DEPR: enforce ufunc, reduction diff deprecations (pandas-dev#49717)

Author: jbrockmendel

doc/source/whatsnew/v2.0.0.rst

@@ -553,6 +553,9 @@ Removal of prior version deprecations/changes
 - Changed behavior of :meth:`DataFrame.apply` with list-like so that any partial failure will raise an error (:issue:`43740`)
 - Changed behavior of :meth:`Series.__setitem__` with an integer key and a :class:`Float64Index` when the key is not present in the index; previously we treated the key as positional (behaving like ``series.iloc[key] = val``), now we treat it is a label (behaving like ``series.loc[key] = val``), consistent with :meth:`Series.__getitem__`` behavior (:issue:`33469`)
 - Removed ``na_sentinel`` argument from :func:`factorize`, :meth:`.Index.factorize`, and :meth:`.ExtensionArray.factorize` (:issue:`47157`)
+- Changed behavior of :meth:`Series.diff` and :meth:`DataFrame.diff` with :class:`ExtensionDtype` dtypes whose arrays do not implement ``diff``, these now raise ``TypeError`` rather than casting to numpy (:issue:`31025`)
+- Enforced deprecation of calling numpy "ufunc"s on :class:`DataFrame` with ``method="outer"``; this now raises ``NotImplementedError`` (:issue:`36955`)
+- Enforced deprecation disallowing passing ``numeric_only=True`` to :class:`Series` reductions (``rank``, ``any``, ``all``, ...) with non-numeric dtype (:issue:`47500`)
 - Changed behavior of :meth:`DataFrameGroupBy.apply` and :meth:`SeriesGroupBy.apply` so that ``group_keys`` is respected even if a transformer is detected (:issue:`34998`)
 - Enforced deprecation ``numeric_only=None`` (the default) in DataFrame reductions that would silently drop columns that raised; ``numeric_only`` now defaults to ``False`` (:issue:`41480`)
 - Changed default of ``numeric_only`` to ``False`` in all DataFrame methods with that argument (:issue:`46096`, :issue:`46906`)

pandas/core/algorithms.py

@@ -1616,14 +1616,10 @@ def diff(arr, n: int, axis: AxisInt = 0):
  raise ValueError(f"cannot diff {type(arr).__name__} on axis={axis}")
  return op(arr, arr.shift(n))
  else:
- warnings.warn(
- "dtype lost in 'diff()'. In the future this will raise a "
- "TypeError. Convert to a suitable dtype prior to calling 'diff'.",
- FutureWarning,
- stacklevel=find_stack_level(),
+ raise TypeError(
+ f"{type(arr).__name__} has no 'diff' method. "
+ "Convert to a suitable dtype prior to calling 'diff'."
  )
- arr = np.asarray(arr)
- dtype = arr.dtype
 
  is_timedelta = False
  if needs_i8_conversion(arr.dtype):

pandas/core/arraylike.py

@@ -341,19 +341,6 @@ def _reconstruct(result):
 
  if result.ndim != self.ndim:
  if method == "outer":
- if self.ndim == 2:
- # we already deprecated for Series
- msg = (
- "outer method for ufunc {} is not implemented on "
- "pandas objects. Returning an ndarray, but in the "
- "future this will raise a 'NotImplementedError'. "
- "Consider explicitly converting the DataFrame "
- "to an array with '.to_numpy()' first."
- )
- warnings.warn(
- msg.format(ufunc), FutureWarning, stacklevel=find_stack_level()
- )
- return result
  raise NotImplementedError
  return result
  if isinstance(result, BlockManager):

pandas/core/generic.py

@@ -9000,12 +9000,9 @@ def ranker(data):
  if numeric_only:
  if self.ndim == 1 and not is_numeric_dtype(self.dtype):
  # GH#47500
- warnings.warn(
- f"Calling Series.rank with numeric_only={numeric_only} and dtype "
- f"{self.dtype} is deprecated and will raise a TypeError in a "
- "future version of pandas",
- category=FutureWarning,
- stacklevel=find_stack_level(),
+ raise TypeError(
+ "Series.rank does not allow numeric_only=True with "
+ "non-numeric dtype."
  )
  data = self._get_numeric_data()
  else:
@@ -10946,6 +10943,7 @@ def _stat_function(
  FutureWarning,
  stacklevel=find_stack_level(),
  )
+
  if axis is lib.no_default:
  axis = None
 

