[ARROW-13081] [C++] Comparison kernels should not allow to compare tz-naive and tz-aware timestamps

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 7.0.0
• Component/s: C++
• Labels: kernel pull-request-available timestamp
• External issue URL: https://github.com/apache/arrow/issues/28787
• Assignee: David Li
• Reporter: Joris Van den Bossche
• Created: 15/Jun/21 09:54
• Updated: 11/Jan/23 08:30
• Resolved: 30/Oct/21 18:12

Description

Currently, when comparing the values of two timestamp arrays, we ignore whether the types have a timezone set or not. For example:

>>> arr1 = pa.array([1], pa.timestamp("s"))
>>> arr2 = pa.array([1], pa.timestamp("s", tz="UTC"))
>>> pc.equal(arr1, arr2)
<pyarrow.lib.BooleanArray object at 0x7ff9ec897f40>
[
  true
]
>>> pc.greater(arr1, arr2)
<pyarrow.lib.BooleanArray object at 0x7f3fbc09bd60>
[
  false
]

While absence of a timezone for the tz-naive/local timestamp, we can't compare both values. So I think we should rather raise an error in this case.
(comparing timestamps with timezones but with a different timezone should be fine, since it's comparing the underlying UTC value which will give the correct result)

—

Note: this will probably depend on the outcome of the discussion on the mailing list about the interpretation of timezone-less timestamps: https://lists.apache.org/thread.html/r8216e5de3efd2935e3907ad9bd20ce07e430952f84de69b36337e5eb%40%3Cdev.arrow.apache.org%3E

Issue Links

• links to: GitHub Pull Request #11541