mperrin
commented
6 years ago 
Kinda pleased that all the webbpsf unit tests pass against this with no units-related changes. Guess I got the back-compatibility part right! ( #162 on the other hand did require some minor webbpsf changes for the custom optic classes there.)
And in the course of putting more units into the OpticalElement classes, I found and fixed two long-standing bugs we'd missed until now, with the NgonAperture class and the unit test for MFT forward & inverse transforms cancelling out. So that's a plus.
I will probably go ahead and merge this relatively soon, and then shift development attention more towards webbpsf.
This addresses #145 by making lots of things into astropy Quantities, primarily wavelengths and pixel scales. As a result we can now explicitly specify wavelengths in meters or microns or whatever and things will Just Work.
As proposed in #145, a decorator is added to
utilsthat allows back-compatible function arguments given as floating point values to be automatically converted to astropy Quantity lengths in meters. You can give other units explicitly as e.g.wavelength=2*u.micron. (This decorator is currently named "quantity_input", the same as the related but different function provided by astropy itself. I couldn't think of any better name apart from monstrosities like e.g.@back_compatible_quantity_input_with_slightly_different_behavior_than_the_astropy_one; please feel free to suggest improvements)In any case, this decorator is then applied liberally throughout poppy_core, fresnel, and optics.py. Many string formatting and FITS header writing code lines had to be fixed. I added various casts to floating point values in meters where needed, which is awkward and a bit convoluted but is needed given that not absolutely everything uses Quantities yet.
mperrin included the following code: https://github.com/mperrin/poppy/pull/165/commits