#!/usr/bin/env python
# -*- coding: utf-8 -*-

from __future__ import absolute_import, division, print_function

import binascii
import calendar
import copy
import datetime
import math
import platform
import random
import re
import struct
import sys
import traceback as _traceback
from time import time

# as pytz is optional in thirdparty libs but we need it for good support under
# python2, just test that it's well installed
import pytz  # noqa
from dateutil.relativedelta import relativedelta
from dateutil.tz import tzutc


def is_32bit():
    """
    Detect if Python is running in 32-bit mode.
    Compatible with Python 2.6 and later versions.
    Returns True if running on 32-bit Python, False for 64-bit.
    """
    # Method 1: Check pointer size
    bits = struct.calcsize("P") * 8

    # Method 2: Check platform architecture string
    try:
        architecture = platform.architecture()[0]
    except RuntimeError:
        architecture = None

    # Method 3: Check maxsize (sys.maxint in Python 2)
    try:
        # Python 2
        is_small_maxsize = sys.maxint <= 2 ** 32
    except AttributeError:
        # Python 3
        is_small_maxsize = sys.maxsize <= 2 ** 32

    # Evaluate all available methods
    is_32 = False

    if bits == 32:
        is_32 = True
    elif architecture and "32" in architecture:
        is_32 = True
    elif is_small_maxsize:
        is_32 = True

    return is_32


try:
    # https://github.com/python/cpython/issues/101069 detection
    if is_32bit():
        datetime.datetime.fromtimestamp(3999999999)
    OVERFLOW32B_MODE = False
except OverflowError:
    OVERFLOW32B_MODE = True

try:
    from collections import OrderedDict
except ImportError:
    OrderedDict = dict  # py26 degraded mode, expanders order will not be immutable


try:
    # py3 recent
    UTC_DT = datetime.timezone.utc
except AttributeError:
    UTC_DT = pytz.utc
EPOCH = datetime.datetime.fromtimestamp(0, UTC_DT)

# fmt: off
M_ALPHAS = {
    "jan": 1, "feb": 2,  "mar": 3,  "apr": 4,  # noqa: E241
    "may": 5, "jun": 6,  "jul": 7,  "aug": 8,  # noqa: E241
    "sep": 9, "oct": 10, "nov": 11, "dec": 12,
}
DOW_ALPHAS = {
    "sun": 0, "mon": 1, "tue": 2, "wed": 3, "thu": 4, "fri": 5, "sat": 6
}

MINUTE_FIELD = 0
HOUR_FIELD = 1
DAY_FIELD = 2
MONTH_FIELD = 3
DOW_FIELD = 4
SECOND_FIELD = 5
YEAR_FIELD = 6

UNIX_FIELDS =   (MINUTE_FIELD, HOUR_FIELD, DAY_FIELD, MONTH_FIELD, DOW_FIELD)  # noqa: E222
SECOND_FIELDS = (MINUTE_FIELD, HOUR_FIELD, DAY_FIELD, MONTH_FIELD, DOW_FIELD, SECOND_FIELD)  # noqa: E222
YEAR_FIELDS =   (MINUTE_FIELD, HOUR_FIELD, DAY_FIELD, MONTH_FIELD, DOW_FIELD, SECOND_FIELD, YEAR_FIELD)  # noqa: E222
# fmt: on

step_search_re = re.compile(r"^([^-]+)-([^-/]+)(/(\d+))?$")
only_int_re = re.compile(r"^\d+$")

WEEKDAYS = "|".join(DOW_ALPHAS.keys())
MONTHS = "|".join(M_ALPHAS.keys())
star_or_int_re = re.compile(r"^(\d+|\*)$")
special_dow_re = re.compile(
    (r"^(?P<pre>((?P<he>(({WEEKDAYS})(-({WEEKDAYS}))?)").format(WEEKDAYS=WEEKDAYS)
    + (r"|(({MONTHS})(-({MONTHS}))?)|\w+)#)|l)(?P<last>\d+)$").format(MONTHS=MONTHS)
)
re_star = re.compile("[*]")
hash_expression_re = re.compile(
    r"^(?P<hash_type>h|r)(\((?P<range_begin>\d+)-(?P<range_end>\d+)\))?(\/(?P<divisor>\d+))?$"
)

CRON_FIELDS = {
    "unix": UNIX_FIELDS,
    "second": SECOND_FIELDS,
    "year": YEAR_FIELDS,
    len(UNIX_FIELDS): UNIX_FIELDS,
    len(SECOND_FIELDS): SECOND_FIELDS,
    len(YEAR_FIELDS): YEAR_FIELDS,
}
UNIX_CRON_LEN = len(UNIX_FIELDS)
SECOND_CRON_LEN = len(SECOND_FIELDS)
YEAR_CRON_LEN = len(YEAR_FIELDS)
# retrocompat
VALID_LEN_EXPRESSION = set(a for a in CRON_FIELDS if isinstance(a, int))
TIMESTAMP_TO_DT_CACHE = {}
EXPRESSIONS = {}
MARKER = object()


def timedelta_to_seconds(td):
    return (td.microseconds + (td.seconds + td.days * 24 * 3600) * 10 ** 6) / 10 ** 6


def datetime_to_timestamp(d):
    if d.tzinfo is not None:
        d = d.replace(tzinfo=None) - d.utcoffset()

    return timedelta_to_seconds(d - datetime.datetime(1970, 1, 1))


class CroniterError(ValueError):
    """General top-level Croniter base exception"""


class CroniterBadTypeRangeError(TypeError):
    """."""


class CroniterBadCronError(CroniterError):
    """Syntax, unknown value, or range error within a cron expression"""


class CroniterUnsupportedSyntaxError(CroniterBadCronError):
    """Valid cron syntax, but likely to produce inaccurate results"""

    # Extending CroniterBadCronError, which may be contridatory, but this allows
    # catching both errors with a single exception.  From a user perspective
    # these will likely be handled the same way.


