DEPR: indexing (#49412)

jbrockmendel · web-flow · commit 9820edc17473 · 2022-11-02T11:56:50.000-07:00
* DEPR: disallow Series.__getitem__ with a single-element list containing slice

* DEPR: disallow slicing with positional slicer and .loc

* DEPR: disallow positional indexing with float key

* move whatsnew

* DEPR: disallow multi-dimensional indexing

* fix matplotlib tests

* update install.rst
diff --git a/ci/deps/actions-38-minimum_versions.yaml b/ci/deps/actions-38-minimum_versions.yaml
@@ -32,7 +32,7 @@ dependencies:
   - gcsfs=2021.07.0
   - jinja2=3.0.0
   - lxml=4.6.3
-  - matplotlib=3.3.2
+  - matplotlib=3.6.0
   - numba=0.53.1
   - numexpr=2.7.3
   - odfpy=1.4.1
diff --git a/doc/source/getting_started/install.rst b/doc/source/getting_started/install.rst
@@ -310,7 +310,7 @@ Can be managed as optional_extra with ``pandas[plot, output_formatting]``, depen
 ========================= ================== ================== =============================================================
 Dependency                Minimum Version    optional_extra     Notes
 ========================= ================== ================== =============================================================
-matplotlib                3.3.2              plot               Plotting library
+matplotlib                3.6.0              plot               Plotting library
 Jinja2                    3.0.0              output_formatting  Conditional formatting with DataFrame.style
 tabulate                  0.8.9              output_formatting  Printing in Markdown-friendly format (see `tabulate`_)
 ========================= ================== ================== =============================================================
diff --git a/doc/source/whatsnew/v2.0.0.rst b/doc/source/whatsnew/v2.0.0.rst
@@ -124,6 +124,8 @@ Optional libraries below the lowest tested version may still work, but are not c
 +=================+=================+=========+
 | pyarrow         | 6.0.0           |    X    |
 +-----------------+-----------------+---------+
+| matplotlib      | 3.6.0           |    X    |
++-----------------+-----------------+---------+
 | fastparquet     | 0.6.3           |    X    |
 +-----------------+-----------------+---------+
 
@@ -282,6 +284,8 @@ Removal of prior version deprecations/changes
 - Enforced disallowing the use of ``**kwargs`` in :class:`.ExcelWriter`; use the keyword argument ``engine_kwargs`` instead (:issue:`40430`)
 - Enforced disallowing a tuple of column labels into :meth:`.DataFrameGroupBy.__getitem__` (:issue:`30546`)
 - Enforced disallowing setting values with ``.loc`` using a positional slice. Use ``.loc`` with labels or ``.iloc`` with positions instead (:issue:`31840`)
+- Enforced disallowing positional indexing with a ``float`` key even if that key is a round number, manually cast to integer instead (:issue:`34193`)
+- Enforced disallowing indexing on a :class:`Index` or positional indexing on a :class:`Series` producing multi-dimensional objects e.g. ``obj[:, None]``, convert to numpy before indexing instead (:issue:`35141`)
 - Enforced disallowing ``dict`` or ``set`` objects in ``suffixes`` in :func:`merge` (:issue:`34810`)
 - Enforced disallowing :func:`merge` to produce duplicated columns through the ``suffixes`` keyword and already existing columns (:issue:`22818`)
 - Enforced disallowing using :func:`merge` or :func:`join` on a different number of levels (:issue:`34862`)
diff --git a/pandas/compat/_optional.py b/pandas/compat/_optional.py
@@ -23,7 +23,7 @@
     "gcsfs": "2021.07.0",
     "jinja2": "3.0.0",
     "lxml.etree": "4.6.3",
-    "matplotlib": "3.3.2",
+    "matplotlib": "3.6.0",
     "numba": "0.53.1",
     "numexpr": "2.7.3",
     "odfpy": "1.4.1",
diff --git a/pandas/core/common.py b/pandas/core/common.py
@@ -148,30 +148,25 @@ def is_bool_indexer(key: Any) -> bool:
     return False
 
 
-def cast_scalar_indexer(val, warn_float: bool = False):
+def cast_scalar_indexer(val):
     """
-    To avoid numpy DeprecationWarnings, cast float to integer where valid.
+    Disallow indexing with a float key, even if that key is a round number.
 
     Parameters
     ----------
     val : scalar
-    warn_float : bool, default False
-        If True, issue deprecation warning for a float indexer.
 
