Unit.sqrt( ) consistency error about openmm

[peastman]

A lot of numpy operation are implemented by calling methods on objects. As long as an object implements those methods, many things will "just work". You could actually make your code work by attaching units to a numpy array instead of to each element individually.

>>> unit_list1 = np.array([10, 20])*unit.femtoseconds
>>> np.std(unit_list1)
Quantity(value=5.0, unit=femtosecond)
>>> unit_list1 = np.array([10])*unit.femtoseconds
>>> np.std(unit_list1)
Quantity(value=0.0, unit=femtosecond)

[peastman]

Numpy isn't compatible with OpenMM units. It would be nice if you could pass quantities with units into arbitrary functions and have them always work, but that's often not the case. If you want to call Numpy functions, you should first convert the quantities to ordinary numbers in a chosen set of units.

[SchroedingersLion]

Thanks Peter, so mixing the two isn't best practice then.
To be frank I was surprised that one could even use numpy functions on OpenMM objects. How does numpy know what to do with the units anyway?