Vocabulary base class / interface etc

[barthanssens]

Maybe over-engineering / over-thinking this a bit:

• new vocabulary abstract class or interface
• methods getNamespace(), getPrefix(), getClasses(), getProperties(), getIndividuals()
• separate packages for
  • W3C: (almost) all of current vocabularies
  • Schema.org (if needed)
  • EU Core Vocabularies (perhaps a bit niche, but I'd be interested in using these)
• "official" tool to generate Java class from OWL

[abrokenjester]

We have had several attempts to create some abstract class or interface for vocabularies in the past, including a design using enums, but they all had various downsides and were never really adopted (see a.o. https://openrdf.atlassian.net/browse/SES-1850). One problem is that static constants and inheritance don't really mix well, so you often end up with an abstract class that gives a template, but then every concrete vocabulary just reimplements/hides the whole lot again to make it work.

That being said, happy to discuss a fresh design attempt. Perhaps more modern Java techniques open up new options.

[abrokenjester]

As for a tool to generate Java classes from OWL: I assume you are "merely" talking about creating vocabulary constants here, and not about some sort of OO mapper?

[hmottestad]

If we are introducing new packages I think we should consider also moving away from the ALL_CAPS naming scheme. So we can have RDF.type and RDF.Property. I don't think we should go as far as to lowercase the class names, e.g rdf.type.

[barthanssens]

As for a tool to generate Java classes from OWL: I assume you are "merely" talking about creating vocabulary constants here, and not about some sort of OO mapper?

Correct, just creating a Java class (using a template engine or just string concatenation) with final static URIs for properties and classes, not an OO mapper

[barthanssens]

If we are introducing new packages I think we should consider also moving away from the ALL_CAPS naming scheme. So we can have RDF.type and RDF.Property.

It would help for vocabularies that have similar named properties and classes (which is a questionable practice IMHO), e.g. DCAT with dcat:dataset (property) and dcat:Dataset (class).

Currently we have to use something like HAS_DATASET and DATASET to avoid name clashes...

I don't think we should go as far as to lowercase the class names, e.g rdf.type.

Indeed, that would look a bit... weird ...

[barthanssens]

The vocabulary generator could (also) be part of the Console tool