ensure timestamps are timezone-aware

- use UTC timestamps everywhere
- use common Z format instead of +00:00 for UTC
- naïve timestamps are given local timezone on serialize/deserialize
- use python-dateutil for parsing, local timezone

Author: minrk

jupyter_client/adapter.py

@@ -6,11 +6,8 @@
 import re
 import json
 
-from datetime import datetime
-
 from jupyter_client import protocol_version_info
 
-
 def code_to_line(code, cursor_pos):
     """Turn a multiline code block and cursor position into a single line
     and new cursor position.
@@ -386,9 +383,10 @@ def adapt(msg, to_version=protocol_version_info[0]):
     msg : dict
         A Jupyter message appropriate in the new version.
     """
+    from .session import utcnow
     header = msg['header']
     if 'date' not in header:
-        header['date'] = datetime.now().isoformat()
+        header['date'] = utcnow()
     if 'version' in header:
         from_version = int(header['version'].split('.')[0])
     else:

jupyter_client/jsonutil.py

@@ -6,6 +6,8 @@
 import re
 from datetime import datetime
 
+from dateutil.parser import parse as _dateutil_parse
+from dateutil.tz import tzlocal
 
 from ipython_genutils import py3compat
 from ipython_genutils.py3compat import string_types, iteritems
@@ -17,7 +19,7 @@
 
 # timestamp formats
 ISO8601 = "%Y-%m-%dT%H:%M:%S.%f"
-ISO8601_PAT=re.compile(r"^(\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2})(\.\d{1,6})?Z?([\+\-]\d{2}:?\d{2})?$")
+ISO8601_PAT = re.compile(r"^(\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2})(\.\d{1,6})?(Z|([\+\-]\d{2}:?\d{2}))?$")
 
 # holy crap, strptime is not threadsafe.
 # Calling it once at import seems to help.
@@ -27,6 +29,16 @@
 # Classes and functions
 #-----------------------------------------------------------------------------
 
+def _ensure_tzinfo(dt):
+    """Ensure a datetime object has tzinfo
+    
+    If no tzinfo is present, add tzlocal
+    """
+    if not dt.tzinfo:
+        # No more naïve datetime objects!
+        dt = dt.replace(tzinfo=tzlocal())
+    return dt
+
 def parse_date(s):
     """parse an ISO8601 date string
     
@@ -38,13 +50,8 @@ def parse_date(s):
         return s
     m = ISO8601_PAT.match(s)
     if m:
-        # FIXME: add actual timezone support
-        # this just drops the timezone info
-        notz, ms, tz = m.groups()
-        if not ms:
-            ms = '.0'
-        notz = notz + ms
-        return datetime.strptime(notz, ISO8601)
+        dt = _dateutil_parse(s)
+        return _ensure_tzinfo(dt)
     return s
 
 def extract_dates(obj):
@@ -75,7 +82,8 @@ def squash_dates(obj):
 def date_default(obj):
     """default function for packing datetime objects in JSON."""
     if isinstance(obj, datetime):
-        return obj.isoformat()
+        obj = _ensure_tzinfo(obj)
+        return obj.isoformat().replace('+00:00', 'Z')
     else:
-        raise TypeError("%r is not JSON serializable"%obj)
+        raise TypeError("%r is not JSON serializable" % obj)
 

jupyter_client/session.py

@@ -43,6 +43,13 @@
     # limiting the surface of attack
     def compare_digest(a,b): return a == b
 
+try:
+    from datetime import timezone
+    utc = timezone.utc
+except ImportError:
+    from dateutil.tz import tzutc
+    utc = tzutc()
+
 import zmq
 from zmq.utils import jsonapi
 from zmq.eventloop.ioloop import IOLoop
@@ -158,6 +165,9 @@ def default_secure(cfg):
     # key/keyfile not specified, generate new UUID:
     cfg.Session.key = new_id_bytes()
 
+def utcnow():
+    """Return timezone-aware UTC timestamp"""
+    return datetime.utcnow().replace(tzinfo=utc)
 
 #-----------------------------------------------------------------------------
 # Classes
@@ -223,7 +233,8 @@ def __getitem__(self, k):
 
 
 def msg_header(msg_id, msg_type, username, session):
-    date = datetime.now()
+    """Create a new message header"""
+    date = utcnow()
     version = protocol_version
     return locals()
 
@@ -519,7 +530,7 @@ def _check_packers(self):
             )
 