     Returns
     -------
     outval : scalar
     """
     # assumes lib.is_scalar(val)
     if lib.is_float(val) and val.is_integer():
-        if warn_float:
-            warnings.warn(
-                "Indexing with a float is deprecated, and will raise an IndexError "
-                "in pandas 2.0. You can manually convert to an integer key instead.",
-                FutureWarning,
-                stacklevel=find_stack_level(),
-            )
-        return int(val)
+        raise IndexError(
+            # GH#34193
+            "Indexing with a float is no longer supported. Manually convert "
+            "to an integer key instead."
+        )
     return val
 
 
diff --git a/pandas/core/indexers/__init__.py b/pandas/core/indexers/__init__.py
@@ -2,7 +2,7 @@
     check_array_indexer,
     check_key_length,
     check_setitem_lengths,
-    deprecate_ndim_indexing,
+    disallow_ndim_indexing,
     is_empty_indexer,
     is_list_like_indexer,
     is_scalar_indexer,
@@ -23,7 +23,7 @@
     "validate_indices",
     "maybe_convert_indices",
     "length_of_indexer",
-    "deprecate_ndim_indexing",
+    "disallow_ndim_indexing",
     "unpack_1tuple",
     "check_key_length",
     "check_array_indexer",
diff --git a/pandas/core/indexers/utils.py b/pandas/core/indexers/utils.py
@@ -7,12 +7,10 @@
     TYPE_CHECKING,
     Any,
 )
-import warnings
 
 import numpy as np
 
 from pandas._typing import AnyArrayLike
-from pandas.util._exceptions import find_stack_level
 
 from pandas.core.dtypes.common import (
     is_array_like,
@@ -333,22 +331,18 @@ def length_of_indexer(indexer, target=None) -> int:
     raise AssertionError("cannot find the length of the indexer")
 
 
-def deprecate_ndim_indexing(result, stacklevel: int = 3) -> None:
+def disallow_ndim_indexing(result) -> None:
     """
-    Helper function to raise the deprecation warning for multi-dimensional
-    indexing on 1D Series/Index.
+    Helper function to disallow multi-dimensional indexing on 1D Series/Index.
 
     GH#27125 indexer like idx[:, None] expands dim, but we cannot do that
-    and keep an index, so we currently return ndarray, which is deprecated
-    (Deprecation GH#30588).
+    and keep an index, so we used to return ndarray, which was deprecated
+    in GH#30588.
     """
     if np.ndim(result) > 1:
-        warnings.warn(
-            "Support for multi-dimensional indexing (e.g. `obj[:, None]`) "
-            "is deprecated and will be removed in a future "
-            "version.  Convert to a numpy array before indexing instead.",
-            FutureWarning,
-            stacklevel=find_stack_level(),
+        raise ValueError(
+            "Multi-dimensional indexing (e.g. `obj[:, None]`) is no longer "
+            "supported. Convert to a numpy array before indexing instead."
         )
 
 
diff --git a/pandas/core/indexes/base.py b/pandas/core/indexes/base.py
@@ -161,7 +161,7 @@
     extract_array,
     sanitize_array,
 )
-from pandas.core.indexers import deprecate_ndim_indexing
+from pandas.core.indexers import disallow_ndim_indexing
 from pandas.core.indexes.frozen import FrozenList
 from pandas.core.ops import get_op_result_name
 from pandas.core.ops.invalid import make_invalid_op
@@ -5213,7 +5213,7 @@ def __getitem__(self, key):
 
         if is_integer(key) or is_float(key):
             # GH#44051 exclude bool, which would return a 2d ndarray
-            key = com.cast_scalar_indexer(key, warn_float=True)
+            key = com.cast_scalar_indexer(key)
             return getitem(key)
 
         if isinstance(key, slice):
@@ -5236,15 +5236,7 @@ def __getitem__(self, key):
         result = getitem(key)
         # Because we ruled out integer above, we always get an arraylike here
         if result.ndim > 1:
-            deprecate_ndim_indexing(result)
-            if hasattr(result, "_ndarray"):
-                # i.e. NDArrayBackedExtensionArray
-                # Unpack to ndarray for MPL compat
-                # error: Item "ndarray[Any, Any]" of
-                # "Union[ExtensionArray, ndarray[Any, Any]]"
-                # has no attribute "_ndarray"
-                return result._ndarray  # type: ignore[union-attr]
-            return result
+            disallow_ndim_indexing(result)
 
