krolib 1.2.1

DSL to handle complex schedules

⏰ Krolib: Schedules for Humans

Magic library and DSL to handle complex schedules.

🚀 Installation

As easy as usual.

$ pip install krolib

Why?

Cron is not enough

Yes, you can create almost any kind of schedule rotation or delay for you app but there are some very special cases when Cron can't help you, like "do something monthly, every second weekend at 6am until 2020". Also, you can create only app-level job, what if you need periodical function execution inside your application?

Declarative schedule format

Create your schedules programmatically, serialize them, store in DB, modify, update in runtime and always have a whole picture of what is the final result.

Readable and flexible structure like:

{
    'timezone': 'Europe/Kiev',
    'start': {
        'relative_timeshift': {
            'delay': 2,
            'time_units': TimeUnits.MONTHS,
        }
    }
}

Supports asyncio out of the box

Works like magic and very helpful to create periodical coroutines:

from krolib.asyncio import scheduler
from krolib.structs import TimeUnits, PeriodicalUnits


@scheduler({
    'start': {
        'relative_timeshift': {
            'delay': 1,
            'time_units': TimeUnits.SECONDS,
        }
    },
    'periodical': {
        'repeats': PeriodicalUnits.SECONDLY,
        'every': 1,
    },
    'stop': {
        'never': False,
        'after_num_repeats': 2
    }
})
async def some_coroutine():
    print('PING')

await some_coroutine()  # will be delayed and called twice!

Delay several coroutines concurrently:

from krolib.asyncio import scheduler
from krolib.structs import TimeUnits, PeriodicalUnits


@scheduler({
    'start': {
        'relative_timeshift': {
            'delay': 1,
            'time_units': TimeUnits.SECONDS,
        }
    },
    'periodical': {
        'repeats': PeriodicalUnits.SECONDLY,
        'every': 1,
    },
    'stop': {
        'never': False,
        'after_num_repeats': 2
    }
})
async def some_coroutine():
    print('PING')

@scheduler({
    'start': {
        'relative_timeshift': {
            'delay': 1,
            'time_units': TimeUnits.SECONDS,
        }
    },
    'periodical': {
        'repeats': PeriodicalUnits.SECONDLY,
        'every': 1,
    },
    'stop': {
        'never': False,
        'after_num_repeats': 2
    }
})
async def another_coroutine():
    print('PONG')

# concurrent execution, get your PING PONG twice!
await asyncio.gather(some_coroutine(), another_coroutine())

More examples

Delay for 1 hour:

from krolib.utils import just_now
from krolib.parser import schedule_parser
from krolib.structs import TimeUnits

now = just_now()
schedule = {
    'start': {
        'relative_timeshift': {
            'delay': 1,
            'time_units': TimeUnits.HOURS,
        }
    },
}
schedule_gen = schedule_parser(schedule)
result = next(schedule_gen)
assert (result - now).total_seconds() == 3600  # seconds

Every last day of the month, infinite:

import datetime
import pytz

from krolib.structs import (
    PeriodicalUnits,
    RelativeUnits,
    RelativeIndexUnits,
)
from krolib.parser import schedule_parser

# today is this date, for example
now = datetime.datetime(year=2018, month=1, day=1)

schedule = {
    'periodical': {
        'repeats': PeriodicalUnits.MONTHLY,
        'relative_day': RelativeUnits.DAY,
        'relative_day_index': RelativeIndexUnits.LAST,
    }
}
schedule_gen = schedule_parser(schedule, now_dt=now)
results = [next(schedule_gen) for _ in range(3)]
assert results == [
    datetime.datetime(2018, 1, 31, 0, 0, tzinfo=pytz.UTC),
    datetime.datetime(2018, 2, 28, 0, 0, tzinfo=pytz.UTC),
    datetime.datetime(2018, 3, 31, 0, 0, tzinfo=pytz.UTC),
]

Since tomorrow, every week with stop after the second one:

import datetime

from krolib.utils import just_now
from krolib.parser import schedule_parser
from krolib.structs import PeriodicalUnits

now = just_now()
start_on = now + datetime.timedelta(days=1)
schedule = {
    'start': {
        'on': start_on,
    },
    'periodical': {
        'repeats': PeriodicalUnits.WEEKLY,
        'every': 1,
    },
    'stop': {
        'never': False,
        'after_num_repeats': 2
    }
}
schedule_gen = schedule_parser(schedule)
one_dt, two_dt = list(schedule_gen)
assert one_dt == start_on
assert (two_dt - one_dt).days == 7

Schedule Format

The general schema structure which is currently supported by Krolib is presented here as a pseudo-Python code.

