Nishuuzz opened a new pull request, #51016:
URL: https://github.com/apache/arrow/pull/51016
`pyarrow.types.is_temporal()` returns True for interval types, but its
docstring doesn't say so:
```python
>>> import pyarrow as pa, pyarrow.types as t
>>> t.is_temporal(pa.month_day_nano_interval())
True
>>> t.is_temporal.__doc__.strip().splitlines()[0]
'Return True if value is an instance of type: date, time, timestamp or
duration.'
```
That's the documented contract being narrower than the behaviour, and since
these docstrings are generated from the `@doc(is_null, datatype=...)` decorator
they're what ends up on the API docs page.
Including interval is deliberate rather than accidental — `_TEMPORAL_TYPES`
is built as
```python
_TEMPORAL_TYPES = ({lib.Type_TIMESTAMP,
lib.Type_DURATION} | _TIME_TYPES | _DATE_TYPES |
_INTERVAL_TYPES)
```
and `test_is_temporal_date_time_timestamp` already asserts
`types.is_temporal(pa.month_day_nano_interval())`. So this is a docs fix, not a
behaviour question; I've only changed the `datatype` string.
After the change the rendered line is
```
Return True if value is an instance of type: date, time, timestamp, duration
or interval.
```
which is exactly the set that returns True — I checked each one (date32,
time32, timestamp, duration, interval all True; int32 False) by applying the
same edit to an installed pyarrow and reading the generated docstring back.
I found this while comparing every predicate set in `types.py` against the
corresponding `arrow::is_*()` in `cpp/src/arrow/type_traits.h`, the same way
GH-50847 turned up. Worth noting the two `is_temporal` definitions genuinely
differ — the C++ one covers only date/time/timestamp, while Python also counts
duration and interval — but that's a semantic question rather than something to
change quietly in a MINOR PR, so I've left it alone and only made the Python
docstring match the Python behaviour.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
