Found reviewing #60.
Which prefixes count as citations rather than vocabularies is now written down in two places that cannot see each other:
scripts/fleet/build_subsets.py:172 skips them with a literal if p in ("DOI","PMID"): continue, which is why they have no record counts.scripts/fleet/build_data.py declares CITATION=["PMID","DOI"] to pin them to the right of the heatmap.
Add a third citation prefix to one and the other keeps treating it as a vocabulary. In the heatmap that shows up as a column that sorts to the far left on Mech count with a record total of zero, which is exactly the failure the pin exists to prevent.
There is a related split in the same area. PREF in build_subsets.py lists DOI but not PMID, while VOC in build_data.py lists both, so PMID occurrences reach the heatmap only through prefix_census.py. Two scripts disagree about whether PMID is a prefix they handle.
Fix would be a single declaration the later scripts import, most naturally in build_subsets.py since it owns the skip.
Found reviewing #60.
Which prefixes count as citations rather than vocabularies is now written down in two places that cannot see each other:
scripts/fleet/build_subsets.py:172skips them with a literalif p in ("DOI","PMID"): continue, which is why they have no record counts.scripts/fleet/build_data.pydeclaresCITATION=["PMID","DOI"]to pin them to the right of the heatmap.Add a third citation prefix to one and the other keeps treating it as a vocabulary. In the heatmap that shows up as a column that sorts to the far left on Mech count with a record total of zero, which is exactly the failure the pin exists to prevent.
There is a related split in the same area.
PREFinbuild_subsets.pylists DOI but not PMID, whileVOCinbuild_data.pylists both, so PMID occurrences reach the heatmap only throughprefix_census.py. Two scripts disagree about whether PMID is a prefix they handle.Fix would be a single declaration the later scripts import, most naturally in
build_subsets.pysince it owns the skip.