feat(python): allow use of StringCache object as a function decorator

[alexander-beedie]

Ref: #9307.

The same trick (upgrading our context manager to do double-duty as a decorator) works nicely here too...

import polars as pl

@pl.StringCache()
def lets_get_schwifty() -> pl.DataFrame:
    df1 = pl.LazyFrame([
        pl.Series("key", ["bar"], dtype=pl.Categorical),
        pl.Series("val", [1]),
    ])
    df2 = pl.LazyFrame([
        pl.Series("key", ["bar"], dtype=pl.Categorical),
        pl.Series("val", [2]),
    ])
    return df1.join(df2, on=["key"], how="outer").collect()

print( lets_get_schwifty() )
# shape: (1, 3)
# ┌─────┬─────┬───────────┐
# │ key ┆ val ┆ val_right │
# │ --- ┆ --- ┆ ---       │
# │ cat ┆ i64 ┆ i64       │
# ╞═════╪═════╪═══════════╡
# │ bar ┆ 1   ┆ 2         │
# └─────┴─────┴───────────┘

[alexander-beedie]

Hmm, that's odd - the typos linter has suddenly become very slightly stricter by upgrading itself to 1.15, though we have it pinned it at 1.14.11 in requirements-lint.txt 🤔 (Ref: the initially failing lint test).

@stinodego, any idea why CI is out of sync with requirements?
In the meantime I'll see if I can tweak what it found so we pass the updated validation.

[stinodego]

@stinodego, any idea why CI is out of sync with requirements?

The CI doesn't use the Python package, it uses the typos Action. So it can be out of sync with the requirements.txt.

[ritchie46]

Related to some of these fixes. Can we prevent the typos check to check parameters and data in dataframes? That's a bit too constraining IMO.

[alexander-beedie]

The CI doesn't use the Python package, it uses the typos Action.

Ahh, gotcha; not a big problem, it only complained about up a small handful of extra stuff.
I'll bump our requirements-lint.txt to match the latest.

[stinodego]

Related to some of these fixes. Can we prevent the typos check to check parameters and data in dataframes? That's a bit too constraining IMO.

I think @ghuls set it up, maybe he knows some of the settings. I could look into it sometime.