DEPR: int slicing always positional (#53338)

* DEPR: int slicing always positional

* update doc

* okwarning

Author: jbrockmendel

doc/source/user_guide/missing_data.rst

@@ -436,7 +436,7 @@ at the new values.
    # interpolate at new_index
    new_index = ser.index.union(pd.Index([49.25, 49.5, 49.75, 50.25, 50.5, 50.75]))
    interp_s = ser.reindex(new_index).interpolate(method="pchip")
-   interp_s[49:51]
+   interp_s.loc[49:51]
 
 .. _scipy: https://scipy.org/
 .. _documentation: https://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation

doc/source/whatsnew/v0.13.0.rst

@@ -343,6 +343,7 @@ Float64Index API change
   Slicing is ALWAYS on the values of the index, for ``[],ix,loc`` and ALWAYS positional with ``iloc``
 
   .. ipython:: python
+     :okwarning:
 
      s[2:4]
      s.loc[2:4]

doc/source/whatsnew/v2.1.0.rst

@@ -553,6 +553,7 @@ Other Deprecations
 - Deprecated the "method" and "limit" keywords in ``ExtensionArray.fillna``, implement and use ``pad_or_backfill`` instead (:issue:`53621`)
 - Deprecated the "method" and "limit" keywords on :meth:`Series.fillna`, :meth:`DataFrame.fillna`, :meth:`SeriesGroupBy.fillna`, :meth:`DataFrameGroupBy.fillna`, and :meth:`Resampler.fillna`, use ``obj.bfill()`` or ``obj.ffill()`` instead (:issue:`53394`)
 - Deprecated the ``method`` and ``limit`` keywords in :meth:`DataFrame.replace` and :meth:`Series.replace` (:issue:`33302`)
+- Deprecated the behavior of :meth:`Series.__getitem__`, :meth:`Series.__setitem__`, :meth:`DataFrame.__getitem__`, :meth:`DataFrame.__setitem__` with an integer slice on objects with a floating-dtype index, in a future version this will be treated as *positional* indexing (:issue:`49612`)
 - Deprecated the use of non-supported datetime64 and timedelta64 resolutions with :func:`pandas.array`. Supported resolutions are: "s", "ms", "us", "ns" resolutions (:issue:`53058`)
 - Deprecated values "pad", "ffill", "bfill", "backfill" for :meth:`Series.interpolate` and :meth:`DataFrame.interpolate`, use ``obj.ffill()`` or ``obj.bfill()`` instead (:issue:`53581`)
 - Deprecated the behavior of :meth:`Index.argmax`, :meth:`Index.argmin`, :meth:`Series.argmax`, :meth:`Series.argmin` with either all-NAs and skipna=True or any-NAs and skipna=False returning -1; in a future version this will raise ``ValueError`` (:issue:`33941`, :issue:`33942`)

pandas/core/indexes/base.py

@@ -4212,16 +4212,30 @@ def _convert_slice_indexer(self, key: slice, kind: Literal["loc", "getitem"]):
         # potentially cast the bounds to integers
         start, stop, step = key.start, key.stop, key.step
 
+        # figure out if this is a positional indexer
+        is_index_slice = is_valid_positional_slice(key)
+
         # TODO(GH#50617): once Series.__[gs]etitem__ is removed we should be able
         #  to simplify this.
         if lib.is_np_dtype(self.dtype, "f"):
             # We always treat __getitem__ slicing as label-based
             # translate to locations
+            if kind == "getitem" and is_index_slice and not start == stop and step != 0:
+                # exclude step=0 from the warning because it will raise anyway
+                # start/stop both None e.g. [:] or [::-1] won't change.
+                # exclude start==stop since it will be empty either way, or
+                # will be [:] or [::-1] which won't change
+                warnings.warn(
+                    # GH#49612
+                    "The behavior of obj[i:j] with a float-dtype index is "
+                    "deprecated. In a future version, this will be treated as "
+                    "positional instead of label-based. For label-based slicing, "
+                    "use obj.loc[i:j] instead",
+                    FutureWarning,
+                    stacklevel=find_stack_level(),
+                )
             return self.slice_indexer(start, stop, step)
 
-        # figure out if this is a positional indexer
-        is_index_slice = is_valid_positional_slice(key)
-
         if kind == "getitem":
             # called from the getitem slicers, validate that we are in fact integers
             if is_index_slice:

pandas/tests/frame/indexing/test_indexing.py

@@ -755,7 +755,9 @@ def test_getitem_setitem_float_labels(self, using_array_manager):
         tm.assert_frame_equal(result, expected)
 
         df.loc[1:2] = 0
-        result = df[1:2]
+        msg = r"The behavior of obj\[i:j\] with a float-dtype index"
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            result = df[1:2]
         assert (result == 0).all().all()
 
