BUG: date_range issue with millisecond resolution (#24129)

Fixes #24110, by avoid floating-point rounding issues with
millisecond resolution or higher timestamps when creating a date
range.

Author: gabrielreid

doc/source/whatsnew/v0.24.0.rst

@@ -1285,6 +1285,7 @@ Datetimelike
 - Bug in :class:`DatetimeIndex` where calling ``np.array(dtindex, dtype=object)`` would incorrectly return an array of ``long`` objects (:issue:`23524`)
 - Bug in :class:`Index` where passing a timezone-aware :class:`DatetimeIndex` and `dtype=object` would incorrectly raise a ``ValueError`` (:issue:`23524`)
 - Bug in :class:`Index` where calling ``np.array(dtindex, dtype=object)`` on a timezone-naive :class:`DatetimeIndex` would return an array of ``datetime`` objects instead of :class:`Timestamp` objects, potentially losing nanosecond portions of the timestamps (:issue:`23524`)
+- Bug in :func:`date_range` where using dates with millisecond resolution or higher could return incorrect values or the wrong number of values in the index (:issue:`24110`)
 
 Timedelta
 ^^^^^^^^^

pandas/core/arrays/datetimes.py

@@ -307,7 +307,12 @@ def _generate_range(cls, start, end, periods, freq, tz=None,
                     end = end.tz_localize(tz).asm8
         else:
             # Create a linearly spaced date_range in local time
-            arr = np.linspace(start.value, end.value, periods)
+            # Nanosecond-granularity timestamps aren't always correctly
+            # representable with doubles, so we limit the range that we
+            # pass to np.linspace as much as possible
+            arr = np.linspace(
+                0, end.value - start.value,
+                periods, dtype='int64') + start.value
             index = cls._simple_new(
                 arr.astype('M8[ns]', copy=False), freq=None, tz=tz
             )

pandas/tests/indexes/datetimes/test_date_range.py

@@ -769,3 +769,14 @@ def test_all_custom_freq(self, freq):
         msg = 'invalid custom frequency string: {freq}'
         with pytest.raises(ValueError, match=msg.format(freq=bad_freq)):
             bdate_range(START, END, freq=bad_freq)
+
+    @pytest.mark.parametrize('start_end', [
+        ('2018-01-01T00:00:01.000Z', '2018-01-03T00:00:01.000Z'),
+        ('2018-01-01T00:00:00.010Z', '2018-01-03T00:00:00.010Z'),
+        ('2001-01-01T00:00:00.010Z', '2001-01-03T00:00:00.010Z')])
+    def test_range_with_millisecond_resolution(self, start_end):
+        # https://github.com/pandas-dev/pandas/issues/24110
+        start, end = start_end
+        result = pd.date_range(start=start, end=end, periods=2, closed='left')
+        expected = DatetimeIndex([start])
+        tm.assert_index_equal(result, expected)