as term type (@DylanVanAssche)
• Test case change: add test to verify diff between rml:IRI and rml:URI (@DylanVanAssche)