It seems that the new generalization of the contouring for both structured and unstructured arrays has broken the ability to contour an array that uses np.nan to mask it.
This can reproduced by assigning a single element of any array as np.nan and then trying to call contour_array(). The result is an empty plot. I believe this is a difference from using plt.tricontour() to generate the contours rather than plt.contour().
One solution would be to bring back plt.contour() in the context of structured grids. It seems plt.tricontour() has a known issue with np.nan for masking https://stackoverflow.com/questions/28964453/set-mask-for-matplotlib-tricontourf
It seems that the new generalization of the contouring for both structured and unstructured arrays has broken the ability to contour an array that uses
np.nanto mask it.This can reproduced by assigning a single element of any array as
np.nanand then trying to callcontour_array(). The result is an empty plot. I believe this is a difference from usingplt.tricontour()to generate the contours rather thanplt.contour().One solution would be to bring back
plt.contour()in the context of structured grids. It seemsplt.tricontour()has a known issue withnp.nanfor masking https://stackoverflow.com/questions/28964453/set-mask-for-matplotlib-tricontourf