nmwsharp
Bug..fix?
It seems like the log map computation did not actually take the orientation of a tangent vector into account. Tried this out with the vector heat demo and this appears to adjust with the correct tangent vector propagation...but not entirely sure why the sign inversion was necessary / why the sweep started from the other side if the real axis (presumably) is oriented in the tv direction.
Adding defaulting to not make other stuff blow up :)
Feel free to reject if unwarranted!
Thanks for submitting! This seems like a reasonable extension to me.
I think the sign inversion should not actually be needed. You might be getting tricked by the Polyscope visualization in the vector heat demo. There, the "branch cut" discontinuity in the angular color ends up at the -x axis, since it's computed via atan.
eh...why it not like 1i? and ah makes sense - will remove the inversion and resubmit
I think i is not defined as a constant on all platforms... std::complex<double>{real_part, imag_part} always works