API: Allow ordered=None in CategoricalDtype

Author: jschendel

doc/source/whatsnew/v0.23.0.txt

@@ -200,6 +200,7 @@ Other API Changes
 - Rearranged the order of keyword arguments in :func:`read_excel()` to align with :func:`read_csv()` (:issue:`16672`)
 - :func:`pandas.merge` now raises a ``ValueError`` when trying to merge on incompatible data types (:issue:`9780`)
 - :func:`wide_to_long` previously kept numeric-like suffixes as ``object`` dtype. Now they are cast to numeric if possible (:issue:`17627`)
+- The default value of the ``ordered`` parameter for :class:`~pandas.api.types.CategoricalDtype` has changed from ``False`` to ``None``.  Behavior should remain consistent for downstream objects, such as :class:`Categorical` (:issue:`18790`)
 
 .. _whatsnew_0230.deprecations:
 

pandas/core/categorical.py

@@ -242,7 +242,7 @@ class Categorical(PandasObject):
     # For comparisons, so that numpy uses our implementation if the compare
     # ops, which raise
     __array_priority__ = 1000
-    _dtype = CategoricalDtype()
+    _dtype = CategoricalDtype(ordered=False)
     _deprecations = frozenset(['labels'])
     _typ = 'categorical'
 
@@ -293,7 +293,7 @@ def __init__(self, values, categories=None, ordered=None, dtype=None,
 
         if fastpath:
             self._codes = coerce_indexer_dtype(values, categories)
-            self._dtype = dtype
+            self._dtype = self._dtype._update_dtype(dtype)
             return
 
         # null_mask indicates missing values we want to exclude from inference.
@@ -357,7 +357,7 @@ def __init__(self, values, categories=None, ordered=None, dtype=None,
             full_codes[~null_mask] = codes
             codes = full_codes
 
-        self._dtype = dtype
+        self._dtype = self._dtype._update_dtype(dtype)
         self._codes = coerce_indexer_dtype(codes, dtype.categories)
 
     @property

pandas/core/dtypes/dtypes.py

@@ -159,11 +159,11 @@ class CategoricalDtype(ExtensionDtype):
     _metadata = ['categories', 'ordered']
     _cache = {}
 
-    def __init__(self, categories=None, ordered=False):
+    def __init__(self, categories=None, ordered=None):
         self._finalize(categories, ordered, fastpath=False)
 
     @classmethod
-    def _from_fastpath(cls, categories=None, ordered=False):
+    def _from_fastpath(cls, categories=None, ordered=None):
         self = cls.__new__(cls)
         self._finalize(categories, ordered, fastpath=True)
         return self
@@ -180,9 +180,7 @@ def _from_categorical_dtype(cls, dtype, categories=None, ordered=None):
 
     def _finalize(self, categories, ordered, fastpath=False):
 
-        if ordered is None:
-            ordered = False
-        else:
+        if ordered is not None:
             self._validate_ordered(ordered)
 
         if categories is not None:
@@ -220,10 +218,10 @@ def __eq__(self, other):
             # CDT(., .) = CDT(None, False) and *all*
             # CDT(., .) = CDT(None, True).
             return True
-        elif self.ordered:
-            return other.ordered and self.categories.equals(other.categories)
-        elif other.ordered:
-            return False
+        elif self.ordered or other.ordered:
+            # at least one ordered
+            return ((self.ordered == other.ordered) and
+                    self.categories.equals(other.categories))
         else:
             # both unordered; this could probably be optimized / cached
             return hash(self) == hash(other)
@@ -361,11 +359,16 @@ def _update_dtype(self, dtype):
                    'got {dtype!r}').format(dtype=dtype)
             raise ValueError(msg)
 
-        # dtype is CDT: keep current categories if None (ordered can't be None)
+        # dtype is CDT: keep current categories/ordered if None
         new_categories = dtype.categories
         if new_categories is None:
             new_categories = self.categories
-        return CategoricalDtype(new_categories, dtype.ordered)
+
+        new_ordered = dtype.ordered
+        if new_ordered is None:
+            new_ordered = self.ordered
+
+        return CategoricalDtype(new_categories, new_ordered)
 
     @property
     def categories(self):

pandas/tests/dtypes/test_dtypes.py

@@ -9,7 +9,6 @@
 from pandas import (
     Series, Categorical, CategoricalIndex, IntervalIndex, date_range)
 
-from pandas.compat import string_types
 from pandas.core.dtypes.dtypes import (
     DatetimeTZDtype, PeriodDtype,
     IntervalDtype, CategoricalDtype)
@@ -24,6 +23,11 @@
 import pandas.util.testing as tm
 
 
+@pytest.fixture(scope='class', params=[True, False, None])
+def ordered(request):
+    return request.param
+
+
 class Base(object):
 
     def setup_method(self, method):
@@ -124,41 +128,6 @@ def test_tuple_categories(self):
         result = CategoricalDtype(categories)
         assert all(result.categories == categories)
 
