ericagol
commented
6 years ago It's useful to have analytic tests to compare with, and one that could be tried is a limb-brightened star. I derived the analytic formulae for this paper:
http://adsabs.harvard.edu/abs/2010ApJ...722L..75S
Of course, this is circularly symmetric, and so would involve a series of m=0 spherical harmonics (I think), but may still be an adequate test.
Another potentially useful test may be that the l -> \infinity case gives a flat light curve (since adjacent bright and dark regions will cancel out). Of course for finite l this will lead to rapid up and down fluctuations during the occultation, but of some amplitude $\propto l^{-2}$, so I'm not sure, whether a large enough l can be reached numerically to make this test useful.
Finally, a delta function expansion (http://mathworld.wolfram.com/SphericalHarmonicClosureRelations.html) can be computed analytically for a bright point on a sphere, and thus may prove to be a useful test as the occultation light curve can be easily computed analytically (it will be a step function which is unity when the point is on the visible hemisphere and unocculted). Again, I'm not sure how high of an l one would have to expand to in order to achieve sufficient numerical accuracy, but at least this series should converge, I expect, as l^{-2}.
rodluger
I tested the code a ton, but we still need a rigorous set of benchmarks in the tests/ folder