What's missing
northness() and eastness() in xrspatial/aspect.py take a method argument and pass it to aspect(). Every test in xrspatial/tests/test_northness_eastness.py uses the default method='planar', so the method='geodesic' branch through these two wrappers has never run in CI on any backend.
The planar side is well covered: correctness, output range, NaN handling, custom names, and all four backends. The gap is only the geodesic branch of the two wrappers.
Why it matters
northness/eastness forward method, z_unit, and boundary to aspect(). If that wiring broke for method='geodesic', nothing would catch it. Geodesic aspect runs a different code path (ECEF projection, lat/lon coordinate grids) than planar, so the wrapper behavior should be pinned on its own rather than assumed from the planar tests.
Proposed fix
Add tests that assert:
northness(agg, method='geodesic') equals cos(deg2rad(...)) of geodesic aspect, with flat cells (aspect -1) mapped to NaNeastness(agg, method='geodesic') equals sin(deg2rad(...)) likewise- numpy, dask+numpy, cupy, and dask+cupy agree, reusing the geodesic raster helpers from test_geodesic_aspect.py
Test-only. No source change expected.
What's missing
northness()andeastness()inxrspatial/aspect.pytake amethodargument and pass it toaspect(). Every test inxrspatial/tests/test_northness_eastness.pyuses the defaultmethod='planar', so themethod='geodesic'branch through these two wrappers has never run in CI on any backend.The planar side is well covered: correctness, output range, NaN handling, custom names, and all four backends. The gap is only the geodesic branch of the two wrappers.
Why it matters
northness/eastnessforwardmethod,z_unit, andboundarytoaspect(). If that wiring broke formethod='geodesic', nothing would catch it. Geodesic aspect runs a different code path (ECEF projection, lat/lon coordinate grids) than planar, so the wrapper behavior should be pinned on its own rather than assumed from the planar tests.Proposed fix
Add tests that assert:
northness(agg, method='geodesic')equalscos(deg2rad(...))of geodesic aspect, with flat cells (aspect -1) mapped to NaNeastness(agg, method='geodesic')equalssin(deg2rad(...))likewiseTest-only. No source change expected.