-    @pytest.mark.parametrize('dtype', [
-        CategoricalDtype(list('abc'), False),
-        CategoricalDtype(list('abc'), True)])
-    @pytest.mark.parametrize('new_dtype', [
-        'category',
-        CategoricalDtype(None, False),
-        CategoricalDtype(None, True),
-        CategoricalDtype(list('abc'), False),
-        CategoricalDtype(list('abc'), True),
-        CategoricalDtype(list('cba'), False),
-        CategoricalDtype(list('cba'), True),
-        CategoricalDtype(list('wxyz'), False),
-        CategoricalDtype(list('wxyz'), True)])
-    def test_update_dtype(self, dtype, new_dtype):
-        if isinstance(new_dtype, string_types) and new_dtype == 'category':
-            expected_categories = dtype.categories
-            expected_ordered = dtype.ordered
-        else:
-            expected_categories = new_dtype.categories
-            if expected_categories is None:
-                expected_categories = dtype.categories
-            expected_ordered = new_dtype.ordered
-
-        result = dtype._update_dtype(new_dtype)
-        tm.assert_index_equal(result.categories, expected_categories)
-        assert result.ordered is expected_ordered
-
-    @pytest.mark.parametrize('bad_dtype', [
-        'foo', object, np.int64, PeriodDtype('Q'), IntervalDtype(object)])
-    def test_update_dtype_errors(self, bad_dtype):
-        dtype = CategoricalDtype(list('abc'), False)
-        msg = 'a CategoricalDtype must be passed to perform an update, '
-        with tm.assert_raises_regex(ValueError, msg):
-            dtype._update_dtype(bad_dtype)
-
 
 class TestDatetimeTZDtype(Base):
 
@@ -568,39 +537,37 @@ def test_caching(self):
 
 class TestCategoricalDtypeParametrized(object):
 
-    @pytest.mark.parametrize('categories, ordered', [
-        (['a', 'b', 'c', 'd'], False),
-        (['a', 'b', 'c', 'd'], True),
-        (np.arange(1000), False),
-        (np.arange(1000), True),
-        (['a', 'b', 10, 2, 1.3, True], False),
-        ([True, False], True),
-        ([True, False], False),
-        (pd.date_range('2017', periods=4), True),
-        (pd.date_range('2017', periods=4), False),
-    ])
+    @pytest.mark.parametrize('categories', [
+        list('abcd'),
+        np.arange(1000),
+        ['a', 'b', 10, 2, 1.3, True],
+        [True, False],
+        pd.date_range('2017', periods=4)])
     def test_basic(self, categories, ordered):
         c1 = CategoricalDtype(categories, ordered=ordered)
         tm.assert_index_equal(c1.categories, pd.Index(categories))
         assert c1.ordered is ordered
 
     def test_order_matters(self):
         categories = ['a', 'b']
-        c1 = CategoricalDtype(categories, ordered=False)
-        c2 = CategoricalDtype(categories, ordered=True)
+        c1 = CategoricalDtype(categories, ordered=True)
+        c2 = CategoricalDtype(categories, ordered=False)
+        c3 = CategoricalDtype(categories, ordered=None)
         assert c1 is not c2
+        assert c1 is not c3
 
-    def test_unordered_same(self):
-        c1 = CategoricalDtype(['a', 'b'])
-        c2 = CategoricalDtype(['b', 'a'])
+    @pytest.mark.parametrize('ordered', [False, None])
+    def test_unordered_same(self, ordered):
+        c1 = CategoricalDtype(['a', 'b'], ordered=ordered)
+        c2 = CategoricalDtype(['b', 'a'], ordered=ordered)
         assert hash(c1) == hash(c2)
 
     def test_categories(self):
         result = CategoricalDtype(['a', 'b', 'c'])
         tm.assert_index_equal(result.categories, pd.Index(['a', 'b', 'c']))
-        assert result.ordered is False
+        assert result.ordered is None
 
-    def test_equal_but_different(self):
+    def test_equal_but_different(self, ordered):
         c1 = CategoricalDtype([1, 2, 3])
         c2 = CategoricalDtype([1., 2., 3.])
         assert c1 is not c2
@@ -611,9 +578,11 @@ def test_equal_but_different(self):
         ([1, 2, 3], [3, 2, 1]),
     ])
     def test_order_hashes_different(self, v1, v2):
-        c1 = CategoricalDtype(v1)
+        c1 = CategoricalDtype(v1, ordered=False)
         c2 = CategoricalDtype(v2, ordered=True)
+        c3 = CategoricalDtype(v1, ordered=None)
         assert c1 is not c2
+        assert c1 is not c3
 
     def test_nan_invalid(self):
         with pytest.raises(ValueError):
@@ -628,26 +597,35 @@ def test_same_categories_different_order(self):
         c2 = CategoricalDtype(['b', 'a'], ordered=True)
         assert c1 is not c2
 
