QC workflow does not make any sense

[gouttegd]

I’ve just observed an unexpected failure of the qc workflow.

The failure is caused by the fact that the latest version of SSSOM-Py requires RDFLib 7.1.2 at least, but the package still declares to only require RDFLib 6.0.0. Because of that, when sssom is installed in the QC workflow, its rdflib dependency is assumed to be satisfied by the rdflib-7.1.1 package provided by the ODK, but when the sssom-py command is invoked it then predictably fails to load.

This can easily (and should) be fixed on SSSOM-Py side by correctly declaring the minimum required version of RDFLib, but looking at this failure made me realise that the QC workflow does not make sense at all, for two reasons:

(1) The workflow installs the latest published version of both sssom and sssom_schema:

      - name: Install latest SSSOM
        env:
          DEFAULT_BRANCH: master
        run: pip install -U sssom sssom-schema --break-system-packages

This means that we run on every single PR a workflow that does not actually test any of the changes made in the PR: we only test already published versions of sssom and sssom_schema. What is even the point of that?

(2) sssom validate exits with a zero (EXIT_SUCCESS) return code even when a mapping set fails to validate, meaning that the make validate_mappings step will never, in fact, validate anything: it will succeed even if some of the mapping sets are invalid.

[matentzn]

Pointless indeed.. 🥶 I added it to my things to look at next week but have some deadlines looming atm.

[gouttegd]

Until the workflow is revised to do something useful, I’ve disabled it, since it is currently broken and fails systematically (because of the incorrect dependencies on SSSOM-Py).