`from_pandas` incorrect type for all null columns

[Usernamelesss]

`from_pandas` incorrect type for all null columns

[stinodego]

This is a conscious design choice. See the linked PR for discussion. Ritchie's response:

I am not sure here. Why not Object type then? We have one.

I am also not sure because often pandas strings are objects. Pandas is ambiguous. We cannot fix that so we have to be a little more lenient here. You will now fix #14601 and for that fix, tons of other issues will open about strings not being correctly read.

To avoid the issue in the original description, explicitly specify a schema if you are working with non-string object columns that could be empty, e.g.:

import polars as pl
import pandas as pd

pd_df = pd.DataFrame({"A": [None]})
df = pl.from_pandas(pd_df, schema_overrides={"A": pl.Time})
print(df)

shape: (1, 1)
┌──────┐
│ A    │
│ ---  │
│ time │
╞══════╡
│ null │
└──────┘