rio.write_crs() deletes the 'grid_mapping' attribute

[hershalpandya]

I have inherited a code which has many functions that do not expect CRS to be present in the xarray dataset. So the first thing the function does is to do ds.rio.write_crs() and after that other data manipulations happen . Whenever this is done, however, the CRS is written under coordinates group with the name specified in the ds.attrs['grid_mapping'] .

Dockerfile.txt
environmentdotyml.txt

The side effect of this is however that the attribute 'grid_mapping' attribute gets deleted from the ds. Is this an intended behavior? Because for my purpose, I need this attribute to stay. Has this been introduced in newer versions of rioxarray ?

[snowman2]

grid_mapping should be in encoding, not attrs. See: in #282

It is recommended to use rio.grid_mapping and rio.write_grid_mapping.

[hershalpandya]

Thanks!