Component array ordering flaky behavior

[giswqs]

One of the Binder examples is not working as expected.

import rioxarray

da = rioxarray.open_rasterio("bahamas_rgb.tif")
da = da.rio.reproject("EPSG:3857")

# Grab the mesh object for use with PyVista
mesh = da.pyvista.mesh(x="x", y="y", component="band")

# plot in 3D
mesh.plot(rgb=True, cpos="xy")

[banesullivan]

This is a known, flaky issue. I'll see what I can do to track it down

[banesullivan]

This test captures it and I currently have it marked as xfail until I figure out what is going awry:

pyvista-xarray/tests/test_rectilinear.py

Lines 103 to 114 in 4efc9eb

	@pytest.mark.xfail(strict=False)
	def test_rioxarray_multicomponent(bahamas_rgb):
	da = rioxarray.open_rasterio(bahamas_rgb)
	with pytest.warns(DataCopyWarning):
	mesh = da.pyvista.mesh(x="x", y="y", component="band")
	assert np.array_equal(mesh.x, da.x.values)
	assert np.may_share_memory(mesh.x, da.x.values)
	assert np.array_equal(mesh.y, da.y.values)
	assert np.may_share_memory(mesh.y, da.y.values)
	# Check multicomponent array
	values = da.values.swapaxes(0, 2).swapaxes(0, 1).reshape(-1, 3)
	assert np.allclose(mesh["data"], values) # TODO: this check is flaky which is very concerning

[banesullivan]

For now, I have removed that example from the notebook. - components should be accessed one at a time for now

[banesullivan]

I think this is still an issue -- keeping open

[giswqs]

No problem. I was cleaning up the issues that I opened long time ago on many repos