Add an option to convert timezone.

Do it while parsing, so (for example) day splitting happens at midnight in
chosen timezone instead of midnight in source timezone.
When no timezone is provided, use input's times verbatim.
DST changes cause artefacts in graphs (double or zeroed hour).
Also, display used timezone in generated page.

apachedex/__init__.py

@@ -30,7 +30,7 @@ from __future__ import print_function, division, absolute_import, \
   unicode_literals
 from cgi import escape
 from collections import defaultdict, Counter
-from datetime import datetime, timedelta, date
+from datetime import datetime, timedelta, date, tzinfo
 from functools import partial
 from operator import itemgetter
 from urllib import splittype, splithost, unquote
@@ -52,6 +52,11 @@ import shlex
 import sys
 import time
 import traceback
+try:
+    import pytz
+except ImportError:
+    pytz = None
+
 
 def getResource(name, encoding='utf-8'):
   return pkgutil.get_data(__name__, name).decode(encoding)
@@ -876,19 +881,12 @@ class ShlexArgumentParser(argparse.ArgumentParser):
     return super(ShlexArgumentParser, self).parse_known_args(args=args,
       namespace=namespace)
 
-def _asMonthString(timestamp):
-  dt, _ = timestamp.split(' ')
-  _, month, year = dt.split(':', 1)[0].split('/')
-  return '%s/%02i' % (year, MONTH_VALUE_DICT[month])
-
 _month_offset_cache = {}
 
-def _asWeekString(timestamp):
-  dt, _ = timestamp.split(' ')
-  day, month, year = dt.split(':', 1)[0].split('/')
-  year = int(year)
-  month = MONTH_VALUE_DICT[month]
-  day = int(day)
+def _asWeekString(dt):
+  year = dt.year
+  month = dt.month
+  day = dt.day
   key = (year, month)
   try:
     offset = _month_offset_cache[key]
@@ -902,7 +900,7 @@ def _asWeekString(timestamp):
   if day < 1:
     month -= 1
     day += calendar.monthrange(year, month)[1]
-    assert day > 0 and month > 0, (timestamp, year, month, day)
+    assert day > 0 and month > 0, (dt, year, month, day)
   return '%04i/%02i/%02i' % (year, month, day)
 
 def _weekStringAsQuarterString(timestamp):
@@ -920,18 +918,6 @@ def _getWeekCoefficient(dt):
   # and return value is 7.
   return max(1, 7. / (32 - dt.day))
 
-def _asDayString(timestamp):
-  dt, _ = timestamp.split(' ')
-  day, month, year = dt.split(':', 1)[0].split('/')
-  return '%s/%02i/%s' % (year, MONTH_VALUE_DICT[month], day)
-
-def _as6HourString(timestamp):
-  dt, _ = timestamp.split(' ')
-  dt_date, hour, _ = dt.split(':', 2)
-  day, month, year = dt_date.split('/')
-  return '%s/%02i/%s %02i' % (year, MONTH_VALUE_DICT[month], day,
-    int(hour) // 6 * 6)
-
 def _round6Hour(dt):
   return dt.replace(hour=dt.hour // 6 * 6)
 
@@ -939,29 +925,11 @@ def _hourAsWeekString(timestamp):
   dt = datetime.strptime(timestamp, '%Y/%m/%d %H')
   return (dt - timedelta(dt.weekday())).date().strftime('%Y/%m/%d')
 
-def _asHourString(timestamp):
-  dt, _ = timestamp.split(' ')
-  dt_date, hour, _ = dt.split(':', 2)
-  day, month, year = dt_date.split('/')
-  return '%s/%02i/%s %s' % (year, MONTH_VALUE_DICT[month], day, hour)
-
 def _asHalfDayString(timestamp):
   prefix, _ = timestamp.rsplit(':', 1)
   prefix, hours = prefix.split(' ')
   return '%s %02i' % (prefix, int(hours) // 12 * 12)
 
