BUG: Fixed strange behaviour of pd.DataFrame.drop() with inplace argu…

doc/source/whatsnew/v1.0.0.rst

@@ -567,13 +567,13 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - :meth:`DataFrame.hist` and :meth:`Series.hist` no longer allows ``figsize="default"``, specify figure size by passinig a tuple instead (:issue:`30003`)
 - Floordiv of integer-dtyped array by :class:`Timedelta` now raises ``TypeError`` (:issue:`21036`)
 - :class:`TimedeltaIndex` and :class:`DatetimeIndex` no longer accept non-nanosecond dtype strings like "timedelta64" or "datetime64", use "timedelta64[ns]" and "datetime64[ns]" instead (:issue:`24806`)
-- Changed the default "skipna" argument in :func:`pandas.api.types.infer_dtype` from ``False`` to ``True`` (:issue:`24050`)
+- :func:`pandas.api.types.infer_dtype` argument ``skipna`` defaults to ``True`` instead of ``False`` (:issue:`24050`)
 - Removed :attr:`Series.ix` and :attr:`DataFrame.ix` (:issue:`26438`)
 - Removed :meth:`Index.summary` (:issue:`18217`)
 - Removed the previously deprecated keyword "fastpath" from the :class:`Index` constructor (:issue:`23110`)
 - Removed :meth:`Series.get_value`, :meth:`Series.set_value`, :meth:`DataFrame.get_value`, :meth:`DataFrame.set_value` (:issue:`17739`)
 - Removed :meth:`Series.compound` and :meth:`DataFrame.compound` (:issue:`26405`)
-- Changed the default "inplace" argument in :meth:`DataFrame.set_index` and :meth:`Series.set_axis` from ``None`` to ``False`` (:issue:`27600`)
+- Changed the the default value of `inplace` in :meth:`DataFrame.set_index` and :meth:`Series.set_axis`. It now defaults to ``False`` (:issue:`27600`)
 - Removed :attr:`Series.cat.categorical`, :attr:`Series.cat.index`, :attr:`Series.cat.name` (:issue:`24751`)
 - Removed the previously deprecated keyword "box" from :func:`to_datetime` and :func:`to_timedelta`; in addition these now always returns :class:`DatetimeIndex`, :class:`TimedeltaIndex`, :class:`Index`, :class:`Series`, or :class:`DataFrame` (:issue:`24486`)
 - :func:`to_timedelta`, :class:`Timedelta`, and :class:`TimedeltaIndex` no longer allow "M", "y", or "Y" for the "unit" argument (:issue:`23264`)
@@ -613,7 +613,7 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - :func:`read_excel` no longer allows an integer value for the parameter ``usecols``, instead pass a list of integers from 0 to ``usecols`` inclusive (:issue:`23635`)
 - Removed the previously deprecated keyword "convert_datetime64" from :meth:`DataFrame.to_records` (:issue:`18902`)
 - Removed :meth:`IntervalIndex.from_intervals` in favor of the :class:`IntervalIndex` constructor (:issue:`19263`)
-- Changed the default "keep_tz" argument in :meth:`DatetimeIndex.to_series` from ``None`` to ``True`` (:issue:`23739`)
+- Changed the default value for the "keep_tz" argument in :meth:`DatetimeIndex.to_series` to ``True`` (:issue:`23739`)
 - Removed :func:`api.types.is_period` and :func:`api.types.is_datetimetz` (:issue:`23917`)
 - Ability to read pickles containing :class:`Categorical` instances created with pre-0.16 version of pandas has been removed (:issue:`27538`)
 - Removed :func:`pandas.tseries.plotting.tsplot` (:issue:`18627`)
@@ -622,7 +622,7 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - Removed the previously deprecated ``FrozenNDArray`` class in ``pandas.core.indexes.frozen`` (:issue:`29335`)
 - Removed the previously deprecated keyword "nthreads" from :func:`read_feather`, use "use_threads" instead (:issue:`23053`)
 - Removed :meth:`Index.is_lexsorted_for_tuple` (:issue:`29305`)
-- Removed support for nested renaming in :meth:`DataFrame.aggregate`, :meth:`Series.aggregate`, :meth:`DataFrameGroupBy.aggregate`, :meth:`SeriesGroupBy.aggregate`, :meth:`Rolling.aggregate` (:issue:`29608`)
+- Removed support for nexted renaming in :meth:`DataFrame.aggregate`, :meth:`Series.aggregate`, :meth:`DataFrameGroupBy.aggregate`, :meth:`SeriesGroupBy.aggregate`, :meth:`Rolling.aggregate` (:issue:`29608`)
 - Removed :meth:`Series.valid`; use :meth:`Series.dropna` instead (:issue:`18800`)
 - Removed :attr:`DataFrame.is_copy`, :attr:`Series.is_copy` (:issue:`18812`)
 - Removed :meth:`DataFrame.get_ftype_counts`, :meth:`Series.get_ftype_counts` (:issue:`18243`)
