Frame-level all_horizontal / any_horizontal

[hericks]

Description

I was recently looking at a question on stackoverflow concerned with checking if all values of a dataframe are True. While answering, I noticed that there are no frame-level methods to compute bitwise AND / OR horizontally across columns (pl.DataFrame.all_horizontal / pl.DataFrame.any_horizontal). This leads to the unpleasant side-effect that the most precise answer to the aforementioned question relies on pl.DataFrame.min_horizontal as follows.

import polars as pl

df = pl.DataFrame({
    "a": [True, True, None],
    "b": [True, True, True],
})

df.fill_null(False).min_horizontal().min()  

In contrast, a more explicit answer using pl.all_horizontal is a bit clumsy as one cannot rely on frame-level methods.

df.fill_null(False).select(pl.all_horizontal(pl.all().all())).item()

Apart from increased API consistency, adding pl.DataFrame.all_horizontal would enable the user to write the following code.

df.fill_null(False).all_horizontal().all()

Hence, I'd propose adding pl.DataFrame.all_horizontal and pl.DataFrame.any_horizontal to polars' public API.

Please let me know what you think. Once signed off, I'd be happy to take a shot at the implementation and open a PR. Thanks!

[orlp]

Just FYI, the optimal solution should never include fill_null but rather use ignore_nulls=False on all().

[hericks]

@orlp Thanks! Then, the aggregated result becomes None. Would you propose the following?

df.select(pl.all_horizontal(pl.all().all(ignore_nulls=False))).item() != None