blob: 90ae45e331e19157479439d263f89ec9c992cbd2 [file] [log] [blame]
'''
crontab.py
Written July 15, 2011 by Josiah Carlson
Released under the GNU LGPL v2.1 and v3
available:
http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html
http://www.gnu.org/licenses/lgpl.html
Other licenses may be available upon request.
'''
from collections import namedtuple
from datetime import datetime, timedelta
import sys
import warnings
_ranges = [
(0, 59),
(0, 59),
(0, 23),
(1, 31),
(1, 12),
(0, 6),
(1970, 2099),
]
ENTRIES = len(_ranges)
SECOND_OFFSET, MINUTE_OFFSET, HOUR_OFFSET, DAY_OFFSET, MONTH_OFFSET, WEEK_OFFSET, YEAR_OFFSET = range(ENTRIES)
_attribute = [
'second',
'minute',
'hour',
'day',
'month',
'isoweekday',
'year'
]
_alternate = {
MONTH_OFFSET: {'jan': 1, 'feb': 2, 'mar': 3, 'apr': 4, 'may': 5, 'jun': 6,
'jul': 7, 'aug': 8, 'sep': 9, 'oct': 10, 'nov':11, 'dec':12},
WEEK_OFFSET: {'sun': 0, 'mon': 1, 'tue': 2, 'wed': 3, 'thu': 4, 'fri': 5,
'sat': 6},
}
_aliases = {
'@yearly': '0 0 1 1 *',
'@annually': '0 0 1 1 *',
'@monthly': '0 0 1 * *',
'@weekly': '0 0 * * 0',
'@daily': '0 0 * * *',
'@hourly': '0 * * * *',
}
WARNING_CHANGE_MESSAGE = '''\
Version 0.22.0+ of crontab will use datetime.utcnow() and
datetime.utcfromtimestamp() instead of datetime.now() and
datetime.fromtimestamp() as was previous. This had been a bug, which will be
remedied. If you would like to keep the *old* behavior:
`ct.next(..., default_utc=False)` . If you want to use the new behavior *now*:
`ct.next(..., default_utc=True)`. If you pass a datetime object with a tzinfo
attribute that is not None, timezones will *just work* to the best of their
ability. There are tests...'''
if sys.version_info >= (3, 0):
_number_types = (int, float)
xrange = range
else:
_number_types = (int, long, float)
SECOND = timedelta(seconds=1)
MINUTE = timedelta(minutes=1)
HOUR = timedelta(hours=1)
DAY = timedelta(days=1)
WEEK = timedelta(days=7)
MONTH = timedelta(days=28)
YEAR = timedelta(days=365)
WARN_CHANGE = object()
# find the next scheduled time
def _end_of_month(dt):
ndt = dt + DAY
while dt.month == ndt.month:
dt += DAY
return ndt.replace(day=1) - DAY
def _month_incr(dt, m):
odt = dt
dt += MONTH
while dt.month == odt.month:
dt += DAY
# get to the first of next month, let the backtracking handle it
dt = dt.replace(day=1)
return dt - odt
def _year_incr(dt, m):
# simple leapyear stuff works for 1970-2099 :)
mod = dt.year % 4
if mod == 0 and (dt.month, dt.day) < (2, 29):
return YEAR + DAY
if mod == 3 and (dt.month, dt.day) > (2, 29):
return YEAR + DAY
return YEAR
_increments = [
lambda *a: SECOND,
lambda *a: MINUTE,
lambda *a: HOUR,
lambda *a: DAY,
_month_incr,
lambda *a: DAY,
_year_incr,
lambda dt,x: dt.replace(second=0),
lambda dt,x: dt.replace(minute=0),
lambda dt,x: dt.replace(hour=0),
lambda dt,x: dt.replace(day=1) if x > DAY else dt,
lambda dt,x: dt.replace(month=1) if x > DAY else dt,
lambda dt,x: dt,
]
# find the previously scheduled time
def _day_decr(dt, m):
if m.day.input != 'l':
return -DAY
odt = dt
ndt = dt = dt - DAY
while dt.month == ndt.month:
dt -= DAY
return dt - odt
def _month_decr(dt, m):
odt = dt
# get to the last day of last month, let the backtracking handle it
dt = dt.replace(day=1) - DAY
return dt - odt
def _year_decr(dt, m):
# simple leapyear stuff works for 1970-2099 :)
mod = dt.year % 4
if mod == 0 and (dt.month, dt.day) > (2, 29):
return -(YEAR + DAY)
if mod == 1 and (dt.month, dt.day) < (2, 29):
return -(YEAR + DAY)
return -YEAR
def _day_decr_reset(dt, x):
if x >= -DAY:
return dt
cur = dt.month
while dt.month == cur:
dt += DAY
return dt - DAY
_decrements = [
lambda *a: -SECOND,
lambda *a: -MINUTE,
lambda *a: -HOUR,
_day_decr,
_month_decr,
lambda *a: -DAY,
_year_decr,
lambda dt,x: dt.replace(second=59),
lambda dt,x: dt.replace(minute=59),
lambda dt,x: dt.replace(hour=23),
_day_decr_reset,
lambda dt,x: dt.replace(month=12) if x < -DAY else dt,
lambda dt,x: dt,
_year_decr,
]
Matcher = namedtuple('Matcher', 'second, minute, hour, day, month, weekday, year')
def _assert(condition, message, *args):
if not condition:
raise ValueError(message%args)
class _Matcher(object):
__slots__ = 'allowed', 'end', 'any', 'input', 'which', 'split'
def __init__(self, which, entry):
