Suppress invalid argument warnings in inverse Mollweide projection #3283
Conversation
The Mollweide inverse projection generates Numpy invalid value warnings
for pixels outside of the projection. For example, the following code:
from matplotlib import pyplot as plt
plt.subplot(111, projection='mollweide')
plt.savefig('test.png')
produces this warning:
.../matplotlib/projections/geo.py:489:
RuntimeWarning: invalid value encountered in arcsin
theta = np.arcsin(y / np.sqrt(2))
This patch silences the warning by wrapping the inverse transformation
in a `np.errstate` context.
|
I also fixed the issue in #3282 I think this is a slightly better place to fix it since in only silences the warnings thrown when calculating the ticks which causes the warning above. Matploblib tries to calculate ticks slightly outside the plot limits which causes the warning above since the limits are equal to the allowed limits. If you silence it here you might also silence it for a transformation where it actually is a valid issue. But perhaps the use of numpys warning filters is better that the generic python one. |
I agree. Thanks for sharing your solution. Closing this PR.
See PR jenshnielsen#2, which you can merge into your own PR against matplotlib itself. |
|
Great, I have merged your pull request |
The Mollweide inverse projection generates Numpy invalid value warnings for pixels outside of the projection. For example, the following code:
produces this warning:
This patch silences the warning by wrapping the inverse transformation in a
np.errstatecontext.