         # NB: Using _constructor._simple_new would break if MultiIndex
         #  didn't override __getitem__
diff --git a/pandas/core/indexes/multi.py b/pandas/core/indexes/multi.py
@@ -2031,7 +2031,7 @@ def __reduce__(self):
 
     def __getitem__(self, key):
         if is_scalar(key):
-            key = com.cast_scalar_indexer(key, warn_float=True)
+            key = com.cast_scalar_indexer(key)
 
             retval = []
             for lev, level_codes in zip(self.levels, self.codes):
diff --git a/pandas/core/series.py b/pandas/core/series.py
@@ -123,7 +123,7 @@
 )
 from pandas.core.generic import NDFrame
 from pandas.core.indexers import (
-    deprecate_ndim_indexing,
+    disallow_ndim_indexing,
     unpack_1tuple,
 )
 from pandas.core.indexes.accessors import CombinedDatetimelikeProperties
@@ -1003,7 +1003,7 @@ def _get_values_tuple(self, key: tuple):
             #  see tests.series.timeseries.test_mpl_compat_hack
             # the asarray is needed to avoid returning a 2D DatetimeArray
             result = np.asarray(self._values[key])
-            deprecate_ndim_indexing(result, stacklevel=find_stack_level())
+            disallow_ndim_indexing(result)
             return result
 
         if not isinstance(self.index, MultiIndex):
diff --git a/pandas/tests/indexes/common.py b/pandas/tests/indexes/common.py
@@ -699,20 +699,16 @@ def test_engine_reference_cycle(self, simple_index):
     def test_getitem_2d_deprecated(self, simple_index):
         # GH#30588, GH#31479
         idx = simple_index
-        msg = "Support for multi-dimensional indexing"
-        with tm.assert_produces_warning(FutureWarning, match=msg):
-            res = idx[:, None]
-
-        assert isinstance(res, np.ndarray), type(res)
+        msg = "Multi-dimensional indexing"
+        with pytest.raises(ValueError, match=msg):
+            idx[:, None]
 
         if not isinstance(idx, RangeIndex):
-            # GH#44051 RangeIndex already raises
-            with tm.assert_produces_warning(FutureWarning, match=msg):
-                res = idx[True]
-            assert isinstance(res, np.ndarray), type(res)
-            with tm.assert_produces_warning(FutureWarning, match=msg):
-                res = idx[False]
-            assert isinstance(res, np.ndarray), type(res)
+            # GH#44051 RangeIndex already raised pre-2.0 with a different message
+            with pytest.raises(ValueError, match=msg):
+                idx[True]
+            with pytest.raises(ValueError, match=msg):
+                idx[False]
         else:
             msg = "only integers, slices"
             with pytest.raises(IndexError, match=msg):
diff --git a/pandas/tests/indexes/datetimes/test_indexing.py b/pandas/tests/indexes/datetimes/test_indexing.py
@@ -98,11 +98,9 @@ def test_dti_business_getitem(self, freq):
     @pytest.mark.parametrize("freq", ["B", "C"])
     def test_dti_business_getitem_matplotlib_hackaround(self, freq):
         rng = bdate_range(START, END, freq=freq)
-        with tm.assert_produces_warning(FutureWarning):
+        with pytest.raises(ValueError, match="Multi-dimensional indexing"):
             # GH#30588 multi-dimensional indexing deprecated
-            values = rng[:, None]
-        expected = rng.values[:, None]
-        tm.assert_numpy_array_equal(values, expected)
+            rng[:, None]
 
     def test_getitem_int_list(self):
         dti = date_range(start="1/1/2005", end="12/1/2005", freq="M")
diff --git a/pandas/tests/indexes/test_base.py b/pandas/tests/indexes/test_base.py
@@ -62,11 +62,11 @@ def test_can_hold_identifiers(self, simple_index):
 
     @pytest.mark.parametrize("index", ["datetime"], indirect=True)
     def test_new_axis(self, index):
-        with tm.assert_produces_warning(FutureWarning):
+        # TODO: a bunch of scattered tests check this deprecation is enforced.
+        #  de-duplicate/centralize them.
+        with pytest.raises(ValueError, match="Multi-dimensional indexing"):
             # GH#30588 multi-dimensional indexing deprecated
-            new_index = index[None, :]
-        assert new_index.ndim == 2
-        assert isinstance(new_index, np.ndarray)
+            index[None, :]
 
     def test_argsort(self, index):
         with tm.maybe_produces_warning(
@@ -1532,15 +1532,15 @@ def test_deprecated_fastpath():
 
