BUG: Fix series.round() handling of EA

pandas/compat/numpy/__init__.py

@@ -1,9 +1,9 @@
 """ support numpy compatiblitiy across versions """
 
-import re
-import numpy as np
 from distutils.version import LooseVersion
+import re
 
+import numpy as np
 
 # numpy versioning
 _np_version = np.__version__
@@ -13,7 +13,10 @@
 _np_version_under1p16 = _nlv < LooseVersion('1.16')
 _np_version_under1p17 = _nlv < LooseVersion('1.17')
 _is_numpy_dev = '.dev' in str(_nlv)
-
+try:
+    _NEP18_enabled = np.core.overrides.ENABLE_ARRAY_FUNCTION
+except Exception:
+    _NEP18_enabled = False
 
 if _nlv < '1.13.3':
     raise ImportError('this version of pandas is incompatible with '
@@ -62,6 +65,7 @@ def np_array_datetime64_compat(arr, *args, **kwargs):
 
 
 __all__ = ['np',
+           '_np_version',
            '_np_version_under1p14',
            '_np_version_under1p15',
            '_np_version_under1p16',

pandas/core/series.py

@@ -2098,7 +2098,7 @@ def round(self, decimals=0, *args, **kwargs):
         dtype: float64
         """
         nv.validate_round(args, kwargs)
-        result = com.values_from_object(self).round(decimals)
+        result = np.round(self.array, decimals=decimals)
         result = self._constructor(result, index=self.index).__finalize__(self)
 
         return result

pandas/tests/arrays/sparse/test_array.py

@@ -6,7 +6,6 @@
 import pytest
 
 from pandas._libs.sparse import IntIndex
-from pandas.compat.numpy import _np_version_under1p16
 import pandas.util._test_decorators as td
 
 import pandas as pd
@@ -175,8 +174,8 @@ def test_constructor_inferred_fill_value(self, data, fill_value):
     @pytest.mark.parametrize('format', ['coo', 'csc', 'csr'])
     @pytest.mark.parametrize('size', [
         pytest.param(0,
-                     marks=pytest.mark.skipif(_np_version_under1p16,
-                                              reason='NumPy-11383')),
+                     marks=td.skip_if_np_lt("1.16",
+                                            reason='NumPy-11383')),
         10
     ])
     @td.skip_if_no_scipy
@@ -870,7 +869,7 @@ def test_all(self, data, pos, neg):
         ([1, 2, 1], 1, 0),
         ([1.0, 2.0, 1.0], 1.0, 0.0)
     ])
-    @td.skip_if_np_lt_115  # prior didn't dispatch
+    @td.skip_if_np_lt("1.15")  # prior didn't dispatch
     def test_numpy_all(self, data, pos, neg):
         # GH 17570
         out = np.all(SparseArray(data))
@@ -916,7 +915,7 @@ def test_any(self, data, pos, neg):
         ([0, 2, 0], 2, 0),
         ([0.0, 2.0, 0.0], 2.0, 0.0)
     ])
-    @td.skip_if_np_lt_115  # prior didn't dispatch
+    @td.skip_if_np_lt("1.15")  # prior didn't dispatch
     def test_numpy_any(self, data, pos, neg):
         # GH 17570
         out = np.any(SparseArray(data))

pandas/tests/extension/base/methods.py

@@ -323,7 +323,7 @@ def test_repeat(self, data, repeats, as_series, use_numpy):
         self.assert_equal(result, expected)
 
     @pytest.mark.parametrize('repeats, kwargs, error, msg', [
-        (2, dict(axis=1), ValueError, "'axis"),
+        (2, dict(axis=1), ValueError, "'?axis"),
         (-1, dict(), ValueError, "negative"),
         ([1, 2], dict(), ValueError, "shape"),
         (2, dict(foo='bar'), TypeError, "'foo'")])

pandas/tests/extension/decimal/array.py

@@ -11,6 +11,8 @@
 from pandas.api.extensions import register_extension_dtype
 from pandas.core.arrays import ExtensionArray, ExtensionScalarOpsMixin
 
