non-internals changes

Author: TomAugspurger

pandas/compat/__init__.py

@@ -451,3 +451,6 @@ def is_platform_mac():
 
 def is_platform_32bit():
     return struct.calcsize("P") * 8 < 64
+
+
+_default_fill_value = object()

pandas/core/algorithms.py

@@ -30,6 +30,7 @@
     _ensure_platform_int, _ensure_object,
     _ensure_float64, _ensure_uint64,
     _ensure_int64)
+from pandas.compat import _default_fill_value
 from pandas.compat.numpy import _np_version_under1p10
 from pandas.core.dtypes.missing import isna, na_value_for_dtype
 
@@ -1441,6 +1442,27 @@ def func(arr, indexer, out, fill_value=np.nan):
     return func
 
 
+def take(arr, indexer, fill_value=_default_fill_value):
+    indexer = np.asarray(indexer)
+
+    if fill_value is _default_fill_value:
+        # NumPy style
+        result = arr.take(indexer)
+    else:
+        # bounds checking
+        if (indexer < -1).any():
+            raise ValueError("Invalid value in 'indexer'. All values "
+                             "must be non-negative or -1. When "
+                             "'fill_value' is specified.")
+
+        # # take on empty array not handled as desired by numpy
+        # # in case of -1 (all missing take)
+        # if not len(arr) and mask.all():
+        #     return arr._from_sequence([fill_value] * len(indexer))
+        result = take_1d(arr, indexer, fill_value=fill_value)
+    return result
+
+
 def take_nd(arr, indexer, axis=0, out=None, fill_value=np.nan, mask_info=None,
             allow_fill=True):
     """

pandas/core/arrays/base.py

@@ -10,11 +10,11 @@
 import numpy as np
 
 from pandas.errors import AbstractMethodError
+from pandas.compat import _default_fill_value
 from pandas.compat.numpy import function as nv
 from pandas.util._decorators import Appender, Substitution
 
 _not_implemented_message = "{} does not implement {}."
-_default_fill_value = object()
 
 
 _take_docstring = textwrap.dedent("""\
@@ -537,13 +537,12 @@ def _values_for_take(self):
     @Appender(_take_docstring)
     def take(self, indexer, fill_value=_default_fill_value):
         # type: (Sequence[int], Optional[Any]) -> ExtensionArray
-        from pandas.core.missing import isna
-
-        if isna(fill_value):
-            fill_value = self.dtype.na_value
+        if fill_value is np.nan:
+            import pdb; pdb.set_trace()
+        from pandas.core.algorithms import take
 
         data = self._values_for_take()
-        result = take_ea(data, indexer, fill_value=fill_value)
+        result = take(data, indexer, fill_value=fill_value)
         return self._from_sequence(result)
 
     def copy(self, deep=False):
@@ -609,30 +608,3 @@ def _ndarray_values(self):
         """
         return np.array(self)
 
-
-@Substitution(arr=textwrap.dedent("""\
-arr : array-like
-    Must satisfy NumPy's indexing sematnics, including `take`
-    and boolean masking.
-"""))
-@Appender(_take_docstring)
-def take_ea(arr, indexer, fill_value=_default_fill_value):
-    indexer = np.asarray(indexer)
-    if fill_value is _default_fill_value:
-        # NumPy style
-        result = arr.take(indexer)
-    else:
-        mask = indexer == -1
-        if (indexer < -1).any():
-            raise ValueError("Invalid value in 'indexer'. All values "
-                             "must be non-negative or -1. When "
-                             "'fill_value' is specified.")
-
-        # take on empty array not handled as desired by numpy
-        # in case of -1 (all missing take)
-        if not len(arr) and mask.all():
-            return arr._from_sequence([fill_value] * len(indexer))
-
-        result = arr.take(indexer)
-        result[mask] = fill_value
-    return result

pandas/core/frame.py

@@ -77,7 +77,7 @@
 from pandas.core.arrays import Categorical, ExtensionArray
 import pandas.core.algorithms as algorithms
 from pandas.compat import (range, map, zip, lrange, lmap, lzip, StringIO, u,
-                           OrderedDict, raise_with_traceback)
+                           OrderedDict, raise_with_traceback, _default_fill_value)
 from pandas import compat
 from pandas.compat import PY36
 from pandas.compat.numpy import function as nv
@@ -3504,7 +3504,7 @@ def _reindex_multi(self, axes, copy, fill_value):
 