@@ -634,7 +634,7 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - Removed :meth:`DatetimeIndex.asobject`, :meth:`TimedeltaIndex.asobject`, :meth:`PeriodIndex.asobject`, use ``astype(object)`` instead (:issue:`29801`)
 - Removed the previously deprecated keyword "order" from :func:`factorize` (:issue:`19751`)
 - Removed the previously deprecated keyword "encoding" from :func:`read_stata` and :meth:`DataFrame.to_stata` (:issue:`21400`)
-- Changed the default "sort" argument in :func:`concat` from ``None`` to ``False`` (:issue:`20613`)
+- In :func:`concat` the default value for ``sort`` has been changed from ``None`` to ``False`` (:issue:`20613`)
 - Removed the previously deprecated keyword "raise_conflict" from :meth:`DataFrame.update`, use "errors" instead (:issue:`23585`)
 - Removed the previously deprecated keyword "n" from :meth:`DatetimeIndex.shift`, :meth:`TimedeltaIndex.shift`, :meth:`PeriodIndex.shift`, use "periods" instead (:issue:`22458`)
 - Removed the previously deprecated keywords "how", "fill_method", and "limit" from :meth:`DataFrame.resample` (:issue:`30139`)
@@ -650,15 +650,16 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - :meth:`Categorical.ravel` returns a :class:`Categorical` instead of a ``ndarray`` (:issue:`27199`)
 - The 'outer' method on Numpy ufuncs, e.g. ``np.subtract.outer`` operating on :class:`Series` objects is no longer supported, and will raise ``NotImplementedError`` (:issue:`27198`)
 - Removed :meth:`Series.get_dtype_counts` and :meth:`DataFrame.get_dtype_counts` (:issue:`27145`)
-- Changed the default "fill_value" argument in :meth:`Categorical.take` from ``True`` to ``False`` (:issue:`20841`)
-- Changed the default value for the `raw` argument in :func:`Series.rolling().apply() <pandas.core.window.Rolling.apply>`, :func:`DataFrame.rolling().apply() <pandas.core.window.Rolling.apply>`, :func:`Series.expanding().apply() <pandas.core.window.Expanding.apply>`, and :func:`DataFrame.expanding().apply() <pandas.core.window.Expanding.apply>` from ``None`` to ``False`` (:issue:`20584`)
+- Changed the default ``fill_value`` in :meth:`Categorical.take` from ``True`` to ``False`` (:issue:`20841`)
+- Changed the default value for the `raw` argument in :func:`Series.rolling().apply() <pandas.core.window.Rolling.apply>`, :func:`DataFrame.rolling().apply() <pandas.core.window.Rolling.apply>`,
+- :func:`Series.expanding().apply() <pandas.core.window.Expanding.apply>`, and :func:`DataFrame.expanding().apply() <pandas.core.window.Expanding.apply>` to ``False`` (:issue:`20584`)
 - Removed deprecated behavior of :meth:`Series.argmin` and :meth:`Series.argmax`, use :meth:`Series.idxmin` and :meth:`Series.idxmax` for the old behavior (:issue:`16955`)
 - Passing a tz-aware ``datetime.datetime`` or :class:`Timestamp` into the :class:`Timestamp` constructor with the ``tz`` argument now raises a ``ValueError`` (:issue:`23621`)
 - Removed :attr:`Series.base`, :attr:`Index.base`, :attr:`Categorical.base`, :attr:`Series.flags`, :attr:`Index.flags`, :attr:`PeriodArray.flags`, :attr:`Series.strides`, :attr:`Index.strides`, :attr:`Series.itemsize`, :attr:`Index.itemsize`, :attr:`Series.data`, :attr:`Index.data` (:issue:`20721`)
 - Changed :meth:`Timedelta.resolution` to match the behavior of the standard library ``datetime.timedelta.resolution``, for the old behavior, use :meth:`Timedelta.resolution_string` (:issue:`26839`)
 - Removed :attr:`Timestamp.weekday_name`, :attr:`DatetimeIndex.weekday_name`, and :attr:`Series.dt.weekday_name` (:issue:`18164`)
 - Removed the previously deprecated keyword "errors" in :meth:`Timestamp.tz_localize`, :meth:`DatetimeIndex.tz_localize`, and :meth:`Series.tz_localize` (:issue:`22644`)
-- Changed the default "ordered" argument in :class:`CategoricalDtype` from ``None`` to ``False`` (:issue:`26336`)
+- Changed the default value for ``ordered`` in :class:`CategoricalDtype` from ``None`` to ``False`` (:issue:`26336`)
 - :meth:`Series.set_axis` and :meth:`DataFrame.set_axis` now require "labels" as the first argument and "axis" as an optional named parameter (:issue:`30089`)
 - Removed :func:`to_msgpack`, :func:`read_msgpack`, :meth:`DataFrame.to_msgpack`, :meth:`Series.to_msgpack` (:issue:`27103`)
 - Removed :meth:`Series.compress` (:issue:`21930`)

