Add support for polars

[kklein]

As of now, covariates X, treatment assignments w and observey outcomes y can be provided as numpy datastructures (np.ndarray) or as pandas datastructures (pd.Series and pd.DataFrame) respectively.

A PR to allow for X to be scipy.sparse.csr_matrix is in the making: PR #86

It might be beneficial to allow for polars datastructures, too.

One question that might arise is how we deal with a potential additional dependency. Do we want to wrap every polars-dependent piece of code in a try-block that tries to import? Do we want to make polars a run dependency of metalearners?

If you'd like to use metalearners with polars please let us know. :)

[kklein]

One question that might arise is how we deal with a potential additional dependency. Do we want to wrap every polars-dependent piece of code in a try-block that tries to import? Do we want to make polars a run dependency of metalearners?

In this comment, Marco Gorelli suggests the following approach to avoid introducing a run-time dependency:

if (pl := sys.modules.get('polars', None) is not None and isinstance(data, pl.DataFrame):
    # Polars-related logic
    ...

[baggiponte]

Hey there! Sick project. I would suggest you look into narwhals by @MarcoGorelli and @FBruzzesi others for support for any dataframe backend.

narwhals is a very thin library that handles compatibility with a bunch of other dataframe backends (polars, dask, modin...). it's currently used, among others, by Altair, scikit-lego, and it's being integrated into Plotly.

Marco and Francesco are super nice and can provide a lot of support for this.