+_should_cast_results = [np.repeat]
+
 
 @register_extension_dtype
 class DecimalDtype(ExtensionDtype):
@@ -59,14 +61,28 @@ def __init__(self, values, dtype=None, copy=False, context=None):
         values = np.asarray(values, dtype=object)
 
         self._data = values
-        # Some aliases for common attribute names to ensure pandas supports
-        # these
-        self._items = self.data = self._data
-        # those aliases are currently not working due to assumptions
-        # in internal code (GH-20735)
-        # self._values = self.values = self.data
         self._dtype = DecimalDtype(context)
 
+    # aliases for common attribute names, to ensure pandas supports these
+    @property
+    def _items(self):
+        return self._data
+
+    @property
+    def data(self):
+        return self._data
+
+    # those aliases are currently not working due to assumptions
+    # in internal code (GH-20735)
+    # @property
+    # def _values(self):
+    #     return self._data
+
+    # @property
+    # def values(self):
+    #     return self._data
+    # end aliases
+
     @property
     def dtype(self):
         return self._dtype
@@ -153,6 +169,40 @@ def _reduce(self, name, skipna=True, **kwargs):
                                       "the {} operation".format(name))
         return op(axis=0)
 
+    # numpy experimental NEP-18 (opt-in numpy 1.16, enabled in in 1.17)
+    def __array_function__(self, func, types, args, kwargs):
+        def coerce_EA(coll):
+            # In order to delegate to numpy, we have to coerce any
+            # ExtensionArrays to the best numpy-friendly dtype approximation
+            # Different functions take different arguments, which may be
+            # nested collections, so we look at everything. Sigh.
+            for i in range(len(coll)):
+                if isinstance(coll[i], (tuple, list)):
+                    coll[i] = coerce_EA(list(coll[i]))
+                else:
+                    if isinstance(coll[i], DecimalArray):
+                        # TODO: how to check for any ndarray-like with
+                        # non-numpy dtype?
+                        coll[i] = np.array(coll[i], dtype=object)
+
+            return coll
+
+        if func is np.round_:
+            values = [decimal.Decimal(round(_))
+                      for _ in self._data]
+            return DecimalArray(values, dtype=self.dtype)
+
+        elif True:  # just assume we can handle all functions
+            args = coerce_EA(list(args))
+            result = func(*args, **kwargs)
+
+            if func in _should_cast_results:
+                result = pd.array(result, dtype=self.dtype)
+
+            return result
+        else:
+            return NotImplemented
+
 
 def to_decimal(values, context=None):
     return DecimalArray([decimal.Decimal(x) for x in values], context=context)

pandas/tests/extension/decimal/test_decimal.py

@@ -5,6 +5,8 @@
 import numpy as np
 import pytest
 
+import pandas.util._test_decorators as td
+
 import pandas as pd
 from pandas.tests.extension import base
 import pandas.util.testing as tm
@@ -380,6 +382,26 @@ def test_divmod_array(reverse, expected_div, expected_mod):
     tm.assert_extension_array_equal(mod, expected_mod)
 
 
+# numpy 1.17 has NEP-18 on by default
+# for numpy 1.16 set shell variable with
+# "export NUMPY_EXPERIMENTAL_ARRAY_FUNCTION=1"
+# before running pytest/python.
+# verify by checking value of `np.core.overrides.ENABLE_ARRAY_FUNCTION`
+@td.skip_if_no_NEP18
+def test_series_round():
+    ser = pd.Series(to_decimal([1.1, 2.4, 3.1])).round()
+    expected = pd.Series(to_decimal([1, 2, 3]))
+    tm.assert_extension_array_equal(ser.array, expected.array)
+    tm.assert_series_equal(ser, expected)
+
+
+@td.skip_if_no_NEP18
+def test_series_round_then_sum():
+    result = pd.Series(to_decimal([1.1, 2.4, 3.1])).round().sum(skipna=False)
+    expected = decimal.Decimal("6")
+    assert result == expected
+
+
 def test_formatting_values_deprecated():
     class DecimalArray2(DecimalArray):
         def _formatting_values(self):

