Convert an RFC 3339 time to a standard Python timestamp

Question

Is there an easy way to convert an RFC 3339 time into a regular Python timestamp?

I've got a script which is reading an ATOM feed and I'd like to be able to compare the timestamp of an item in the ATOM feed to the modification time of a file.

I notice from the ATOM spec, that ATOM dates include a time zone offset (Z<a number>) but, in my case, there's nothing after the Z so I guess we can assume GMT.

I suppose I could parse the time with a regex of some sort but I was hoping Python had a built-in way of doing it that I just haven't been able to find.

Solution 1

No builtin, afaik.

feed.date.rfc3339 This is a Python library module with functions for converting timestamp strings in RFC 3339 format to Python time float values, and vice versa. RFC 3339 is the timestamp format used by the Atom feed syndication format.

It is BSD-licensed.

http://home.blarg.net/~steveha/pyfeed.html

(Edited so it's clear I didn't write it. :-)

Solution 2

You don't include an example, but if you don't have a Z-offset or timezone, and assuming you don't want durations but just the basic time, then maybe this will suit you:

import datetime as dt
>>> dt.datetime.strptime('1985-04-12T23:20:50.52', '%Y-%m-%dT%H:%M:%S.%f')
datetime.datetime(1985, 4, 12, 23, 20, 50, 520000)

The strptime() function was added to the datetime module in Python 2.5 so some people don't yet know it's there.

Edit: The time.strptime() function has existed for a while though, and works about the same to give you a struct_time value:

>>> ts = time.strptime('1985-04-12T23:20:50.52', '%Y-%m-%dT%H:%M:%S.%f')
>>> ts
time.struct_time(tm_year=1985, tm_mon=4, tm_mday=12, tm_hour=23, tm_min=20, tm_sec=50, tm_wday=4, tm_yday=102, tm_isdst=-1)
>>> time.mktime(ts)
482210450.0

Solution 3

I struggled with RFC3339 datetime format a lot, but I found a suitable solution to convert date_string <=> datetime_object in both directions.

You need two different external modules, because one of them is is only able to do the conversion in one direction (unfortunately):

first install:

sudo pip install rfc3339
sudo pip install iso8601

then include:

import datetime     # for general datetime object handling
import rfc3339      # for date object -> date string
import iso8601      # for date string -> date object

For not needing to remember which module is for which direction, I wrote two simple helper functions:

def get_date_object(date_string):
  return iso8601.parse_date(date_string)

def get_date_string(date_object):
  return rfc3339.rfc3339(date_object)

which inside your code you can easily use like this:

input_string = '1989-01-01T00:18:07-05:00'
test_date = get_date_object(input_string)
# >>> datetime.datetime(1989, 1, 1, 0, 18, 7, tzinfo=<FixedOffset '-05:00' datetime.timedelta(-1, 68400)>)

test_string = get_date_string(test_date)
# >>> '1989-01-01T00:18:07-05:00'

test_string is input_string # >>> True

Heureka! Now you can easily (haha) use your date strings and date strings in a useable format.

Solution 4

If you're using Django, you could use Django's function parse_datetime:

>>> from django.utils.dateparse import parse_datetime
>>> parse_datetime("2016-07-19T07:30:36+05:00")
datetime.datetime(2016, 7, 19, 7, 30, 36, tzinfo=<django.utils.timezone.FixedOffset object at 0x101c0c1d0>)

Solution 5

http://pypi.python.org/pypi/iso8601/ seems to be able to parse iso 8601, which RFC 3339 is a subset of, maybe this could be useful, but again, not built-in.

View more solutions

Comments

[steveha]

If you read the RFC 3339 spec, you will see that the only valid time zone offset values are: (0) a bare "Z", meaning the time stamp is UTC; or (1) an offset of the form "[+-]\d\d:\d\d" such as "+02:00" or "-08:00". Note that an offset of "+00:00" would mean the same thing as "Z". For more details, read the RFC 3339 spec: tools.ietf.org/html/rfc3339

[jfs]

Python bug: issue15873: datetime: add ability to parse RFC 3339 dates and times

[jfs]

related: Convert timestamps with offset to datetime obj using strptime

[steveha]

Also, I wrote the PyFeed (and Xe) libraries, and I hang out here on StackOverflow, so if you have any questions about it, I would be happy to answer them.

[Mark Biek]

PyFeed does exactly what I need, courtesy of the tf_from_timestamp() function in feed.date.rfc3339

[steveha]

Note that PyFeed can be used to parse an Atom feed. It uses xml.dom.minidom to do the actual parsing, and then unpacks the XML tree structure into nice convenient classes. Hmm, I ought to put Xe and PyFeed up on PyPI.

[Mark Biek]

@steveha Excellent, thanks for the offer. The libraries seem pretty easy to use so far but I'll remember you're here if I run into anything weird.

[steveha]

@Alex Brasetvik, it was already clear that you were not claiming to have written it. If you were claiming credit you wouldn't have included the direct link to my web page! P.S. I was happy to see you recommending my library; thank you.

[Ram Rachum]

This seems much nicer than the "time floats" offered by PyFeed.

[steveha]

cool-NR, if you have a time float value, you can call time.gmtime() and get the struct_time value. And it is a lot easier to do relative times with a time float value; two days from now is simply tf + 2 * seconds_per_day (where seconds_per_day is 24 * 60 * 60). Python's struct_time is great for inspecting (what day of the week is it?) but terribly inconvenient for computing.

[Tobu]

Time floats and struct_time aren't timezone aware. Since RFC 3339 requires UTC-compatible time zones, which in Python means non-naive datetime objects, this is the only sane option so far.

[Yarin]

link's broken- should it be code.google.com/p/feedparser/source/browse/trunk/feedparser/‌​…?

[Yarin]

This won't work- Methods that are not time-zone aware are not RFC 3339 compatible.

[Peter Hansen]

Yarin, clearly, but your complaint should be with the original question's use of "RFC 3339" then, as my answer did address his actual question, where he notes he doesn't have a time zone...

[Yarin]

Peter- The way I read his question he's trying to compare an ATOM feed RFC 3999 date with timezone to another date with assumed GMT, but maybe i don't get it

[Romuald Brunet]

The only problem being that it also parses non-date values like "now"

[jfs]

@Yarin: "there's nothing after the Z" -- it means the input is UTC and strptime() could be used. Though mktime() that expects time in the local timezone should not be used. calendar.timegm() could be used instead (to compare with the result of os.path.getmtime(): "I'd like to be able to compare the timestamp of an item in the ATOM feed to the modification time of a file.")

[jfs]

it doesn't support numeric utc offset: +HHMM. The second example is not rfc 3339; it is rfc 5322

[Lennart Rolland]

It does not keep milliseconds?

[jfs]

@LennartRolland: yes. time.struct_time doesn't store fractions of a second.

[Ross Smith II]

Python 3.10.4 reported Invalid isoformat string: '2022-05-08T02:47:35.839Z'