BUG: Series/Frame invert dtypes (#31183)

Author: TomAugspurger

doc/source/whatsnew/v1.0.0.rst

@@ -1019,6 +1019,7 @@ Numeric
 - Bug in :meth:`DataFrame.round` where a :class:`DataFrame` with a :class:`CategoricalIndex` of :class:`IntervalIndex` columns would incorrectly raise a ``TypeError`` (:issue:`30063`)
 - Bug in :meth:`Series.pct_change` and :meth:`DataFrame.pct_change` when there are duplicated indices (:issue:`30463`)
 - Bug in :class:`DataFrame` cumulative operations (e.g. cumsum, cummax) incorrect casting to object-dtype (:issue:`19296`)
+- Bug in dtypes being lost in ``DataFrame.__invert__`` (``~`` operator) with mixed dtypes (:issue:`31183`)
 - Bug in :class:`~DataFrame.diff` losing the dtype for extension types (:issue:`30889`)
 - Bug in :class:`DataFrame.diff` raising an ``IndexError`` when one of the columns was a nullable integer dtype (:issue:`30967`)
 
@@ -1169,6 +1170,7 @@ ExtensionArray
 - Bug in :class:`arrays.PandasArray` when setting a scalar string (:issue:`28118`, :issue:`28150`).
 - Bug where nullable integers could not be compared to strings (:issue:`28930`)
 - Bug where :class:`DataFrame` constructor raised ``ValueError`` with list-like data and ``dtype`` specified (:issue:`30280`)
+- Bug in dtype being lost in ``__invert__``  (``~`` operator) for extension-array backed ``Series`` and ``DataFrame`` (:issue:`23087`)
 
 
 Other

pandas/core/arrays/masked.py

@@ -50,6 +50,9 @@ def __iter__(self):
     def __len__(self) -> int:
         return len(self._data)
 
+    def __invert__(self):
+        return type(self)(~self._data, self._mask)
+
     def to_numpy(
         self, dtype=None, copy=False, na_value: "Scalar" = lib.no_default,
     ):

pandas/core/generic.py

@@ -1393,8 +1393,9 @@ def __invert__(self):
             # inv fails with 0 len
             return self
 
-        arr = operator.inv(com.values_from_object(self))
-        return self.__array_wrap__(arr)
+        new_data = self._data.apply(operator.invert)
+        result = self._constructor(new_data).__finalize__(self)
+        return result
 
     def __nonzero__(self):
         raise ValueError(

pandas/core/indexing.py

@@ -2233,7 +2233,7 @@ def check_bool_indexer(index: Index, key) -> np.ndarray:
         result = result.astype(bool)._values
     else:
         if is_sparse(result):
-            result = result.to_dense()
+            result = np.asarray(result)
         result = check_bool_array_indexer(index, result)
 
     return result

pandas/tests/arrays/sparse/test_arithmetics.py

@@ -476,6 +476,14 @@ def test_invert(fill_value):
     expected = SparseArray(~arr, fill_value=not fill_value)
     tm.assert_sp_array_equal(result, expected)
 
+    result = ~pd.Series(sparray)
+    expected = pd.Series(expected)
+    tm.assert_series_equal(result, expected)
+
+    result = ~pd.DataFrame({"A": sparray})
+    expected = pd.DataFrame({"A": expected})
+    tm.assert_frame_equal(result, expected)
+
 
 @pytest.mark.parametrize("fill_value", [0, np.nan])
 @pytest.mark.parametrize("op", [operator.pos, operator.neg])

pandas/tests/arrays/test_boolean.py

@@ -471,6 +471,24 @@ def test_ufunc_reduce_raises(values):
         np.add.reduce(a)
 
 
+class TestUnaryOps:
+    def test_invert(self):
+        a = pd.array([True, False, None], dtype="boolean")
+        expected = pd.array([False, True, None], dtype="boolean")
+        tm.assert_extension_array_equal(~a, expected)
+
+        expected = pd.Series(expected, index=["a", "b", "c"], name="name")
+        result = ~pd.Series(a, index=["a", "b", "c"], name="name")
+        tm.assert_series_equal(result, expected)
+
+        df = pd.DataFrame({"A": a, "B": [True, False, False]}, index=["a", "b", "c"])
+        result = ~df
+        expected = pd.DataFrame(
+            {"A": expected, "B": [False, True, True]}, index=["a", "b", "c"]
+        )
+        tm.assert_frame_equal(result, expected)
+
+
 class TestLogicalOps(BaseOpsUtil):
     def test_numpy_scalars_ok(self, all_logical_operators):
         a = pd.array([True, False, None], dtype="boolean")

pandas/tests/extension/base/__init__.py

@@ -49,7 +49,12 @@ class TestMyDtype(BaseDtypeTests):
 from .io import BaseParsingTests  # noqa
 from .methods import BaseMethodsTests  # noqa
 from .missing import BaseMissingTests  # noqa
-from .ops import BaseArithmeticOpsTests, BaseComparisonOpsTests, BaseOpsUtil  # noqa
+from .ops import (  # noqa
+    BaseArithmeticOpsTests,
+    BaseComparisonOpsTests,
+    BaseOpsUtil,
+    BaseUnaryOpsTests,
+)
 from .printing import BasePrintingTests  # noqa
 from .reduce import (  # noqa
     BaseBooleanReduceTests,

pandas/tests/extension/base/ops.py

@@ -168,3 +168,11 @@ def test_direct_arith_with_series_returns_not_implemented(self, data):
             assert result is NotImplemented
         else:
             raise pytest.skip(f"{type(data).__name__} does not implement __eq__")
+
+
+class BaseUnaryOpsTests(BaseOpsUtil):
+    def test_invert(self, data):
+        s = pd.Series(data, name="name")
+        result = ~s
+        expected = pd.Series(~data, name="name")
+        self.assert_series_equal(result, expected)

pandas/tests/extension/test_boolean.py

@@ -342,6 +342,10 @@ class TestPrinting(base.BasePrintingTests):
     pass
 
 
+class TestUnaryOps(base.BaseUnaryOpsTests):
+    pass
+
+
 # TODO parsing not yet supported
 # class TestParsing(base.BaseParsingTests):
 #     pass

pandas/tests/frame/test_operators.py

@@ -61,6 +61,27 @@ def test_invert(self, float_frame):
 
         tm.assert_frame_equal(-(df < 0), ~(df < 0))
 
+    def test_invert_mixed(self):
+        shape = (10, 5)
+        df = pd.concat(
+            [
+                pd.DataFrame(np.zeros(shape, dtype="bool")),
+                pd.DataFrame(np.zeros(shape, dtype=int)),
+            ],
+            axis=1,
+            ignore_index=True,
+        )
+        result = ~df
+        expected = pd.concat(
+            [
+                pd.DataFrame(np.ones(shape, dtype="bool")),
+                pd.DataFrame(-np.ones(shape, dtype=int)),
+            ],
+            axis=1,
+            ignore_index=True,
+        )
+        tm.assert_frame_equal(result, expected)
+
     @pytest.mark.parametrize(
         "df",
         [