Bug Fix: pd.DataFrame.sum with min_count changes dtype if result contains NaNs

doc/source/whatsnew/v1.5.0.rst

@@ -780,7 +780,7 @@ Indexing
 - Bug in setting large integer values into :class:`Series` with ``float32`` or ``float16`` dtype incorrectly altering these values instead of coercing to ``float64`` dtype (:issue:`45844`)
 - Bug in :meth:`Series.asof` and :meth:`DataFrame.asof` incorrectly casting bool-dtype results to ``float64`` dtype (:issue:`16063`)
 - Bug in :meth:`NDFrame.xs`, :meth:`DataFrame.iterrows`, :meth:`DataFrame.loc` and :meth:`DataFrame.iloc` not always propagating metadata (:issue:`28283`)
--
+- Bug in :meth:`DataFrame.sum` min_count changes dtype if input contains NaNs (:issue:`46947`)
 
 Missing
 ^^^^^^^

pandas/core/nanops.py

@@ -1471,8 +1471,8 @@ def _maybe_null_out(
             if is_numeric_dtype(result):
                 if np.iscomplexobj(result):
                     result = result.astype("c16")
-                else:
-                    result = result.astype("f8")
+                elif not is_float_dtype(result):
+                    result = result.astype("f8", copy=False)
                 result[null_mask] = np.nan
             else:
                 # GH12941, use None to auto cast null

pandas/tests/frame/test_reductions.py

@@ -778,6 +778,23 @@ def test_sum_nanops_min_count(self):
         expected = Series([np.nan, np.nan], index=["x", "y"])
         tm.assert_series_equal(result, expected)
 
+    @pytest.mark.parametrize("float_type", ["float16", "float32", "float64"])
+    @pytest.mark.parametrize(
+        "kwargs, expected_result",
+        [
+            ({"axis": 1, "min_count": 2}, [3.2, 5.3, np.NaN]),
+            ({"axis": 1, "min_count": 3}, [np.NaN, np.NaN, np.NaN]),
+            ({"axis": 1, "skipna": False}, [3.2, 5.3, np.NaN]),
+        ],
+    )
+    def test_sum_nanops_dtype_min_count(self, float_type, kwargs, expected_result):
+        # GH#46947
+        # pass
+        df = DataFrame({"a": [1.0, 2.3, 4.4], "b": [2.2, 3, np.nan]}, dtype=float_type)
+        result = df.sum(**kwargs)
+        expected = Series(expected_result).astype(float_type)
+        tm.assert_series_equal(result, expected)
+
     def test_sum_object(self, float_frame):
         values = float_frame.values.astype(int)
         frame = DataFrame(values, index=float_frame.index, columns=float_frame.columns)