Incorrectly identifies list[f64] column as Object when one of the inner lists contains Int about polars HOT 4 CLOSED

@stinodego this seems to be a regression.

from polars.

Indeed this is a regression. Simpler reproducer:

import polars as pl

s = pl.Series([[0.1, 1]])
print(s)

shape: (1,)
Series: '' [o][object]
[
        [0.1, 1]
]

There must be something in the Python-side constructor logic that falls back to objects in this case.

from polars.

I think this is high priority. Creating objects is really bad and leads to lots of unexpected things.

from polars.

I'm not sure if it's the exact same regression, but there was a similar example in #14324

import polars as pl

s = pl.Series([[{'B': 1}], [{'B': 2}, {'B': 'N/A'}]])
print(s)

shape: (2,)
Series: '' [o][object]
[
	[{'B': 1}]
	[{'B': 2}, {'B': 'N/A'}]
]

from polars.