Merge pull request #9177 from ledmonster/fix/9049

BUG: Fix not to reindex on non-Categorical groups (GH9049)

doc/source/whatsnew/v0.16.0.txt

@@ -267,3 +267,4 @@ Bug Fixes
 - ``SparseSeries`` and ``SparsePanel`` now accept zero argument constructors (same as their non-sparse counterparts) (:issue:`9272`).
 
 - Bug in ``read_csv`` with buffer overflows with certain malformed input files (:issue:`9205`)
+- Bug in groupby MultiIndex with missing pair (:issue:`9049`, :issue:`9344`)

pandas/core/groupby.py

@@ -1862,7 +1862,6 @@ def __init__(self, index, grouper=None, obj=None, name=None, level=None,
             self.grouper = grouper.values
 
         # pre-computed
-        self._was_factor = False
         self._should_compress = True
 
         # we have a single grouper which may be a myriad of things, some of which are
@@ -1887,8 +1886,6 @@ def __init__(self, index, grouper=None, obj=None, name=None, level=None,
                 level_values = index.levels[level].take(inds)
                 self.grouper = level_values.map(self.grouper)
             else:
-                self._was_factor = True
-
                 # all levels may not be observed
                 labels, uniques = algos.factorize(inds, sort=True)
 
@@ -1913,17 +1910,10 @@ def __init__(self, index, grouper=None, obj=None, name=None, level=None,
 
             # a passed Categorical
             elif isinstance(self.grouper, Categorical):
-
-                factor = self.grouper
-                self._was_factor = True
-
-                # Is there any way to avoid this?
-                self.grouper = np.asarray(factor)
-
-                self._labels = factor.codes
-                self._group_index = factor.categories
+                self._labels = self.grouper.codes
+                self._group_index = self.grouper.categories
                 if self.name is None:
-                    self.name = factor.name
+                    self.name = self.grouper.name
 
             # a passed Grouper like
             elif isinstance(self.grouper, Grouper):
@@ -1936,8 +1926,8 @@ def __init__(self, index, grouper=None, obj=None, name=None, level=None,
                     self.name = grouper.name
 
             # no level passed
-            if not isinstance(self.grouper, (Series, Index, np.ndarray)):
-                if getattr(self.grouper,'ndim', 1) != 1:
+            if not isinstance(self.grouper, (Series, Index, Categorical, np.ndarray)):
+                if getattr(self.grouper, 'ndim', 1) != 1:
                     t = self.name or str(type(self.grouper))
                     raise ValueError("Grouper for '%s' not 1-dimensional" % t)
                 self.grouper = self.index.map(self.grouper)
@@ -1988,21 +1978,15 @@ def group_index(self):
         return self._group_index
 
     def _make_labels(self):
-        if self._was_factor:  # pragma: no cover
-            raise Exception('Should not call this method grouping by level')
-        else:
+        if self._labels is None or self._group_index is None:
             labels, uniques = algos.factorize(self.grouper, sort=self.sort)
             uniques = Index(uniques, name=self.name)
             self._labels = labels
             self._group_index = uniques
 
-    _groups = None
-
-    @property
+    @cache_readonly
     def groups(self):
-        if self._groups is None:
-            self._groups = self.index.groupby(self.grouper)
-        return self._groups
+        return self.index.groupby(self.grouper)
 
 def _get_grouper(obj, key=None, axis=0, level=None, sort=True):
     """
@@ -3238,10 +3222,11 @@ def _reindex_output(self, result):
             return result
         elif len(groupings) == 1:
             return result
-        elif not any([ping._was_factor for ping in groupings]):
+        elif not any([isinstance(ping.grouper, Categorical)
+                      for ping in groupings]):
             return result
 
-        levels_list = [ ping._group_index for ping in groupings ]
+        levels_list = [ ping.group_index for ping in groupings ]
         index = MultiIndex.from_product(levels_list, names=self.grouper.names)
         d = { self.obj._get_axis_name(self.axis) : index, 'copy' : False }
         return result.reindex(**d).sortlevel(axis=self.axis)

pandas/tests/test_groupby.py

@@ -3297,6 +3297,34 @@ def test_groupby_categorical(self):
         expected.index.names = ['myfactor', None]
         assert_frame_equal(desc_result, expected)
 
+    def test_groupby_datetime_categorical(self):
+        # GH9049: ensure backward compatibility
+        levels = pd.date_range('2014-01-01', periods=4)
+        codes = np.random.randint(0, 4, size=100)
+
+        cats = Categorical.from_codes(codes, levels, name='myfactor')
+
+        data = DataFrame(np.random.randn(100, 4))
+
+        result = data.groupby(cats).mean()
+
+        expected = data.groupby(np.asarray(cats)).mean()
+        expected = expected.reindex(levels)
+        expected.index.name = 'myfactor'
+
+        assert_frame_equal(result, expected)
+        self.assertEqual(result.index.name, cats.name)
+
+        grouped = data.groupby(cats)
+        desc_result = grouped.describe()
+
+        idx = cats.codes.argsort()
+        ord_labels = np.asarray(cats).take(idx)
+        ord_data = data.take(idx)
+        expected = ord_data.groupby(ord_labels, sort=False).describe()
+        expected.index.names = ['myfactor', None]
+        assert_frame_equal(desc_result, expected)
+
     def test_groupby_groups_datetimeindex(self):
         # #1430
         from pandas.tseries.api import DatetimeIndex
@@ -3484,6 +3512,31 @@ def test_groupby_categorical_unequal_len(self):
         # len(bins) != len(series) here
         self.assertRaises(ValueError,lambda : series.groupby(bins).mean())
 
+    def test_groupby_multiindex_missing_pair(self):
+        # GH9049
+        df = DataFrame({'group1': ['a','a','a','b'],
+                        'group2': ['c','c','d','c'],
+                        'value': [1,1,1,5]})
+        df = df.set_index(['group1', 'group2'])
+        df_grouped = df.groupby(level=['group1','group2'], sort=True)
+
+        res = df_grouped.agg('sum')
+        idx = MultiIndex.from_tuples([('a','c'), ('a','d'), ('b','c')], names=['group1', 'group2'])
+        exp = DataFrame([[2], [1], [5]], index=idx, columns=['value'])
+
+        tm.assert_frame_equal(res, exp)
+
+    def test_groupby_levels_and_columns(self):
+        # GH9344, GH9049
+        idx_names = ['x', 'y']
+        idx = pd.MultiIndex.from_tuples([(1, 1), (1, 2), (3, 4), (5, 6)], names=idx_names)
+        df = pd.DataFrame(np.arange(12).reshape(-1, 3), index=idx)
+
+        by_levels = df.groupby(level=idx_names).mean()
+        by_columns = df.reset_index().groupby(idx_names).mean()
+
+        tm.assert_frame_equal(by_levels, by_columns)
+
     def test_gb_apply_list_of_unequal_len_arrays(self):
 
         # GH1738

vb_suite/groupby.py

@@ -390,6 +390,18 @@ def f(g):
 
 groupby_sum_booleans = Benchmark("df.groupby('ii').sum()", setup)
 
+
+#----------------------------------------------------------------------
+# multi-indexed group sum #9049
+
+setup = common_setup + """
+N = 50
+df = DataFrame({'A': range(N) * 2, 'B': range(N*2), 'C': 1}).set_index(["A", "B"])
+"""
+
+groupby_sum_multiindex = Benchmark("df.groupby(level=[0, 1]).sum()", setup)
+
+
 #----------------------------------------------------------------------
 # Transform testing