Why does pytz correctly adjust time & offset when crossing TZ & DST boundaries but not the TZ name?

Question

I have reviewed several pytz-related questions here, but none seems to address the exact problem I'm seeing.

Following the pytz documentation, here's a loop to print the current time in multiple time zones, including time zone offset, time zone name, and whether the datetime object thinks it's DST.

nowDT = datetime.datetime.now()
chicagoTz = pytz.timezone('America/Chicago')
chicagoDT = chicagoTz.normalize(chicagoTz.localize(nowDT))
sys.stdout.write( "%-10s %-35s %s\n" % ('Chicago',
                                        chicagoDT.strftime("%Y/%m/%d %H:%M:%S %Z %z"),
                                        chicagoDT.dst()) )

tzTups = [('New York', 'America/New_York'),
          ('London',   'Europe/London'),
          ('Sydney',   'Australia/Sydney')]

for tzTup in tzTups:
    tz = pytz.timezone(tzTup[1])
    locDT = tz.normalize(chicagoDT.astimezone(tz))
    sys.stdout.write( "%-10s %-35s %s\n" % (tzTup[0],
                                            locDT.strftime("%Y/%m/%d %H:%M:%S %Z %z"),
                                            locDT.dst()) )

Here's the output:

Chicago    2014/03/12 14:34:53 CDT -0500       1:00:00
New York   2014/03/12 15:34:53 EDT -0400       1:00:00
London     2014/03/12 19:34:53 GMT +0000       0:00:00
Sydney     2014/03/13 06:34:53 EST +1100       1:00:00

Checking with, say, timeanddate.com, we see that all of this information is correct, including the Sydney time, offset, and the 1:00:00 indicating that the datetime object believes that DST is currently in effect in Sydney.

The only problem is that the Sydney time is labeled EST instead of EDT. In fact, I can't get Python to ever declare Sydney in EDT even though it knows about the DST offset:

tz = pytz.timezone('Australia/Sydney')
for i in range(1,13):
    locDT = tz.normalize(tz.localize(datetime.datetime(2013, i, 15)))
    sys.stdout.write("%02d %s %s\n" % (i, locDT.dst(), locDT.tzname()))

Output:

01 1:00:00 EST
02 1:00:00 EST
03 1:00:00 EST
04 0:00:00 EST
05 0:00:00 EST
06 0:00:00 EST
07 0:00:00 EST
08 0:00:00 EST
09 0:00:00 EST
10 1:00:00 EST
11 1:00:00 EST
12 1:00:00 EST

Am I doing something wrong? Is /usr/share/zoneinfo out of date on my system? Is this a known issue corrected in recent versions of pytz or the Olson DB that I might not have? (Mine says it's using OLSON_VERSION = '2010b'.)

use `tz.localize(naive_dt, is_dst=None)` unless you want to default to `is_dst=False` for ambiguous or non-existing local times. — jfs, Mar 17 '14 at 12:59
Thanks. That's what the surrounding `normalize` is meant for -- to set the DST appropriately for the date/time in question. — dg99, Mar 17 '14 at 21:53
if the time is ambiguous; it is ambiguous: `tz.normalize()` can't fix it (if it could then `.localize()` would do the right thing by itself). — jfs, Mar 18 '14 at 04:43
@J.F.Sebastian Oh, I see what you're saying. Thanks for the tip. — dg99, Mar 21 '14 at 17:06

unutbu · Accepted Answer · 2014-03-13T01:11:58.823

The IANA maintains the Olson database. The question of what timezone abbreviation(s) should be used for Australia was discussed in the IANA's tz mailing list here (the discussion spanned two months: March 2013, April 2013).

There seems to be strong opinion on all sides as to what the abbreviations should be and those strong opinions have resulted in gridlock.

Some say the abbreviations are a relic of the past and should not be used and the ambiguity should not be fixed to help discourage its use.

Apparently there is no recognized authority in Australia defining the abbreviations. Some say conflicting organizations use different timezone abbreviations, and so as not to pick political sides, the IANA chose EST for both standard and daylight savings times.

For now, the Olson DB uses EST for all timezones for all dates in Australia/Sydney:

In [60]: import pytz
In [61]: sydney = pytz.timezone('Australia/Sydney')

In [68]: [(date, tzabbrev) for date, (utcoffset, dstoffset, tzabbrev) in zip(sydney._utc_transition_times, sydney._transition_info)]
Out[68]: 
[(datetime.datetime(1, 1, 1, 0, 0), 'EST'),
 (datetime.datetime(1916, 12, 31, 14, 1), 'EST'),
 (datetime.datetime(1917, 3, 24, 15, 0), 'EST'),
 (datetime.datetime(1941, 12, 31, 16, 0), 'EST'),
 (datetime.datetime(1942, 3, 28, 15, 0), 'EST'),
 (datetime.datetime(1942, 9, 26, 16, 0), 'EST'),
 (datetime.datetime(1943, 3, 27, 15, 0), 'EST'),
 ...]

In [69]: set([tzabbrev for utcoffset, dstoffset, tzabbrev in sydney._transition_info])
Out[69]: {'EST'}

This shows that in the Australia/Sydney timezone, EST is used across every transition boundary.

Good luck to a bunch of database maintainers, trying to reform common public practice. Assuming that the practice in Australia is as widespread as it is in the US. — Mark Ransom, Mar 12 '14 at 21:25
+1. And for clarity (if you can call it that), in the summer - Australia's Eastern time zone uses "Eastern Summer Time" instead of "Eastern Standard Time", hence the two "EST" abbreviations. The continual debate is whether to call it EST, EDT, AEST, or AEDT. — Matt Johnson-Pint, Mar 13 '14 at 15:57

Why does pytz correctly adjust time & offset when crossing TZ & DST boundaries but not the TZ name?

1 Answers1

Linked