DEPR: Deprecate Series.view (#56054)

Co-authored-by: Joris Van den Bossche <[email protected]>

Author: phofl

doc/source/whatsnew/v2.2.0.rst

@@ -362,6 +362,7 @@ Other Deprecations
 - Deprecated :func:`read_gbq` and :meth:`DataFrame.to_gbq`. Use ``pandas_gbq.read_gbq`` and ``pandas_gbq.to_gbq`` instead https://pandas-gbq.readthedocs.io/en/latest/api.html (:issue:`55525`)
 - Deprecated :meth:`.DataFrameGroupBy.fillna` and :meth:`.SeriesGroupBy.fillna`; use :meth:`.DataFrameGroupBy.ffill`, :meth:`.DataFrameGroupBy.bfill` for forward and backward filling or :meth:`.DataFrame.fillna` to fill with a single value (or the Series equivalents) (:issue:`55718`)
 - Deprecated :meth:`Index.format`, use ``index.astype(str)`` or ``index.map(formatter)`` instead (:issue:`55413`)
+- Deprecated :meth:`Series.view`, use ``astype`` instead to change the dtype (:issue:`20251`)
 - Deprecated ``core.internals`` members ``Block``, ``ExtensionBlock``, and ``DatetimeTZBlock``, use public APIs instead (:issue:`55139`)
 - Deprecated ``year``, ``month``, ``quarter``, ``day``, ``hour``, ``minute``, and ``second`` keywords in the :class:`PeriodIndex` constructor, use :meth:`PeriodIndex.from_fields` instead (:issue:`55960`)
 - Deprecated allowing non-integer ``periods`` argument in :func:`date_range`, :func:`timedelta_range`, :func:`period_range`, and :func:`interval_range` (:issue:`56036`)

pandas/core/series.py

@@ -877,6 +877,10 @@ def view(self, dtype: Dtype | None = None) -> Series:
         """
         Create a new view of the Series.
 
+        .. deprecated:: 2.2.0
+            ``Series.view`` is deprecated and will be removed in a future version.
+            Use :meth:`Series.astype` as an alternative to change the dtype.
+
         This function will return a new Series with a view of the same
         underlying values in memory, optionally reinterpreted with a new data
         type. The new data type must preserve the same size in bytes as to not
@@ -907,38 +911,14 @@ def view(self, dtype: Dtype | None = None) -> Series:
 
         Examples
         --------
-        >>> s = pd.Series([-2, -1, 0, 1, 2], dtype='int8')
-        >>> s
-        0   -2
-        1   -1
-        2    0
-        3    1
-        4    2
-        dtype: int8
-
-        The 8 bit signed integer representation of `-1` is `0b11111111`, but
-        the same bytes represent 255 if read as an 8 bit unsigned integer:
-
-        >>> us = s.view('uint8')
-        >>> us
-        0    254
-        1    255
-        2      0
-        3      1
-        4      2
-        dtype: uint8
-
-        The views share the same underlying values:
-
-        >>> us[0] = 128
-        >>> s
-        0   -128
-        1     -1
-        2      0
-        3      1
-        4      2
-        dtype: int8
+        Use ``astype`` to change the dtype instead.
         """
+        warnings.warn(
+            "Series.view is deprecated and will be removed in a future version. "
+            "Use ``astype`` as an alternative to change the dtype.",
+            FutureWarning,
+            stacklevel=2,
+        )
         # self.array instead of self._values so we piggyback on NumpyExtensionArray
         #  implementation
         res_values = self.array.view(dtype)

pandas/core/window/ewm.py

@@ -15,6 +15,7 @@
     is_datetime64_ns_dtype,
     is_numeric_dtype,
 )
+from pandas.core.dtypes.generic import ABCSeries
 from pandas.core.dtypes.missing import isna
 
 from pandas.core import common
@@ -118,6 +119,8 @@ def _calculate_deltas(
     np.ndarray
         Diff of the times divided by the half-life
     """
+    if isinstance(times, ABCSeries):
+        times = times._values
     _times = np.asarray(times.view(np.int64), dtype=np.float64)
     # TODO: generalize to non-nano?
     _halflife = float(Timedelta(halflife).as_unit("ns")._value)

pandas/io/stata.py

@@ -429,9 +429,9 @@ def parse_dates_safe(
                 d["year"] = date_index._data.year
                 d["month"] = date_index._data.month
             if days:
-                days_in_ns = dates.view(np.int64) - to_datetime(
+                days_in_ns = dates._values.view(np.int64) - to_datetime(
                     d["year"], format="%Y"
-                ).view(np.int64)
+                )._values.view(np.int64)
                 d["days"] = days_in_ns // NS_PER_DAY
 
         elif infer_dtype(dates, skipna=False) == "datetime":

pandas/tests/copy_view/test_methods.py

@@ -1934,7 +1934,8 @@ def test_series_view(using_copy_on_write, warn_copy_on_write):
     ser = Series([1, 2, 3])
     ser_orig = ser.copy()
 