class CroniterBadDateError(CroniterError):
    """Unable to find next/prev timestamp match"""


class CroniterNotAlphaError(CroniterBadCronError):
    """Cron syntax contains an invalid day or month abbreviation"""


class croniter(object):
    MONTHS_IN_YEAR = 12

    # This helps with expanding `*` fields into `lower-upper` ranges. Each item
    # in this tuple maps to the corresponding field index
    RANGES = (
        (0, 59),
        (0, 23),
        (1, 31),
        (1, 12),
        (0, 6),
        (0, 59),
        (1970, 2099),
    )
    DAYS = (31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31)

    ALPHACONV = (
        {},  # 0: min
        {},  # 1: hour
        {"l": "l"},  # 2: dom
        # 3: mon
        copy.deepcopy(M_ALPHAS),
        # 4: dow
        copy.deepcopy(DOW_ALPHAS),
        # 5: second
        {},
        # 6: year
        {},
    )

    LOWMAP = (
        {},
        {},
        {0: 1},
        {0: 1},
        {7: 0},
        {},
        {},
    )

    LEN_MEANS_ALL = (
        60,
        24,
        31,
        12,
        7,
        60,
        130,
    )

    def __init__(
        self,
        expr_format,
        start_time=None,
        ret_type=float,
        day_or=True,
        max_years_between_matches=None,
        is_prev=False,
        hash_id=None,
        implement_cron_bug=False,
        second_at_beginning=None,
        expand_from_start_time=False,
    ):
        self._ret_type = ret_type
        self._day_or = day_or
        self._implement_cron_bug = implement_cron_bug
        self.second_at_beginning = bool(second_at_beginning)
        self._expand_from_start_time = expand_from_start_time

        if hash_id:
            if not isinstance(hash_id, (bytes, str)):
                raise TypeError("hash_id must be bytes or UTF-8 string")
            if not isinstance(hash_id, bytes):
                hash_id = hash_id.encode("UTF-8")

        self._max_years_btw_matches_explicitly_set = max_years_between_matches is not None
        if not self._max_years_btw_matches_explicitly_set:
            max_years_between_matches = 50
        self._max_years_between_matches = max(int(max_years_between_matches), 1)

        if start_time is None:
            start_time = time()

        self.tzinfo = None

        self.start_time = None
        self.dst_start_time = None
        self.cur = None
        self.set_current(start_time, force=False)

        self.expanded, self.nth_weekday_of_month = self.expand(
            expr_format,
            hash_id=hash_id,
            from_timestamp=self.dst_start_time if self._expand_from_start_time else None,
            second_at_beginning=second_at_beginning,
        )
        self.fields = CRON_FIELDS[len(self.expanded)]
        self.expressions = EXPRESSIONS[(expr_format, hash_id, second_at_beginning)]
        self._is_prev = is_prev

    @classmethod
    def _alphaconv(cls, index, key, expressions):
        try:
            return cls.ALPHACONV[index][key]
        except KeyError:
            raise CroniterNotAlphaError("[{0}] is not acceptable".format(" ".join(expressions)))

    def get_next(self, ret_type=None, start_time=None, update_current=True):
        if start_time and self._expand_from_start_time:
            raise ValueError("start_time is not supported when using expand_from_start_time = True.")
        return self._get_next(
            ret_type=ret_type,
            start_time=start_time,
            is_prev=False,
            update_current=update_current,
        )

    def get_prev(self, ret_type=None, start_time=None, update_current=True):
        return self._get_next(
            ret_type=ret_type,
            start_time=start_time,
            is_prev=True,
            update_current=update_current,
        )

    def get_current(self, ret_type=None):
        ret_type = ret_type or self._ret_type
        if issubclass(ret_type, datetime.datetime):
            return self.timestamp_to_datetime(self.cur)
        return self.cur

    def set_current(self, start_time, force=True):
        if (force or (self.cur is None)) and start_time is not None:
            if isinstance(start_time, datetime.datetime):
                self.tzinfo = start_time.tzinfo
                start_time = self.datetime_to_timestamp(start_time)

            self.start_time = start_time
            self.dst_start_time = start_time
            self.cur = start_time
        return self.cur

    @staticmethod
    def datetime_to_timestamp(d):
        """
        Converts a `datetime` object `d` into a UNIX timestamp.
        """
        return datetime_to_timestamp(d)

    _datetime_to_timestamp = datetime_to_timestamp  # retrocompat

    def timestamp_to_datetime(self, timestamp, tzinfo=MARKER):
        """
        Converts a UNIX `timestamp` into a `datetime` object.
        """
        if tzinfo is MARKER:  # allow to give tzinfo=None even if self.tzinfo is set
            tzinfo = self.tzinfo
        k = timestamp
        if tzinfo:
            k = (timestamp, repr(tzinfo))
        try:
            return TIMESTAMP_TO_DT_CACHE[k]
        except KeyError:
            pass
        if OVERFLOW32B_MODE:
            # degraded mode to workaround Y2038
            # see https://github.com/python/cpython/issues/101069
            result = EPOCH.replace(tzinfo=None) + datetime.timedelta(seconds=timestamp)
        else:
            result = datetime.datetime.fromtimestamp(timestamp, tz=tzutc()).replace(tzinfo=None)
        if tzinfo:
            result = result.replace(tzinfo=UTC_DT).astimezone(tzinfo)
        TIMESTAMP_TO_DT_CACHE[(result, repr(result.tzinfo))] = result
        return result

    _timestamp_to_datetime = timestamp_to_datetime  # retrocompat

    @staticmethod
    def timedelta_to_seconds(td):
        """
        Converts a 'datetime.timedelta' object `td` into seconds contained in
        the duration.
        Note: We cannot use `timedelta.total_seconds()` because this is not
        supported by Python 2.6.
        """
        return timedelta_to_seconds(td)

    _timedelta_to_seconds = timedelta_to_seconds  # retrocompat

    def _get_next(
        self,
        ret_type=None,
        start_time=None,
        is_prev=None,
        update_current=None,
    ):
        if update_current is None:
            update_current = True
        self.set_current(start_time, force=True)
        if is_prev is None:
            is_prev = self._is_prev
        self._is_prev = is_prev
        expanded = self.expanded[:]
        nth_weekday_of_month = self.nth_weekday_of_month.copy()

        ret_type = ret_type or self._ret_type

        if not issubclass(ret_type, (float, datetime.datetime)):
            raise TypeError("Invalid ret_type, only 'float' or 'datetime' is acceptable.")

