Its worth drawing attention to the awesome work pandas are doing building in support for arrow data types right now. This actually means you can have zero copy

Its worth drawing attention to the awesome work pandas are doing building in support for arrow data types right now. This actually means you can have zero copy swaps between polars, pandas and duckdb with very little runtime impact. Definitely a big help if your using a non-pandas df library since the conversion to pandas when needed is now really cheap.