     @Appender(_shared_docs['align'] % _shared_doc_kwargs)
     def align(self, other, join='outer', axis=None, level=None, copy=True,
-              fill_value=None, method=None, limit=None, fill_axis=0,
+              fill_value=_default_fill_value, method=None, limit=None, fill_axis=0,
               broadcast_axis=None):
         return super(DataFrame, self).align(other, join=join, axis=axis,
                                             level=level, copy=copy,

pandas/core/generic.py

@@ -50,7 +50,8 @@
 from pandas import compat
 from pandas.compat.numpy import function as nv
 from pandas.compat import (map, zip, lzip, lrange, string_types, to_str,
-                           isidentifier, set_function_name, cPickle as pkl)
+                           isidentifier, set_function_name, cPickle as pkl,
+                           _default_fill_value)
 from pandas.core.ops import _align_method_FRAME
 import pandas.core.nanops as nanops
 from pandas.util._decorators import (Appender, Substitution,
@@ -3660,7 +3661,7 @@ def reindex(self, *args, **kwargs):
         copy = kwargs.pop('copy', True)
         limit = kwargs.pop('limit', None)
         tolerance = kwargs.pop('tolerance', None)
-        fill_value = kwargs.pop('fill_value', np.nan)
+        fill_value = kwargs.pop('fill_value', _default_fill_value)
 
         # Series.reindex doesn't use / need the axis kwarg
         # We pop and ignore it here, to make writing Series/Frame generic code
@@ -3790,7 +3791,9 @@ def reindex_axis(self, labels, axis=0, method=None, level=None, copy=True,
         return self._reindex_with_indexers({axis: [new_index, indexer]},
                                            fill_value=fill_value, copy=copy)
 
-    def _reindex_with_indexers(self, reindexers, fill_value=np.nan, copy=False,
+    def _reindex_with_indexers(self, reindexers,
+                               fill_value=_default_fill_value,
+                               copy=False,
                                allow_dups=False):
         """allow_dups indicates an internal call here """
 
@@ -7209,7 +7212,7 @@ def ranker(data):
 
     @Appender(_shared_docs['align'] % _shared_doc_kwargs)
     def align(self, other, join='outer', axis=None, level=None, copy=True,
-              fill_value=None, method=None, limit=None, fill_axis=0,
+              fill_value=_default_fill_value, method=None, limit=None, fill_axis=0,
               broadcast_axis=None):
         from pandas import DataFrame, Series
         method = missing.clean_fill_method(method)
@@ -7359,6 +7362,9 @@ def _align_series(self, other, join='outer', axis=None, level=None,
                 right = other.reindex(join_index, level=level)
 
         # fill
+        if fill_value is _default_fill_value:
+            fill_value = None
+
         fill_na = notna(fill_value) or (method is not None)
         if fill_na:
             left = left.fillna(fill_value, method=method, limit=limit,

pandas/core/series.py

@@ -56,7 +56,7 @@
 from pandas import compat
 from pandas.io.formats.terminal import get_terminal_size
 from pandas.compat import (
-    zip, u, OrderedDict, StringIO, range, get_range_parameters, PY36)
+    zip, u, OrderedDict, StringIO, range, get_range_parameters, PY36, _default_fill_value)
 from pandas.compat.numpy import function as nv
 
 import pandas.core.ops as ops
@@ -3216,7 +3216,10 @@ def _reindex_indexer(self, new_index, indexer, copy):
                 return self.copy()
             return self
 
-        new_values = algorithms.take_1d(self._values, indexer)
+        from pandas.core.dtypes.missing import na_value_for_dtype
+        fill_value = na_value_for_dtype(self.dtype)
+        new_values = algorithms.take(self._values, indexer,
+                                     fill_value=fill_value)
         return self._constructor(new_values, index=new_index)
 
     def _needs_reindex_multi(self, axes, method, level):
@@ -3227,7 +3230,7 @@ def _needs_reindex_multi(self, axes, method, level):
 
     @Appender(generic._shared_docs['align'] % _shared_doc_kwargs)
     def align(self, other, join='outer', axis=None, level=None, copy=True,
-              fill_value=None, method=None, limit=None, fill_axis=0,
+              fill_value=_default_fill_value, method=None, limit=None, fill_axis=0,
               broadcast_axis=None):
         return super(Series, self).align(other, join=join, axis=axis,
                                          level=level, copy=copy,

pandas/tests/extension/decimal/test_decimal.py

@@ -108,7 +108,25 @@ class TestReshaping(BaseDecimal, base.BaseReshapingTests):
 
 
 class TestGetitem(BaseDecimal, base.BaseGetitemTests):
-    pass
+
+    def test_take_basic(self):
+        ea = DecimalArray([decimal.Decimal('1'),
+                           decimal.Decimal('2'),
+                           decimal.Decimal('3')])
+        result = ea.take([1, 2, -1])
+        expected = DecimalArray([decimal.Decimal('2'),
+                                 decimal.Decimal('3'),
+                                 decimal.Decimal('3')])
+        self.assert_extension_array_equal(result, expected)
+
+        result = ea.take([1, 2, -1], fill_value=ea.dtype.na_value)
+        expected = DecimalArray([decimal.Decimal('2'),
+                                 decimal.Decimal('3'),
+                                 decimal.Decimal('NaN')])
+        self.assert_extension_array_equal(result, expected)
+
+        result = pd.Series(ea).reindex([1, 2, -1]).values
+        self.assert_extension_array_equal(result, expected)
 
 
 class TestMissing(BaseDecimal, base.BaseMissingTests):