Incorrect Parsing of dates with underscore delimiter

[danhamill]

If I have the following list of strings:

a = ['Loc_RaffertytoLong_2004_02_21',
 'Loc_RaffertytoLong_2004_02_22',
 'Loc_RaffertytoLong_2004_02_23',
 'Loc_RaffertytoLong_2004_02_24',
 'Loc_RaffertytoLong_2004_02_26',
 'Loc_RaffertytoLong_2004_02_27',
 'Loc_RaffertytoLong_2004_02_28',
 'Loc_RaffertytoLong_2004_02_29']

And I try to parse the date using dateutil:

from dateutil import parse as dparse
for i in a:
    print(dparse.parse(i,fuzzy=True))

I get the printout:

2019-02-21 00:00:00
2019-02-22 00:00:00
2019-02-23 00:00:00
2019-02-24 00:00:00
2019-02-26 00:00:00
2019-02-27 00:00:00
2019-02-28 00:00:00

And the error:

ValueError: ('Unknown string format:', 'Loc_RaffertytoLong_2004_02_29')

From the printout, it is not correctly parsing the year from the string. It seems to be returning dates in the year 2019, which explains why it fails to parse a 29 Feb in 2019.

I can get the correct behavior is i replace all underscores with hashes.

for i in a:
    print(dparse.parse('-'.join(i.split('_')),fuzzy=True))

Is there a reason why dateutil doesn't recognize a underscore delimiter?

System information:

# Name                    Version                   Build  Channel
python-dateutil           2.8.0                    py36_0
Windows 10
python 3.6.8

[pganssle]

I'm not quite sure what's going on here, a minimal reproducing case would be:

from dateutil.parser import parse
from datetime import datetime

dt = parse("2004_09_17")
assert dt == datetime(2004, 9, 17), dt

When you use fuzzy_with_tokens, we get:

>>> parse('2004_02_04', fuzzy_with_tokens=True)
(datetime.datetime(2019, 2, 4, 0, 0), ('_', '_'))

This is one of a class of bugs where part of a datetime string gets eaten in some way - not included in the datetime or in the fuzzy tokens. Last time I saw this I think it was silently eating things it considered to be time zones, but we added a warning for that. @jbrockmendel any ideas?

[pganssle]

By the way @danhamill, if your datetimes are actually in a fixed format like that, I strongly recommend using datetime.strftime instead of dateutil.parser.parse. The dateutil parser is really best only for when parsing strings known to be datetimes but in an unknown format.

[jbrockmendel]

Not obvious whats going on here; I'll take a closer look

[Cheukting]

I have fixed the problem by adding "_" to the list of JUMP. It passes this test and all other tests. I still don't fully understand why it got eat away sometimes. I can tidy up and send a PR soon.

[cyberyu]

I ran into the same issue today:

import dateutil.parser as dparser dt=dparser.parse("2019_01_26",fuzzy=True) print(dt)

-->2022-01-26 00:00:00

Installed version is python_dateutil-2.8.2-py2.py3-none-any.whl