CLN: Stopped dtype inference in sanitize_array with Index[object] (#58655)

Author: mroeschke

doc/source/whatsnew/v3.0.0.rst

@@ -244,6 +244,7 @@ Removal of prior version deprecations/changes
 - Removed extension test classes ``BaseNoReduceTests``, ``BaseNumericReduceTests``, ``BaseBooleanReduceTests`` (:issue:`54663`)
 - Removed the "closed" and "normalize" keywords in :meth:`DatetimeIndex.__new__` (:issue:`52628`)
 - Require :meth:`SparseDtype.fill_value` to be a valid value for the :meth:`SparseDtype.subtype` (:issue:`53043`)
+- Stopped performing dtype inference when setting a :class:`Index` into a :class:`DataFrame` (:issue:`56102`)
 - Stopped performing dtype inference with in :meth:`Index.insert` with object-dtype index; this often affects the index/columns that result when setting new entries into an empty :class:`Series` or :class:`DataFrame` (:issue:`51363`)
 - Removed the "closed" and "unit" keywords in :meth:`TimedeltaIndex.__new__` (:issue:`52628`, :issue:`55499`)
 - All arguments in :meth:`Index.sort_values` are now keyword only (:issue:`56493`)

pandas/core/construction.py

@@ -38,7 +38,6 @@
     ensure_object,
     is_list_like,
     is_object_dtype,
-    is_string_dtype,
     pandas_dtype,
 )
 from pandas.core.dtypes.dtypes import NumpyEADtype
@@ -555,9 +554,7 @@ def sanitize_array(
         # Avoid ending up with a NumpyExtensionArray
         dtype = dtype.numpy_dtype
 
-    object_index = False
-    if isinstance(data, ABCIndex) and data.dtype == object and dtype is None:
-        object_index = True
+    data_was_index = isinstance(data, ABCIndex)
 
     # extract ndarray or ExtensionArray, ensure we have no NumpyExtensionArray
     data = extract_array(data, extract_numpy=True, extract_range=True)
@@ -610,15 +607,8 @@ def sanitize_array(
 
         if dtype is None:
             subarr = data
-            if data.dtype == object:
+            if data.dtype == object and not data_was_index:
                 subarr = maybe_infer_to_datetimelike(data)
-                if (
-                    object_index
-                    and using_pyarrow_string_dtype()
-                    and is_string_dtype(subarr)
-                ):
-                    # Avoid inference when string option is set
-                    subarr = data
             elif data.dtype.kind == "U" and using_pyarrow_string_dtype():
                 from pandas.core.arrays.string_ import StringDtype
 

pandas/core/frame.py

@@ -5059,22 +5059,7 @@ def _sanitize_column(self, value) -> tuple[ArrayLike, BlockValuesRefs | None]:
 
         if is_list_like(value):
             com.require_length_match(value, self.index)
-        arr = sanitize_array(value, self.index, copy=True, allow_2d=True)
-        if (
-            isinstance(value, Index)
-            and value.dtype == "object"
-            and arr.dtype != value.dtype
-        ):  #
-            # TODO: Remove kludge in sanitize_array for string mode when enforcing
-            # this deprecation
-            warnings.warn(
-                "Setting an Index with object dtype into a DataFrame will stop "
-                "inferring another dtype in a future version. Cast the Index "
-                "explicitly before setting it into the DataFrame.",
-                FutureWarning,
-                stacklevel=find_stack_level(),
-            )
-        return arr, None
+        return sanitize_array(value, self.index, copy=True, allow_2d=True), None
 
     @property
     def _series(self):

pandas/tests/frame/indexing/test_setitem.py

@@ -782,20 +782,18 @@ def test_loc_setitem_ea_dtype(self):
         df.iloc[:, 0] = Series([11], dtype="Int64")
         tm.assert_frame_equal(df, expected)
 
-    def test_setitem_object_inferring(self):
+    def test_setitem_index_object_dtype_not_inferring(self):
         # GH#56102
         idx = Index([Timestamp("2019-12-31")], dtype=object)
         df = DataFrame({"a": [1]})
-        with tm.assert_produces_warning(FutureWarning, match="infer"):
-            df.loc[:, "b"] = idx
-        with tm.assert_produces_warning(FutureWarning, match="infer"):
-            df["c"] = idx
+        df.loc[:, "b"] = idx
+        df["c"] = idx
 
         expected = DataFrame(
             {
                 "a": [1],
-                "b": Series([Timestamp("2019-12-31")], dtype="datetime64[ns]"),
-                "c": Series([Timestamp("2019-12-31")], dtype="datetime64[ns]"),
+                "b": idx,
+                "c": idx,
             }
         )
         tm.assert_frame_equal(df, expected)