Follow-up from #72.
The current height-function spacing derives deltaX from cell volume (sqrt(V/thickness) in 2D and cbrt(V) in 3D). That is only the directional cell spacing for square/cubic cells. On stretched or anisotropic cells it becomes a geometric-mean size rather than the spacing along the selected height-function direction, which can bias curvature.
The spacing should instead be derived from local cell geometry in cols.direction() (or an equivalent directional measure).
A regression should use an anisotropic mesh and compare the height-function result against the expected directional spacing.
Follow-up from #72.
The current height-function spacing derives
deltaXfrom cell volume (sqrt(V/thickness)in 2D andcbrt(V)in 3D). That is only the directional cell spacing for square/cubic cells. On stretched or anisotropic cells it becomes a geometric-mean size rather than the spacing along the selected height-function direction, which can bias curvature.The spacing should instead be derived from local cell geometry in
cols.direction()(or an equivalent directional measure).A regression should use an anisotropic mesh and compare the height-function result against the expected directional spacing.