ENH: Parse %z directive in format for to_datetime

return timedeltas as list return timedeltas in a numpy array some flake fixes

Author: mroeschke

pandas/_libs/tslibs/strptime.pyx

@@ -29,7 +29,8 @@ cimport cython
 import numpy as np
 from numpy cimport ndarray, int64_t
 
-from datetime import date as datetime_date
+from datetime import (date as datetime_date, timedelta as datetime_timedelta,
+ timezone as datetime_timezone)
 from cpython.datetime cimport datetime
 
 from np_datetime cimport (check_dts_bounds,
@@ -58,6 +59,7 @@ def array_strptime(ndarray[object] values, object fmt,
  Py_ssize_t i, n = len(values)
  pandas_datetimestruct dts
  ndarray[int64_t] iresult
+ ndarray[object] results_tz
  int year, month, day, minute, hour, second, weekday, julian, tz
  int week_of_year, week_of_year_start
  int64_t us, ns
@@ -109,6 +111,8 @@ def array_strptime(ndarray[object] values, object fmt,
  result = np.empty(n, dtype='M8[ns]')
  iresult = result.view('i8')
 
+ results_tz = np.empty(n, dtype='object')
+
  dts.us = dts.ps = dts.as = 0
 
  cdef dict _parse_code_table = {
@@ -130,7 +134,8 @@ def array_strptime(ndarray[object] values, object fmt,
  'U': 15,
  'W': 16,
  'Z': 17,
- 'p': 18 # just an additional key, works only with I
+ 'p': 18, # just an additional key, works only with I
+ 'z': 19,
  }
  cdef int parse_code
 
@@ -177,6 +182,8 @@ def array_strptime(ndarray[object] values, object fmt,
  month = day = 1
  hour = minute = second = ns = us = 0
  tz = -1
+ gmtoff = None
+ gmtoff_fraction = 0
  # Default to -1 to signify that values not known; not critical to have,
  # though
  week_of_year = -1
@@ -281,6 +288,32 @@ def array_strptime(ndarray[object] values, object fmt,
  else:
  tz = value
  break
+ elif parse_code == 19:
+ z = found_dict['z']
+ if z == 'Z':
+ gmtoff = 0
+ else:
+ if z[3] == ':':
+ z = z[:3] + z[4:]
+ if len(z) > 5:
+ if z[5] != ':':
+ msg = "Unconsistent use of : in {0}"
+ raise ValueError(msg.format(found_dict['z']))
+ z = z[:5] + z[6:]
+ hours = int(z[1:3])
+ minutes = int(z[3:5])
+ seconds = int(z[5:7] or 0)
+ gmtoff = (hours * 60 * 60) + (minutes * 60) + seconds
+ gmtoff_remainder = z[8:]
+ # Pad to always return microseconds.
+ pad_number = 6 - len(gmtoff_remainder)
+ gmtoff_remainder_padding = "0" * pad_number
+ gmtoff_fraction = int(gmtoff_remainder +
+ gmtoff_remainder_padding)
+ if z.startswith("-"):
+ gmtoff = -gmtoff
+ gmtoff_fraction = -gmtoff_fraction
+
  # If we know the wk of the year and what day of that wk, we can figure
  # out the Julian day of the year.
  if julian == -1 and week_of_year != -1 and weekday != -1:
@@ -330,7 +363,17 @@ def array_strptime(ndarray[object] values, object fmt,
  continue
  raise
 
- return result
+ if gmtoff is not None:
+ tzdelta = datetime_timedelta(seconds=gmtoff,
+ microseconds=gmtoff_fraction)
+ tzname = found_dict.get('Z')
+ if tzname:
+ tzinfo = datetime_timezone(tzdelta, tzname)
+ else:
+ tzinfo = datetime_timezone(tzdelta, tzname)
+ results_tz[i] = tzinfo
+
+ return result, results_tz
 
 
 """_getlang, LocaleTime, TimeRE, _calc_julian_from_U_or_W are vendored
@@ -538,6 +581,7 @@ class TimeRE(dict):
  # XXX: Does 'Y' need to worry about having less or more than
  # 4 digits?
  'Y': r"(?P<Y>\d\d\d\d)",
+ 'z': r"(?P<z>[+-]\d\d:?[0-5]\d(:?[0-5]\d(\.\d{1,6})?)?|Z)",
  'A': self.__seqToRE(self.locale_time.f_weekday, 'A'),
  'a': self.__seqToRE(self.locale_time.a_weekday, 'a'),
  'B': self.__seqToRE(self.locale_time.f_month[1:], 'B'),

pandas/core/tools/datetimes.py

@@ -344,7 +344,7 @@ def _convert_listlike(arg, box, format, name=None, tz=tz):
  if result is None:
  try:
  result = array_strptime(arg, format, exact=exact,
- errors=errors)
+ errors=errors)[0]
  except tslib.OutOfBoundsDatetime:
  if errors == 'raise':
  raise