BUG: fix infer frequency for business daily (#16683)

Author: AdamGleave

Diff for: asv_bench/benchmarks/timeseries.py

@@ -53,7 +53,11 @@ def setup(self):
         self.rng6 = date_range(start='1/1/1', periods=self.N, freq='B')
 
         self.rng7 = date_range(start='1/1/1700', freq='D', periods=100000)
-        self.a = self.rng7[:50000].append(self.rng7[50002:])
+        self.no_freq = self.rng7[:50000].append(self.rng7[50002:])
+        self.d_freq = self.rng7[:50000].append(self.rng7[50000:])
+
+        self.rng8 = date_range(start='1/1/1700', freq='B', periods=100000)
+        self.b_freq = self.rng8[:50000].append(self.rng8[50000:])
 
     def time_add_timedelta(self):
         (self.rng + dt.timedelta(minutes=2))
@@ -94,8 +98,14 @@ def time_infer_dst(self):
     def time_timeseries_is_month_start(self):
         self.rng6.is_month_start
 
-    def time_infer_freq(self):
-        infer_freq(self.a)
+    def time_infer_freq_none(self):
+        infer_freq(self.no_freq)
+
+    def time_infer_freq_daily(self):
+        infer_freq(self.d_freq)
+
+    def time_infer_freq_business(self):
+        infer_freq(self.b_freq)
 
 
 class TimeDatetimeConverter(object):

Diff for: doc/source/whatsnew/v0.20.3.txt

@@ -74,7 +74,6 @@ Groupby/Resample/Rolling
 ^^^^^^^^^^^^^^^^^^^^^^^^
 
 
-
 Sparse
 ^^^^^^
 

Diff for: doc/source/whatsnew/v0.21.0.txt

@@ -169,7 +169,7 @@ Groupby/Resample/Rolling
 ^^^^^^^^^^^^^^^^^^^^^^^^
 - Bug in ``DataFrame.resample().size()`` where an empty ``DataFrame`` did not return a ``Series`` (:issue:`14962`)
 
-
+- Bug in ``infer_freq`` causing indices with 2-day gaps during the working week to be wrongly inferred as business daily (:issue:`16624`)
 
 Sparse
 ^^^^^^

Diff for: pandas/tests/indexes/timedeltas/test_timedelta.py

@@ -564,15 +564,23 @@ def test_freq_conversion(self):
 
 
 class TestSlicing(object):
+    @pytest.mark.parametrize('freq', ['B', 'D'])
+    def test_timedelta(self, freq):
+        index = date_range('1/1/2000', periods=50, freq=freq)
 
-    def test_timedelta(self):
-        # this is valid too
-        index = date_range('1/1/2000', periods=50, freq='B')
         shifted = index + timedelta(1)
         back = shifted + timedelta(-1)
-        assert tm.equalContents(index, back)
-        assert shifted.freq == index.freq
-        assert shifted.freq == back.freq
+        tm.assert_index_equal(index, back)
+
+        if freq == 'D':
+            expected = pd.tseries.offsets.Day(1)
+            assert index.freq == expected
+            assert shifted.freq == expected
+            assert back.freq == expected
+        else:  # freq == 'B'
+            assert index.freq == pd.tseries.offsets.BusinessDay(1)
+            assert shifted.freq is None
+            assert back.freq == pd.tseries.offsets.BusinessDay(1)
 
         result = index - timedelta(1)
         expected = index + timedelta(-1)

Diff for: pandas/tests/tseries/test_frequencies.py

@@ -504,9 +504,14 @@ def test_raise_if_too_few(self):
         pytest.raises(ValueError, frequencies.infer_freq, index)
 
     def test_business_daily(self):
-        index = _dti(['12/31/1998', '1/3/1999', '1/4/1999'])
+        index = _dti(['01/01/1999', '1/4/1999', '1/5/1999'])
         assert frequencies.infer_freq(index) == 'B'
 
+    def test_business_daily_look_alike(self):
+        # GH 16624, do not infer 'B' when 'weekend' (2-day gap) in wrong place
+        index = _dti(['12/31/1998', '1/3/1999', '1/4/1999'])
+        assert frequencies.infer_freq(index) is None
+
     def test_day(self):
         self._check_tick(timedelta(1), 'D')
 

Diff for: pandas/tseries/frequencies.py

@@ -975,8 +975,7 @@ def _infer_daily_rule(self):
             else:
                 return _maybe_add_count('D', days)
 
-        # Business daily. Maybe
-        if self.day_deltas == [1, 3]:
+        if self._is_business_daily():
             return 'B'
 
         wom_rule = self._get_wom_rule()
@@ -1012,6 +1011,19 @@ def _get_monthly_rule(self):
         return {'cs': 'MS', 'bs': 'BMS',
                 'ce': 'M', 'be': 'BM'}.get(pos_check)
 
+    def _is_business_daily(self):
+        # quick check: cannot be business daily
+        if self.day_deltas != [1, 3]:
+            return False
+
+        # probably business daily, but need to confirm
+        first_weekday = self.index[0].weekday()
+        shifts = np.diff(self.index.asi8)
+        shifts = np.floor_divide(shifts, _ONE_DAY)
+        weekdays = np.mod(first_weekday + np.cumsum(shifts), 7)
+        return np.all(((weekdays == 0) & (shifts == 3)) |
+                      ((weekdays > 0) & (weekdays <= 4) & (shifts == 1)))
+
     def _get_wom_rule(self):
         #         wdiffs = unique(np.diff(self.index.week))
         # We also need -47, -49, -48 to catch index spanning year boundary