Get the Olson TZ name for the local timezone?

问题:

How do I get the Olson timezone name (such as Australia/Sydney) corresponding to the value given by C's localtime call?

This is the value overridden via TZ, by symlinking /etc/localtime, or setting a TIMEZONE variable in time-related system configuration files.

回答1:

I think best bet is to go thru all pytz timezones and check which one matches local timezone, each pytz timezone object contains info about utcoffset and tzname like CDT, EST, same info about local time can be obtained from time.timezone/altzone and time.tzname, and I think that is enough to correctly match local timezone in pytz database e.g.

import time
import pytz
import datetime

local_names = []
if time.daylight:
    local_offset = time.altzone
    localtz = time.tzname[1]
else:
    local_offset = time.timezone
    localtz = time.tzname[0]

local_offset = datetime.timedelta(seconds=-local_offset)

for name in pytz.all_timezones:
    timezone = pytz.timezone(name)
    if not hasattr(timezone, '_tzinfos'):
        continue#skip, if some timezone doesn't have info
    # go thru tzinfo and see if short name like EDT and offset matches
    for (utcoffset, daylight, tzname), _ in timezone._tzinfos.iteritems():
        if utcoffset == local_offset and tzname == localtz:
            local_names.append(name)

print local_names

output:

['America/Atikokan', 'America/Bahia_Banderas', 'America/Bahia_Banderas', 'America/Belize', 'America/Cambridge_Bay', 'America/Cancun', 'America/Chicago', 'America/Chihuahua', 'America/Coral_Harbour', 'America/Costa_Rica', 'America/El_Salvador', 'America/Fort_Wayne', 'America/Guatemala', 'America/Indiana/Indianapolis', 'America/Indiana/Knox', 'America/Indiana/Marengo', 'America/Indiana/Marengo', 'America/Indiana/Petersburg', 'America/Indiana/Tell_City', 'America/Indiana/Vevay', 'America/Indiana/Vincennes', 'America/Indiana/Winamac', 'America/Indianapolis', 'America/Iqaluit', 'America/Kentucky/Louisville', 'America/Kentucky/Louisville', 'America/Kentucky/Monticello', 'America/Knox_IN', 'America/Louisville', 'America/Louisville', 'America/Managua', 'America/Matamoros', 'America/Menominee', 'America/Merida', 'America/Mexico_City', 'America/Monterrey', 'America/North_Dakota/Beulah', 'America/North_Dakota/Center', 'America/North_Dakota/New_Salem', 'America/Ojinaga', 'America/Pangnirtung', 'America/Rainy_River', 'America/Rankin_Inlet', 'America/Resolute', 'America/Resolute', 'America/Tegucigalpa', 'America/Winnipeg', 'CST6CDT', 'Canada/Central', 'Mexico/General', 'US/Central', 'US/East-Indiana', 'US/Indiana-Starke']

In production you can create such a mapping beforehand and save it instead of iterating always.

Testing script after changing timezone:

$ export TZ='Australia/Sydney'
$ python get_tz_names.py
['Antarctica/Macquarie', 'Australia/ACT', 'Australia/Brisbane', 'Australia/Canberra', 'Australia/Currie', 'Australia/Hobart', 'Australia/Lindeman', 'Australia/Melbourne', 'Australia/NSW', 'Australia/Queensland', 'Australia/Sydney', 'Australia/Tasmania', 'Australia/Victoria']

回答2:

This is kind of cheating, I know, but getting from '/etc/localtime' doesn't work for you? Like following:

>>>  import os
>>> '/'.join(os.readlink('/etc/localtime').split('/')[-2:])
'Australia/Sydney'

Hope it helps.

Edit: I liked @A.H.'s idea, in case '/etc/localtime' isn't a symlink. Translating that into Python:

#!/usr/bin/env python

from hashlib import sha224
import os

def get_current_olsonname():
    tzfile = open('/etc/localtime')
    tzfile_digest = sha224(tzfile.read()).hexdigest()
    tzfile.close()

    for root, dirs, filenames in os.walk("/usr/share/zoneinfo/"):
        for filename in filenames:
            fullname = os.path.join(root, filename)
            f = open(fullname)
            digest = sha224(f.read()).hexdigest()
            if digest == tzfile_digest:
                return '/'.join((fullname.split('/'))[-2:])
            f.close()
        return None

if __name__ == '__main__':
    print get_current_olsonname()

回答3:

One problem is that there are multiple "pretty names" , like "Australia/Sydney" , which point to the same time zone (e.g. CST).

So you will need to get all the possible names for the local time zone, and then select the name you like.

e.g.: for Australia, there are 5 time zones, but way more time zone identifiers:

     "Australia/Lord_Howe", "Australia/Hobart", "Australia/Currie", 
     "Australia/Melbourne", "Australia/Sydney", "Australia/Broken_Hill", 
     "Australia/Brisbane", "Australia/Lindeman", "Australia/Adelaide", 
     "Australia/Darwin", "Australia/Perth", "Australia/Eucla"

---

you should check if there is a library which wraps TZinfo , to handle the time zone API.

e.g.: for Python, check the pytz library:

http://pytz.sourceforge.net/

and

http://pypi.python.org/pypi/pytz/

in Python you can do:

from pytz import timezone
import pytz

In [56]: pytz.country_timezones('AU')
Out[56]: 
[u'Australia/Lord_Howe',
 u'Australia/Hobart',
 u'Australia/Currie',
 u'Australia/Melbourne',
 u'Australia/Sydney',
 u'Australia/Broken_Hill',
 u'Australia/Brisbane',
 u'Australia/Lindeman',
 u'Australia/Adelaide',
 u'Australia/Darwin',
 u'Australia/Perth',
 u'Australia/Eucla']

but the API for Python seems to be pretty limited, e.g. it doesn't seem to have a call like Ruby's all_linked_zone_names -- which can find all the synonym names for a given time zone.