        # exception to support day of month and day of week as defined in cron
        dom_dow_exception_processed = False
        if (expanded[DAY_FIELD][0] != "*" and expanded[DOW_FIELD][0] != "*") and self._day_or:
            # If requested, handle a bug in vixie cron/ISC cron where day_of_month and day_of_week form
            # an intersection (AND) instead of a union (OR) if either field is an asterisk or starts with an asterisk
            # (https://crontab.guru/cron-bug.html)
            if self._implement_cron_bug and (
                re_star.match(self.expressions[DAY_FIELD]) or re_star.match(self.expressions[DOW_FIELD])
            ):
                # To produce a schedule identical to the cron bug, we'll bypass the code that
                # makes a union of DOM and DOW, and instead skip to the code that does an intersect instead
                pass
            else:
                bak = expanded[DOW_FIELD]
                expanded[DOW_FIELD] = ["*"]
                t1 = self._calc(self.cur, expanded, nth_weekday_of_month, is_prev)
                expanded[DOW_FIELD] = bak
                expanded[DAY_FIELD] = ["*"]

                t2 = self._calc(self.cur, expanded, nth_weekday_of_month, is_prev)
                if not is_prev:
                    result = t1 if t1 < t2 else t2
                else:
                    result = t1 if t1 > t2 else t2
                dom_dow_exception_processed = True

        if not dom_dow_exception_processed:
            result = self._calc(self.cur, expanded, nth_weekday_of_month, is_prev)

        # DST Handling for cron job spanning across days
        dtstarttime = self._timestamp_to_datetime(self.dst_start_time)
        dtstarttime_utcoffset = dtstarttime.utcoffset() or datetime.timedelta(0)
        dtresult = self.timestamp_to_datetime(result)
        lag = lag_hours = 0
        # do we trigger DST on next crontab (handle backward changes)
        dtresult_utcoffset = dtstarttime_utcoffset
        if dtresult and self.tzinfo:
            dtresult_utcoffset = dtresult.utcoffset()
            lag_hours = self._timedelta_to_seconds(dtresult - dtstarttime) / (60 * 60)
            lag = self._timedelta_to_seconds(dtresult_utcoffset - dtstarttime_utcoffset)
        hours_before_midnight = 24 - dtstarttime.hour
        if dtresult_utcoffset != dtstarttime_utcoffset:
            if (lag > 0 and abs(lag_hours) >= hours_before_midnight) or (
                lag < 0 and ((3600 * abs(lag_hours) + abs(lag)) >= hours_before_midnight * 3600)
            ):
                dtresult_adjusted = dtresult - datetime.timedelta(seconds=lag)
                result_adjusted = self._datetime_to_timestamp(dtresult_adjusted)
                # Do the actual adjust only if the result time actually exists
                if self._timestamp_to_datetime(result_adjusted).tzinfo == dtresult_adjusted.tzinfo:
                    dtresult = dtresult_adjusted
                    result = result_adjusted
                self.dst_start_time = result
        if update_current:
            self.cur = result
        if issubclass(ret_type, datetime.datetime):
            result = dtresult
        return result

    # iterator protocol, to enable direct use of croniter
    # objects in a loop, like "for dt in croniter("5 0 * * *'): ..."
    # or for combining multiple croniters into single
    # dates feed using 'itertools' module
    def all_next(self, ret_type=None, start_time=None, update_current=None):
        """
        Returns a generator yielding consecutive dates.

        May be used instead of an implicit call to __iter__ whenever a
        non-default `ret_type` needs to be specified.
        """
        # In a Python 3.7+ world:  contextlib.suppress and contextlib.nullcontext could be used instead
        try:
            while True:
                self._is_prev = False
                yield self._get_next(
                    ret_type=ret_type,
                    start_time=start_time,
                    update_current=update_current,
                )
                start_time = None
        except CroniterBadDateError:
            if self._max_years_btw_matches_explicitly_set:
                return
            raise

    def all_prev(self, ret_type=None, start_time=None, update_current=None):
        """
        Returns a generator yielding previous dates.
        """
        try:
            while True:
                self._is_prev = True
                yield self._get_next(
                    ret_type=ret_type,
                    start_time=start_time,
                    update_current=update_current,
                )
                start_time = None
        except CroniterBadDateError:
            if self._max_years_btw_matches_explicitly_set:
                return
            raise

    def iter(self, *args, **kwargs):
        return self.all_prev if self._is_prev else self.all_next

    def __iter__(self):
        return self

    __next__ = next = _get_next

    def _calc(self, now, expanded, nth_weekday_of_month, is_prev):
        if is_prev:
            now = math.ceil(now)
            nearest_diff_method = self._get_prev_nearest_diff
            sign = -1
            offset = 1 if (len(expanded) > UNIX_CRON_LEN or now % 60 > 0) else 60
        else:
            now = math.floor(now)
            nearest_diff_method = self._get_next_nearest_diff
            sign = 1
            offset = 1 if (len(expanded) > UNIX_CRON_LEN) else 60

        dst = now = self.timestamp_to_datetime(now + sign * offset)

        month, year = dst.month, dst.year
        current_year = now.year
        DAYS = self.DAYS

        def proc_year(d):
            if len(expanded) == YEAR_CRON_LEN:
                try:
                    expanded[YEAR_FIELD].index("*")
                except ValueError:
                    # use None as range_val to indicate no loop
                    diff_year = nearest_diff_method(d.year, expanded[YEAR_FIELD], None)
                    if diff_year is None:
                        return None, d
                    if diff_year != 0:
                        if is_prev:
                            d += relativedelta(
                                years=diff_year,
                                month=12,
                                day=31,
                                hour=23,
                                minute=59,
                                second=59,
                            )
                        else:
                            d += relativedelta(
                                years=diff_year,
                                month=1,
                                day=1,
                                hour=0,
                                minute=0,
                                second=0,
                            )
                        return True, d
            return False, d

        def proc_month(d):
            try:
                expanded[MONTH_FIELD].index("*")
            except ValueError:
                diff_month = nearest_diff_method(d.month, expanded[MONTH_FIELD], self.MONTHS_IN_YEAR)
                reset_day = 1

                if diff_month is not None and diff_month != 0:
                    if is_prev:
                        d += relativedelta(months=diff_month)
                        reset_day = DAYS[d.month - 1]
                        if d.month == 2 and self.is_leap(d.year) is True:
                            reset_day += 1
                        d += relativedelta(day=reset_day, hour=23, minute=59, second=59)
                    else:
                        d += relativedelta(months=diff_month, day=reset_day, hour=0, minute=0, second=0)
                    return True, d
            return False, d

        def proc_day_of_month(d):
            try:
                expanded[DAY_FIELD].index("*")
            except ValueError:
                days = DAYS[month - 1]
                if month == 2 and self.is_leap(year) is True:
                    days += 1
                if "l" in expanded[DAY_FIELD] and days == d.day:
                    return False, d

                if is_prev:
                    days_in_prev_month = DAYS[(month - 2) % self.MONTHS_IN_YEAR]
                    diff_day = nearest_diff_method(d.day, expanded[DAY_FIELD], days_in_prev_month)
                else:
                    diff_day = nearest_diff_method(d.day, expanded[DAY_FIELD], days)

                if diff_day is not None and diff_day != 0:
                    if is_prev:
                        d += relativedelta(days=diff_day, hour=23, minute=59, second=59)
                    else:
                        d += relativedelta(days=diff_day, hour=0, minute=0, second=0)
                    return True, d
            return False, d

        def proc_day_of_week(d):
            try:
                expanded[DOW_FIELD].index("*")
            except ValueError:
                diff_day_of_week = nearest_diff_method(d.isoweekday() % 7, expanded[DOW_FIELD], 7)
                if diff_day_of_week is not None and diff_day_of_week != 0:
                    if is_prev:
                        d += relativedelta(days=diff_day_of_week, hour=23, minute=59, second=59)
                    else:
                        d += relativedelta(days=diff_day_of_week, hour=0, minute=0, second=0)
                    return True, d
            return False, d

        def proc_day_of_week_nth(d):
            if "*" in nth_weekday_of_month:
                s = nth_weekday_of_month["*"]
                for i in range(0, 7):
                    if i in nth_weekday_of_month:
                        nth_weekday_of_month[i].update(s)
                    else:
                        nth_weekday_of_month[i] = s
                del nth_weekday_of_month["*"]

            candidates = []
            for wday, nth in nth_weekday_of_month.items():
                c = self._get_nth_weekday_of_month(d.year, d.month, wday)
                for n in nth:
                    if n == "l":
                        candidate = c[-1]
                    elif len(c) < n:
                        continue
                    else:
                        candidate = c[n - 1]
                    if (is_prev and candidate <= d.day) or (not is_prev and d.day <= candidate):
                        candidates.append(candidate)

            if not candidates:
                if is_prev:
                    d += relativedelta(days=-d.day, hour=23, minute=59, second=59)
                else:
                    days = DAYS[month - 1]
                    if month == 2 and self.is_leap(year) is True:
                        days += 1
                    d += relativedelta(days=(days - d.day + 1), hour=0, minute=0, second=0)
                return True, d

            candidates.sort()
            diff_day = (candidates[-1] if is_prev else candidates[0]) - d.day
            if diff_day != 0:
                if is_prev:
                    d += relativedelta(days=diff_day, hour=23, minute=59, second=59)
                else:
                    d += relativedelta(days=diff_day, hour=0, minute=0, second=0)
                return True, d
            return False, d

        def proc_hour(d):
            try:
                expanded[HOUR_FIELD].index("*")
            except ValueError:
                diff_hour = nearest_diff_method(d.hour, expanded[HOUR_FIELD], 24)
                if diff_hour is not None and diff_hour != 0:
                    if is_prev:
                        d += relativedelta(hours=diff_hour, minute=59, second=59)
                    else:
                        d += relativedelta(hours=diff_hour, minute=0, second=0)
                    return True, d
            return False, d

        def proc_minute(d):
            try:
                expanded[MINUTE_FIELD].index("*")
            except ValueError:
                diff_min = nearest_diff_method(d.minute, expanded[MINUTE_FIELD], 60)
                if diff_min is not None and diff_min != 0:
                    if is_prev:
                        d += relativedelta(minutes=diff_min, second=59)
                    else:
                        d += relativedelta(minutes=diff_min, second=0)
                    return True, d
            return False, d

        def proc_second(d):
            if len(expanded) > UNIX_CRON_LEN:
                try:
                    expanded[SECOND_FIELD].index("*")
                except ValueError:
                    diff_sec = nearest_diff_method(d.second, expanded[SECOND_FIELD], 60)
                    if diff_sec is not None and diff_sec != 0:
                        d += relativedelta(seconds=diff_sec)
                        return True, d
            else:
                d += relativedelta(second=0)
            return False, d

        procs = [
            proc_year,
            proc_month,
            proc_day_of_month,
            (proc_day_of_week_nth if nth_weekday_of_month else proc_day_of_week),
            proc_hour,
            proc_minute,
            proc_second,
        ]

        while abs(year - current_year) <= self._max_years_between_matches:
            next = False
            stop = False
            for proc in procs:
                (changed, dst) = proc(dst)
                # `None` can be set mostly for year processing
                # so please see proc_year / _get_prev_nearest_diff / _get_next_nearest_diff
                if changed is None:
                    stop = True
                    break
                if changed:
                    month, year = dst.month, dst.year
                    next = True
                    break
            if stop:
                break
            if next:
                continue
            return self.datetime_to_timestamp(dst.replace(microsecond=0))

        if is_prev:
            raise CroniterBadDateError("failed to find prev date")
        raise CroniterBadDateError("failed to find next date")

