I didn’t realize that with Pandas 2.0, Pandas-to-Polars conversions can be zero copy if you use Arrow as the data frame format.