Fix date_range(start, end, freq) when end-start is divisible by freq (

#16516)

xref #16507

`date_range` generates its dates via `range`, and the end of this range was calculated via `math.ceil((end - start) / freq)`. If `(end - start) / freq` did not produce a remainder, `math.ceil` would not correctly increment this value by `1` to capture the last date.

Instead, this PR uses `math.floor((end - start) / freq) + 1` to always ensure the last date is captured

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - Bradley Dice (https://github.com/bdice)

URL: #16516

python/cudf/cudf/core/index.py

@@ -2414,11 +2414,13 @@ def day_name(self, locale: str | None = None) -> Index:
         >>> datetime_index = cudf.date_range("2016-12-31", "2017-01-08", freq="D")
         >>> datetime_index
         DatetimeIndex(['2016-12-31', '2017-01-01', '2017-01-02', '2017-01-03',
-                       '2017-01-04', '2017-01-05', '2017-01-06', '2017-01-07'],
+                       '2017-01-04', '2017-01-05', '2017-01-06', '2017-01-07',
+                       '2017-01-08'],
                       dtype='datetime64[ns]', freq='D')
         >>> datetime_index.day_name()
         Index(['Saturday', 'Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday',
-               'Friday', 'Saturday'], dtype='object')
+               'Friday', 'Saturday', 'Sunday'],
+              dtype='object')
         """
         day_names = self._column.get_day_names(locale)
         return Index._from_data({self.name: day_names})

python/cudf/cudf/core/series.py

@@ -801,14 +801,17 @@ def dt(self):
         >>> s.dt.hour
         0    12
         1    13
+        2    14
         dtype: int16
         >>> s.dt.second
         0    0
         1    0
+        2    0
         dtype: int16
         >>> s.dt.day
         0    3
         1    3
+        2    3
         dtype: int16
 
         Returns

python/cudf/cudf/core/tools/datetimes.py

@@ -951,17 +951,18 @@ def date_range(
         end = cudf.Scalar(end, dtype=dtype)
         _is_increment_sequence = end >= start
 
-        periods = math.ceil(
+        periods = math.floor(
             int(end - start) / _offset_to_nanoseconds_lower_bound(offset)
         )
 
         if periods < 0:
             # Mismatched sign between (end-start) and offset, return empty
             # column
             periods = 0
-        elif periods == 0:
-            # end == start, return exactly 1 timestamp (start)
-            periods = 1
+        else:
+            # If end == start, periods == 0 and we return exactly 1 timestamp (start).
+            # Otherwise, since closed="both", we ensure the end point is included.
+            periods += 1
 
     # We compute `end_estim` (the estimated upper bound of the date
     # range) below, but don't always use it.  We do this to ensure

python/cudf/cudf/tests/test_datetime.py

@@ -2536,3 +2536,9 @@ def test_dti_methods(method, kwargs):
     result = getattr(cudf_dti, method)(**kwargs)
     expected = getattr(pd_dti, method)(**kwargs)
     assert_eq(result, expected)
+
+
+def test_date_range_start_end_divisible_by_freq():
+    result = cudf.date_range("2011-01-01", "2011-01-02", freq="h")
+    expected = pd.date_range("2011-01-01", "2011-01-02", freq="h")
+    assert_eq(result, expected)