BUG: Copy categorical codes if empty (fixes #18051)

rebased to remove conflicts in whats new

Author: Ghasem Naddaf

doc/source/whatsnew/v0.21.1.txt

@@ -130,6 +130,7 @@ Categorical
 - Error messages in the testing module have been improved when items have
   different ``CategoricalDtype`` (:issue:`18069`)
 - ``CategoricalIndex`` can now correctly take a ``pd.api.types.CategoricalDtype`` as its dtype (:issue:`18116`)
+- Bug in ``Categorical.unique()`` returning read-only ``codes``  array when all categories were ``NaN`` (:issue:`18051`)
 
 Other
 ^^^^^

pandas/core/categorical.py

@@ -2250,7 +2250,7 @@ def _recode_for_categories(codes, old_categories, new_categories):
 
     if len(old_categories) == 0:
         # All null anyway, so just retain the nulls
-        return codes
+        return codes.copy()
     indexer = coerce_indexer_dtype(new_categories.get_indexer(old_categories),
                                    new_categories)
     new_codes = take_1d(indexer, codes.copy(), fill_value=-1)

pandas/tests/test_categorical.py

@@ -1673,6 +1673,10 @@ def test_unique(self):
         exp_cat = Categorical(["b", np.nan, "a"], categories=["b", "a"])
         tm.assert_categorical_equal(res, exp_cat)
 
+        # GH 18051 unique()._codes should be writeable
+        cat_nan = Categorical([np.nan])
+        assert cat_nan.unique()._codes.flags.writeable
+
     def test_unique_ordered(self):
         # keep categories order when ordered=True
         cat = Categorical(['b', 'a', 'b'], categories=['a', 'b'], ordered=True)