{
   'start': {
       'on': datetime.datetime,
       'relative_timeshift': {
           'delay': int,
           'time_units': 'seconds/minutes/hours/days/weeks/months',
       }
   },
   'periodical': {
       'repeats': 'yearly/monthly/weekly/daily/hourly/minutely/secondly',
       'every': int,
       'month': 1..12 or None,
       'day': 1..31 or None,
       'weekday': [0..6] or None,
       'hour': 0..23 or None,
       'minute': 0..59 or None,
       'second': 0..59 or None,
   },
   'stop': {
       'never': True or False,
       'on': datetime.datetime,
       'after_num_repeats': int
   },
   'timezone': str or None,
}

Also, there is special form for relative periodical rotations:

{
   'start': {
       'on': datetime.datetime,
       'relative_timeshift': {
           'delay': int,
           'time_units': 'seconds/minutes/hours/days/weeks/months',
       }
   },
   'periodical': {
       'repeats': 'yearly/monthly',
       'relative_day': 'day/weekday/weekend/sunday/monday/tuesday/wednesday/thursday/friday/saturday',
       'relative_day_index': 'first/second/third/fourth/last',
       'every': int,
       'hour': 0..23 or None,
       'minute': 0..59 or None,
       'second': 0..59 or None,
   },
   'stop': {
       'never': True or False,
       'on': datetime.datetime,
       'after_num_repeats': int
   },
   'timezone': str or None,
}

Schedule structure

The schedule data structure has three main sections:

• start
• periodical
• stop

The schedule structure also holds a timezone attribute, which sets the timezone for all the sections of the schedule. It should hold a string with the timezone identifier, for example Europe/Kiev. Supported list of the timezones could be found here

Please take into account, that timezone validation and processing utilizes pytz library, so finally you might want to dig into the sources of this library in case if you face any troubles.

Start section

Start section is used to specify the exact time or delay before the rule or block starts/continues it's execution (processing).

on key

This key holds the exact time (datetime object) when to start the execution/processing. When this key is set, values in the relative_timeshift section should not be filled with any values.

{
    'start': {
        'on': datetime.datetime.utcnow(),
    }
}

relative_timeshift key

This key holds an object with two keys — delay and time_units.

delay holds a string with an integer in it for the amount of time units.

time_units holds a string with one of the values:

• seconds
• minutes
• hours
• days
• weeks
• months

When the relative timeshift is set, both delay and time_units should hold the proper values.

Example of this kind of start section with the delay for 3 days:

{
    'start': {
        'relative_timeshift': {
            'delay': '3',
            'time_units': 'days',
        }
    }
}

Periodical section

The periodical section is used to describe a schedule with the recurring events. If you use this section, you should also set the values in the stop section of the schedule data structure.

repeats key

This key holds the type of repeats:

• yearly
• monthly
• weekly
• daily
• hourly
• minulety
• secondly

Example of yearly repeats on September, 3-rd at 20:30:

{
    'periodical': {
        'repeats': 'yearly',
        'every': 1,
        'month': 9,
        'day': 3,
        'hour': 20,
        'minute': 30,
    }
}

every key

This key gives additional flexibility for repeats. It holds a non-negative integer greater than zero (a natural number) and is used to calculate skips.

For example, for daily repeats the every key could be set to 1 — this would mean every day repeats, to 3 — this would mean skip two days, then execute every third day, skip two days again, etc.

{
    'periodical': {
        'repeats': 'daily',
        'every': 5,
        'hour': 20,
        'minute': 30,
    }
}

month, day, hour, minute and second keys

These keys hold integers which correspond to appropriate key. For example, 1-12 (inclusive) for month, 1-31 (inclusive) for day and so on.

weekday key

This key holds the information about the weekdays for the recurring event. Valid weekday range is from 0 to 6 where 0 is Monday and 6 is Sunday.

Example of weekly repeats (every week without skipping) on Mondays and Fridays at 15.00:

{
    'periodical': {
        'repeats': 'weekly',
        'every': 1,
        'weekday': [0, 4],
        'hour': 15,
        'minute': 0,
    }
}

Relative days case

For monthly and yearly values of the repeats key, altenative relative days schedule could be set.

In this case the day and weekday keys should not be set and relative_day and relative_day_index should be. These keys indicate some relative day of the month.

relative_day key

This field should hold the string with one of the following values:

• day
• weekday
• weekend
• sunday
• monday
• tuesday
• wednesday
• thursday
• friday
• saturday

relative_day_index key

This field should hold a string with one of the following values:

• first
• second
• third
• fourth
• last

Example of monthly repeats on the third Friday of the month at 15.00:

{
    'periodical': {
        'repeats': 'monthly',
        'every': 1,
        'hour': 15,
        'minute': 0,
        'relative_day': 'friday',
        'relative_day_index': 'third'
    }
}

Stop section

This section is obligatory. It holds information about when to stop the recurring event execution. Has never, on and after_num_repeats keys.

never key

This key holds a boolean value. If set to True — other keys of this section should not hold any values, because this means that recurring event will never stop it's execution. If set to False — other keys should be also set.

Stop section with never ending repeats:

{
    'stop': {
        'never': True,
    }
}

on key

This key holds exact time (datetime object) when to stop the recurring:

{
    'stop': {
        'never': False,
        'on': datetime.datetime.utcnow() + datetime.timedelta(days=1)
    }
}

after_num_repeats key

This key holds an integer indicating amount of times to repeat the recurring schedule.

Stop section with stop after 25 times executing/processing smth:

{
    'stop': {
        'never': False,
        'after_num_repeats': 25
    }
}

🤝 Special Thanks

• Alexander Omyshev @akalex
• Dmitry Nikonenko @ndmytro
• Phil Steitz @psteitz
• Ralph Goers @rgoers
• Vitalii Iaskal @vavilon17

📝 License

Published under Apache Software License 2.0, see LICENSE file.