回答4:

If evaluating /etc/localtime is OK for you, the following trick might work - after translating it to python:

> md5sum /etc/localtime
abcdefabcdefabcdefabcdefabcdefab /etc/localtime
> find /usr/share/zoneinfo -type f |xargs md5sum | grep abcdefabcdefabcdefabcdefabcdefab
abcdefabcdefabcdefabcdefabcdefab /usr/share/zoneinfo/Europe/London
abcdefabcdefabcdefabcdefabcdefab /usr/share/zoneinfo/posix/Europe/London
...

The duplicates could be filtered using only the official region names "Europe", "America" ... If there are still duplicates, you could take the shortest name :-)

回答5:

Install pytz

import pytz
import time
#import locale
import urllib2

yourOlsonTZ = None
#yourCountryCode = locale.getdefaultlocale()[0].split('_')[1]
yourCountryCode = urllib2.urlopen('http://api.hostip.info/country.php').read()

for olsonTZ in [pytz.timezone(olsonTZ) for olsonTZ in pytz.all_timezones]:
    if (olsonTZ._tzname in time.tzname) and (str(olsonTZ) in pytz.country_timezones[yourCountryCode]):
        yourOlsonTZ = olsonTZ
        break

print yourOlsonTZ

This code will take a best-guess crack at your Olson Timezone based both on your Timezone Name (as according to Python's time module), and your Country Code (as according to Python's locale module the hostip.info project, which references your IP address and geolocates you accordingly).

For example, simply matching the Timzone Names could result in America/Moncton, America/Montreal, or America/New_York for EST (GMT-5). If your country is the US, however, it will limit the answer to America/New_York.

However, if your country is Canada, the script will simply default to the topmost Canadian result (America/Moncton). If there is a way to further refine this, please feel free to leave suggestions in comments.

回答6:

The tzlocal module for Python is aimed at exactly this problem. It produces consistent results under both Linux and Windows, properly converting from Windows time zone ids to Olson using the CLDR mappings.

回答7:

This will get you the time zone name, according to what's in the TZ variable, or localtime file if unset:

#! /usr/bin/env python

import time

time.tzset
print time.tzname

回答8:

Here's another posibility, using PyICU instead; which is working for my purposes:

>>> from PyICU import ICUtzinfo
>>> from datetime import datetime
>>> datetime(2012, 1, 1, 12, 30, 18).replace(tzinfo=ICUtzinfo.getDefault()).isoformat()
'2012-01-01T12:30:18-05:00'
>>> datetime(2012, 6, 1, 12, 30, 18).replace(tzinfo=ICUtzinfo.getDefault()).isoformat()
'2012-06-01T12:30:18-04:00'

Here it is interpreting niave datetimes (as would be returned by a database query) in the local timezone.

回答9:

I prefer following a slightly better than poking around _xxx values

import time, pytz, os

cur_name=time.tzname
cur_TZ=os.environ.get("TZ")

def is_current(name):
   os.environ["TZ"]=name
   time.tzset()
   return time.tzname==cur_name

print "Possible choices:", filter(is_current, pytz.all_timezones)

# optional tz restore
if cur_TZ is None: del os.environ["TZ"]
else: os.environ["TZ"]=cur_TZ
time.tzset()

回答10:

I changed tcurvelo's script to find the right form of time zone (Continent/..../City), in most of cases, but return all of them if fails

#!/usr/bin/env python

from hashlib import sha224
import os
from os import listdir
from os.path import join, isfile, isdir

infoDir = '/usr/share/zoneinfo/'

def get_current_olsonname():
    result = []
    tzfile_digest = sha224(open('/etc/localtime').read()).hexdigest()

    test_match = lambda filepath: sha224(open(filepath).read()).hexdigest() == tzfile_digest

    def walk_over(dirpath):
        for root, dirs, filenames in os.walk(dirpath):
            for fname in filenames:
                fpath = join(root, fname)
                if test_match(fpath):
                    result.append(tuple(root.split('/')[4:]+[fname]))

    for dname in listdir(infoDir):
        if dname in ('posix', 'right', 'SystemV', 'Etc'):
            continue
        dpath = join(infoDir, dname)
        if not isdir(dpath):
            continue
        walk_over(dpath)

    if not result:
        walk_over(join(infoDir))

    return result


if __name__ == '__main__':
    print get_current_olsonname()

回答11:

This JavaScript project attempts to solve the same issue in the browser client-side. It works by playing "twenty questions" with the locale, asking for the UTC offset of certain past times (to test for summer time boundaries, etc.) and using those results to deduce what the local time zone must be. I am not aware of any equivalent Python package unfortunately, so if someone wanted to use this solution it would have to be ported to Python.

While this formula will require updating every time (at worst) the TZ database is updated, a combination of this algorithm and the solution proposed by Anurag Uniyal (keeping only possibilities returned by both methods) sounds to me like the surest way to compute the effective local timezone. As long as there is some difference between the UTC offset of at least one local time in any two time zones, such a system can correctly choose between them.