-    @pytest.mark.parametrize('ordered, other, expected', [
-        (True, CategoricalDtype(['a', 'b'], True), True),
-        (False, CategoricalDtype(['a', 'b'], False), True),
-        (True, CategoricalDtype(['a', 'b'], False), False),
-        (False, CategoricalDtype(['a', 'b'], True), False),
-        (True, CategoricalDtype([1, 2], False), False),
-        (False, CategoricalDtype([1, 2], True), False),
-        (False, CategoricalDtype(None, True), True),
-        (True, CategoricalDtype(None, True), True),
-        (False, CategoricalDtype(None, False), True),
-        (True, CategoricalDtype(None, False), True),
-        (True, 'category', True),
-        (False, 'category', True),
-        (True, 'not a category', False),
-        (False, 'not a category', False),
-    ])
-    def test_categorical_equality(self, ordered, other, expected):
-        c1 = CategoricalDtype(['a', 'b'], ordered)
+    @pytest.mark.parametrize('ordered1', [True, False, None])
+    @pytest.mark.parametrize('ordered2', [True, False, None])
+    def test_categorical_equality(self, ordered1, ordered2):
+        # same categories
+        c1 = CategoricalDtype(list('abc'), ordered1)
+        c2 = CategoricalDtype(list('abc'), ordered2)
+        result = c1 == c2
+        expected = (ordered1 is ordered2) or not any([ordered1, ordered2])
+        assert result is expected
+
+        # different categories
+        c2 = CategoricalDtype([1, 2, 3], ordered2)
+        assert c1 != c2
+
+        # none categories
+        c1 = CategoricalDtype(list('abc'), ordered1)
+        c2 = CategoricalDtype(None, ordered2)
+        c3 = CategoricalDtype(None, ordered1)
+        assert c1 == c2
+        assert c2 == c1
+        assert c2 == c3
+
+    @pytest.mark.parametrize('categories', [list('abc'), None])
+    @pytest.mark.parametrize('other', ['category', 'not a category'])
+    def test_categorical_equality_strings(self, categories, ordered, other):
+        c1 = CategoricalDtype(categories, ordered)
         result = c1 == other
-        assert result == expected
+        expected = other == 'category'
+        assert result is expected
 
     def test_invalid_raises(self):
         with tm.assert_raises_regex(TypeError, 'ordered'):
@@ -688,16 +666,51 @@ def test_from_categorical_dtype_both(self):
             c1, categories=[1, 2], ordered=False)
         assert result == CategoricalDtype([1, 2], ordered=False)
 
-    def test_str_vs_repr(self):
-        c1 = CategoricalDtype(['a', 'b'])
+    def test_str_vs_repr(self, ordered):
+        c1 = CategoricalDtype(['a', 'b'], ordered=ordered)
         assert str(c1) == 'category'
         # Py2 will have unicode prefixes
-        pat = r"CategoricalDtype\(categories=\[.*\], ordered=False\)"
-        assert re.match(pat, repr(c1))
+        pat = r"CategoricalDtype\(categories=\[.*\], ordered={ordered}\)"
+        assert re.match(pat.format(ordered=ordered), repr(c1))
 
     def test_categorical_categories(self):
         # GH17884
         c1 = CategoricalDtype(Categorical(['a', 'b']))
         tm.assert_index_equal(c1.categories, pd.Index(['a', 'b']))
         c1 = CategoricalDtype(CategoricalIndex(['a', 'b']))
         tm.assert_index_equal(c1.categories, pd.Index(['a', 'b']))
+
+    @pytest.mark.parametrize('new_categories', [
+        list('abc'), list('cba'), list('wxyz'), None])
+    @pytest.mark.parametrize('new_ordered', [True, False, None])
+    def test_update_dtype(self, ordered, new_categories, new_ordered):
+        dtype = CategoricalDtype(list('abc'), ordered)
+        new_dtype = CategoricalDtype(new_categories, new_ordered)
+
+        expected_categories = new_dtype.categories
+        if expected_categories is None:
+            expected_categories = dtype.categories
+
+        expected_ordered = new_dtype.ordered
+        if expected_ordered is None:
+            expected_ordered = dtype.ordered
+
+        result = dtype._update_dtype(new_dtype)
+        tm.assert_index_equal(result.categories, expected_categories)
+        assert result.ordered is expected_ordered
+
+    def test_update_dtype_string(self, ordered):
+        dtype = CategoricalDtype(list('abc'), ordered)
+        expected_categories = dtype.categories
+        expected_ordered = dtype.ordered
+        result = dtype._update_dtype('category')
+        tm.assert_index_equal(result.categories, expected_categories)
+        assert result.ordered is expected_ordered
+
+    @pytest.mark.parametrize('bad_dtype', [
+        'foo', object, np.int64, PeriodDtype('Q'), IntervalDtype(object)])
+    def test_update_dtype_errors(self, bad_dtype):
+        dtype = CategoricalDtype(list('abc'), False)
+        msg = 'a CategoricalDtype must be passed to perform an update, '
+        with tm.assert_raises_regex(ValueError, msg):
+            dtype._update_dtype(bad_dtype)