Strange Timezone offset

[hasandiwan]

Issue Description

In [28]:divmod((Arrow.utcnow() - Arrow.now()).days, 1)[0]
< Out[28]:-1

$ echo TZ
< America/Los_Angeles

• 🐍 Python version:

3.11

• 🏹 Arrow version:
  In [3]: arrow.version
  Out[3]: '1.3.0'

[satveerbrar]

@hasandiwan The difference you're experiencing is due to a very minor delta between when the utcnow and now objects are created

For example, if you print both timestamps, you'll notice a small difference (around 59 milliseconds) between them.

print(f"{arrow.utcnow()} and {arrow.now('America/Los_Angeles')}")
2025-01-20T17:48:51.138943+00:00 and 2025-01-20T09:48:51.198620-08:00

[hasandiwan]

The difference in days between utcnow and now, given that I'm in PST8PDT, should not be -1.

[ariebovenberg]

The reason you're getting such a strange result is probably due to the quirk of how timedelta stores negative durations:

>>> timedelta(seconds=-1)
datetime.timedelta(days=-1, seconds=86399)

the .days property only gets the days component, to get the actual "total" value you need to call total_seconds()

[hasandiwan]

In [3]: (Arrow.utcnow() - Arrow.now()).total_seconds()
Out[3]: -5.2e-05

Seems unlikely that an 8 hour time difference would be less than a second.

[ariebovenberg]

It's always frustrating when code does something unexpected. Try running each part of the code one-by-one to see where the surprising result comes from 🙂. It looks like your system timezone may be set to UTC.

edit: the result of both the statements is aware, and thus the result is measured in absolute terms. So a few milliseconds seems likely

[hasandiwan]

% env TZ=PST8PDT ~/.virtualenvs/data/bin/ipython
Python 3.12.8 (main, Dec 3 2024, 18:42:41) [Clang 16.0.0 (clang-1600.0.26.4)]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.20.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: from arrow import Arrow

In [2]: (Arrow.utcnow()-Arrow.now()).total_seconds()
Out[2]: -4.7e-05

[ariebovenberg]

Note: if you want the UTC offset of the local timezone at the current moment, just do:

>>> arrow.now().utcoffset().total_seconds()

---

Now on to why your earlier result isn't expected.

>>> a = arrow.utcnow()
<Arrow [2025-01-20T20:45:18.461830+00:00]>
>>> b = arrow.now()
<Arrow [2025-01-20T21:45:20.316133+01:00]>

The subtraction operator determines the difference in absolute terms—i.e. as if both datetimes were normalized to UTC.

If you want to compare datetimes "naively", use the naive attribute:

(a.naive - b.naive).total_seconds()