BUG: added series type to wrap_result for empty DataFrame

Author: discort

doc/source/whatsnew/v0.19.0.txt

@@ -1563,3 +1563,4 @@ Bug Fixes
 - ``PeriodIndex`` can now accept ``list`` and ``array`` which contains ``pd.NaT`` (:issue:`13430`)
 - Bug in ``df.groupby`` where ``.median()`` returns arbitrary values if grouped dataframe contains empty bins (:issue:`13629`)
 - Bug in ``Index.copy()`` where ``name`` parameter was ignored (:issue:`14302`)
+- Bug in ``_downsample()``. Inconsistent return type on resample of empty DataFrame (:issue:`14962`)

pandas/tests/tseries/test_resample.py

@@ -25,8 +25,7 @@
                                      DatetimeIndexResampler)
 from pandas.tseries.tdi import timedelta_range, TimedeltaIndex
 from pandas.util.testing import (assert_series_equal, assert_almost_equal,
-                                 assert_frame_equal, assert_index_equal,
-                                 assertIsInstance)
+                                 assert_frame_equal, assert_index_equal)
 from pandas._period import IncompatibleFrequency
 
 bday = BDay()
@@ -761,7 +760,7 @@ def test_resample_empty_series(self):
                     assert_series_equal(result, expected, check_dtype=False,
                                         check_names=False)
                     # this assert will break when fixed
-                    self.assertTrue(result.name is None)
+                    # self.assertTrue(result.name is None)
                 else:
                     assert_series_equal(result, expected, check_dtype=False)
 
@@ -772,15 +771,15 @@ def test_resample_empty_dataframe(self):
 
         for freq in ['M', 'D', 'H']:
             # count retains dimensions too
-            methods = downsample_methods + ['count']
+            methods = downsample_methods + upsample_methods
             for method in methods:
                 result = getattr(f.resample(freq), method)()
 
-                expected = f.copy()
+                expected = pd.Series([])
                 expected.index = f.index._shallow_copy(freq=freq)
                 assert_index_equal(result.index, expected.index)
                 self.assertEqual(result.index.freq, expected.index.freq)
-                assert_frame_equal(result, expected, check_dtype=False)
+                assert_series_equal(result, expected, check_dtype=False)
 
             # test size for GH13212 (currently stays as df)
 
@@ -839,11 +838,13 @@ def test_resample_loffset_arg_type(self):
 
     def test_resample_empty_dataframe_with_size(self):
         # GH 14962
-        df1 = pd.DataFrame(dict(a=range(100)),
-                           index=pd.date_range('1/1/2000', periods=100, freq="M"))
-        df2 = df1[df1.a < 0]
-        result = df2.resample("Q").size()
-        assertIsInstance(result, pd.Series)
+        index = pd.DatetimeIndex([], freq='M')
+        df = pd.DataFrame([], index=index)
+
+        for freq in ['M', 'D', 'H']:
+            result = df.resample(freq).size()
+            expected = pd.Series([], index=index, dtype='int64')
+            assert_series_equal(result, expected)
 
 
 class TestDatetimeIndex(Base, tm.TestCase):

pandas/tseries/resample.py

@@ -701,12 +701,8 @@ def _downsample(self, how, **kwargs):
         if not len(ax):
             # reset to the new freq
             obj = obj.copy()
-            if how == "size" and isinstance(obj, pd.DataFrame):
-                obj = obj.groupby(
-                    self.grouper, axis=self.axis).aggregate(how, **kwargs)
-
             obj.index.freq = self.freq
-            return obj
+            return self._wrap_result(obj)
 
         # do we have a regular frequency
         if ax.freq is not None or ax.inferred_freq is not None:
@@ -777,6 +773,15 @@ def _wrap_result(self, result):
         # convert if needed
         if self.kind == 'period' and not isinstance(result.index, PeriodIndex):
             result.index = result.index.to_period(self.freq)
+
+        # Make consistent type of result. GH14962
+        if not len(self.ax):
+            grouper = BinGrouper([], result.index)
+            grouped = self._selected_obj.groupby(grouper)
+            result = pd.Series([],
+                               index=result.index,
+                               name=grouped.name,
+                               dtype='int64')
         return result