Interoperability forms and unit dialects: openff-units, ASDF, CF/UDUNITS, UCUM, QUDT

[dprada]

What. Reviewing how others serialize and exchange quantities (#83, "Alternatives and prior experience") surfaced conversions PyUnitWizard could offer. They would let a user move quantities between ecosystems in one call, and let PyUnitWizard's own serialization codec (#82) interoperate instead of standing alone.

Target	Kind	What conversion would mean	Why
openff-units	form (a pint subclass with its own registry)	convert(q, to_form="openff.units") and back	Molecular force-field ecosystem. Blocked today by cross-registry support (#84).
ASDF quantity tag	serialized form	read and write the value + unit node, with ndarray blocks	Astronomy's archival format; asdf-astropy already maps it for astropy, and a PyUnitWizard bridge would reach every backend
CF / UDUNITS unit strings	string dialect	parse and emit the units attribute of NetCDF and Zarr	Lets the codec write a CF units attribute next to its verified manifest, so CF tools keep working
UCUM codes	string dialect	parse and emit UCUM (mol/L, %, case-sensitive; M is mega, not molar)	The portable spelling used in clinical and biomedical exchange (HL7/FHIR, LOINC); proposed as the third-party spelling in #83
QUDT (and OBO UO) identifiers	semantic annotation	map a unit to its QUDT unit IRI and an optional quantity-kind IRI	Distinguishes Hz/Bq and J/N·m, which dimensions cannot; pint does not model kinds (pint#676, closed)
OpenMM (existing)	form	already supported	Listed for completeness; openff-units has its own to_openmm/from_openmm

How. Each item should be justified by a measured consumer case before it is implemented, following the lesson of #83's prior-art section. They can be optional extras loaded through DepDigest. Suggested order:

1. cross-registry pint (a prerequisite, and a defect today);
2. CF/UDUNITS and UCUM string dialects (needed by QuantityRecord: PyUnitWizard's native inert interchange form (verified quantity records) #82's bindings);
3. openff-units;
4. ASDF;
5. QUDT kinds.

Why here. PyUnitWizard is already the interoperability layer across pint, openmm.unit, unyt, astropy, physipy and quantities (devguide/interop_future_directions.md, "Ecosystem integrations"). Conversions written in each consumer diverge, and that divergence is where units get lost.

Related: #83 (review and design record), #82 (codec), #81.

[dprada]

Priority update (Diego, 2026-09-24): OpenFF compatibility is in scope. openff-units moves up to second place, right after the cross-registry prerequisite (#84):

1. Quantities from another pint UnitRegistry are not recognized and cannot be converted #84: recognize and convert quantities from another pint registry.
2. openff-units form: convert(q, to_form="openff.units") and back. Check that unit definitions agree (openff-units ships its own CODATA-based registry) and fail loudly if they do not. Add a conformance test that round-trips through OpenFF → PyUnitWizard → OpenMM and compares with openff-units' own to_openmm.
3. CF/UDUNITS and UCUM dialects.
4. ASDF.
5. QUDT kinds.

Why: OpenFF is the force-field ecosystem closest to MolSysMT and MolSysSuite. Its per-value {"val", "unit"} serialization is also the nearest precedent to #82 in our domain, so the codec should read and write OpenFF's representation as well.

[dprada]

OpenFF compatibility is now tracked on its own in #86. Item 2 of the priority list above lives there; this issue keeps the other forms and dialects (CF/UDUNITS, UCUM, ASDF, QUDT).