BUG: OHLC-upsampling of PeriodIndex now returns DataFrame (GH 13083)

Author: winklerand

pandas/core/resample.py

@@ -883,6 +883,12 @@ def _downsample(self, how, **kwargs):
             # Downsampling
             return self._groupby_and_aggregate(how, grouper=self.grouper)
         elif is_superperiod(ax.freq, self.freq):
+            if how == 'ohlc':
+                # upsampling to subperiods is handled as an asfreq, which works
+                # for pure aggregating/reducing methods
+                # OHLC reduces along the time dimension, but creates multiple
+                # values for each period -> handle by _groupby_and_aggregate()
+                return self._groupby_and_aggregate(how, grouper=self.grouper)
             return self.asfreq()
         elif ax.freq == self.freq:
             return self.asfreq()
@@ -1278,7 +1284,10 @@ def _get_period_bins(self, ax):
         memb = ax.asfreq(self.freq, how=self.convention)
         i8 = memb.asi8
         freq_mult = self.freq.n
-        rng = np.arange(i8[0], i8[-1] + 1, freq_mult)
+        # when upsampling to subperiods, we need to generate enough bins
+        expected_bins_count = len(binner) * freq_mult
+        i8_extend = expected_bins_count - (i8[-1] - i8[0])
+        rng = np.arange(i8[0], i8[-1] + i8_extend, freq_mult)
         rng += freq_mult
         bins = memb.searchsorted(rng, side='left')
 

pandas/tests/test_resample.py

@@ -2890,6 +2890,29 @@ def test_loffset_returns_datetimeindex(self):
                 assert_frame_equal(result_agg, expected)
                 assert_frame_equal(result_how, expected)
 
+    def test_upsampling_ohlc(self):
+        # GH 13083
+        pi = PeriodIndex(start='2000', freq='D', periods=10)
+        s = Series(range(len(pi)), index=pi)
+        expected = s.to_timestamp().resample('H').ohlc().to_period()
+        # timestamp-based resampling doesn't include all sub-periods
+        # of the last original period, so extend accordingly:
+        pi_ext = PeriodIndex(start='2000', freq='H', periods=24 * len(pi))
+        expected = expected.reindex(pi_ext)
+        result = s.resample('H').ohlc()
+        assert_frame_equal(result, expected)
+
+    def test_upsampling_ohlc_freq_multiples(self):
+        pi = PeriodIndex(start='2000', freq='D', periods=10)
+        s = pd.Series(range(len(pi)), index=pi)
+        expected = s.to_timestamp().resample('12H').ohlc().to_period('12H')
+        # timestamp-based resampling doesn't include all sub-periods
+        # of the last original period, so extend accordingly:
+        pi_ext = PeriodIndex(start='2000', freq='12H', periods=2 * len(pi))
+        expected = expected.reindex(pi_ext)
+        result = s.resample('12H', kind='period').ohlc()
+        assert_frame_equal(result, expected)
+
 
 class TestTimedeltaIndex(Base):
     _index_factory = lambda x: timedelta_range