Data Model Integrity Checks

[twalmsley]

HQDM has a lot of flexibility in how a data model is created, with none of the relationships being mandatory in the current code. This is not a problem in itself, but it raises the issue of how should HQDM be used and how can we check that the data model is not missing some required triple in the Jena database. Data Integration between databases will be easier if they all use HQDM in the same way.

One way is to implement a series of SPARQL data integrity checks that can be run against a database to check whether it complies with the guidelines agreed on by some interested community. For example, this SPARQL query CONSTRUCTS some triples to indicate when some required triples are missing. It checks that every participant_in an association is also a member_of_kind of at least one role (the constructed triples are only in the query results, not the database):

PREFIX hqdm: <http://www.semanticweb.org/hqdm#>

construct {
  ?s hqdm:error_participant_with_no_role "Should be a member_of_kind of a role.".
  ?s a ?type.
}
where {
    ?s hqdm:participant_in ?association;
       a ?type.
  
    ?association a hqdm:association.
  
    OPTIONAL {
      ?s hqdm:member_of_kind ?kind.
      ?kind a hqdm:role.
    }

    FILTER(!bound(?kind))
}

This might need to be a separate data-integrity or data-validation module with options to run it against local or remote SPARQL endpoints or against in-memory databases as part of unit tests in core.

[twalmsley]

Theses checks were needed to get #38 working with the right test data, so they will be included in the same PR.