[jira] [Commented] (ARROW-5248) [Python] support dateutil timezones

["Matt (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARROW-5248?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17252394#comment-17252394 ] 

Matt commented on ARROW-5248:
-----------------------------

I believe another instance of the same bug can be seen rather simply with the following:
{code:java}
In [1]: import pyarrow as pa

In [2]: from dateutil.tz import tzlocal

In [3]: pa.timestamp('ns', tzlocal())
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-3-95e5f76d00ff> in <module>
----> 1 pa.timestamp('ns', tzlocal())

~/Development/jiracli/venv/lib/python3.8/site-packages/pyarrow/types.pxi in pyarrow.lib.timestamp()

~/Development/jiracli/venv/lib/python3.8/site-packages/pyarrow/types.pxi in pyarrow.lib.tzinfo_to_string()

~/Development/jiracli/venv/lib/python3.8/site-packages/pyarrow/error.pxi in pyarrow.lib.pyarrow_internal_check_status()

~/Development/jiracli/venv/lib/python3.8/site-packages/dateutil/tz/tz.py in tzname(self, dt)
    236     @tzname_in_python2
    237     def tzname(self, dt):
--> 238         return self._tznames[self._isdst(dt)]
    239
    240     def is_ambiguous(self, dt):

~/Development/jiracli/venv/lib/python3.8/site-packages/dateutil/tz/tz.py in _isdst(self, dt, fold_naive)
    289
    290         # Check for ambiguous times:
--> 291         dstval = self._naive_is_dst(dt)
    292         fold = getattr(dt, 'fold', None)
    293

~/Development/jiracli/venv/lib/python3.8/site-packages/dateutil/tz/tz.py in _naive_is_dst(self, dt)
    257
    258     def _naive_is_dst(self, dt):
--> 259         timestamp = _datetime_to_timestamp(dt)
    260         return time.localtime(timestamp + time.timezone).tm_isdst
    261

~/Development/jiracli/venv/lib/python3.8/site-packages/dateutil/tz/tz.py in _datetime_to_timestamp(dt)
   1812     seconds since January 1, 1970, ignoring the time zone.
   1813     """
-> 1814     return (dt.replace(tzinfo=None) - EPOCH).total_seconds()
   1815
   1816

AttributeError: 'NoneType' object has no attribute 'replace'{code}

> [Python] support dateutil timezones
> -----------------------------------
>
>                 Key: ARROW-5248
>                 URL: https://issues.apache.org/jira/browse/ARROW-5248
>             Project: Apache Arrow
>          Issue Type: Bug
>          Components: Python
>            Reporter: Joris Van den Bossche
>            Priority: Minor
>              Labels: beginner
>             Fix For: 3.0.0
>
>
> The {{dateutil}} packages also provides a set of timezone objects (https://dateutil.readthedocs.io/en/stable/tz.html) in addition to {{pytz}}. In pyarrow, we only support pytz timezones (and the stdlib datetime.timezone fixed offset):
> {code}
> In [2]: import dateutil.tz                                                                                                                                    
> In [3]: import pyarrow as pa                                                                                                                                  
> In [5]: pa.timestamp('us', dateutil.tz.gettz('Europe/Brussels'))                                                                                              
> ...
> ~/miniconda3/envs/dev37/lib/python3.7/site-packages/pyarrow/types.pxi in pyarrow.lib.tzinfo_to_string()
> ValueError: Unable to convert timezone `tzfile('/usr/share/zoneinfo/Europe/Brussels')` to string
> {code}
> But pandas also supports dateutil timezones. As a consequence, when having a pandas DataFrame that uses a dateutil timezone, you get an error when converting to an arrow table.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)