    @staticmethod
    def _get_next_nearest(x, to_check):
        small = [item for item in to_check if item < x]
        large = [item for item in to_check if item >= x]
        large.extend(small)
        return large[0]

    @staticmethod
    def _get_prev_nearest(x, to_check):
        small = [item for item in to_check if item <= x]
        large = [item for item in to_check if item > x]
        small.reverse()
        large.reverse()
        small.extend(large)
        return small[0]

    @staticmethod
    def _get_next_nearest_diff(x, to_check, range_val):
        """
        `range_val` is the range of a field.
        If no available time, we can move to next loop(like next month).
        `range_val` can also be set to `None` to indicate that there is no loop.
        ( Currently, should only used for `year` field )
        """
        for i, d in enumerate(to_check):
            if d == "l" and range_val is not None:
                # if 'l' then it is the last day of month
                # => its value of range_val
                d = range_val
            if d >= x:
                return d - x
        # When range_val is None and x not exists in to_check,
        # `None` will be returned to suggest no more available time
        if range_val is None:
            return None
        return to_check[0] - x + range_val

    @staticmethod
    def _get_prev_nearest_diff(x, to_check, range_val):
        """
        `range_val` is the range of a field.
        If no available time, we can move to previous loop(like previous month).
        Range_val can also be set to `None` to indicate that there is no loop.
        ( Currently should only used for `year` field )
        """
        candidates = to_check[:]
        candidates.reverse()
        for d in candidates:
            if d != "l" and d <= x:
                return d - x
        if "l" in candidates:
            return -x
        # When range_val is None and x not exists in to_check,
        # `None` will be returned to suggest no more available time
        if range_val is None:
            return None
        candidate = candidates[0]
        for c in candidates:
            # fixed: c < range_val
            # this code will reject all 31 day of month, 12 month, 59 second,
            # 23 hour and so on.
            # if candidates has just a element, this will not harmful.
            # but candidates have multiple elements, then values equal to
            # range_val will rejected.
            if c <= range_val:
                candidate = c
                break
        # fix crontab "0 6 30 3 *" condidates only a element, then get_prev error return 2021-03-02 06:00:00
        if candidate > range_val:
            return -range_val
        return candidate - x - range_val

    @staticmethod
    def _get_nth_weekday_of_month(year, month, day_of_week):
        """For a given year/month return a list of days in nth-day-of-month order.
        The last weekday of the month is always [-1].
        """
        w = (day_of_week + 6) % 7
        c = calendar.Calendar(w).monthdayscalendar(year, month)
        if c[0][0] == 0:
            c.pop(0)
        return tuple(i[0] for i in c)

    @staticmethod
    def is_leap(year):
        return bool(year % 400 == 0 or (year % 4 == 0 and year % 100 != 0))

    @classmethod
    def value_alias(cls, val, field_index, len_expressions=UNIX_CRON_LEN):
        if isinstance(len_expressions, (list, dict, tuple, set)):
            len_expressions = len(len_expressions)
        if val in cls.LOWMAP[field_index] and not (
            # do not support 0 as a month either for classical 5 fields cron,
            # 6fields second repeat form or 7 fields year form
            # but still let conversion happen if day field is shifted
            (field_index in [DAY_FIELD, MONTH_FIELD] and len_expressions == UNIX_CRON_LEN)
            or (field_index in [MONTH_FIELD, DOW_FIELD] and len_expressions == SECOND_CRON_LEN)
            or (field_index in [DAY_FIELD, MONTH_FIELD, DOW_FIELD] and len_expressions == YEAR_CRON_LEN)
        ):
            val = cls.LOWMAP[field_index][val]
        return val

    @classmethod
    def _expand(
        cls,
        expr_format,
        hash_id=None,
        second_at_beginning=False,
        from_timestamp=None,
    ):
        # Split the expression in components, and normalize L -> l, MON -> mon,
        # etc. Keep expr_format untouched so we can use it in the exception
        # messages.
        expr_aliases = {
            "@midnight": ("0 0 * * *", "h h(0-2) * * * h"),
            "@hourly": ("0 * * * *", "h * * * * h"),
            "@daily": ("0 0 * * *", "h h * * * h"),
            "@weekly": ("0 0 * * 0", "h h * * h h"),
            "@monthly": ("0 0 1 * *", "h h h * * h"),
            "@yearly": ("0 0 1 1 *", "h h h h * h"),
            "@annually": ("0 0 1 1 *", "h h h h * h"),
        }

        efl = expr_format.lower()
        hash_id_expr = 1 if hash_id is not None else 0
        try:
            efl = expr_aliases[efl][hash_id_expr]
        except KeyError:
            pass

        expressions = efl.split()

        if len(expressions) not in VALID_LEN_EXPRESSION:
            raise CroniterBadCronError("Exactly 5, 6 or 7 columns has to be specified for iterator expression.")