         # check datetime support
-        msg = dict(t=datetime.now())
+        msg = dict(t=utcnow())
         try:
             unpacked = unpack(pack(msg))
             if isinstance(unpacked['t'], datetime):

jupyter_client/tests/test_jsonutil.py

@@ -5,32 +5,37 @@
 # Distributed under the terms of the Modified BSD License.
 
 import datetime
+from datetime import timedelta
 import json
 
 import nose.tools as nt
 
+from dateutil.tz import tzlocal, gettz
 from jupyter_client import jsonutil
-from ipython_genutils.py3compat import unicode_to_str, str_to_bytes, iteritems
+from jupyter_client.session import utcnow
 
 
 def test_extract_dates():
     timestamps = [
         '2013-07-03T16:34:52.249482',
         '2013-07-03T16:34:52.249482Z',
-        '2013-07-03T16:34:52.249482Z-0800',
-        '2013-07-03T16:34:52.249482Z+0800',
-        '2013-07-03T16:34:52.249482Z+08:00',
-        '2013-07-03T16:34:52.249482Z-08:00',
         '2013-07-03T16:34:52.249482-0800',
         '2013-07-03T16:34:52.249482+0800',
-        '2013-07-03T16:34:52.249482+08:00',
         '2013-07-03T16:34:52.249482-08:00',
+        '2013-07-03T16:34:52.249482+08:00',
     ]
     extracted = jsonutil.extract_dates(timestamps)
     ref = extracted[0]
     for dt in extracted:
         nt.assert_true(isinstance(dt, datetime.datetime))
-        nt.assert_equal(dt, ref)
+        nt.assert_not_equal(dt.tzinfo, None)
+
+    nt.assert_equal(extracted[0].tzinfo.utcoffset(ref), tzlocal().utcoffset(ref))
+    nt.assert_equal(extracted[1].tzinfo.utcoffset(ref), timedelta(0))
+    nt.assert_equal(extracted[2].tzinfo.utcoffset(ref), timedelta(hours=-8))
+    nt.assert_equal(extracted[3].tzinfo.utcoffset(ref), timedelta(hours=8))
+    nt.assert_equal(extracted[4].tzinfo.utcoffset(ref), timedelta(hours=-8))
+    nt.assert_equal(extracted[5].tzinfo.utcoffset(ref), timedelta(hours=8))
 
 def test_parse_ms_precision():
     base = '2013-07-03T16:34:52'
@@ -47,27 +52,15 @@ def test_parse_ms_precision():
             nt.assert_is_instance(parsed, str)
 
 
-ZERO = datetime.timedelta(0)
-
-class tzUTC(datetime.tzinfo):
-    """tzinfo object for UTC (zero offset)"""
-
-    def utcoffset(self, d):
-        return ZERO
-
-    def dst(self, d):
-        return ZERO
-
-UTC = tzUTC()
 
 def test_date_default():
-    now = datetime.datetime.now()
-    utcnow = now.replace(tzinfo=UTC)
-    data = dict(now=now, utcnow=utcnow)
+    naive = datetime.datetime.now()
+    data = dict(naive=naive, utc=utcnow(), withtz=naive.replace(tzinfo=gettz('CEST')))
     jsondata = json.dumps(data, default=jsonutil.date_default)
-    nt.assert_in("+00", jsondata)
-    nt.assert_equal(jsondata.count("+00"), 1)
+    nt.assert_in("Z", jsondata)
+    nt.assert_equal(jsondata.count("Z"), 1)
     extracted = jsonutil.extract_dates(json.loads(jsondata))
     for dt in extracted.values():
         nt.assert_is_instance(dt, datetime.datetime)
+        nt.assert_not_equal(dt.tzinfo, None)
 

jupyter_client/tests/test_session.py

@@ -258,8 +258,8 @@ def test_bad_roundtrip(self):
             session = ss.Session(unpack=lambda b: 5)
 
     def _datetime_test(self, session):
-        content = dict(t=datetime.now())
-        metadata = dict(t=datetime.now())
+        content = dict(t=ss.utcnow())
+        metadata = dict(t=ss.utcnow())
         p = session.msg('msg')
         msg = session.msg('msg', content=content, metadata=metadata, parent=p['header'])
         smsg = session.serialize(msg)

setup.py

@@ -78,6 +78,7 @@
     'traitlets',
     'jupyter_core',
     'pyzmq>=13',
+    'python-dateutil>=2.1',
 ]
 
 extras_require = setuptools_args['extras_require'] = {