-def _as30MinutesString(timestamp):
-  dt, _ = timestamp.split(' ')
-  dt_date, hour, minute, _ = dt.split(':', 3)
-  day, month, year = dt_date.split('/')
-  return '%s/%02i/%s %s:%02i' % (year, MONTH_VALUE_DICT[month], day, hour, int(minute) // 30 * 30)
-
-def _asMinuteString(timestamp):
-  dt, _ = timestamp.split(' ')
-  dt_date, hour, minute, _ = dt.split(':', 3)
-  day, month, year = dt_date.split('/')
-  return '%s/%02i/%s %s:%s' % (year, MONTH_VALUE_DICT[month], day, hour, minute)
-
 def _asQuarterHourString(timestamp):
   prefix, minute = timestamp.rsplit(':', 1)
   return '%s:%02i' % (prefix, int(minute) // 15 * 15)
@@ -984,7 +952,7 @@ def _asQuarterHourString(timestamp):
 #   than what it should be.
 period_parser = {
   'year': (
-    _asMonthString,
+    lambda x: x.strftime('%Y/%m'),
     lambda x: x.split('/', 1)[0],
     'month',
     '%Y/%m',
@@ -1007,7 +975,7 @@ period_parser = {
     _getWeekCoefficient,
   ),
   'month': (
-    _asDayString,
+    lambda x: x.strftime('%Y/%m/%d'),
     lambda x: '/'.join(x.split('/', 2)[:2]),
     'day',
     '%Y/%m/%d',
@@ -1018,7 +986,7 @@ period_parser = {
     lambda x: 1,
   ),
   'week': (
-    _as6HourString,
+    lambda x: x.strftime('%Y/%m/%d ') + '%02i' % (x.hour // 6 * 6),
     _hourAsWeekString,
     '6 hours',
     '%Y/%m/%d %H',
@@ -1028,7 +996,7 @@ period_parser = {
     lambda x: 1,
   ),
   'day': (
-    _asHourString,
+    lambda x: x.strftime('%Y/%m/%d %H'),
     lambda x: x.split(' ')[0],
     'hour',
     '%Y/%m/%d %H',
@@ -1039,7 +1007,7 @@ period_parser = {
     lambda x: 1,
   ),
   'halfday': (
-    _as30MinutesString,
+    lambda x: x.strftime('%Y/%m/%d %H:') + '%02i' % (x.minute // 30 * 30),
     _asHalfDayString,
     '30 minutes',
     '%Y/%m/%d %H:%M',
@@ -1048,7 +1016,7 @@ period_parser = {
     lambda x: 1,
   ),
   'quarterhour': (
-    _asMinuteString,
+    lambda x: x.strftime('%Y/%m/%d %H:%M'),
     _asQuarterHourString,
     'minute',
     '%Y/%m/%d %H:%M',
@@ -1121,6 +1089,7 @@ def asHTML(out, encoding, per_site, args, default_site, period_parameter_dict,
   for caption, value in (
         ('apdex threshold', '%.2fs' % args.apdex),
         ('period', args.period or (period + ' (auto)')),
+        ('timezone', args.to_timezone or "(input's)")
       ):
     out.write('<tr><th class="text">%s</th><td>%s</td></tr>' % (
       caption, value))
@@ -1215,6 +1184,44 @@ format_generator = {
   'json': (asJSON, 'ascii'),
 }
 
+ZERO_TIMEDELTA = timedelta(0, 0)
+
+class AutoTZInfo(tzinfo):
+  """
+  Only for fixed UTC offsets ([+-]HHMM)
+  Because datetime.strptime doesn't support %z.
+  """
+  def __init__(self, name):
+    assert len(name) == 5, repr(name)
+    sign = name[0]
+    assert sign in '+-', sign
+    hour = int(name[1:3])
+    assert 0 <= hour <= 12, hour
+    minute = int(name[3:])
+    assert 0 <= minute < 60, minute
+    if sign == '-':
+      hour = -hour
+      minute = -minute
+    self.offset = timedelta(hours=hour, minutes=minute)
+    self.name = name
+
+  def utcoffset(self, dt):
+    return self.offset
+
+  def dst(self, dt):
+    return ZERO_TIMEDELTA
+
+  def tzname(self, dt):
+    return self.name
+
+_tz_cache = {}
+def getTZInfo(tz):
+  try:
+    return _tz_cache[tz]
+  except KeyError:
+    _tz_cache[tz] = tzi = AutoTZInfo(tz)
+    return tzi
+
 def main():
   parser = ShlexArgumentParser(description='Compute Apdex out of '
     'apache-style log files', fromfile_prefix_chars='@')
@@ -1233,6 +1240,12 @@ def main():
       'Does not imply -q.')
   parser.add_argument('--state-file', nargs='+', default=[],
     help='Use given JSON files as initial state. Use - for stdin.')
+  parser.add_argument('--to-timezone', help='Timezone to convert log '
+    'timestamps to before splitting days. If not provided, no conversion '
+    'happens. In addition to "Continent/City" format which know about DST '
+    'but requires pytz module, fixed UTC offsets can be provided in the '
+    '+hhmm form (ex: -0700 for UTC-7). This form does not require pytz '
+    'module.')
 
   group = parser.add_argument_group('generated content (all formats)')
   group.add_argument('-a', '--apdex', default=1.0, type=float,
@@ -1346,6 +1359,24 @@ def main():
   else:
     to_next_period = None
     period = args.period
+  def _matchToDateTime(match):
+    dt, tz = match.group('timestamp').split()
+    day, month, rest = dt.split('/', 2)
+    return datetime.strptime(
+      '%s/%02i/%s' % (day, MONTH_VALUE_DICT[month], rest),
+      '%d/%m/%Y:%H:%M:%S').replace(tzinfo=getTZInfo(tz))
+  if args.to_timezone:
+    to_timezone = args.to_timezone
+    if re.match(r'^[+-]\d{4}$', to_timezone):
+      getTimezoneInfo = getTZInfo
+    else:
+      if pytz is None:
+        raise ValueError('pytz is not available, cannot convert timezone.')
+      getTimezoneInfo = pytz.timezone
+    tz_info = getTimezoneInfo(to_timezone)
+    matchToDateTime = lambda x: _matchToDateTime(x).astimezone(tz_info)
+  else:
+    matchToDateTime = _matchToDateTime
   asDate, decimator, graph_period, date_format, placeholder_delta, \
     round_date, graph_coefficient = period_parser[period]
   site_list, site_caption_dict = args.path
@@ -1459,7 +1490,7 @@ def main():
       if action is None:
         skipped_lines += 1
         continue
-      hit_date = asDate(match.group('timestamp'))
+      hit_date = asDate(matchToDateTime(match))
       if to_next_period is not None:
         if latest_date is None or latest_date < hit_date:
           latest_date = hit_date
@@ -1488,7 +1519,7 @@ def main():
           if show_progress:
             print('done (%s)' % timedelta(seconds=time.time()
               - period_increase_start), file=sys.stderr)
-          hit_date = asDate(match.group('timestamp'))
+          hit_date = asDate(matchToDateTime(match))
       try:
         site_data = per_site[site]
       except KeyError: