update schema in udfs

[ritchie46]

fixes #4157

[ritchie46]

@stinodego I get a flake8 linting error after a black run? Should we ignore that flake8 lint?

Black should decide the spacing.

[stinodego]

@stinodego I get a flake8 linting error after a black run? Should we ignore that flake8 lint?

Black should decide the spacing.

I'll check in a sec.

EDIT:

@ritchie46 This is a bug in black. There have been similar issues before, like this one. black should reformat when you add a trailing comma to a list/dict/etc. In this case, apparently black doesn't know how to.

We can take this to the black bug tracker. Easiest fix for now is to manually remove the trailing commas in both dicts, so you get:

def test_update_schema_after_projection_pd_4157() -> None:
    assert pl.DataFrame({"c0": [], "c1": [], "c2": []}).lazy().rename(
        {"c2": "c2_"}
    ).drop("c2_").select(pl.col("c0")).collect().columns == ["c0"]

Or actually this would be prettier, in my opinion (again with trailing commas removed):

def test_update_schema_after_projection_pd_4157() -> None:
    result = (
        pl.DataFrame({"c0": [], "c1": [], "c2": []})
        .lazy()
        .rename({"c2": "c2_"})
        .drop("c2_")
        .select(pl.col("c0"))
        .collect()
    )
    assert result.columns == ["c0"]

[ritchie46]

Alright. I'll see if we can update black. Maybe this bug is already resolved in a later version. Thanks for the explanation.

[stinodego]

Alright. I'll see if we can update black. Maybe this bug is already resolved in a later version. Thanks for the explanation.

I tried the latest black version, but that doesn't help, unfortunately.

Updating black is a good idea anyway, though. Does not seem to impact any code.