Robin Deits

Ok, thanks! I think I may have a fix for this...

Duplicate of https://discourse.julialang.org/t/reversediff-resultst-do-not-fit-theory/11549

Having geometry stored inside the mechanism would make RigidBodyTreeInspector's job much easier 👍

Also, the GJK implementation from GeometryTypes will need to be improved before it can be useful here. It's got two important features missing: - it repeatedly searches over all of...

Makes sense. I agree that having ambiguity about whether a copy is made is undesirable. What about having the constructors that take q, v, and s just be simple wrappers...

So I think I prefer option 3

The explicit check for zero seems to not be too expensive: ```julia julia> function f1(r) sqrt(r.sx * r.sx + r.sy * r.sy + r.sz * r.sz) end f1 (generic function...

> Interestingly @rdeits thought the whole concept of Flatten was questionable... That's entirely possible, but I don't actually recall. Can I help out here?

Aha, I see. Wow, that was one of the first packages I ever wrote in Julia, so I wouldn't be surprised if it's not entirely sound :slightly_smiling_face: . It was...

Yeah, Julia's pretty good at this stuff (but the only way to be sure is to try it and benchmark for yourself). But I think the rule of thumb is...