One of the steps that is performed when preparing an import module is the execution of the src/sparql/preprocess-module.ru SPARQL query, which simply sets a dc:source ontology annotation:

INSERT { 
    ?ontology dc:source ?version_iri .
}

WHERE {
  ?ontology rdf:type owl:Ontology ;
        owl:versionIRI ?version_iri .
}

Running a SPARQL query with ROBOT requires (1) converting the entire ontology in memory from the OWLAPI model to the JENA model, (2) actually running the query on the JENA model, (3) serialising the JENA model to Turtle, and (4) parsing the serialised Turtle back into a OWLAPI model. Doing all of that for a task that could be done in 3 lines of Java code is a huge waste of resources.

In fact, the cost of that process is a big reason why some import modules need to be marked as “large”, so that we can selectively avoid that step for those modules:

$(ROBOT) \
    {% if ont.is_large %}extract -i $< -T $(IMPORTDIR)/{{ ont.id }}_terms_combined.txt --copy-ontology-annotations true --force true --method BOT \
    {% else %}query -i $< --update ../sparql/preprocess-module.ru extract -T $(IMPORTDIR)/{{ ont.id }}_terms_combined.txt --copy-ontology-annotations true --force true --method BOT {% endif %}

A simple ROBOT plugin could easily replace the SPARQL query. This would make the preprocessing much faster for all modules (since we avoid the whole roundtrip to JENA and back), and would remove the need for treating “large” imports differently, which would in turn make the Makefile template easier to read and maintain.

Likewise for the post-processing step, which simply strips all ontology annotations except the dc:source annotation.