fix (ontologies): Use Knora-specific properties instead of rdfs properties to check consistency during updates

[benjamingeer]

This pull request replaces rdfs:domain and rdfs:range, which we were misusing for consistency checking, with Knora-specific properties, following the discussion in #34 and #33.

• rdfs:domain is replaced with knora-base:subjectClassConstraint.
• rdfs:range is replaced with knora-base:objectClassConstraint and knora-base:objectDatatypeConstraint.

A subsequent pull request will enable us to use GraphDB's built-in consistency checking for both knora-base:subjectClassConstraint and knora-base:objectClassConstraint.

I've also cleaned up the ontologies a bit in general.

The old SALSAH export script will also need to be changed to output these Knora-specific properties instead of rdfs:domain and rdfs:range.

[tobiasschweizer]

In knora-base.ttl, shouldn't there be a knora-base:objectDataTypeConstraint on resourceIcon replacing the original rdfs:range (xsd:string)? As far as I understood, knora-base:objectClassConstraint is used to replace rdfs:range in case of an OWL-class, whereas knora-base::objectDataTypeConstraint should be used for a datatype like xsd:string. I understood that the latter case is not involved in consistency checking, but should still be there for documentation purposes.

[tobiasschweizer]

Could you update the knora-base documentation (docs/latex/knora-base/knora-base.tex)?

[benjamingeer]

In knora-base.ttl, shouldn't there be a knora-base:objectDataTypeConstraint on resourceIcon replacing the original rdfs:range (xsd:string)?

Yes. Fixed.

Could you update the knora-base documentation (docs/latex/knora-base/knora-base.tex)?

Done.

[benjamingeer]

I missed a few rdfs:range in the LaTeX doc, fixed now.

[tobiasschweizer]

Did you replace rdfs:domain and rdfs_range manually using a text editor?

I am concerned that - given the multitude of changes made to the ontology files - we cannot really assure that we actually did not change the ontology by accident.

[benjamingeer]

Did you replace rdfs:domain and rdfs_range manually using a text editor?

Yes.

I am concerned that - given the multitude of changes made to the ontology files - we cannot really assure that we actually did not change the ontology by accident.

If I made a mistake, it's your job as reviewer to find it. :) But all the tests still pass.

[benjamingeer]

Other things I changed, besides rdfs:domain and rdfs:range:

• Shorten a lot of rdfs:comment strings (and correct the English in them). We don't need such long comments in the ontologies now that we have separate documentation.
• Add copyrights to the ontology files.
• Use the knora-base prefix in knora-dc.ttl instead of full IRIs.

[benjamingeer]

And I took out the statements that imported ontologies into other ontologies. These seemed to be there only because of some limitation of Protégé, and that didn't seem like a good enough reason to keep them.

[tobiasschweizer]

We should be sure that when we make a lot changes that they were actually done in a consistent way.

What about counting all the original occurrences of rdfs:range and rdfs:domain and comparing them to the occurrences of subjectClassConstraint as well as objectClassConstraint and objectDataTypeConstraintsummed up?

Since it is a search and replace operation, the numbers should be the same.

In the actual branch:

grep "subjectClassConstraint" knora-base.ttl | wc -l

On development branch:

grep "rdfs:domain" knora-base.ttl | wc -l

As a reviewer, it is my job to mistakes and inconsistencies. But I cannot go through every line you changed.

[benjamingeer]

What about counting all the original occurrences... Since it is a search and replace operation, the numbers should be the same.

It wasn't just a search and replace operation. I had to look at every case manually, and I corrected a lot of things that were either incorrect or impractical, so just counting the changes won't work. For example, I removed owl:unionOf, which was used like this:

                   rdfs:domain [ rdf:type owl:Class ;
                                 owl:unionOf ( :Resource
                                               :User
                                               :Value
                                             )
                               ] .

I removed these because I think they probably won't work with constraint checking. Once constraint checking works, I can try to put them back.

I've just gone through all the changes in knora-base manually again.

These properties had owl:unionOf in rdfs:domain, which I removed:

• attachedToProject
• attachedToUser
• hasPermission
• hasRestrictedViewPermission
• hasViewPermission
• hasModifyPermission
• hasDeletePermission
• hasChangeRightsPermission
• seqnum
• isDeleted
• preferredLanguage

These properties had owl:unionOf in rdfs:range, which I removed:

• hasValue
• hasExtResValue

The property deleteDate had no rdfs:domain, but the comment says it only applies to resources, so I gave it a subjectClassConstraint.

The property phone had no rdfs:range, but is clearly a string, so I gave it an objectDatatypeConstraint.

These properties had an rdfs:domain of owl:Thing, which is meaningless, so I removed it:

• description
• name

The property protocol had a list of literals as an rdfs:range. I don't think this can work with constraint checking, and I don't see the point of having a list of all possible Internet protocols in the ontology, so I replaced it with an objectDatatypeConstraint of xsd:string.

Other changes:

I changed owl:qualifiedCardinality, owl:minQualifiedCardinality, and owl:maxQualifiedCardinality to owl:cardinality, owl:minCardinality, and owl:maxCardinality, because Knora doesn't support qualified cardinalities.

I changed some cardinalities that were clearly wrong (e.g. in LinkValue, where subject and object were optional but should be required).

I removed knora-base:mimetypes, which was taking a huge amount of space in the ontology, but isn't used in the application.

I removed the root user, which doesn't belong in knora-base.

As a reviewer, it is my job to mistakes and inconsistencies. But I cannot go through every line you changed.

What do you think I do when I review your changes? :)

[tobiasschweizer]

Ok, let me have a final look at it :-)

[tobiasschweizer]

In knora-base.tex, the examples still use rdfs:domain in chapter 3.

[tobiasschweizer]

Otherwise, the changes seem fine to me. In the Scala code, you simply had to replace one constant by another and adapt method, parameter and variable names.

[benjamingeer]

In knora-base.tex, the examples still use rdfs:domain in chapter 3.

Thanks, fixed now.