pandas/core/series.py

@@ -18,7 +18,6 @@
  cast,
  overload,
 )
-import warnings
 import weakref
 
 import numpy as np
@@ -69,7 +68,6 @@
  Substitution,
  doc,
 )
-from pandas.util._exceptions import find_stack_level
 from pandas.util._validators import (
  validate_ascending,
  validate_bool_kwarg,
@@ -4579,14 +4577,9 @@ def _reduce(
  if name in ["any", "all"]:
  kwd_name = "bool_only"
  # GH#47500 - change to TypeError to match other methods
- warnings.warn(
- f"Calling Series.{name} with {kwd_name}={numeric_only} and "
- f"dtype {self.dtype} will raise a TypeError in the future",
- FutureWarning,
- stacklevel=find_stack_level(),
- )
- raise NotImplementedError(
- f"Series.{name} does not implement {kwd_name}."
+ raise TypeError(
+ f"Series.{name} does not allow {kwd_name}={numeric_only} "
+ "with non-numeric dtypes."
  )
  with np.errstate(all="ignore"):
  return op(delegate, skipna=skipna, **kwds)

pandas/tests/arrays/categorical/test_algos.py

@@ -69,15 +69,12 @@ def test_isin_empty(empty):
 
 
 def test_diff():
- s = pd.Series([1, 2, 3], dtype="category")
- with tm.assert_produces_warning(FutureWarning):
- result = s.diff()
- expected = pd.Series([np.nan, 1, 1])
- tm.assert_series_equal(result, expected)
-
- expected = expected.to_frame(name="A")
- df = s.to_frame(name="A")
- with tm.assert_produces_warning(FutureWarning):
- result = df.diff()
-
- tm.assert_frame_equal(result, expected)
+ ser = pd.Series([1, 2, 3], dtype="category")
+
+ msg = "Convert to a suitable dtype"
+ with pytest.raises(TypeError, match=msg):
+ ser.diff()
+
+ df = ser.to_frame(name="A")
+ with pytest.raises(TypeError, match=msg):
+ df.diff()

pandas/tests/frame/test_reductions.py

@@ -1489,6 +1489,7 @@ def test_median_categorical_dtype_nuisance_column(self):
  # TODO: np.median(df, axis=0) gives np.array([2.0, 2.0]) instead
  # of expected.values
 
+ @pytest.mark.filterwarnings("ignore:.*will return a scalar.*:FutureWarning")
  @pytest.mark.parametrize("method", ["min", "max"])
  def test_min_max_categorical_dtype_non_ordered_nuisance_column(self, method):
  # GH#28949 DataFrame.min should behave like Series.min

pandas/tests/frame/test_ufunc.py

@@ -200,9 +200,10 @@ def test_unary_accumulate_axis():
  tm.assert_frame_equal(result, expected)
 
 
-def test_frame_outer_deprecated():
+def test_frame_outer_disallowed():
  df = pd.DataFrame({"A": [1, 2]})
- with tm.assert_produces_warning(FutureWarning):
+ with pytest.raises(NotImplementedError, match=""):
+ # deprecation enforced in 2.0
  np.subtract.outer(df, df)
 
 

pandas/tests/groupby/test_function.py

@@ -1528,8 +1528,10 @@ def test_deprecate_numeric_only_series(dtype, groupby_func, request):
  err_category = TypeError
  err_msg = f"{groupby_func} is not supported for object dtype"
  elif groupby_func == "skew":
- warn_category = FutureWarning
- warn_msg = "will raise a TypeError in the future"
+ warn_category = None
+ warn_msg = ""
+ err_category = TypeError
+ err_msg = "Series.skew does not allow numeric_only=True with non-numeric"
  else:
  warn_category = FutureWarning
  warn_msg = "This will raise a TypeError"

pandas/tests/groupby/transform/test_transform.py

@@ -495,9 +495,9 @@ def test_transform_coercion():
 
  expected = g.transform(np.mean)
 
- msg = "will return a scalar mean"
- with tm.assert_produces_warning(FutureWarning, match=msg, check_stacklevel=False):
-  result = g.transform(lambda x: np.mean(x))
+ # in 2.0 np.mean on a DataFrame is equivalent to frame.mean(axis=None)
+ # which not gives a scalar instead of Series
+ result = g.transform(lambda x: np.mean(x))
  tm.assert_frame_equal(result, expected)
 
  with tm.assert_produces_warning(None):