Comment by minimaxir

You can do to_pandas() and from_pandas() which doesn't have much overhead, or to_numpy() directly which is typically zero-copy and thus no overhead: https://docs.pola.rs/api/python/stable/reference/dataframe/a...