Expand description
IfcBSplineSurface and its knotted / rational refinements.
§The same job as crate::curve::bspline, one dimension up
Read the parameters, check the invariants, evaluate nothing. What changes
in 2D is that every invariant now has a u form and a v form, and the
control points are a grid rather than a list. The grid is where files go
wrong: ControlPointsList is LIST OF LIST OF IfcCartesianPoint, and a
ragged inner list means the surface is not a tensor product at all.
§Which index is which
ControlPointsList[i][j] has i running along u and j along v.
So the outer list length is the u control point count and the inner length
the v count. Transposing them yields a surface that is plausible, is not
the one in the file, and passes every count check because the two knot
vectors are usually the same length in test data. This module’s accessors
are named u_* and v_* for exactly that reason, and the row/column
convention is asserted in the tests.
§Invariants checked here
- Control point grid is rectangular (no ragged rows).
sum(UMultiplicities) = u control points + UDegree + 1, and the v form.UKnots/UMultiplicitiesare parallel lists, and the v form.- Weights form a grid of the same shape as the control points, all positive.
Structs§
- BSpline
Surface - A borrowed view of any
IfcBSplineSurfacesubtype. - Control
Point Grid - The control point grid,
[u][v].
Enums§
- BSpline
Surface Form IfcBSplineSurfaceForm: what shape the surface originally was.
Type Aliases§
- Surface
Knot Vector - Marker kept so the curve module’s knot types are visibly reused.