Updates

Author: TomAugspurger

doc/source/whatsnew/v0.23.0.txt

@@ -856,7 +856,7 @@ Other API Changes
 - Constructing a Series from a list of length 1 no longer broadcasts this list when a longer index is specified (:issue:`19714`, :issue:`20391`).
 - :func:`DataFrame.to_dict` with ``orient='index'`` no longer casts int columns to float for a DataFrame with only int and float columns (:issue:`18580`)
 - A user-defined-function that is passed to :func:`Series.rolling().aggregate() <pandas.core.window.Rolling.aggregate>`, :func:`DataFrame.rolling().aggregate() <pandas.core.window.Rolling.aggregate>`, or its expanding cousins, will now *always* be passed a ``Series``, rather than a ``np.array``; ``.apply()`` only has the ``raw`` keyword, see :ref:`here <whatsnew_0230.enhancements.window_raw>`. This is consistent with the signatures of ``.aggregate()`` across pandas (:issue:`20584`)
-- The previous default behavior of negative indices in ``Categorical.take`` is deprecated. In a future version it will change from meaning missing values to meaning positional indicies from the right (:issue:`20664`)
+- The previous default behavior of negative indices in ``Categorical.take`` is deprecated. In a future version it will change from meaning missing values to meaning positional indices from the right. The future behavior is consistent with :meth:`Series.take` (:issue:`20664`).
 
 .. _whatsnew_0230.deprecations:
 
@@ -1025,6 +1025,7 @@ Categorical
 - Bug in ``Categorical.__iter__`` not converting to Python types (:issue:`19909`)
 - Bug in :func:`pandas.factorize` returning the unique codes for the ``uniques``. This now returns a ``Categorical`` with the same dtype as the input (:issue:`19721`)
 - Bug in :func:`pandas.factorize` including an item for missing values in the ``uniques`` return value (:issue:`19721`)
+- Bug in :meth:`Series.take` with categorical data interpreting ``-1`` in `indicies` as missing value markers, rather than the last element of the Series (:issue:`20664`)
 
 Datetimelike
 ^^^^^^^^^^^^

pandas/core/arrays/categorical.py

@@ -1746,9 +1746,33 @@ def fillna(self, value=None, method=None, limit=None):
 
     def take_nd(self, indexer, allow_fill=None, fill_value=None):
         """
-        Return the indices
+        Take elements from the Categorical.
 
-        For internal compatibility with numpy arrays.
+        Parameters
+        ----------
+        indexer : sequence of integers
+        allow_fill : bool, default None.
+            How to handle negative values in `indexer`.
+
+            * False: negative values in `indices` indicate positional indices
+              from the right. This is similar to
+              :func:`numpy.take`.
+
+            * True: negative values in `indices` indicate missing values
+              (the default). These values are set to `fill_value`. Any other
+              other negative values raise a ``ValueError``.
+
+            .. versionchanged:: 0.23.0
+
+               Deprecated the default value of `allow_fill`. The deprecated
+               default is ``True``. In the future, this will change to
+               ``False``.
+
+        Returns
+        -------
+        Categorical
+            This Categorical will have the same categories and ordered as
+            `self`.
         """
         indexer = np.asarray(indexer, dtype=np.intp)
         if allow_fill is None:
@@ -1757,8 +1781,8 @@ def take_nd(self, indexer, allow_fill=None, fill_value=None):
                 allow_fill = True
 
         if fill_value is None or isna(fill_value):
-            # For backwards compatability, we have to override
-            # any na values for `fill_value`
+            # The isna(fill_value) is included for backwards compatability.
+            # Categorical overrides any NA value with -1.
             fill_value = -1
 
         codes = take(self._codes, indexer, allow_fill=allow_fill,

pandas/tests/extension/category/test_categorical.py

@@ -111,7 +111,7 @@ def test_take_non_na_fill_value(self):
     def test_take_out_of_bounds_raises(self):
         pass
 
-    @skip_take
+    @pytest.mark.skip(reason="GH-20747. Unobserved categories.")
     def test_take_series(self):
         pass