BUG: Copy categorical codes if empty (fixes #18051) (#18436)

(cherry picked from commit b45325e)

Author: topper-123

doc/source/whatsnew/v0.21.1.txt

@@ -138,6 +138,7 @@ 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

@@ -2268,7 +2268,7 @@ 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/series/test_analytics.py

@@ -848,6 +848,12 @@ def test_value_counts_nunique(self):
         result = series.nunique()
         assert result == 11
 
+        # GH 18051
+        s = pd.Series(pd.Categorical([]))
+        assert s.nunique() == 0
+        s = pd.Series(pd.Categorical([np.nan]))
+        assert s.nunique() == 0
+
     def test_unique(self):
 
         # 714 also, dtype=float
@@ -920,6 +926,14 @@ def test_drop_duplicates(self):
             sc.drop_duplicates(keep=False, inplace=True)
             assert_series_equal(sc, s[~expected])
 
+        # GH 18051
+        s = pd.Series(pd.Categorical([]))
+        tm.assert_categorical_equal(s.unique(), pd.Categorical([]),
+                                    check_dtype=False)
+        s = pd.Series(pd.Categorical([np.nan]))
+        tm.assert_categorical_equal(s.unique(), pd.Categorical([np.nan]),
+                                    check_dtype=False)
+
     def test_clip(self):
         val = self.ts.median()