pandas/tests/frame/test_analytics.py

@@ -1565,21 +1565,21 @@ def test_any_all_bool_only(self):
         (np.all, {'A': pd.Series([0, 1], dtype=int)}, False),
         (np.any, {'A': pd.Series([0, 1], dtype=int)}, True),
         pytest.param(np.all, {'A': pd.Series([0, 1], dtype='M8[ns]')}, False,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.any, {'A': pd.Series([0, 1], dtype='M8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.all, {'A': pd.Series([1, 2], dtype='M8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.any, {'A': pd.Series([1, 2], dtype='M8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.all, {'A': pd.Series([0, 1], dtype='m8[ns]')}, False,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.any, {'A': pd.Series([0, 1], dtype='m8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.all, {'A': pd.Series([1, 2], dtype='m8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         pytest.param(np.any, {'A': pd.Series([1, 2], dtype='m8[ns]')}, True,
-                     marks=[td.skip_if_np_lt_115]),
+                     marks=[td.skip_if_np_lt("1.15")]),
         (np.all, {'A': pd.Series([0, 1], dtype='category')}, False),
         (np.any, {'A': pd.Series([0, 1], dtype='category')}, True),
         (np.all, {'A': pd.Series([1, 2], dtype='category')}, True),

pandas/tests/series/test_analytics.py

@@ -1105,7 +1105,7 @@ def test_value_counts_categorical_not_ordered(self):
         dict(keepdims=True),
         dict(out=object()),
     ])
-    @td.skip_if_np_lt_115
+    @td.skip_if_np_lt("1.15")
     def test_validate_any_all_out_keepdims_raises(self, kwargs, func):
         s = pd.Series([1, 2])
         param = list(kwargs)[0]
@@ -1117,7 +1117,7 @@ def test_validate_any_all_out_keepdims_raises(self, kwargs, func):
         with pytest.raises(ValueError, match=msg):
             func(s, **kwargs)
 
-    @td.skip_if_np_lt_115
+    @td.skip_if_np_lt("1.15")
     def test_validate_sum_initial(self):
         s = pd.Series([1, 2])
         msg = (r"the 'initial' parameter is not "
@@ -1136,7 +1136,7 @@ def test_validate_median_initial(self):
             # method instead of the ufunc.
             s.median(overwrite_input=True)
 
-    @td.skip_if_np_lt_115
+    @td.skip_if_np_lt("1.15")
     def test_validate_stat_keepdims(self):
         s = pd.Series([1, 2])
         msg = (r"the 'keepdims' parameter is not "

pandas/util/_test_decorators.py

@@ -23,14 +23,15 @@ def test_foo():
 
 For more information, refer to the ``pytest`` documentation on ``skipif``.
 """
+from distutils.version import LooseVersion
 import locale
 from typing import Optional
 
 from _pytest.mark.structures import MarkDecorator
 import pytest
 
 from pandas.compat import is_platform_32bit, is_platform_windows
-from pandas.compat.numpy import _np_version_under1p15
+from pandas.compat.numpy import _NEP18_enabled, _np_version
 
 from pandas.core.computation.expressions import (
     _NUMEXPR_INSTALLED, _USE_NUMEXPR)
@@ -142,8 +143,6 @@ def skip_if_no(
 
 skip_if_no_mpl = pytest.mark.skipif(_skip_if_no_mpl(),
                                     reason="Missing matplotlib dependency")
-skip_if_np_lt_115 = pytest.mark.skipif(_np_version_under1p15,
-                                       reason="NumPy 1.15 or greater required")
 skip_if_mpl = pytest.mark.skipif(not _skip_if_no_mpl(),
                                  reason="matplotlib is present")
 skip_if_32bit = pytest.mark.skipif(is_platform_32bit(),
@@ -166,6 +165,15 @@ def skip_if_no(
                                    "installed->{installed}".format(
                                        enabled=_USE_NUMEXPR,
                                        installed=_NUMEXPR_INSTALLED))
+skip_if_no_NEP18 = pytest.mark.skipif(not _NEP18_enabled,
+                                      reason="numpy NEP-18 disabled")
+
+
+def skip_if_np_lt(ver_str, reason=None, *args, **kwds):
+    if reason is None:
+        reason = "NumPy %s or greater required" % ver_str
+    return pytest.mark.skipif(_np_version < LooseVersion(ver_str),
+                              reason=reason, *args, **kwds)
 
 
 def parametrize_fixture_doc(*args):