Feature request for colorbar support on geographic axes

[lwq-star]

Hello,

I am using the Ultraplot library to create a 2×2 grid of subplots. The first row consists of geographic axes, while the second row contains regular (non-geographic) axes.

I created a colorbar for the geographic plots, but I would like to position the colorbar between the two rows of subplots, spanning across both columns.

However, in my attempts, the colorbar always appears below the first subplot instead of being centered between the two rows.

How can I correctly place the colorbar in the middle between the two rows and have it span across the columns?

This issue is similar to #392.

import ultraplot as uplt
import numpy as np

fig, axs = uplt.subplots(nrows=2, ncols=2, proj=["npstere", "npstere", None, None])

rng = np.random.default_rng()
data = rng.random((100, 100))

cm = axs[0, 0].pcolormesh(data)
fig.colorbar(cm, loc="b", ax=axs[0, :], span=(1, 2))

[cvanelteren]

Yeah no problemo!

[gepcel]

Have pr #394 stopped working?

[cvanelteren]

From my read, I think the implementation is not at the level that GeoAxes can see the implementation. Tried running the code also thinking this would work but ran into an issue with my env and then didn't have time to look at it.

[gepcel]

It may not be an issue with GeoAxes. I tried it briefly and it doesn't work under normal conditions either

[cvanelteren]

@gepcel that's strange. I found the culprit locally. I am not tracing where I made the mistake. THe issue is caused by forcing the colorbar to the width of the parent.

I think it was caused by commit #58996bd1

[gepcel]

Just want to leave a note here: pr #681 doesn't actually close this issue. It just fix a regresstion that spanning colorbar should work since pr 394, but stopped working somehow. From what I test, spanning colorbar works again for default projection, but not for "npstere" like this example.

[cvanelteren]

Ah I see. I think this is tricky as the projection is not rectilinear -- need to think about how to deal with that logic.

[lwq-star]

It is also ineffective for the default projection.

import ultraplot as uplt
import numpy as np

fig, axs = uplt.subplots(nrows=2, ncols=2, proj=["cyl", "cyl", None, None])

rng = np.random.default_rng()
data = rng.random((100, 100))

cm = axs[0].pcolormesh(data)
fig.colorbar(cm, loc="b", ax=axs[0, :], span=(1, 2))

print(uplt.__version__)
2.1.9

[gepcel]

Hah, as of "default projection", I mean: fig, axs = uplt.subplots(nrows=2, ncols=2).