Panel not reshaping to main plot for GeoAxes with Cartopy backend

[lixuan-code]

I don't know if you saw this issue, but when I drew the image on the right, the first thing that came to my mind was the panel() method in the example code. But it seems that the parameter 'high' is derived from the main axes. Although the top panel did a good job in completing the task in this image, I believe there may also be corresponding issues in certain specific images. I don't know how to solve this problem, although I can use the fig.add_axes() method to implement it, I believe someone will encounter corresponding problems and questions.

[lixuan-code]

I don't know if you saw this issue, but when I drew the image on the right, the first thing that came to my mind was the panel() method in the example code. But it seems that the parameter 'high' is derived from the main axes. Although the top panel did a good job in completing the task in this image, I believe there may also be corresponding issues in certain specific images. I don't know how to solve this problem, although I can use the fig.add_axes() method to implement it, I believe someone will encounter corresponding problems and questions.

Also, the feature you added last time was really useful and very helpful!

[cvanelteren]

Ah yes I will have to look into this as the cartopy backend will change the size of the main plot if you interect if you provide custom ranges.

I will be traveling pretty soon, probably have to push this until the new year.

[cvanelteren]

Can you provide a minimal and verifiable example for your code?

[lixuan-code]

Can you provide a minimal and verifiable example for your code?
Due to the time difference, I fell asleep just now. Sorry for only providing you with the code now. I don't know if it can.

import numpy as np
import xarray as xr

import cartopy.crs as ccrs
import ultraplot as uplt


TEST_DATA = xr.DataArray(
    data=np.cos(np.linspace(-np.pi * 4, np.pi * 4, 1800).reshape((10, 180))),
    coords=dict(y=np.linspace(30, 90, 10), x=np.linspace(-180, 180, 180)),
    dims=["y", "x"],
)
PROJ = ccrs.PlateCarree()

fig = uplt.figure(refwidth=4, share=0, figsize=("15cm", "10cm"))

main_map_ax = fig.subplot(proj=PROJ)  # type: ignore
main_map_ax.format(
    land=True,
    lonlines=np.arange(-120, 121, 60),
    latlines=np.arange(50, 81, 30),
    labels=True,
    latlim=(30, 90),
    latlabelrotation=90,
)

Meridional_diff_ax = main_map_ax.panel("top", pad=0.4)
Meridional_diff = TEST_DATA.median(dim="y").rolling(x=5).mean().squeeze()
Meridional_diff.plot.line(ax=Meridional_diff_ax)
Meridional_diff_ax.format(
    xticks=[-120, -60, -0, 60, 120],
    xtickloc="top",
    ytickloc="right",
    xlabel="",
    ylabel="",
    grid=False,
)
Latitudinal_diff_ax = main_map_ax.panel("right", width=0.4)
Latitudinal_diff = TEST_DATA.median(dim="x").rolling(y=5).mean().squeeze()
Latitudinal_diff.plot.line(ax=Latitudinal_diff_ax, y="y")
Latitudinal_diff_ax.format(
    xtickloc="top",
    ytickloc="right",
    xlabel="",
    ylabel="",
    grid=False,
)
uplt.show()

[cvanelteren]

The issue is caused by the figure settings. This overrides our gridspec that aligns the layouts. For example

import cartopy.crs as ccrs
import mpltern
import numpy as np
import xarray as xr

import ultraplot as uplt

TEST_DATA = xr.DataArray(
    data=np.cos(np.linspace(-np.pi * 4, np.pi * 4, 1800).reshape((10, 180))),
    coords=dict(y=np.linspace(30, 90, 10), x=np.linspace(-180, 180, 180)),
    dims=["y", "x"],
)
PROJ = ccrs.PlateCarree()

fig, main_map_ax = uplt.subplots(refwidth=4, share=0, projection=PROJ)

Meridional_diff_ax = main_map_ax.panel("top", pad=0.4)
Latitudinal_diff_ax = main_map_ax.panel("right", width=0.4)
main_map_ax.format(
    land=True,
    lonlines=np.arange(-120, 121, 60),
    latlines=np.arange(50, 81, 30),
    labels=True,
    latlim=(30, 90),
    latlabelrotation=90,
)

Meridional_diff = TEST_DATA.median(dim="y").rolling(x=5).mean().squeeze()
Meridional_diff.plot.line(ax=Meridional_diff_ax)
Meridional_diff_ax.format(
    xticks=[-120, -60, -0, 60, 120],
    xtickloc="top",
    ytickloc="right",
    xlabel="",
    ylabel="",
    grid=False,
)
Latitudinal_diff = TEST_DATA.median(dim="x").rolling(y=5).mean().squeeze()
Latitudinal_diff.plot.line(ax=Latitudinal_diff_ax, y="y")
Latitudinal_diff_ax.format(
    xtickloc="top",
    ytickloc="right",
    xlabel="",
    ylabel="",
    grid=False,
)
uplt.show()

Produces:

[cvanelteren]

I suspect that the cartopy axis has a way of redrawing the main plot. Not sure if we can address this directly. Will need to look at this in the future (new year). PR welcome.

[lixuan-code]

I suspect that the cartopy axis has a way of redrawing the main plot. Not sure if we can address this directly. Will need to look at this in the future (new year). PR welcome.

I chose this way of organising things because I also need to draw other subgraphs. As you can see, the whole graph is subgraph a. Additionally, I hope you have a good holiday and a pleasant journey.

[cvanelteren]

I did some digging. It is indeed the case that cartopy adjusts the size of the axes to ensure that the projection is visualized in an accurate manner. A fix should therefore be to adjust the panel locations for the geo axis, rather than stretching the main plot