Replace NaNs in images, log warning

Closes #419

Simple example to illustrate:

import numpy as np
from matplotlib import pyplot as plt
import scopesim as sim
import scopesim_templates as sim_tp

src=sim_tp.galaxy("brown/NGC5866")
src.fields[0].data[70:80, 70:80] = np.nan
np.isnan(src.fields[0].data).sum()

sim.download_packages(["ELT", "Armazones", "MICADO"])
cmds=sim.UserCommands(use_instrument="MICADO")
opt=sim.OpticalTrain(cmds)

opt.observe(src)
plt.imshow(opt.readout()[0][1].data)

shows:

hole

and prints:

astar.scopesim.optics.fov - WARNING: Source contained NaN values, which were replaced by 0.000000.
astar.scopesim.optics.fov - The replacement value for NaNs in sources can be set in '!SIM.computing.nan_fill_value'.

The resulting image has a clear hole in the middle where the NaNs were, but still has some noise in there, as would be expected when setting the pixel values in the source to zero.

Setting the fill value to something intentionally ridiculously high: cmds["!SIM.computing.nan_fill_value"] = 1e20 results in:

brighthole

Also added a test with a similar setup to at least run that new code and make sure no NaNs remain.

I picked this particular place ScopeSim to put this because that's where the actually used part of the image is created. Any NaNs outside the FOV can be ignored anyway as they don't affect the final result. Note that this now applies only to ImageHDU-type source fields, so Tables are not checked. However, one would have to be particularly dense to create a source table with NaNs, I guess. Most of the NaN cases will come from users loading mythical FITS files from God-knows-where...

AstarVienna / ScopeSim

Replace NaNs in images, log warning #466

Codecov Report