geo.rotate_in_cylindrical_cs performance issue

[xcast-ppi]

Hello devs,

I have developed a PyDPF tool to extract stress fields from a mechanical RST file and also rotate these stresses to local cylindrical coordinate systems. I've noticed a slow performance of the operators to rotate fields to another coordinate systems. As I need to process thousands of time steps for dozens of scoped element selections, this is an issue (1-2 sec per rotation alone). The other operations such as stress filed creation or averages are much faster (0.01 sec or less). To mitigate this issue. I could replace the cartesian coordinate systems rotation by plain Python rotation (very fast) using the coordinate system matrix, but not for cylindrical rotations because my results do not match Ansys Mechanical results. I don't know what operations are done internally for cylindrical cases and can't find it documented.

Is there a documentation about the math behind the cylindrical rotation (rotate_in_cylindrical_cs)? It does not appear to be only R @ Sigma @ R.T matrix operation.
Is there any reason for the low performance of this operator?

Despite this problem, PyDPF is a great resource. Thanks for developing it!