BUG,TST: Remove case where vectorization fails in pct_change groupby method

Author: simonariddell

pandas/core/groupby/generic.py

@@ -1222,13 +1222,14 @@ def _apply_to_column_groupbys(self, func):
 
     def pct_change(self, periods=1, fill_method='pad', limit=None, freq=None):
         """Calcuate pct_change of each value to previous entry in group"""
-        with _group_selection_context(self) as new:
-            if fill_method:
-                new = copy.copy(new)
-                new.obj = getattr(new, fill_method)(limit=limit)
-                new._reset_cache('_selected_obj')
-            shifted = new.shift(periods=periods, freq=freq)
-            return (new.obj / shifted) - 1
+        if freq:
+            return self.apply(lambda x: x.pct_change(periods=periods,
+                                                     fill_method=fill_method,
+                                                     limit=limit, freq=freq))
+        filled = getattr(self, fill_method)(limit=limit)
+        fill_grp = filled.groupby(self.grouper.labels)
+        shifted = fill_grp.shift(periods=periods, freq=freq)
+        return (filled / shifted) - 1
 
 
 class DataFrameGroupBy(NDFrameGroupBy):

pandas/core/groupby/groupby.py

@@ -1997,13 +1997,11 @@ def pct_change(self, periods=1, fill_method='pad', limit=None, freq=None,
                                                      fill_method=fill_method,
                                                      limit=limit, freq=freq,
                                                      axis=axis))
-        with _group_selection_context(self) as new:
-            if fill_method:
-                new = copy.copy(new)
-                new.obj = getattr(new, fill_method)(limit=limit)
-            obj = new.obj.drop(self.grouper.names, axis=1)
-            shifted = new.shift(periods=periods, freq=freq)
-            return (obj / shifted) - 1
+        filled = getattr(self, fill_method)(limit=limit)
+        filled = filled.drop(self.grouper.names, axis=1)
+        fill_grp = filled.groupby(self.grouper.labels)
+        shifted = fill_grp.shift(periods=periods, freq=freq)
+        return (filled/shifted)-1
 
     @Substitution(name='groupby')
     @Appender(_doc_template)

pandas/tests/groupby/test_transform.py

@@ -765,20 +765,28 @@ def test_pad_stable_sorting(fill_method):
 
 
 @pytest.mark.parametrize("test_series", [True, False])
+@pytest.mark.parametrize("test_freq", [True, False])
 @pytest.mark.parametrize("periods,fill_method,limit", [
-    (1, None, None), (1, None, 1),
     (1, 'ffill', None), (1, 'ffill', 1),
     (1, 'bfill', None), (1, 'bfill', 1),
     (-1, 'ffill', None), (-1, 'ffill', 1),
     (-1, 'bfill', None), (-1, 'bfill', 1),
 ])
-def test_pct_change(test_series, periods, fill_method, limit):
+def test_pct_change(test_series, test_freq, periods, fill_method, limit):
     # GH  21200, 21621
     vals = [3, np.nan, np.nan, np.nan, 1, 2, 4, 10, np.nan, 4]
     keys = ['a', 'b']
     key_v = np.repeat(keys, len(vals))
     df = DataFrame({'key': key_v, 'vals': vals * 2})
 
+    if test_freq:
+        freq = 'D'
+        dt_idx = pd.DatetimeIndex(start='2010-01-01', freq=freq,
+                                  periods=len(vals))
+        df.index = np.concatenate([dt_idx.values]*2)
+    else:
+        freq = None
+
     if fill_method:
         df_g = getattr(df.groupby('key'), fill_method)(limit=limit)
         grp = df_g.groupby('key')
@@ -790,16 +798,17 @@ def test_pct_change(test_series, periods, fill_method, limit):
     if test_series:
         result = df.groupby('key')['vals'].pct_change(periods=periods,
                                                       fill_method=fill_method,
-                                                      limit=limit)
+                                                      limit=limit,
+                                                      freq=freq)
         tm.assert_series_equal(result, expected)
     else:
         result = df.groupby('key').pct_change(periods=periods,
                                               fill_method=fill_method,
-                                              limit=limit)
+                                              limit=limit,
+                                              freq=freq)
         tm.assert_frame_equal(result, expected.to_frame('vals'))
 
 
-
 @pytest.mark.parametrize("func", [np.any, np.all])
 def test_any_all_np_func(func):
     # GH 20653