 
 def test_shape_of_invalid_index():
-    # Currently, it is possible to create "invalid" index objects backed by
+    # Pre-2.0, it was possible to create "invalid" index objects backed by
     # a multi-dimensional array (see https://github.com/pandas-dev/pandas/issues/27125
     # about this). However, as long as this is not solved in general,this test ensures
     # that the returned shape is consistent with this underlying array for
     # compat with matplotlib (see https://github.com/pandas-dev/pandas/issues/27775)
     idx = Index([0, 1, 2, 3])
-    with tm.assert_produces_warning(FutureWarning):
+    with pytest.raises(ValueError, match="Multi-dimensional indexing"):
         # GH#30588 multi-dimensional indexing deprecated
-        assert idx[:, None].shape == (4, 1)
+        idx[:, None]
 
 
 def test_validate_1d_input():
diff --git a/pandas/tests/indexes/test_indexing.py b/pandas/tests/indexes/test_indexing.py
@@ -290,11 +290,9 @@ def test_putmask_with_wrong_mask(self, index):
 def test_getitem_deprecated_float(idx):
     # https://github.com/pandas-dev/pandas/issues/34191
 
-    with tm.assert_produces_warning(FutureWarning):
-        result = idx[1.0]
-
-    expected = idx[1]
-    assert result == expected
+    msg = "Indexing with a float is no longer supported"
+    with pytest.raises(IndexError, match=msg):
+        idx[1.0]
 
 
 @pytest.mark.parametrize(
diff --git a/pandas/tests/series/indexing/test_getitem.py b/pandas/tests/series/indexing/test_getitem.py
@@ -269,19 +269,10 @@ def test_getitem_partial_str_slice_high_reso_with_timedeltaindex(self):
 
     def test_getitem_slice_2d(self, datetime_series):
         # GH#30588 multi-dimensional indexing deprecated
-
-        with tm.assert_produces_warning(
-            FutureWarning, match="Support for multi-dimensional indexing"
-        ):
-            # GH#30867 Don't want to support this long-term, but
-            # for now ensure that the warning from Index
-            # doesn't comes through via Series.__getitem__.
-            result = datetime_series[:, np.newaxis]
-        expected = datetime_series.values[:, np.newaxis]
-        tm.assert_almost_equal(result, expected)
+        with pytest.raises(ValueError, match="Multi-dimensional indexing"):
+            datetime_series[:, np.newaxis]
 
     # FutureWarning from NumPy.
-    @pytest.mark.filterwarnings("ignore:Using a non-tuple:FutureWarning")
     def test_getitem_median_slice_bug(self):
         index = date_range("20090415", "20090519", freq="2B")
         ser = Series(np.random.randn(13), index=index)
@@ -291,6 +282,10 @@ def test_getitem_median_slice_bug(self):
         with pytest.raises(ValueError, match=msg):
             # GH#31299
             ser[indexer]
+        # but we're OK with a single-element tuple
+        result = ser[(indexer[0],)]
+        expected = ser[indexer[0]]
+        tm.assert_series_equal(result, expected)
 
     @pytest.mark.parametrize(
         "slc, positions",
@@ -554,14 +549,8 @@ def test_getitem_generator(string_series):
     ],
 )
 def test_getitem_ndim_deprecated(series):
-    with tm.assert_produces_warning(
-        FutureWarning,
-        match="Support for multi-dimensional indexing",
-    ):
-        result = series[:, None]
-
-    expected = np.asarray(series)[:, None]
-    tm.assert_numpy_array_equal(result, expected)
+    with pytest.raises(ValueError, match="Multi-dimensional indexing"):
+        series[:, None]
 
 
 def test_getitem_multilevel_scalar_slice_not_implemented(
diff --git a/pandas/tests/series/indexing/test_indexing.py b/pandas/tests/series/indexing/test_indexing.py
@@ -184,8 +184,6 @@ def test_setslice(datetime_series):
     assert sl.index.is_unique is True
 
 
-# FutureWarning from NumPy about [slice(None, 5).
-@pytest.mark.filterwarnings("ignore:Using a non-tuple:FutureWarning")
 def test_basic_getitem_setitem_corner(datetime_series):
     # invalid tuples, e.g. td.ts[:, None] vs. td.ts[:, 2]
     msg = "key of type tuple not found and not a MultiIndex"
@@ -200,6 +198,11 @@ def test_basic_getitem_setitem_corner(datetime_series):
         # GH#31299
         datetime_series[[slice(None, 5)]]
 
+    # but we're OK with a single-element tuple
+    result = datetime_series[(slice(None, 5),)]
+    expected = datetime_series[:5]
+    tm.assert_series_equal(result, expected)
+
     # OK
     msg = r"unhashable type(: 'slice')?"
     with pytest.raises(TypeError, match=msg):