_assert(0 <= which <= YEAR_OFFSET,
"improper number of cron entries specified")
self.input = entry.lower()
self.split = self.input.split(',')
self.which = which
self.allowed = set()
self.end = None
self.any = '*' in self.split or '?' in self.split
for it in self.split:
al, en = self._parse_crontab(which, it)
if al is not None:
self.allowed.update(al)
self.end = en
_assert(self.end is not None,
"improper item specification: %r", entry.lower()
)
self.allowed = frozenset(self.allowed)
def __call__(self, v, dt):
for i, x in enumerate(self.split):
if x == 'l':
if v == _end_of_month(dt).day:
return True
elif x.startswith('l'):
# We have to do this in here, otherwise we can end up, for
# example, accepting *any* Friday instead of the *last* Friday.
if dt.month == (dt + WEEK).month:
continue
x = x[1:]
if x.isdigit():
x = int(x) if x != '7' else 0
if v == x:
return True
continue
start, end = map(int, x.partition('-')[::2])
allowed = set(range(start, end+1))
if 7 in allowed:
allowed.add(0)
if v in allowed:
return True
return self.any or v in self.allowed
def __lt__(self, other):
if self.any:
return self.end < other
return all(item < other for item in self.allowed)
def __gt__(self, other):
if self.any:
return _ranges[self.which][0] > other
return all(item > other for item in self.allowed)
def __eq__(self, other):
if self.any:
return other.any
return self.allowed == other.allowed
def __hash__(self):
return hash((self.any, self.allowed))
def _parse_crontab(self, which, entry):
'''
This parses a single crontab field and returns the data necessary for
this matcher to accept the proper values.
See the README for information about what is accepted.
'''
# this handles day of week/month abbreviations
def _fix(it):
if which in _alternate and not it.isdigit():
if it in _alternate[which]:
return _alternate[which][it]
_assert(it.isdigit(),
"invalid range specifier: %r (%r)", it, entry)
it = int(it, 10)
_assert(_start <= it <= _end_limit,
"item value %r out of range [%r, %r]",
it, _start, _end_limit)
return it
# this handles individual items/ranges
def _parse_piece(it):
if '-' in it:
start, end = map(_fix, it.split('-'))
# Allow "sat-sun"
if which in (DAY_OFFSET, WEEK_OFFSET) and end == 0:
end = 7
elif it == '*':
start = _start
end = _end
else:
start = _fix(it)
end = _end
if increment is None:
return set([start])
_assert(_start <= start <= _end_limit,
"%s range start value %r out of range [%r, %r]",
_attribute[which], start, _start, _end_limit)
_assert(_start <= end <= _end_limit,
"%s range end value %r out of range [%r, %r]",
_attribute[which], end, _start, _end_limit)
_assert(start <= end,
"%s range start value %r > end value %r",
_attribute[which], start, end)
return set(range(start, end+1, increment or 1))
_start, _end = _ranges[which]
_end_limit = _end
# wildcards
if entry in ('*', '?'):
if entry == '?':
_assert(which in (DAY_OFFSET, WEEK_OFFSET),
"cannot use '?' in the %r field", _attribute[which])
return None, _end
# last day of the month
if entry == 'l':
_assert(which == DAY_OFFSET,
"you can only specify a bare 'L' in the 'day' field")
return None, _end
# for the last 'friday' of the month, for example
elif entry.startswith('l'):
_assert(which == WEEK_OFFSET,
"you can only specify a leading 'L' in the 'weekday' field")
es, _, ee = entry[1:].partition('-')
_assert((entry[1:].isdigit() and 0 <= int(es) <= 7) or
(_ and es.isdigit() and ee.isdigit() and 0 <= int(es) <= 7 and 0 <= int(ee) <= 7),
"last <day> specifier must include a day number or range in the 'weekday' field, you entered %r", entry)
return None, _end
increment = None
# increments
if '/' in entry:
entry, increment = entry.split('/')
increment = int(increment, 10)
_assert(increment > 0,
"you can only use positive increment values, you provided %r",
increment)
# allow Sunday to be specified as weekday 7
if which == WEEK_OFFSET:
_end_limit = 7
# handle singles and ranges
good = _parse_piece(entry)
# change Sunday to weekday 0
if which == WEEK_OFFSET and 7 in good:
good.discard(7)
good.add(0)
return good, _end
class CronTab(object):
__slots__ = 'matchers',
def __init__(self, crontab):
self.matchers = self._make_matchers(crontab)
def _make_matchers(self, crontab):
'''
This constructs the full matcher struct.
