Try to efficiently evaluate relative R-factor with NumPy.
It is always safe to pass Ji argument. It will be ignored if it’s not needed. This is because even fully general R-factors sometimes do not depend on J_i.
