EnhancedMeshVariable ± UWQuantity fails for compatible units (blocks units-active buoyancy T − T_ref)

[lmoresi]

Summary

Subtracting (or adding) a UWQuantity to/from an EnhancedMeshVariable raises even when the units are compatible:

TypeError: Cannot subtract UWQuantity from EnhancedMeshVariable. Check dimensional compatibility.

(with an upstream SympifyError: 0.633... [kilokelvin] — the quantity can't be sympified into the variable's expression path).

Minimal repro

m = uw.get_default_model(); m.set_reference_quantities(...)   # units active
T = uw.discretisation.MeshVariable('T', mesh, 1, degree=2, units='K')
T - uw.quantity(500.0, 'K')    # FAILS (TypeError) — units ARE compatible
T - uw.quantity(0.5,  'kK')    # FAILS too (not a unit-mismatch)
T.sym[0,0] - 500.0             # OK (workaround: drop to .sym + model-unit value)

Both compatible-unit cases fail, so it is not a dimensional-mismatch — the variable's arithmetic (MathematicalMixin / EnhancedMeshVariable) just doesn't accept a UWQuantity operand.

Why it matters

T − T_ref (a reference-temperature offset) is fundamental to buoyancy: F = ρ₀ α g (T − T_ref) ẑ. This blocks any units-active convection model, and is the current blocker for the thermal-convection units tutorial (PR #263) now that the SL trace-back crash (#267) is fixed.

Expected

meshvar ± UWQuantity (and the reverse) should work for compatible units — non-dimensionalise the quantity into the variable's ND expression and compose — relates to [[bare-variable composition]] (#137) / the MathematicalMixin units handling.

Exposed by PR #263.

Underworld development team with AI support from Claude Code

[lmoresi]

Fixed by #283 (merged): meshvar ± / * UWQuantity now reduces the quantity to its non-dimensional value before composing with the variable's ND .sym (exact: T - 500K == T.sym - 0.5; kK prefix consistent). Scope is variable-on-the-left; 'quantity - meshvar' (quantity on the left, handled by the UWQuantity class) is a separate follow-up.