        if len(expressions) > UNIX_CRON_LEN and second_at_beginning:
            # move second to it's own(6th) field to process by same logical
            expressions.insert(SECOND_FIELD, expressions.pop(0))

        expanded = []
        nth_weekday_of_month = {}

        for field_index, expr in enumerate(expressions):
            for expanderid, expander in EXPANDERS.items():
                expr = expander(cls).expand(
                    efl,
                    field_index,
                    expr,
                    hash_id=hash_id,
                    from_timestamp=from_timestamp,
                )

            if "?" in expr:
                if expr != "?":
                    raise CroniterBadCronError(
                        "[{0}] is not acceptable. Question mark can not used with other characters".format(expr_format)
                    )
                if field_index not in [DAY_FIELD, DOW_FIELD]:
                    raise CroniterBadCronError(
                        "[{0}] is not acceptable. Question mark can only used in day_of_month or day_of_week".format(
                            expr_format
                        )
                    )
                # currently just trade `?` as `*`
                expr = "*"

            e_list = expr.split(",")
            res = []

            while len(e_list) > 0:
                e = e_list.pop()
                nth = None

                if field_index == DOW_FIELD:
                    # Handle special case in the dow expression: 2#3, l3
                    special_dow_rem = special_dow_re.match(str(e))
                    if special_dow_rem:
                        g = special_dow_rem.groupdict()
                        he, last = g.get("he", ""), g.get("last", "")
                        if he:
                            e = he
                            try:
                                nth = int(last)
                                assert 5 >= nth >= 1
                            except (KeyError, ValueError, AssertionError):
                                raise CroniterBadCronError(
                                    "[{0}] is not acceptable. Invalid day_of_week value: '{1}'".format(
                                        expr_format, nth
                                    )
                                )
                        elif last:
                            e = last
                            nth = g["pre"]  # 'l'

                # Before matching step_search_re, normalize "*" to "{min}-{max}".
                # Example: in the minute field, "*/5" normalizes to "0-59/5"
                t = re.sub(
                    r"^\*(\/.+)$",
                    r"%d-%d\1" % (cls.RANGES[field_index][0], cls.RANGES[field_index][1]),
                    str(e),
                )
                m = step_search_re.search(t)

                if not m:
                    # Before matching step_search_re,
                    # normalize "{start}/{step}" to "{start}-{max}/{step}".
                    # Example: in the minute field, "10/5" normalizes to "10-59/5"
                    t = re.sub(
                        r"^(.+)\/(.+)$",
                        r"\1-%d/\2" % (cls.RANGES[field_index][1]),
                        str(e),
                    )
                    m = step_search_re.search(t)

                if m:
                    # early abort if low/high are out of bounds
                    (low, high, step) = m.group(1), m.group(2), m.group(4) or 1
                    if field_index == DAY_FIELD and high == "l":
                        high = "31"

                    if not only_int_re.search(low):
                        low = "{0}".format(cls._alphaconv(field_index, low, expressions))

                    if not only_int_re.search(high):
                        high = "{0}".format(cls._alphaconv(field_index, high, expressions))

                    # normally, it's already guarded by the RE that should not accept not-int values.
                    if not only_int_re.search(str(step)):
                        raise CroniterBadCronError(
                            "[{0}] step '{2}' in field {1} is not acceptable".format(expr_format, field_index, step)
                        )
                    step = int(step)

                    for band in low, high:
                        if not only_int_re.search(str(band)):
                            raise CroniterBadCronError(
                                "[{0}] bands '{2}-{3}' in field {1} are not acceptable".format(
                                    expr_format, field_index, low, high
                                )
                            )

                    low, high = [cls.value_alias(int(_val), field_index, expressions) for _val in (low, high)]

                    if max(low, high) > max(cls.RANGES[field_index][0], cls.RANGES[field_index][1]):
                        raise CroniterBadCronError("{0} is out of bands".format(expr_format))

                    if from_timestamp:
                        low = cls._get_low_from_current_date_number(field_index, int(step), int(from_timestamp))

                    # Handle when the second bound of the range is in backtracking order:
                    # eg: X-Sun or X-7 (Sat-Sun) in DOW, or X-Jan (Apr-Jan) in MONTH
                    if low > high:
                        whole_field_range = list(
                            range(
                                cls.RANGES[field_index][0],
                                cls.RANGES[field_index][1] + 1,
                                1,
                            )
                        )
                        # Add FirstBound -> ENDRANGE, respecting step
                        rng = list(range(low, cls.RANGES[field_index][1] + 1, step))
                        # Then 0 -> SecondBound, but skipping n first occurences according to step
                        # EG to respect such expressions : Apr-Jan/3
                        to_skip = 0
                        if rng:
                            already_skipped = list(reversed(whole_field_range)).index(rng[-1])
                            curpos = whole_field_range.index(rng[-1])
                            if ((curpos + step) > len(whole_field_range)) and (already_skipped < step):
                                to_skip = step - already_skipped
                        rng += list(range(cls.RANGES[field_index][0] + to_skip, high + 1, step))
                    # if we include a range type: Jan-Jan, or Sun-Sun,
                    #  it means the whole cycle (all days of week, # all monthes of year, etc)
                    elif low == high:
                        rng = list(
                            range(
                                cls.RANGES[field_index][0],
                                cls.RANGES[field_index][1] + 1,
                                step,
                            )
                        )
                    else:
                        try:
                            rng = list(range(low, high + 1, step))
                        except ValueError as exc:
                            raise CroniterBadCronError("invalid range: {0}".format(exc))

                    rng = (
                        ["{0}#{1}".format(item, nth) for item in rng]
                        if field_index == DOW_FIELD and nth and nth != "l"
                        else rng
                    )
                    e_list += [a for a in rng if a not in e_list]
                else:
                    if t.startswith("-"):
                        raise CroniterBadCronError(
                            "[{0}] is not acceptable," "negative numbers not allowed".format(expr_format)
                        )
                    if not star_or_int_re.search(t):
                        t = cls._alphaconv(field_index, t, expressions)