'''
crontab = _aliases.get(crontab, crontab)
ct = crontab.split()
if len(ct) == 5:
ct.insert(0, '0')
ct.append('*')
elif len(ct) == 6:
ct.insert(0, '0')
_assert(len(ct) == 7,
"improper number of cron entries specified; got %i need 5 to 7"%(len(ct,)))
matchers = [_Matcher(which, entry) for which, entry in enumerate(ct)]
return Matcher(*matchers)
def _test_match(self, index, dt):
'''
This tests the given field for whether it matches with the current
datetime object passed.
'''
at = _attribute[index]
attr = getattr(dt, at)
if index == WEEK_OFFSET:
attr = attr() % 7
return self.matchers[index](attr, dt)
def next(self, now=None, increments=_increments, delta=True, default_utc=WARN_CHANGE):
'''
How long to wait in seconds before this crontab entry can next be
executed.
'''
if default_utc is WARN_CHANGE and (isinstance(now, _number_types) or (now and not now.tzinfo) or now is None):
warnings.warn(WARNING_CHANGE_MESSAGE, FutureWarning, 2)
default_utc = False
now = now or (datetime.utcnow() if default_utc and default_utc is not WARN_CHANGE else datetime.now())
if isinstance(now, _number_types):
now = datetime.utcfromtimestamp(now) if default_utc else datetime.fromtimestamp(now)
# handle timezones if the datetime object has a timezone and get a
# reasonable future/past start time
onow, now = now, now.replace(tzinfo=None)
tz = onow.tzinfo
future = now.replace(microsecond=0) + increments[0]()
if future < now:
# we are going backwards...
_test = lambda: future.year < self.matchers.year
if now.microsecond:
future = now.replace(microsecond=0)
else:
# we are going forwards
_test = lambda: self.matchers.year < future.year
# Start from the year and work our way down. Any time we increment a
# higher-magnitude value, we reset all lower-magnitude values. This
# gets us performance without sacrificing correctness. Still more
# complicated than a brute-force approach, but also orders of
# magnitude faster in basically all cases.
to_test = ENTRIES - 1
while to_test >= 0:
if not self._test_match(to_test, future):
inc = increments[to_test](future, self.matchers)
future += inc
for i in xrange(0, to_test):
future = increments[ENTRIES+i](future, inc)
try:
if _test():
return None
except:
print(future, type(future), type(inc))
raise
to_test = ENTRIES-1
continue
to_test -= 1
# verify the match
match = [self._test_match(i, future) for i in xrange(ENTRIES)]
_assert(all(match),
"\nYou have discovered a bug with crontab, please notify the\n" \
"author with the following information:\n" \
"crontab: %r\n" \
"now: %r", ' '.join(m.input for m in self.matchers), now)
delay = future - now
if tz:
delay += tz.utcoffset(now)
delay -= tz.utcoffset(future)
if not delta:
begin = datetime(1970, 1, 1)
delay = future - begin
if tz:
delay -= tz.utcoffset(future)
return delay.days * 86400 + delay.seconds + delay.microseconds / 1000000.
def previous(self, now=None, delta=True, default_utc=WARN_CHANGE):
return self.next(now, _decrements, delta, default_utc)
def test(self, entry):
if isinstance(entry, _number_types):
entry = datetime.utcfromtimestamp(entry)
for index in xrange(ENTRIES):
if not self._test_match(index, entry):
return False
return True