Fix GroupBy nth Handling with Observed=False

doc/source/whatsnew/v0.25.0.rst

@@ -443,6 +443,7 @@ Groupby/Resample/Rolling
 - Bug in :meth:`pandas.core.groupby.GroupBy.cumsum`, :meth:`pandas.core.groupby.GroupBy.cumprod`, :meth:`pandas.core.groupby.GroupBy.cummin` and :meth:`pandas.core.groupby.GroupBy.cummax` with categorical column having absent categories, would return incorrect result or segfault (:issue:`16771`)
 - Bug in :meth:`pandas.core.groupby.GroupBy.nth` where NA values in the grouping would return incorrect results (:issue:`26011`)
 - Bug in :meth:`pandas.core.groupby.SeriesGroupBy.transform` where transforming an empty group would raise error (:issue:`26208`)
+- Bug in :meth:`pandas.core.groupby.GroupBy.nth` where ``observed=False`` was being ignored for Categorical groupers (:issue:`26385`)
 
 
 Reshaping

pandas/core/groupby/groupby.py

@@ -42,7 +42,7 @@ class providing the base-class of operations.
 from pandas.core.frame import DataFrame
 from pandas.core.generic import NDFrame
 from pandas.core.groupby import base
-from pandas.core.index import Index, MultiIndex
+from pandas.core.index import CategoricalIndex, Index, MultiIndex
 from pandas.core.series import Series
 from pandas.core.sorting import get_group_index_sorter
 
@@ -839,6 +839,7 @@ def _cython_transform(self, how, numeric_only=True, **kwargs):
  def _cython_agg_general(self, how, alt=None, numeric_only=True,
  min_count=-1):
  output = {}
+
  for name, obj in self._iterate_slices():
  is_numeric = is_numeric_dtype(obj.dtype)
  if numeric_only and not is_numeric:
@@ -1707,7 +1708,12 @@ def nth(self,
  if not self.as_index:
  return out
 
- out.index = self.grouper.result_index[ids[mask]]
+ result_index = self.grouper.result_index
+ out.index = result_index[ids[mask]]
+
+ if not self.observed and isinstance(
+ result_index, CategoricalIndex):
+ out = out.reindex(result_index)
 
  return out.sort_index() if self.sort else out
 

pandas/tests/groupby/conftest.py

@@ -76,3 +76,12 @@ def three_group():
  'D': np.random.randn(11),
  'E': np.random.randn(11),
  'F': np.random.randn(11)})
+
+
+AGG_FUNCS = ['sum', 'prod', 'min', 'max', 'mean', 'median', 'var', 'first',
+ 'last', 'nth'] # TODO: ohlc?
+
+
+@pytest.fixture(params=AGG_FUNCS)
+def agg_func(request):
+ return request.param

pandas/tests/groupby/test_categorical.py

@@ -433,6 +433,35 @@ def test_observed_groups_with_nan(observed):
  tm.assert_dict_equal(result, expected)
 
 
+def test_observed_ops(agg_func):
+ cat = pd.Categorical(['a', np.nan, np.nan], categories=['a', 'b', 'c'])
+ ser = pd.Series([1., 2., 3.])
+ df = pd.DataFrame({'cat': cat, 'ser': ser})
+
+ grp = df.groupby('cat', observed=False)['ser']
+ func = getattr(grp, agg_func)
+
+ if agg_func == 'nth': # Need an argument
+ result = func(0)
+ else:
+ result = func()
+
+ if agg_func == 'sum': # TODO: maybe a bug?
+ expected_vals = [1., 0., 0.]
+ elif agg_func == 'prod': # TODO: Definitely seems like a bug
+ expected_vals = [1., 1., 1.]
+ elif agg_func == 'var':
+ expected_vals = [np.nan, np.nan, np.nan]
+ else:
+ expected_vals = [1., np.nan, np.nan]
+
+ index = pd.Categorical(['a', 'b', 'c'], categories=['a', 'b', 'c'])
+ expected = pd.Series(expected_vals, index=index, name='ser')
+ expected.index.name = 'cat'
+
+ tm.assert_series_equal(result, expected)
+
+
 def test_dataframe_categorical_with_nan(observed):
  # GH 21151
  s1 = pd.Categorical([np.nan, 'a', np.nan, 'a'],