Primitive Rotation

[sjkelly]

Maybe use homogenous coordinates?

[sjkelly]

So actually rotating the implicit is kind of difficult. An easier approach is as follows.

f(p::Primitive,x,y,z) = n

We say x,y,z is in p if n < 0, on the surface if n = 0, and outside if n > 0.

So we leave the primitive definition at the origin. We keep a homogenous transform matrix (T) with each object. To then test point membership we apply the inverse of T to the point and use our origin-centered primitive.

Open questions: How to do bounding boxes? I think we will need all 8 points to do it robustly.

I am not sure how this will affect Rvachev functions.