REGR: fillna on f32 column raising for f64

doc/source/whatsnew/v1.3.3.rst

@@ -25,6 +25,8 @@ Fixed regressions
 - Fixed regression in :func:`is_list_like` where objects with ``__iter__`` set to ``None`` would be identified as iterable (:issue:`43373`)
 - Fixed regression in :meth:`.Resampler.aggregate` when used after column selection would raise if ``func`` is a list of aggregation functions (:issue:`42905`)
 - Fixed regression in :meth:`DataFrame.corr` where Kendall correlation would produce incorrect results for columns with repeated values (:issue:`43401`)
+- Fixed regression in :meth:`Series.fillna` raising ``TypeError`` when filling ``float32`` ``Series`` with list-like fill value having a larger dtype (like ``float64``) (:issue:`43424`)
+-
 
 .. ---------------------------------------------------------------------------
 
@@ -43,6 +45,7 @@ Performance improvements
 Bug fixes
 ~~~~~~~~~
 - Bug in :meth:`.DataFrameGroupBy.agg` and :meth:`.DataFrameGroupBy.transform` with ``engine="numba"`` where ``index`` data was not being correctly passed into ``func`` (:issue:`43133`)
+- Bug in setting a ``float32`` :class:`Series` with a larger dtype (like ``float64``) not upcasting (:issue:`43424`)
 -
 
 .. ---------------------------------------------------------------------------

pandas/core/array_algos/putmask.py

@@ -114,7 +114,9 @@ def putmask_smart(values: np.ndarray, mask: np.ndarray, new) -> np.ndarray:
 
     new = np.asarray(new)
 
-    if values.dtype.kind == new.dtype.kind:
+    values_kind = values.dtype.kind
+    new_kind = new.dtype.kind
+    if values_kind == new_kind or (values_kind == "f" and new_kind in ["i", "u"]):
         # preserves dtype if possible
         return _putmask_preserve(values, new, mask)
 

pandas/core/dtypes/cast.py

@@ -2208,10 +2208,11 @@ def can_hold_element(arr: ArrayLike, element: Any) -> bool:
 
     elif dtype.kind == "f":
         if tipo is not None:
-            # TODO: itemsize check?
             if tipo.kind not in ["f", "i", "u"]:
                 # Anything other than float/integer we cannot hold
                 return False
+            elif dtype.itemsize < tipo.itemsize:
+                return False
             elif not isinstance(tipo, np.dtype):
                 # i.e. nullable IntegerDtype or FloatingDtype;
                 #  we can put this into an ndarray losslessly iff it has no NAs

pandas/tests/indexing/test_indexing.py

@@ -200,6 +200,24 @@ def test_setitem_dtype_upcast3(self):
         assert is_float_dtype(left["foo"])
         assert is_float_dtype(left["baz"])
 
+    @pytest.mark.parametrize("dtype", [np.int64, np.uint64, np.float64])
+    def test_setitem_dtype_float_listlike_upcasts(self, dtype, indexer_sli):
+        # GH-43424
+        result = Series([1.1, 1.1], dtype=np.float32)
+        indexer_sli(result)[[0]] = np.array([1], dtype=dtype)
+        expected = Series([1.0, 1.1], dtype=np.float64)
+        tm.assert_series_equal(result, expected)
+
+    @pytest.mark.parametrize(
+        "dtype,new_dtype", [(np.int8, np.int64), (np.uint16, np.uint32)]
+    )
+    def test_setitem_dtype_int_listlike_upcasts(self, dtype, new_dtype, indexer_sli):
+        # GH-43424
+        result = Series([1, 1], dtype=dtype)
+        indexer_sli(result)[[0]] = np.array([2], dtype=new_dtype)
+        expected = Series([2, 1], dtype=new_dtype)
+        tm.assert_series_equal(result, expected)
+
     def test_dups_fancy_indexing(self):
 
         # GH 3455

pandas/tests/series/methods/test_fillna.py

@@ -702,6 +702,22 @@ def test_fillna_categorical_raises(self):
         with pytest.raises(TypeError, match=msg):
             ser.fillna(DataFrame({1: ["a"], 3: ["b"]}))
 
+    @pytest.mark.parametrize("fill_type", [np.float64, np.uint64, np.int64])
+    def test_fillna_f32_upcast(self, fill_type):
+        # GH-43424
+        ser = Series([np.nan, 1.2], dtype=np.float32)
+        fill_values = Series([2, 2], dtype=fill_type)
+        result = ser.fillna(fill_values)
+        expected = Series([2.0, 1.2], dtype="float64")
+        tm.assert_series_equal(result, expected)
+
+    def test_fillna_f32_upcast_with_dict(self):
+        # GH-43424
+        ser = Series([np.nan, 1.2], dtype=np.float32)
+        result = ser.fillna({0: 1})
+        expected = Series([1.0, 1.2])
+        tm.assert_series_equal(result, expected)
+
     # ---------------------------------------------------------------
     # Invalid Usages