I have a custom dtype in Pandas as well as an extension array. I would like to use the pyarrow back end over the default numpy backend.

See: .html#is-this-something-to-worry-about-now See: .DataFrame.convert_dtypes.html

On my extension array, I did implement __arrow_array__.

My question is what do I need to implement on the custom ExtensionDtype to get the convert_dtypes to work with the pyarrow back end?

Thank you