BUG: Avoid AmbiguousTimeError on groupby

doc/source/whatsnew/v0.19.2.txt

@@ -43,3 +43,4 @@ Bug Fixes
 
 
 - Explicit check in ``to_stata`` and ``StataWriter`` for out-of-range values when writing doubles (:issue:`14618`)
+- Fix AmbiguousTimeError exception when resampling a DatetimeIndex in local TZ, covering a DST change (:issue:`14682`)

pandas/tseries/index.py

@@ -439,7 +439,7 @@ def _generate(cls, start, end, periods, name, offset,
                 tz = tz.localize(date.replace(tzinfo=None)).tzinfo
 
         if tz is not None and inferred_tz is not None:
-            if not inferred_tz == tz:
+            if not tslib.get_timezone(inferred_tz) == tslib.get_timezone(tz):
                 raise AssertionError("Inferred time zone not equal to passed "
                                      "time zone")
 

pandas/tseries/resample.py

@@ -1283,9 +1283,18 @@ def _adjust_dates_anchored(first, last, offset, closed='right', base=0):
     #
     # See https://github.com/pandas-dev/pandas/issues/8683
 
+    # 14682 - Since we need to drop the TZ information to perform
+    # the adjustment in the presence of a DST change,
+    # save TZ Info and the DST state of the first and last parameters
+    # so that we can accurately rebuild them at the end.
     first_tzinfo = first.tzinfo
+    last_tzinfo = last.tzinfo
+    first_dst = bool(first.dst())
+    last_dst = bool(last.dst())
+
     first = first.tz_localize(None)
     last = last.tz_localize(None)
+
     start_day_nanos = first.normalize().value
 
     base_nanos = (base % offset.n) * offset.nanos // offset.n
@@ -1320,11 +1329,8 @@ def _adjust_dates_anchored(first, last, offset, closed='right', base=0):
         else:
             lresult = last.value + offset.nanos
 
-#     return (Timestamp(fresult, tz=first.tz),
-#             Timestamp(lresult, tz=last.tz))
-
-    return (Timestamp(fresult).tz_localize(first_tzinfo),
-            Timestamp(lresult).tz_localize(first_tzinfo))
+    return (Timestamp(fresult).tz_localize(first_tzinfo, ambiguous=first_dst),
+            Timestamp(lresult).tz_localize(last_tzinfo, ambiguous=last_dst))
 
 
 def asfreq(obj, freq, method=None, how=None, normalize=False):

pandas/tseries/tests/test_resample.py

@@ -1912,7 +1912,31 @@ def test_resample_size(self):
         right = Series(val, index=ix)
         assert_series_equal(left, right)
 
-    def test_resmaple_dst_anchor(self):
+    def test_resample_across_dst(self):
+        # The test resamples a DatetimeIndex with values before and after a
+        # DST change
+        # Issue: 14682
+
+        # The DatetimeIndex we will start with
+        # (note that DST happens at 03:00+02:00 -> 02:00+01:00)
+        # 2016-10-30 02:23:00+02:00, 2016-10-30 02:23:00+01:00
+        df1 = DataFrame([1477786980, 1477790580], columns=['ts'])
+        dti1 = DatetimeIndex(pd.to_datetime(df1.ts, unit='s')
+                             .dt.tz_localize('UTC')
+                             .dt.tz_convert('Europe/Madrid'))
+
+        # The expected DatetimeIndex after resampling.
+        # 2016-10-30 02:00:00+02:00, 2016-10-30 02:00:00+01:00
+        df2 = DataFrame([1477785600, 1477789200], columns=['ts'])
+        dti2 = DatetimeIndex(pd.to_datetime(df2.ts, unit='s')
+                             .dt.tz_localize('UTC')
+                             .dt.tz_convert('Europe/Madrid'))
+        df = DataFrame([5, 5], index=dti1)
+        assert_frame_equal(
+            df.resample(rule='H').sum(),
+            DataFrame([5, 5], index=dti2))
+
+    def test_resample_dst_anchor(self):
         # 5172
         dti = DatetimeIndex([datetime(2012, 11, 4, 23)], tz='US/Eastern')
         df = DataFrame([5], index=dti)