Closed mgeier closed 1 year ago
Thanks, but as discussed in #41, I think this is actually worse than the current implementation for generic quaternions.
Also, this would be a backwards-incompatible change. (There is actual data out there that relies on the current definition of squad
, which would be objectively wrong and could not be recovered if I changed the meaning of squad
.)
This changes the calculation of the quadrangle points according to https://splines.readthedocs.io/en/0.3.0/rotation/squad.html#Non-Uniform-Parameterization.
This way, centripetal parameterization leads to the nice properties that are expected from it, see the discussion in #41.
This can probably be made more efficient, because e.g. I'm repeating the call to
roll(t_in, -1)
.This is meant as a proof of concept, feel free to tweak.