-    ser2 = ser.view()
+    with tm.assert_produces_warning(FutureWarning, match="is deprecated"):
+        ser2 = ser.view()
     assert np.shares_memory(get_array(ser), get_array(ser2))
     if using_copy_on_write:
         assert not ser2._mgr._has_no_reference(0)

pandas/tests/frame/indexing/test_where.py

@@ -771,7 +771,8 @@ def test_where_datetimelike_noop(self, dtype):
         # GH#45135, analogue to GH#44181 for Period don't raise on no-op
         # For td64/dt64/dt64tz we already don't raise, but also are
         #  checking that we don't unnecessarily upcast to object.
-        ser = Series(np.arange(3) * 10**9, dtype=np.int64).view(dtype)
+        with tm.assert_produces_warning(FutureWarning, match="is deprecated"):
+            ser = Series(np.arange(3) * 10**9, dtype=np.int64).view(dtype)
         df = ser.to_frame()
         mask = np.array([False, False, False])
 

pandas/tests/frame/test_constructors.py

@@ -1101,8 +1101,8 @@ def test_constructor_maskedarray_nonfloat(self):
         mat2[0, 0] = 1
         mat2[1, 2] = 2
         frame = DataFrame(mat2, columns=["A", "B", "C"], index=[1, 2])
-        assert 1 == frame["A"].view("i8")[1]
-        assert 2 == frame["C"].view("i8")[2]
+        assert 1 == frame["A"].astype("i8")[1]
+        assert 2 == frame["C"].astype("i8")[2]
 
         # masked bool promoted to object
         mat = ma.masked_all((2, 3), dtype=bool)

pandas/tests/generic/test_finalize.py

@@ -31,11 +31,6 @@
 # - Callable: pass the constructed value with attrs set to this.
 
 _all_methods = [
-    (
-        pd.Series,
-        (np.array([0], dtype="float64")),
-        operator.methodcaller("view", "int64"),
-    ),
     (pd.Series, ([0],), operator.methodcaller("take", [])),
     (pd.Series, ([0],), operator.methodcaller("__getitem__", [True])),
     (pd.Series, ([0],), operator.methodcaller("repeat", 2)),

pandas/tests/groupby/test_cumulative.py

@@ -216,7 +216,7 @@ def test_cummax_i8_at_implementation_bound():
     # the minimum value used to be treated as NPY_NAT+1 instead of NPY_NAT
     #  for int64 dtype GH#46382
     ser = Series([pd.NaT._value + n for n in range(5)])
-    df = DataFrame({"A": 1, "B": ser, "C": ser.view("M8[ns]")})
+    df = DataFrame({"A": 1, "B": ser, "C": ser._values.view("M8[ns]")})
     gb = df.groupby("A")
 
     res = gb.cummax()

pandas/tests/groupby/test_timegrouper.py

@@ -726,7 +726,7 @@ def test_groupby_groups_periods(self):
 
     def test_groupby_first_datetime64(self):
         df = DataFrame([(1, 1351036800000000000), (2, 1351036800000000000)])
-        df[1] = df[1].view("M8[ns]")
+        df[1] = df[1].astype("M8[ns]")
 
         assert issubclass(df[1].dtype.type, np.datetime64)
 

pandas/tests/io/json/test_pandas.py

@@ -168,7 +168,7 @@ def test_frame_non_unique_columns(self, orient, data):
                 # in milliseconds; these are internally stored in nanosecond,
                 # so divide to get where we need
                 # TODO: a to_epoch method would also solve; see GH 14772
-                expected.iloc[:, 0] = expected.iloc[:, 0].view(np.int64) // 1000000
+                expected.iloc[:, 0] = expected.iloc[:, 0].astype(np.int64) // 1000000
         elif orient == "split":
             expected = df
             expected.columns = ["x", "x.1"]

pandas/tests/io/test_sql.py

@@ -1908,7 +1908,7 @@ def test_api_timedelta(conn, request):
             name="foo",
         )
     else:
-        expected = df["foo"].view("int64")
+        expected = df["foo"].astype("int64")
     tm.assert_series_equal(result["foo"], expected)
 
 

pandas/tests/series/methods/test_view.py

@@ -9,6 +9,10 @@
 )
 import pandas._testing as tm
 
+pytestmark = pytest.mark.filterwarnings(
+    "ignore:Series.view is deprecated and will be removed in a future version.:FutureWarning"  # noqa: E501
+)
+
 
 class TestView:
     def test_view_i8_to_datetimelike(self):

pandas/tests/series/test_constructors.py

@@ -969,7 +969,7 @@ def test_constructor_dtype_datetime64_10(self):
         # GH3414 related
         expected = Series(pydates, dtype="datetime64[ms]")
 
-        result = Series(Series(dates).view(np.int64) / 1000000, dtype="M8[ms]")
+        result = Series(Series(dates).astype(np.int64) / 1000000, dtype="M8[ms]")
         tm.assert_series_equal(result, expected)
 
         result = Series(dates, dtype="datetime64[ms]")

pandas/tests/test_algos.py

@@ -976,7 +976,7 @@ def test_isin_datetimelike_values_numeric_comps(self, dtype, dtype1):
         # Anything but object and we get all-False shortcut
 
         dta = date_range("2013-01-01", periods=3)._values
-        arr = Series(dta.view("i8")).view(dtype1)._values
+        arr = Series(dta.view("i8")).array.view(dtype1)
 
         comps = arr.view("i8").astype(dtype)
 

pandas/tests/test_nanops.py

@@ -1114,12 +1114,16 @@ def test_nanmean(self, unit):
         expected = dti[1]
 
         for obj in [dti, DatetimeArray(dti), Series(dti)]:
+            if isinstance(obj, Series):
+                obj = obj._values
             result = nanops.nanmean(obj)
             assert result == expected
 
         dti2 = dti.insert(1, pd.NaT)
 
         for obj in [dti2, DatetimeArray(dti2), Series(dti2)]:
+            if isinstance(obj, Series):
+                obj = obj._values
             result = nanops.nanmean(obj)
             assert result == expected