BUG: fix dtype for .resample().size()/count() of empty series/dataframe (pandas-dev#28427)

Author: Gabriel Corona

doc/source/whatsnew/v1.0.0.rst

@@ -424,6 +424,7 @@ Groupby/resample/rolling
 
 -
 - Bug in :meth:`DataFrame.groupby` with multiple groups where an ``IndexError`` would be raised if any group contained all NA values (:issue:`20519`)
+- Bug in :meth:`pandas.core.resample.Resampler.size` and :meth:`pandas.core.resample.Resampler.count` returning wrong dtype when used with an empty series or dataframe (:issue:`28427`)
 - Bug in :meth:`DataFrame.rolling` not allowing for rolling over datetimes when ``axis=1`` (:issue: `28192`)
 - Bug in :meth:`DataFrame.rolling` not allowing rolling over multi-index levels (:issue: `15584`).
 - Bug in :meth:`DataFrame.rolling` not allowing rolling on monotonic decreasing time indexes (:issue: `19248`).

pandas/core/resample.py

@@ -869,13 +869,32 @@ def var(self, ddof=1, *args, **kwargs):
 
     @Appender(GroupBy.size.__doc__)
     def size(self):
-        # It's a special case as higher level does return
-        # a copy of 0-len objects. GH14962
         result = self._downsample("size")
-        if not len(self.ax) and isinstance(self._selected_obj, ABCDataFrame):
+        if not len(self.ax):
             from pandas import Series
 
-            result = Series([], index=result.index, dtype="int64")
+            if self._selected_obj.ndim == 1:
+                name = self._selected_obj.name
+            else:
+                name = None
+            result = Series([], index=result.index, dtype="int64", name=name)
+        return result
+
+    @Appender(GroupBy.count.__doc__)
+    def count(self):
+        result = self._downsample("count")
+        if not len(self.ax):
+            if self._selected_obj.ndim == 1:
+                result = self._selected_obj.__class__(
+                    [], index=result.index, dtype="int64", name=self._selected_obj.name
+                )
+            else:
+                from pandas import DataFrame
+
+                result = DataFrame(
+                    [], index=result.index, columns=result.columns, dtype="int64"
+                )
+
         return result
 
     def quantile(self, q=0.5, **kwargs):
@@ -923,14 +942,6 @@ def g(self, _method=method, *args, **kwargs):
     g.__doc__ = getattr(GroupBy, method).__doc__
     setattr(Resampler, method, g)
 
-# groupby & aggregate methods
-for method in ["count"]:
-
-    def h(self, _method=method):
-        return self._downsample(_method)
-
-    h.__doc__ = getattr(GroupBy, method).__doc__
-    setattr(Resampler, method, h)
 
 # series only methods
 for method in ["nunique"]:

pandas/tests/resample/test_base.py

@@ -112,6 +112,19 @@ def test_resample_empty_series(freq, empty_series, resample_method):
     tm.assert_series_equal(result, expected, check_dtype=False)
 
 
+@pytest.mark.parametrize("freq", ["M", "D", "H"])
+@pytest.mark.parametrize("resample_method", ["count", "size"])
+def test_resample_count_empty_series(freq, resample_method):
+    # GH28427
+
+    empty_series = pd.Series([], dtype="datetime64[ns]", index=pd.DatetimeIndex([]))
+    result = getattr(empty_series.resample(freq), resample_method)()
+
+    expected = pd.Series([], dtype="int64", index=pd.DatetimeIndex([], freq=freq))
+
+    tm.assert_series_equal(result, expected)
+
+
 @all_ts
 @pytest.mark.parametrize("freq", ["M", "D", "H"])
 def test_resample_empty_dataframe(empty_frame, freq, resample_method):
@@ -136,6 +149,36 @@ def test_resample_empty_dataframe(empty_frame, freq, resample_method):
     # test size for GH13212 (currently stays as df)
 
 
+@pytest.mark.parametrize("freq", ["M", "D", "H"])
+def test_resample_count_empty_dataframe(freq):
+    # GH28427
+
+    empty_dataframe = pd.DataFrame(
+        {"a": []}, dtype="datetime64[ns]", index=pd.DatetimeIndex([])
+    )
+    result = empty_dataframe.resample(freq).count()
+
+    expected = pd.DataFrame(
+        {"a": []}, dtype="int64", index=pd.DatetimeIndex([], freq=freq)
+    )
+
+    tm.assert_frame_equal(result, expected)
+
+
+@pytest.mark.parametrize("freq", ["M", "D", "H"])
+def test_resample_size_empty_dataframe(freq):
+    # GH28427
+
+    empty_dataframe = pd.DataFrame(
+        {"a": []}, dtype="datetime64[ns]", index=pd.DatetimeIndex([])
+    )
+    result = empty_dataframe.resample(freq).size()
+
+    expected = pd.Series([], dtype="int64", index=pd.DatetimeIndex([], freq=freq))
+
+    tm.assert_series_equal(result, expected)
+
+
 @pytest.mark.parametrize("index", tm.all_timeseries_index_generator(0))
 @pytest.mark.parametrize("dtype", [np.float, np.int, np.object, "datetime64[ns]"])
 def test_resample_empty_dtypes(index, dtype, resample_method):