pandas/core/generic.py

@@ -3215,7 +3215,10 @@ def _get_cacher(self):
         return cacher
 
     def _maybe_update_cacher(
-        self, clear: bool_t = False, verify_is_copy: bool_t = True
+        self,
+        clear: bool_t = False,
+        verify_is_copy: bool_t = True,
+        change_cache: bool_t = True,
     ) -> None:
         """
         See if we need to update our parent cacher if clear, then clear our
@@ -3227,8 +3230,9 @@ def _maybe_update_cacher(
             Clear the item cache.
         verify_is_copy : bool, default True
             Provide is_copy checks.
+        change_cache: bool, default True
+            If True, then cache may be changed
         """
-
         cacher = getattr(self, "_cacher", None)
         if cacher is not None:
             ref = cacher[1]()
@@ -3237,7 +3241,7 @@ def _maybe_update_cacher(
             # a copy
             if ref is None:
                 del self._cacher
-            else:
+            elif change_cache:
                 # Note: we need to call ref._maybe_cache_changed even in the
                 #  case where it will raise.  (Uh, not clear why)
                 try:
@@ -3931,22 +3935,24 @@ def _drop_axis(self, labels, axis, level=None, errors: str = "raise"):
 
         return result
 
-    def _update_inplace(self, result, verify_is_copy: bool_t = True) -> None:
+    def _update_inplace(self, result, verify_is_copy: bool_t = True, **kwargs) -> None:
         """
         Replace self internals with result.
 
         Parameters
         ----------
         verify_is_copy : bool, default True
             Provide is_copy checks.
+        **kwargs
+            Passed to self._maybe_update_cacher
         """
         # NOTE: This does *not* call __finalize__ and that's an explicit
         # decision that we may revisit in the future.
 
         self._reset_cache()
         self._clear_item_cache()
         self._data = getattr(result, "_data", result)
-        self._maybe_update_cacher(verify_is_copy=verify_is_copy)
+        self._maybe_update_cacher(verify_is_copy=verify_is_copy, **kwargs)
 
     def add_prefix(self, prefix: str):
         """

pandas/core/ops/methods.py

@@ -97,7 +97,9 @@ def f(self, other):
             # this makes sure that we are aligned like the input
             # we are updating inplace so we want to ignore is_copy
             self._update_inplace(
-                result.reindex_like(self, copy=False)._data, verify_is_copy=False
+                result.reindex_like(self, copy=False)._data,
+                verify_is_copy=False,
+                change_cache=False,
             )
 
             return self

pandas/tests/frame/test_operators.py

@@ -829,6 +829,28 @@ def test_inplace_ops_identity2(self, op):
         expected = id(df)
         assert id(df) == expected
 
+    def test_inplace_drop_and_add(self):
+        # GH 30484
+        # Get expected df
+        expected = pd.DataFrame({})
+        expected["x1"] = [1, 2, 3, 4, 5]
+        expected["x2"] = [0, 0, 0, 1, 1]
+        expected["target"] = [10, 20, 30, 40, 50]
+        y_expected = expected["target"]
+        expected = expected.drop("target", axis=1, inplace=False)
+        y_expected += np.min(y_expected)
+        # Get tested df
+        df = pd.DataFrame({})
+        df["x1"] = [1, 2, 3, 4, 5]
+        df["x2"] = [0, 0, 0, 1, 1]
+        df["target"] = [10, 20, 30, 40, 50]
+        y = df["target"]
+        df.drop("target", axis=1, inplace=True)
+        y += np.min(y)
+        # compare
+        tm.assert_frame_equal(df, expected)
+        tm.assert_series_equal(y, y_expected)
+
     def test_alignment_non_pandas(self):
         index = ["A", "B", "C"]
         columns = ["X", "Y", "Z"]