Merge cf6b425 into ff4ac98

Changelog

@@ -6,6 +6,10 @@ version 1.3.0 (release tag v1.3.0rel)
    methods, like dayofwk, dayofyr, __add__ and __sub__, will not work). Fixes issue #198.
    The calendar specific sub-classes are now deprecated, but remain for now
    as stubs that just instantiate the base class and override __repr__.
+ * update regex in _cpdef _parse_date so reference years with more than four
+   digits can be handled. Allow 'calendar=None' inr cftime.date2num (calendar associated with first
+   input datetime object is used - we may want to change the default from 'standard' to None
+   in a future release).
 
 version 1.2.1 (release tag v1.2.1rel)
 =====================================

cftime/_cftime.pyx

@@ -58,7 +58,7 @@ __version__ = '1.3.0'
 # Adapted from http://delete.me.uk/2005/03/iso8601.html
 # Note: This regex ensures that all ISO8601 timezone formats are accepted - but, due to legacy support for other timestrings, not all incorrect formats can be rejected.
 #       For example, the TZ spec "+01:0" will still work even though the minutes value is only one character long.
-ISO8601_REGEX = re.compile(r"(?P<year>[+-]?[0-9]{1,4})(-(?P<month>[0-9]{1,2})(-(?P<day>[0-9]{1,2})"
+ISO8601_REGEX = re.compile(r"(?P<year>[+-]?[0-9]+)(-(?P<month>[0-9]{1,2})(-(?P<day>[0-9]{1,2})"
                            r"(((?P<separator1>.)(?P<hour>[0-9]{1,2}):(?P<minute>[0-9]{1,2})(:(?P<second>[0-9]{1,2})(\.(?P<fraction>[0-9]+))?)?)?"
                            r"((?P<separator2>.?)(?P<timezone>Z|(([-+])([0-9]{2})((:([0-9]{2}))|([0-9]{2}))?)))?)?)?)?"
                            )
@@ -184,38 +184,62 @@ def date2num(dates,units,calendar='standard'):
     returns a numeric time value, or an array of numeric time values
     with approximately 1 microsecond accuracy.
     """
-    calendar = calendar.lower()
-    basedate = _dateparse(units,calendar=calendar)
-    (unit, isostring) = _datesplit(units)
-    # real-world calendars limited to positive reference years.
-    if calendar in ['julian', 'standard', 'gregorian', 'proleptic_gregorian']:
-        if basedate.year == 0:
-            msg='zero not allowed as a reference year, does not exist in Julian or Gregorian calendars'
-            raise ValueError(msg)
-    if unit not in UNIT_CONVERSION_FACTORS:
-        raise ValueError("Unsupported time units provided, {!r}.".format(unit))
-    if unit in ["months", "month"] and calendar != "360_day":
-        raise ValueError("Units of months only valid for 360_day calendar.")
-    factor = UNIT_CONVERSION_FACTORS[unit]
-    can_use_python_basedatetime = _can_use_python_datetime(basedate,calendar)
 
