REGR: unstack on 'int' dtype prevent fillna to work

doc/source/whatsnew/v1.1.5.rst

@@ -20,6 +20,7 @@ Fixed regressions
 - Fixed regression in inplace operations on :class:`Series` with ``ExtensionDtype`` with NumPy dtyped operand (:issue:`37910`)
 - Fixed regression in metadata propagation for ``groupby`` iterator (:issue:`37343`)
 - Fixed regression in :class:`MultiIndex` constructed from a :class:`DatetimeIndex` not retaining frequency (:issue:`35563`)
+- Fixed regression in :meth:`DataFrame.unstack` with columns with integer dtype (:issue:`37115`)
 - Fixed regression in indexing on a :class:`Series` with ``CategoricalDtype`` after unpickling (:issue:`37631`)
 - Fixed regression in :meth:`DataFrame.groupby` aggregation with out-of-bounds datetime objects in an object-dtype column (:issue:`36003`)
 - Fixed regression in ``df.groupby(..).rolling(..)`` with the resulting :class:`MultiIndex` when grouping by a label that is in the index (:issue:`37641`)

pandas/core/internals/blocks.py

@@ -1542,7 +1542,7 @@ def _unstack(self, unstacker, fill_value, new_placement):
         new_values = new_values.T[mask]
         new_placement = new_placement[mask]
 
-        blocks = [self.make_block_same_class(new_values, placement=new_placement)]
+        blocks = [make_block(new_values, placement=new_placement)]
         return blocks, mask
 
     def quantile(self, qs, interpolation="linear", axis: int = 0):

pandas/tests/frame/test_stack_unstack.py

@@ -1880,3 +1880,30 @@ def test_unstack_group_index_overflow(self):
         s = Series(np.arange(1000), index=index)
         result = s.unstack(4)
         assert result.shape == (500, 2)
+
+    def test_unstack_with_missing_int_cast_to_float(self):
+        # https://github.com/pandas-dev/pandas/issues/37115
+        df = DataFrame(
+            {
+                "a": ["A", "A", "B"],
+                "b": ["ca", "cb", "cb"],
+                "v": [10] * 3,
+            }
+        ).set_index(["a", "b"])
+
+        # add another int column to get 2 blocks
+        df["is_"] = 1
+        assert len(df._mgr.blocks) == 2
+
+        result = df.unstack("b")
+        result[("is_", "ca")] = result[("is_", "ca")].fillna(0)
+
+        expected = DataFrame(
+            [[10.0, 10.0, 1.0, 1.0], [np.nan, 10.0, 0.0, 1.0]],
+            index=Index(["A", "B"], dtype="object", name="a"),
+            columns=MultiIndex.from_tuples(
+                [("v", "ca"), ("v", "cb"), ("is_", "ca"), ("is_", "cb")],
+                names=[None, "b"],
+            ),
+        )
+        tm.assert_frame_equal(result, expected)