                    try:
                        t = int(t)
                    except ValueError:
                        pass

                    t = cls.value_alias(t, field_index, expressions)

                    if t not in ["*", "l"] and (
                        int(t) < cls.RANGES[field_index][0] or int(t) > cls.RANGES[field_index][1]
                    ):
                        raise CroniterBadCronError("[{0}] is not acceptable, out of range".format(expr_format))

                    res.append(t)

                    if field_index == DOW_FIELD and nth:
                        if t not in nth_weekday_of_month:
                            nth_weekday_of_month[t] = set()
                        nth_weekday_of_month[t].add(nth)

            res = set(res)
            res = sorted(res, key=lambda i: "{:02}".format(i) if isinstance(i, int) else i)
            if len(res) == cls.LEN_MEANS_ALL[field_index]:
                # Make sure the wildcard is used in the correct way (avoid over-optimization)
                if (field_index == DAY_FIELD and "*" not in expressions[DOW_FIELD]) or (
                    field_index == DOW_FIELD and "*" not in expressions[DAY_FIELD]
                ):
                    pass
                else:
                    res = ["*"]

            expanded.append(["*"] if (len(res) == 1 and res[0] == "*") else res)

        # Check to make sure the dow combo in use is supported
        if nth_weekday_of_month:
            dow_expanded_set = set(expanded[DOW_FIELD])
            dow_expanded_set = dow_expanded_set.difference(nth_weekday_of_month.keys())
            dow_expanded_set.discard("*")
            # Skip: if it's all weeks instead of wildcard
            if dow_expanded_set and len(set(expanded[DOW_FIELD])) != cls.LEN_MEANS_ALL[DOW_FIELD]:
                raise CroniterUnsupportedSyntaxError(
                    "day-of-week field does not support mixing literal values and nth day of week syntax.  "
                    "Cron: '{}'    dow={} vs nth={}".format(expr_format, dow_expanded_set, nth_weekday_of_month)
                )

        EXPRESSIONS[(expr_format, hash_id, second_at_beginning)] = expressions
        return expanded, nth_weekday_of_month

    @classmethod
    def expand(
        cls,
        expr_format,
        hash_id=None,
        second_at_beginning=False,
        from_timestamp=None,
    ):
        """
        Expand a cron expression format into a noramlized format of
        list[list[int | 'l' | '*']]. The first list representing each element
        of the epxression, and each sub-list representing the allowed values
        for that expression component.

        A tuple is returned, the first value being the expanded epxression
        list, and the second being a `nth_weekday_of_month` mapping.

        Examples:

        # Every minute
        >>> croniter.expand('* * * * *')
        ([['*'], ['*'], ['*'], ['*'], ['*']], {})

        # On the hour
        >>> croniter.expand('0 0 * * *')
        ([[0], [0], ['*'], ['*'], ['*']], {})

        # Hours 0-5 and 10 monday through friday
        >>> croniter.expand('0-5,10 * * * mon-fri')
        ([[0, 1, 2, 3, 4, 5, 10], ['*'], ['*'], ['*'], [1, 2, 3, 4, 5]], {})

        Note that some special values such as nth day of week are expanded to a
        special mapping format for later processing:

        # Every minute on the 3rd tuesday of the month
        >>> croniter.expand('* * * * 2#3')
        ([['*'], ['*'], ['*'], ['*'], [2]], {2: {3}})

        # Every hour on the last day of the month
        >>> croniter.expand('0 * l * *')
        ([[0], ['*'], ['l'], ['*'], ['*']], {})

        # On the hour every 15 seconds
        >>> croniter.expand('0 0 * * * */15')
        ([[0], [0], ['*'], ['*'], ['*'], [0, 15, 30, 45]], {})
        """
        try:
            return cls._expand(
                expr_format,
                hash_id=hash_id,
                second_at_beginning=second_at_beginning,
                from_timestamp=from_timestamp,
            )
        except (ValueError,) as exc:
            if isinstance(exc, CroniterError):
                raise
            if int(sys.version[0]) >= 3:
                trace = _traceback.format_exc()
                raise CroniterBadCronError(trace)
            raise CroniterBadCronError("{0}".format(exc))

    @classmethod
    def _get_low_from_current_date_number(cls, field_index, step, from_timestamp):
        dt = datetime.datetime.fromtimestamp(from_timestamp, tz=UTC_DT)
        if field_index == MINUTE_FIELD:
            return dt.minute % step
        if field_index == HOUR_FIELD:
            return dt.hour % step
        if field_index == DAY_FIELD:
            return ((dt.day - 1) % step) + 1
        if field_index == MONTH_FIELD:
            return dt.month % step
        if field_index == DOW_FIELD:
            return (dt.weekday() + 1) % step

        raise ValueError("Can't get current date number for index larger than 4")

    @classmethod
    def is_valid(
        cls,
        expression,
        hash_id=None,
        encoding="UTF-8",
        second_at_beginning=False,
    ):
        if hash_id:
            if not isinstance(hash_id, (bytes, str)):
                raise TypeError("hash_id must be bytes or UTF-8 string")
            if not isinstance(hash_id, bytes):
                hash_id = hash_id.encode(encoding)
        try:
            cls.expand(expression, hash_id=hash_id, second_at_beginning=second_at_beginning)
        except CroniterError:
            return False
        return True

    @classmethod
    def match(cls, cron_expression, testdate, day_or=True, second_at_beginning=False):
        return cls.match_range(cron_expression, testdate, testdate, day_or, second_at_beginning)