+    # input a scale or array-like?
     isscalar = False
     try:
         dates[0]
     except:
         isscalar = True
+
+    # masked array input?
     ismasked = False
     if np.ma.isMA(dates) and np.ma.is_masked(dates):
         mask = dates.mask
         ismasked = True
-    dates = np.asanyarray(dates)
-    shape = dates.shape
-    # are all dates python datetime instances?
+
+    # are all input dates 'real' python datetime objects?
+    dates = np.asanyarray(dates) # convert to numpy array
+    shape = dates.shape # save shape of input
     all_python_datetimes = True
     for date in dates.flat:
         if not isinstance(date,datetime_python):
            all_python_datetimes = False
            break
+
+    # if calendar is None or '', use calendar of first input cftime.datetime instances.
+    # if inputs are 'real' python datetime instances, use propleptic gregorian.
+    if not calendar:
+        if all_python_datetimes:
+            calendar = 'proleptic_gregorian'
+        else:
+            if isscalar:
+                d0 = dates.item()
+            else:
+                d0 = dates.flat[0]
+            if isinstance(d0,datetime_python):
+                calendar = 'proleptic_gregorian' 
+            else:
+                try:
+                    calendar = d0.calendar
+                except AttributeError:
+                    raise ValueError('no calendar specified',type(d0))
+
+    calendar = calendar.lower()
+    basedate = _dateparse(units,calendar=calendar)
+    (unit, isostring) = _datesplit(units)
+    # real-world calendars cannot have zero as a reference year.
+    if calendar in ['julian', 'standard', 'gregorian', 'proleptic_gregorian']:
+        if basedate.year == 0:
+            msg='zero not allowed as a reference year, does not exist in Julian or Gregorian calendars'
+            raise ValueError(msg)
+    if unit not in UNIT_CONVERSION_FACTORS:
+        raise ValueError("Unsupported time units provided, {!r}.".format(unit))
+    if unit in ["months", "month"] and calendar != "360_day":
+        raise ValueError("Units of months only valid for 360_day calendar.")
+    factor = UNIT_CONVERSION_FACTORS[unit]
+    can_use_python_basedatetime = _can_use_python_datetime(basedate,calendar)
+
     if can_use_python_basedatetime and all_python_datetimes:
         use_python_datetime = True
         if not isinstance(basedate, datetime_python):
@@ -1555,7 +1579,7 @@ cdef tuple add_timedelta(datetime dt, delta, bint (*is_leap)(int), bint julian_g
             if month > 12:
                 month = 1
                 year += 1
-                if year == 0:
+                if year == 0 and not has_year_zero:
                     year = 1
                 month_length = month_lengths(is_leap, year)
             day = 1

test/test_cftime.py

@@ -772,6 +772,9 @@ def roundtrip(delta,eps,units):
         dt1 = datetime(1810, 4, 24, 16, 15, 10)
         units = 'days since -4713-01-01 12:00'
         dt2 = num2date(date2num(dt1, units), units)
+        # switch to these if default calendar for date2num changed to None
+        #dt2 = num2date(date2num(dt1, units), units, calendar='proleptic_gregorian')
+        #dt2 = num2date(date2num(dt1, units, calendar='standard'), units)
         assert(dt1 == dt2)
 # issue #189 - leap years calculated incorrectly for negative years in proleptic_gregorian calendar
         dt1 = datetime(2020, 4, 24, 16, 15, 10)
@@ -785,6 +788,26 @@ def roundtrip(delta,eps,units):
             dt = cftime.datetime(2020, 1, 1, calendar=cal)
             dt += timedelta(hours=1)
             assert(str(dt) == '2020-01-01 01:00:00')
+# issue #193 - years with more than four digits in reference date
+        assert(cftime.date2num(cftime.datetime(18000, 12, 1, 0, 0), 'days since 18000-1-1', '360_day') == 330.0)
+        # julian day not including year zero
+        d = cftime.datetime(2020, 12, 1, 12, calendar='julian')
+        units = 'days since -4713-1-1 12:00'
+        jd = cftime.date2num(d,units,calendar='julian')
+        assert(jd == 2459198.0)
+        # if calendar=None, use input date to determine calendar
+        jd = cftime.date2num(d,units,calendar=None)
+        assert(jd == 2459198.0)
+        # if no calendar specified, default assumed 'standard'
+        jd = cftime.date2num(d,units)
+        assert(jd == 2459185.0)
+        # switch to these if default calendar for date2num switched to None
+        # if no calendar specified, use input date to determine calendar
+        #jd = cftime.date2num(d,units)
+        #assert(jd == 2459198.0)
+        ## use 'standard' calendar
+        #jd = cftime.date2num(d,units,calendar='standard')
+        #assert(jd == 2459185.0)
 
 
 class TestDate2index(unittest.TestCase):