Error parsing unambiguous dates

[gazpachoking]

It seems in version 2.5.2, there started being problems parsing unambiguous dates that don't follow the specified dayfirst option:

>>> import dateutil
>>> dateutil.__version__
'2.5.2'
>>> dateutil.parser.parse('2010 09 25', dayfirst=True)
Traceback (most recent call last):
  File "<input>", line 1, in <module>
  File "C:\Users\Chase\projects\flexget\py27\lib\site-packages\dateutil\parser.py", line 1164, in parse
    return DEFAULTPARSER.parse(timestr, **kwargs)
  File "C:\Users\Chase\projects\flexget\py27\lib\site-packages\dateutil\parser.py", line 577, in parse
    ret = default.replace(**repl)
ValueError: month must be in 1..12

In previous versions it would produce a valid date.

>>> import dateutil
>>> dateutil.__version__
'2.5.1'
>>> from dateutil import parser
>>> parser.parse('2010 09 25', dayfirst=True)
datetime.datetime(2010, 9, 25, 0, 0)

Was this an intended change?

[pganssle]

This was most certainly not an ambiguous change, and probably can be patched in a bugfix release. The 2.5.2 release doesn't have much in it (mainly it was just released to update the tzdata update), so I suggest using the 2.5.1 release if possible until we can get out a fix.

[pganssle]

Note - if anyone goes to fix this before I get around to it, please work off the 2.5.x branch, not the the master branch.

[rupert]

The fix in #234 only works when the day part of YYYY-MM-DD is <= 12. I suppose you might want to support YYYY-DD-MM (!) - SQL Server uses this format for some languages.

In 2.5.1:

>>> import dateutil
>>> print dateutil.__version__
2.5.1
>>> from dateutil.parser import parse
>>> parse('2015-01-02')
datetime.datetime(2015, 1, 2, 0, 0)
>>> parse('2015-01-02', dayfirst=True)
datetime.datetime(2015, 1, 2, 0, 0)
>>> parse('2015-01-13', dayfirst=True)
datetime.datetime(2015, 1, 13, 0, 0)

In 2.5.3:

>>> import dateutil
>>> print dateutil.__version__
2.5.3
>>> from dateutil.parser import parse
>>> parse('2015-01-02')
datetime.datetime(2015, 1, 2, 0, 0)
>>> parse('2015-01-02', dayfirst=True)
datetime.datetime(2015, 2, 1, 0, 0)
>>> parse('2015-01-13', dayfirst=True)
datetime.datetime(2015, 1, 13, 0, 0)

Delorean bug: myusuf3/delorean#87

[pganssle]

@rupert I'm not sure I understand the issue. 2015-01-13 is unambiguously YYYY-DD-MM. No other interpretation of those values constructs a valid date. The fundamental assumption of the parser is that you gave it a valid date, and it will be as charitable as possible in interpreting that date.

See my comments on #214 if you want dayfirst or yearfirst to be strict.

[rupert]

Yes 2015-01-13 is handled correctly, the issue is with 2015-01-02 being parsed as YYYY-DD-MM instead of YYYY-MM-DD (i.e. the 2.5.1 behaviour is correct).

Day and month swapped in 2.5.3:

>>> parse('2015-01-02', dayfirst=True)
datetime.datetime(2015, 2, 1, 0, 0)

[rupert]

To be honest this is more a problem with delorean defaulting dayfirst to True: https://github.com/myusuf3/delorean/blob/064bc8d0716e06838f11ac38fbf3e56ec63c4e76/delorean/interface.py#L15

[pganssle]

Just set it to False?

That is the intended behavior change for dateutil version 2.5.2 - the combination of dayfirst and yearfirst are used to disambiguate between ymd, ydm, dmy, mdy as the "default" order for ambiguous triplets. See #81 and #238 for more details on how these work. At some point relatively soon, I will update the documentation with the chart in #238.

[rupert]

Yes setting it to False works. Originally I wasn't sure if this was an intended change. Thanks for clarifiying; I'm actually now using dateutil directly.

delorean has some documentation for dayfirst and yearfirst which needs updating. I'll add a comment to the other issue. Thanks.