You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Found by the verifiers of #304. These exist on dev at 499b1bc after #304's grid fix.
The 1e-12 edge inset is not enough when fractional slice points carry float32 rounding (about 1e-8). BiSb with normal=(1,1,0), 30x30, padding 10, fractional mode leaves 12 NaN points at grid points (28,0) and (29,0). The comment at bandstructure2D.py:155 overstates the fix. A RegularGridInterpolator on the padded fractional mesh would remove the hull edge case.
Planes whose cut through the mesh is not a parallelogram (for example normal (1,1,1)) still leave about half the rectangular grid NaN.
No test uses a plane whose normal is not along kz: replacing the computed across axis with 2 (bandstructure2D.py:159) passes every test, and leaves 900/900 points NaN on a vertical plane.
The grid_interpolation docstrings say "(u, v) directions", but for normal=(0,1,0) axis 0 now runs along v.
PlaneInfo.u_limits/v_limits hold the slice's Cartesian bounding box, not the grid's extent.
A point with NaN height still gets NaN x and y from the 2 pi scale.
get_uv_bands_grid in plotter/bs_2d_plot.py is dead code.
Activity