fix from_pandas object null array (#3733)

py-polars/polars/internals/construction.py

@@ -297,14 +297,14 @@ def _pandas_series_to_arrow(
     """
     dtype = values.dtype
     if dtype == "object" and len(values) > 0:
-        if isinstance(_get_first_non_none(values.values), str):  # type: ignore
+        first_non_none = _get_first_non_none(values.values)  # type: ignore
+
+        if isinstance(first_non_none, str):
             return pa.array(values, pa.large_utf8(), from_pandas=nan_to_none)
+        if first_non_none is None:
+            return pa.nulls(min_len, pa.large_utf8())
 
-        # array is null array, we set to a float64 array
-        if values.values[0] is None and min_len is not None:
-            return pa.nulls(min_len, pa.float64())
-        else:
-            return pa.array(values, from_pandas=nan_to_none)
+        return pa.array(values, from_pandas=nan_to_none)
     else:
         return pa.array(values, from_pandas=nan_to_none)
 

py-polars/tests/test_interop.py

@@ -181,14 +181,15 @@ def test_arrow_list_chunked_array() -> None:
 
 
 def test_from_pandas_null() -> None:
+    # null column is an object dtype, so pl.Utf8 is most close
     df = pd.DataFrame([{"a": None}, {"a": None}])
     out = pl.DataFrame(df)
-    assert out.dtypes == [pl.Float64]
+    assert out.dtypes == [pl.Utf8]
     assert out["a"][0] is None
 
     df = pd.DataFrame([{"a": None, "b": 1}, {"a": None, "b": 2}])
     out = pl.DataFrame(df)
-    assert out.dtypes == [pl.Float64, pl.Int64]
+    assert out.dtypes == [pl.Utf8, pl.Int64]
 
 
 def test_from_pandas_nested_list() -> None: