Working with Hashable type for variable names and mypy

[jorgsk]

Hi We'd like to run mypy as part of CI of an application that uses xarray but ran into a large number of barks from mypy because we perform string-operations on variable names from Datasets and DataArrays. These are type-defined as Hashable.

What would be the best way to work around something like:

with xr.open_dataset('my_ncfile.nc') as ncfile:
    new_names = [var_name.upper() for var_name in ncfile.data_vars]

Without adding # type: ignore to these lines?

[rhkleijn]

I guess str(var_name).upper() would do the trick.

[max-sixty]

Yes, think unavoidable when the values are read externally.

Potentially there could be some way of using generics to allow retaining type information between the ds construction and later code; others will know more as to whether that's possible

[jorgsk]

Thanks! In other places I've used cast('str', var_name) where cast is from the Typing module. I ended up sprinkling these changes in a handful of places and now mypy is happy.