GeoQuadMesh.set_array cannot handle None

[kconnour]

Description

I have an rgb array of pixels that I'd like to project onto a globe to simulate what a spacecraft saw. To do this, I'm using pcolormesh. This requires me (as far as I know) to use set_array(None) in order to plot the rgb values (if I remove that part, matplotlib defaults back to the viridis colormap), which fails in the GeoQuadMesh.set_array method because that method expects A to be an ndarray. When I remove the check that A has dimensionality of at least 1, everything works as expected.

Code to reproduce

import matplotlib.pyplot as plt
import cartopy.crs as ccrs

# Setup the figure
rmars = 3400
fig = plt.figure(figsize=(6, 6), facecolor='k')
globe = ccrs.Globe(semimajor_axis=rmars * 1e3, semiminor_axis=rmars * 1e3)
projection = ccrs.NearsidePerspective(central_latitude=-10, central_longitude=250, satellite_height=6400 * 10 ** 3, globe=globe)
transform = ccrs.PlateCarree(globe=globe)
globe_ax = plt.axes(projection=projection)

# Attempt to plot a dummy set of rgb values at a dummy set of latitude/longitude points
globe_ax.pcolormesh(np.array([[1, 2], [3, 4]]), np.array([[1, 2], [3, 4]]), [[1]], color=[[0.2, 0.5, 0.6]], linewidth=0, edgecolors='none', rasterized=True, transform=transform).set_array(None)

Traceback

Traceback (most recent call last):
  File "/home/me/repos/myrepo/graphics/standard_products.py", line 265, in <module>
    globe_ax.pcolormesh(np.array([[1, 2], [3, 4]]), np.array([[1, 2], [3, 4]]), [[1]], color=[[0.2, 0.5, 0.6]], linewidth=0, edgecolors='none', rasterized=True,
  File "/home/me/myrepo/venv/lib/python3.9/site-packages/cartopy/mpl/geocollection.py", line 29, in set_array
    if A.ndim > 1:
AttributeError: 'NoneType' object has no attribute 'ndim'

Full environment definition

Operating system

Ubuntu 22.04

Cartopy version

0.20.2

conda list

I don't use conda

pip list

matplotlib 3.5.2
numpy 1.22.3

[dopplershift]

Looks like matplotlib does explicitly support set_array(None), so we should too.

[nvogtvincent]

Hi, is there any update on this? I've run into this issue several times because, at present, there is no way to plot an RGBA image using pcolormesh with cartopy (see here).

[kconnour]

@nvogtvincent I've heard nothing about this issue since raising it. However, with the recently released matplotlib v3.7.0, you no longer need to use set_array(None) with pcolormesh; it now natively supports RGB(A) arrays. I haven't tested it with Cartopy yet though.

[rcomer]

I opened #2166 to try to get pcolormesh working directly with RGB(A), but it turned out to be way more complicated than I thought and I don't know how soon we will complete it.

I suspect all that would be needed to support this workaround would be to skip the array-specific checks in set_array if A is None. That seems like an ideal task for a new contributor if any of you want to try it and, if it works, submit a pull request @kconnour, @nvogtvincent, @julietbravo. There is a brief guide to contributing here: https://scitools.org.uk/cartopy/docs/latest/contribute.html#contribute