    @classmethod
    def match_range(
        cls,
        cron_expression,
        from_datetime,
        to_datetime,
        day_or=True,
        second_at_beginning=False,
    ):
        cron = cls(
            cron_expression,
            to_datetime,
            ret_type=datetime.datetime,
            day_or=day_or,
            second_at_beginning=second_at_beginning,
        )
        tdp = cron.get_current(datetime.datetime)
        if not tdp.microsecond:
            tdp += relativedelta(microseconds=1)
        cron.set_current(tdp, force=True)
        try:
            tdt = cron.get_prev()
        except CroniterBadDateError:
            return False
        precision_in_seconds = 1 if len(cron.expanded) > UNIX_CRON_LEN else 60
        duration_in_second = (to_datetime - from_datetime).total_seconds() + precision_in_seconds
        return (max(tdp, tdt) - min(tdp, tdt)).total_seconds() < duration_in_second


def croniter_range(
    start,
    stop,
    expr_format,
    ret_type=None,
    day_or=True,
    exclude_ends=False,
    _croniter=None,
    second_at_beginning=False,
    expand_from_start_time=False,
):
    """
    Generator that provides all times from start to stop matching the given cron expression.
    If the cron expression matches either 'start' and/or 'stop', those times will be returned as
    well unless 'exclude_ends=True' is passed.

    You can think of this function as sibling to the builtin range function for datetime objects.
    Like range(start,stop,step), except that here 'step' is a cron expression.
    """
    _croniter = _croniter or croniter
    auto_rt = datetime.datetime
    # type is used in first if branch for perfs reasons
    if type(start) is not type(stop) and not (isinstance(start, type(stop)) or isinstance(stop, type(start))):
        raise CroniterBadTypeRangeError(
            "The start and stop must be same type.  {0} != {1}".format(type(start), type(stop))
        )
    if isinstance(start, (float, int)):
        start, stop = (datetime.datetime.fromtimestamp(t, tzutc()).replace(tzinfo=None) for t in (start, stop))
        auto_rt = float
    if ret_type is None:
        ret_type = auto_rt
    if not exclude_ends:
        ms1 = relativedelta(microseconds=1)
        if start < stop:  # Forward (normal) time order
            start -= ms1
            stop += ms1
        else:  # Reverse time order
            start += ms1
            stop -= ms1
    year_span = math.floor(abs(stop.year - start.year)) + 1
    ic = _croniter(
        expr_format,
        start,
        ret_type=datetime.datetime,
        day_or=day_or,
        max_years_between_matches=year_span,
        second_at_beginning=second_at_beginning,
        expand_from_start_time=expand_from_start_time,
    )
    # define a continue (cont) condition function and step function for the main while loop
    if start < stop:  # Forward

        def cont(v):
            return v < stop

        step = ic.get_next
    else:  # Reverse

        def cont(v):
            return v > stop

        step = ic.get_prev
    try:
        dt = step()
        while cont(dt):
            if ret_type is float:
                yield ic.get_current(float)
            else:
                yield dt
            dt = step()
    except CroniterBadDateError:
        # Stop iteration when this exception is raised; no match found within the given year range
        return


class HashExpander:
    def __init__(self, cronit):
        self.cron = cronit

    def do(self, idx, hash_type="h", hash_id=None, range_end=None, range_begin=None):
        """Return a hashed/random integer given range/hash information"""
        if range_end is None:
            range_end = self.cron.RANGES[idx][1]
        if range_begin is None:
            range_begin = self.cron.RANGES[idx][0]
        if hash_type == "r":
            crc = random.randint(0, 0xFFFFFFFF)
        else:
            crc = binascii.crc32(hash_id) & 0xFFFFFFFF
        return ((crc >> idx) % (range_end - range_begin + 1)) + range_begin

    def match(self, efl, idx, expr, hash_id=None, **kw):
        return hash_expression_re.match(expr)

    def expand(self, efl, idx, expr, hash_id=None, match="", **kw):
        """Expand a hashed/random expression to its normal representation"""
        if match == "":
            match = self.match(efl, idx, expr, hash_id, **kw)
        if not match:
            return expr
        m = match.groupdict()

        if m["hash_type"] == "h" and hash_id is None:
            raise CroniterBadCronError("Hashed definitions must include hash_id")

        if m["range_begin"] and m["range_end"]:
            if int(m["range_begin"]) >= int(m["range_end"]):
                raise CroniterBadCronError("Range end must be greater than range begin")

        if m["range_begin"] and m["range_end"] and m["divisor"]:
            # Example: H(30-59)/10 -> 34-59/10 (i.e. 34,44,54)
            if int(m["divisor"]) == 0:
                raise CroniterBadCronError("Bad expression: {0}".format(expr))

            return "{0}-{1}/{2}".format(
                self.do(
                    idx,
                    hash_type=m["hash_type"],
                    hash_id=hash_id,
                    range_begin=int(m["range_begin"]),
                    range_end=int(m["divisor"]) - 1 + int(m["range_begin"]),
                ),
                int(m["range_end"]),
                int(m["divisor"]),
            )
        elif m["range_begin"] and m["range_end"]:
            # Example: H(0-29) -> 12
            return str(
                self.do(
                    idx,
                    hash_type=m["hash_type"],
                    hash_id=hash_id,
                    range_end=int(m["range_end"]),
                    range_begin=int(m["range_begin"]),
                )
            )
        elif m["divisor"]:
            # Example: H/15 -> 7-59/15 (i.e. 7,22,37,52)
            if int(m["divisor"]) == 0:
                raise CroniterBadCronError("Bad expression: {0}".format(expr))

            return "{0}-{1}/{2}".format(
                self.do(
                    idx,
                    hash_type=m["hash_type"],
                    hash_id=hash_id,
                    range_begin=self.cron.RANGES[idx][0],
                    range_end=int(m["divisor"]) - 1 + self.cron.RANGES[idx][0],
                ),
                self.cron.RANGES[idx][1],
                int(m["divisor"]),
            )
        else:
            # Example: H -> 32
            return str(
                self.do(
                    idx,
                    hash_type=m["hash_type"],
                    hash_id=hash_id,
                )
            )


EXPANDERS = OrderedDict(
    [
        ("hash", HashExpander),
    ]
)