OBO Format does support prefixes now, but is skipped

[matentzn]

See

robot/robot-core/src/main/java/org/obolibrary/robot/IOHelper.java

Line 232 in c77d862

if (!df.isPrefixOWLOntologyFormat()) {

You cant inject a prefix into a OBO format serialisation pipeline (where OBO format is the last step), but you can inject it into some other format (like ofn), save it, and then convert to OBO in a separate pipeline.

I guess this could be an OWL API issue, falsely claiming OBO is not a isPrefixOWLOntologyFormat().

[gouttegd]

I guess this could be an OWL API issue, falsely claiming OBO is not a isPrefixOWLOntologyFormat().

It might be a OWLAPI issue, but it should not be an issue of the OWLAPI “claiming OBO is not a isPrefixOWLOntologyFormat()”, because the OWLAPI does not in fact do that:

public class OBODocumentFormat extends PrefixDocumentFormatImpl {

    /* ... */

    @Override
    public boolean isPrefixOWLOntologyFormat() {
        return true;
    }

    @Override
    public PrefixDocumentFormat asPrefixOWLOntologyFormat() {
        return this;
    }
}

So at least nominally, the OWLAPI is aware that the OBO format supports prefixes.

[gouttegd]

Pretty sure a problem with prefixes in OBO has been discussed on Slack some months ago (curse me for not immediately creating a ticket instead of letting the discussion disappear in the Slack blackhole).

As I recall, you noticed at some point that prefixes added with the --add-prefix options were ignored when writing to OBO. This is because, even though the OWLAPI was updated to declare the OBO format as a prefix-supporting format, not all parts of the OWLAPI were updated accordingly. In particular, there is nothing in the OWLAPI to automatically take care of providing the prefixes (“contained“ in the document format) to the OWL2OBO translator.

There is an easy fix in ROBOT, but only for the case when we write to OBO using the --clean-obo option of the convert command – because in that case, we are bypassing the OWLManager’s save method and instead instantiating our own OWL2OBO translator, so we can easily pass it the document format as a prefix manager:

// Assuming:
// - 'ontology' is the ontology to write to OBO
// - 'format' is a OBODocumentFormat object to which we already added all the prefixes we want
OWLAPIOwl2Obo oboConverter = new OWLAPIOwl2Obo(ontology.getOWLOntologyManager());
oboConverter.setPrefixManager(format.asPrefixOWLOntologyFormat())

For the general case when we write to OBO without any special option, the fix is out of reach for ROBOT, because in that case ROBOT relies entirely on the ontology manager object to perform the writing and never gets a glimpse of the OWL2OBO converter object, so setting the prefix manager must happen in the OWLAPI.