         # #2727

pandas/tests/indexing/test_floats.py

@@ -488,8 +488,12 @@ def test_floating_misc(self, indexer_sl):
         for fancy_idx in [[5, 0], np.array([5, 0])]:
             tm.assert_series_equal(indexer_sl(s)[fancy_idx], expected)
 
+        warn = FutureWarning if indexer_sl is tm.setitem else None
+        msg = r"The behavior of obj\[i:j\] with a float-dtype index"
+
         # all should return the same as we are slicing 'the same'
-        result1 = indexer_sl(s)[2:5]
+        with tm.assert_produces_warning(warn, match=msg):
+            result1 = indexer_sl(s)[2:5]
         result2 = indexer_sl(s)[2.0:5.0]
         result3 = indexer_sl(s)[2.0:5]
         result4 = indexer_sl(s)[2.1:5]
@@ -498,7 +502,8 @@ def test_floating_misc(self, indexer_sl):
         tm.assert_series_equal(result1, result4)
 
         expected = Series([1, 2], index=[2.5, 5.0])
-        result = indexer_sl(s)[2:5]
+        with tm.assert_produces_warning(warn, match=msg):
+            result = indexer_sl(s)[2:5]
 
         tm.assert_series_equal(result, expected)
 

pandas/tests/series/methods/test_interpolate.py

@@ -129,7 +129,7 @@ def test_interpolate_cubicspline(self):
         new_index = ser.index.union(Index([1.25, 1.5, 1.75, 2.25, 2.5, 2.75])).astype(
             float
         )
-        result = ser.reindex(new_index).interpolate(method="cubicspline")[1:3]
+        result = ser.reindex(new_index).interpolate(method="cubicspline").loc[1:3]
         tm.assert_series_equal(result, expected)
 
     def test_interpolate_pchip(self):
@@ -142,7 +142,7 @@ def test_interpolate_pchip(self):
         ).astype(float)
         interp_s = ser.reindex(new_index).interpolate(method="pchip")
         # does not blow up, GH5977
-        interp_s[49:51]
+        interp_s.loc[49:51]
 
     def test_interpolate_akima(self):
         pytest.importorskip("scipy")
@@ -157,7 +157,7 @@ def test_interpolate_akima(self):
             float
         )
         interp_s = ser.reindex(new_index).interpolate(method="akima")
-        tm.assert_series_equal(interp_s[1:3], expected)
+        tm.assert_series_equal(interp_s.loc[1:3], expected)
 
         # interpolate at new_index where `der` is a non-zero int
         expected = Series(
@@ -168,7 +168,7 @@ def test_interpolate_akima(self):
             float
         )
         interp_s = ser.reindex(new_index).interpolate(method="akima", der=1)
-        tm.assert_series_equal(interp_s[1:3], expected)
+        tm.assert_series_equal(interp_s.loc[1:3], expected)
 
     def test_interpolate_piecewise_polynomial(self):
         pytest.importorskip("scipy")
@@ -183,7 +183,7 @@ def test_interpolate_piecewise_polynomial(self):
             float
         )
         interp_s = ser.reindex(new_index).interpolate(method="piecewise_polynomial")
-        tm.assert_series_equal(interp_s[1:3], expected)
+        tm.assert_series_equal(interp_s.loc[1:3], expected)
 
     def test_interpolate_from_derivatives(self):
         pytest.importorskip("scipy")
@@ -198,7 +198,7 @@ def test_interpolate_from_derivatives(self):
             float
         )
         interp_s = ser.reindex(new_index).interpolate(method="from_derivatives")
-        tm.assert_series_equal(interp_s[1:3], expected)
+        tm.assert_series_equal(interp_s.loc[1:3], expected)
 
     @pytest.mark.parametrize(
         "kwargs",
@@ -218,7 +218,7 @@ def test_interpolate_corners(self, kwargs):
 
     def test_interpolate_index_values(self):
         s = Series(np.nan, index=np.sort(np.random.default_rng(2).random(30)))
-        s[::3] = np.random.default_rng(2).standard_normal(10)
+        s.loc[::3] = np.random.default_rng(2).standard_normal(10)
 
         vals = s.index.values.astype(float)