BUG: Fixed bug when creating new column with missing values when setting a single string value (#56321)

Author: MarcoGorelli

doc/source/whatsnew/v2.2.0.rst

@@ -618,6 +618,7 @@ Indexing
 - Bug in :meth:`DataFrame.loc` when setting :class:`Series` with extension dtype into NumPy dtype (:issue:`55604`)
 - Bug in :meth:`Index.difference` not returning a unique set of values when ``other`` is empty or ``other`` is considered non-comparable (:issue:`55113`)
 - Bug in setting :class:`Categorical` values into a :class:`DataFrame` with numpy dtypes raising ``RecursionError`` (:issue:`52927`)
+- Fixed bug when creating new column with missing values when setting a single string value (:issue:`56204`)
 
 Missing
 ^^^^^^^

pandas/core/indexing.py

@@ -68,6 +68,7 @@
 from pandas.core.construction import (
     array as pd_array,
     extract_array,
+    sanitize_array,
 )
 from pandas.core.indexers import (
     check_array_indexer,
@@ -1876,7 +1877,13 @@ def _setitem_with_indexer(self, indexer, value, name: str = "iloc"):
                                 return
 
                             self.obj[key] = empty_value
-
+                        elif not is_list_like(value):
+                            # Find our empty_value dtype by constructing an array
+                            #  from our value and doing a .take on it
+                            arr = sanitize_array(value, Index(range(1)), copy=False)
+                            taker = -1 * np.ones(len(self.obj), dtype=np.intp)
+                            empty_value = algos.take_nd(arr, taker)
+                            self.obj[key] = empty_value
                         else:
                             # FIXME: GH#42099#issuecomment-864326014
                             self.obj[key] = infer_fill_value(value)

pandas/tests/frame/indexing/test_indexing.py

@@ -1935,6 +1935,26 @@ def test_adding_new_conditional_column() -> None:
     tm.assert_frame_equal(df, expected)
 
 
+@pytest.mark.parametrize(
+    ("dtype", "infer_string"),
+    [
+        (object, False),
+        ("string[pyarrow_numpy]", True),
+    ],
+)
+def test_adding_new_conditional_column_with_string(dtype, infer_string) -> None:
+    # https://github.com/pandas-dev/pandas/issues/56204
+    pytest.importorskip("pyarrow")
+
+    df = DataFrame({"a": [1, 2], "b": [3, 4]})
+    with pd.option_context("future.infer_string", infer_string):
+        df.loc[df["a"] == 1, "c"] = "1"
+    expected = DataFrame({"a": [1, 2], "b": [3, 4], "c": ["1", float("nan")]}).astype(
+        {"a": "int64", "b": "int64", "c": dtype}
+    )
+    tm.assert_frame_equal(df, expected)
+
+
 def test_add_new_column_infer_string():
     # GH#55366
     pytest.importorskip("pyarrow")

pandas/tests/frame/indexing/test_set_value.py

@@ -32,10 +32,7 @@ def test_set_value_resize(self, float_frame, using_infer_string):
         else:
             assert res["baz"].dtype == np.object_
         res = float_frame.copy()
-        with tm.assert_produces_warning(
-            FutureWarning, match="Setting an item of incompatible dtype"
-        ):
-            res._set_value("foobar", "baz", True)
+        res._set_value("foobar", "baz", True)
         assert res["baz"].dtype == np.object_
 
         res = float_frame.copy()