Basic operation results in invalid geometry

[MatthiasJ1]

p = Box(2,2,2) - Pos(-1.5) * Sphere(1)

[gumyr]

This is an OCCT problem with how the Edge of the sphere is aligned. Consider this version:

s = Pos(-1.5) * Sphere(1).rotate(Axis.Y, 180)
p = Box(2, 2, 2) - s

Note how the Edge of the sphere is rotated away from the cube which allows the boolean operation to work as expected.

I've considered constructing a sphere from two hemispheres but an clean operation would fuse them together again which doesn't solve the problem. This really needs to be fixed at the OCCT level.

[MatthiasJ1]

Another issue from this scenario: This creates valid geometry but the exported STEP file gives the union rather than subtraction.

p = Box(2,2,2) - Pos(0.7) * Sphere(0.7)

[gumyr]

I'm guessing the OCCT requires all Solids to have at least one Edge but it breaks so many things. I expect fixing it in OCCT would be difficult but I don't know.

[MatthiasJ1]

OCCT fundamentally requires surfaces to be closed in their parametric space. Removing that requirement in OCCT would be a monumental effort, however, the geometries here should be able to comply within that limitation.

[MatthiasJ1]

I believe it should be possible (it only needs to be closed in the 2d parametric space of the surface and you can create a circle with a single vertex and edge, so I imagine a clipped sphere shell is topologically equivalent) to create a parametrically closed sphere like the one below without any seams on the spherical surface but I have been unsuccessful so far. It would be useful in many scenarios to circumvent the seam issue.