Unsupported aggregation functions return None on String columns

[orlp]

Reproduction:

df = pl.DataFrame({"s": ["test", "foo"], "l": [[1], [2]]})
print(df)
print(df["s"].sum(), df["s"].min(), df["s"].max())
print(df["l"].sum(), df["l"].min(), df["l"].max())

This prints:

shape: (2, 2)
┌──────┬───────────┐
│ s    ┆ l         │
│ ---  ┆ ---       │
│ str  ┆ list[i64] │
╞══════╪═══════════╡
│ test ┆ [1]       │
│ foo  ┆ [2]       │
└──────┴───────────┘
None foo test
None None None

I'd expect sum/min/max to give proper errors for unsupported dtypes instead of silently returning None.

[mcrumiller]

Slightly related, should a utf8 sum be a concatenation of all elements?

pl.Series(["a"]) + pl.Series(["b"])

Series: '' [str]
[
        "ab"
]

[mcrumiller]

@thomasfrederikhoeck yes, those are implemented.

[thomasfrederikhoeck]

@mcrumiller I was a bit to fast on the trigger (I reailised and deleted :-D )

[orlp]

@mcrumiller

Slightly related, should a utf8 sum be a concatenation of all elements?

I'd prefer not, the error message will redirect you to .str.concat which is the appropriate implementation.

[stinodego]

This has been fixed for lists.

We should still update String methods and (possibly?) make String sum to redirect to str.concat.