How to crop 2D array by dropping nan values outside

[zxdawn]

Hi, I'm trying to drop nan values of 2D Array and plot the image. However, I wanna keep the original distribution with nan values inside.

Here's an example:

import numpy as np
import xarray as xr
import matplotlib.pyplot as plt
import matplotlib.patches as patches

data = np.array([[np.nan, np.nan, np.nan, np.nan],
                 [np.nan, np.nan, np.nan, 3],
                 [1, np.nan, np.nan, 4],
                 [2, np.nan, np.nan, 5],
                 [np.nan, np.nan, np.nan, 6],])

da_xarray = xr.DataArray(data, dims=['y', 'x'])

fig, ax = plt.subplots()
da_xarray.plot()

# preferred image region
rect = patches.Rectangle((-0.5, 0.5), 4, 4, linewidth=3, edgecolor='r', facecolor='none')
ax.add_patch(rect)

I wanna plot the Array inside the red box.

I have tried to use dropna:

da_xarray.dropna('y', how='all').dropna('x', how='all').plot()

This will drop nan values between them.

Any idea how to plot valid data and keep the red box part?

Because the nan values are random in my real data, I can not use sel or isel to select them manually.

[dcherian]

da_xarray.where(da_xarray.notnull(), drop=True)?

[zxdawn]

Thanks for the suggestion. It gives the wrong result as same as applying drop on each axis twice:

I wanna get this result:

[dcherian]

oh right, yeah I think you'll have to do it manually.