Drop support for NaN categories in Categorical

Deprecated in 0.17.0.

xref gh-10748

Author: gfyoung

doc/source/whatsnew/v0.20.0.txt

@@ -811,6 +811,7 @@ Removal of prior version deprecations/changes
   in favor of ``iloc`` and ``iat`` as explained :ref:`here <whatsnew_0170.deprecations>` (:issue:`10711`).
 - The deprecated ``DataFrame.iterkv()`` has been removed in favor of ``DataFrame.iteritems()`` (:issue:`10711`)
 - The ``Categorical`` constructor has dropped the ``name`` parameter (:issue:`10632`)
+- ``Categorical`` has dropped support for ``NaN`` categories (:issue:`10748`)
 - The ``take_last`` parameter has been dropped from ``duplicated()``, ``drop_duplicates()``, ``nlargest()``, and ``nsmallest()`` methods (:issue:`10236`, :issue:`10792`, :issue:`10920`)
 - ``Series``, ``Index``, and ``DataFrame`` have dropped the ``sort`` and ``order`` methods (:issue:`10726`)
 - Where clauses in ``pytables`` are only accepted as strings and expressions types and not other data-types (:issue:`12027`)

pandas/core/categorical.py

@@ -545,18 +545,11 @@ def _validate_categories(cls, categories, fastpath=False):
 
         if not fastpath:
 
-            # check properties of the categories
-            # we don't allow NaNs in the categories themselves
-
+            # Categories cannot contain NaN.
             if categories.hasnans:
-                # NaNs in cats deprecated in 0.17
-                # GH 10748
-                msg = ('\nSetting NaNs in `categories` is deprecated and '
-                       'will be removed in a future version of pandas.')
-                warn(msg, FutureWarning, stacklevel=3)
-
-            # categories must be unique
+                raise ValueError('Categorial categories cannot be NaN')
 
+            # Categories must be unique.
             if not categories.is_unique:
                 raise ValueError('Categorical categories must be unique')
 

pandas/tests/test_categorical.py

@@ -222,14 +222,6 @@ def f():
         cat = pd.Categorical([np.nan, 1., 2., 3.])
         self.assertTrue(is_float_dtype(cat.categories))
 
-        # Deprecating NaNs in categoires (GH #10748)
-        # preserve int as far as possible by converting to object if NaN is in
-        # categories
-        with tm.assert_produces_warning(FutureWarning):
-            cat = pd.Categorical([np.nan, 1, 2, 3],
-                                 categories=[np.nan, 1, 2, 3])
-        self.assertTrue(is_object_dtype(cat.categories))
-
         # This doesn't work -> this would probably need some kind of "remember
         # the original type" feature to try to cast the array interface result
         # to...
@@ -418,6 +410,12 @@ def f():
 
         self.assertRaises(ValueError, f)
 
+        # NaN categories included
+        def f():
+            Categorical.from_codes([0, 1, 2], ["a", "b", np.nan])
+
+        self.assertRaises(ValueError, f)
+
         # too negative
         def f():
             Categorical.from_codes([-2, 1, 2], ["a", "b", "c"])