CompoundGeometry.from_points with control_points=None

Describe the bug When using CompoundGeometry.from_points with control_points=None, the class method fails.

To Reproduce

import sectionproperties.pre.sections as sections
from sectionproperties.analysis.cross_section import Section

points = [[0,0], [0,1], [1,1], [1,0], [1,0.25], [1,0.75], [2,0.75], [2,0.25] ]
facets = [[0,1], [1,2], [2,5],[5,4], [4,3], [3,0], [4,5], [5,6], [6,7], [7,4]]
holes = []

geometry = sections.CompoundGeometry.from_points(points, facets, holes)

TypeError: 'NoneType' object is not iterable

originating at here

Expected behaviour With control_points=None and material=None, I would expect Shapely representative points to be generated.

[normanrichardson]

@connorferster The “possible scenarios” you showed, are the exact complexity issue I was beginning to see myself, by allowing control_points=None in CompoundGeometry.

My intent behind this issue, was that I feel that for the user it would be ambiguous to allow control_points=None in Geometry and not allow control_points=None in CompoundGeometry. I feel this way about the behavior of the two functions in general, what the one does the other should do. That said, fundamentally there is nothing wrong with allowing different behaviors, and I see many viable solutions here:

1. CompoundGeometry.from_points is different from Geometry.from_points. We allow control_points=None in Geometry and document it nicely.
2. CompoundGeometry.from_points and Geometry.from_points both don’t allow control_points=None.
3. CompoundGeometry.from_points and Geometry.from_points both allow control_points=None.
4. We deprecate CompoundGeometry.from_points and Geometry.from_points

I think 4 is the most difficult solution as we then have to look hard at the whole library and consider other deprecations.

I see 3 as a doable but perhaps difficult solution. I also agree with Robbie’s feelings questioning scope. I can see another issue looming: a user will put in one set of data (say facets), and different data could come out! I think this would happen in “possible scenarios” 4 (two overlapping rectangles). A facet list goes in, and all of a sudden the facet list is now modified with additional entries, etc. I think a user may find this confusing. If this is not clear, let me know and I can make an example.

I see 1 and 2 as more or less equal, easy enough to implement. 2 would mean the user would have to be 100% explicit about the control_points, while reducing convenience.

If I’m missing solutions let me know.

[connorferster]

Baaaammm! Nailed it.

Will write some new tests for CompoundGeometry.from_points() and add a new section to Advanced Geometry in the docs before submitting the PR but here is the more robust .from_points() algorithm at work: