In Python, how can I put a thread to sleep until a specific time?

I know that I can cause a thread to sleep for a specific amount of time with:

time.sleep(NUM)

How can I make a thread sleep until 2AM? Do I have to do math to determine the number of seconds until 2AM? Or is there some library function?

( Yes, I know about cron and equivalent systems in Windows, but I want to sleep my thread in python proper and not rely on external stimulus or process signals.)

Here's a half-ass solution that doesn't account for clock jitter or adjustment of the clock. See comments for ways to get rid of that.

import time
import datetime

# if for some reason this script is still running
# after a year, we'll stop after 365 days
for i in xrange(0,365):
    # sleep until 2AM
    t = datetime.datetime.today()
    future = datetime.datetime(t.year,t.month,t.day,2,0)
    if t.hour >= 2:
        future += datetime.timedelta(days=1)
    time.sleep((future-t).total_seconds())
    
    # do 2AM stuff

import pause
from datetime import datetime

pause.until(datetime(2015, 8, 12, 2))

One possible approach is to sleep for an hour. Every hour, check if the time is in the middle of the night. If so, proceed with your operation. If not, sleep for another hour and continue.

If the user were to change their clock in the middle of the day, this approach would reflect that change. While it requires slightly more resources, it should be negligible.

Another approach, using sleep, decreasing the timeout logarithmically.

def wait_until(end_datetime):
    while True:
        diff = (end_datetime - datetime.now()).total_seconds()
        if diff < 0: return       # In case end_datetime was in past to begin with
        time.sleep(diff/2)
        if diff <= 0.1: return

Building on the answer of @MZA and the comment of @Mads Y

I tried the "pause" pacakage. It does not work for Python 3.x. From the pause package I extracted the code required to wait until a specific datetime and made the following def.

def wait_until(execute_it_now):
    while True:
        diff = (execute_it_now - datetime.now()).total_seconds()
        if diff <= 0:
            return
        elif diff <= 0.1:
            time.sleep(0.001)
        elif diff <= 0.5:
            time.sleep(0.01)
        elif diff <= 1.5:
            time.sleep(0.1)
        else:
            time.sleep(1)

Slightly more generalized solution (based off of Ross Rogers') in case you'd like to add minutes as well.

def sleepUntil(self, hour, minute):
    t = datetime.datetime.today()
    future = datetime.datetime(t.year, t.month, t.day, hour, minute)
    if t.timestamp() > future.timestamp():
        future += datetime.timedelta(days=1)
    time.sleep((future-t).total_seconds())

Slightly beside the point of the original question:

Even if you don't want to muck around with crontabs, if you can schedule python scripts to those hosts, you might be interested to schedule anacron tasks? anacron's major differentiator to cron is that it does not rely the computer to run continuously. Depending on system configuration you may need admin rights even for such user-scheduled tasks.

A similar, more modern tool is upstart provided by the Ubuntu folks: http://upstart.ubuntu.com/ This does not yet even have the required features. But scheduling jobs and replacing anacron is a planned feature. It has quite some traction due to its usage as Ubuntu default initd replacement. (I am not affiliated with the project)

Of course, with the already provided answer, you can code the same functionality into your python script and it might suit you better in your case.

Still, for others, anacron or similar existing systems might be a better solution. anacron is preinstalled on many current linux distributions (there are portability issues for windows users).

Wikipedia provides a pointer page: https://en.wikipedia.org/wiki/Anacron

If you do go for a python version I'd look at the asynchronous aspect, and ensure the script works even if the time is changed (daylight savings, etc) as others have commented already. Instead of waiting til a pre-calculated future, I'd always at maximum wait one hour, then re-check the time. The compute cycles invested should be negligible even on mobile, embedded systems.

adapt this:

from datetime import datetime, timedelta
from time import sleep

now = datetime.utcnow
to = (now() + timedelta(days = 1)).replace(hour=1, minute=0, second=0)
sleep((to-now()).seconds)

I know is way late for this, but I wanted to post an answer (inspired on the marked answer) considering systems that might have - incorrect - desired timezone + include how to do this threaded for people wondering how.

It looks big because I'm commenting every step to explain the logic.

import pytz #timezone lib
import datetime
import time
from threading import Thread

# using this as I am, check list of timezone strings at:
## https://en.wikipedia.org/wiki/List_of_tz_database_time_zones
TIMEZONE = pytz.timezone("America/Sao_Paulo")

# function to return desired seconds, even if it's the next day
## check the bkp_time variable (I use this for a bkp thread)
## to edit what time you want to execute your thread
def get_waiting_time_till_two(TIMEZONE):
    # get current time and date as our timezone
    ## later we remove the timezone info just to be sure no errors
    now = datetime.datetime.now(tz=TIMEZONE).replace(tzinfo=None)
    curr_time = now.time()
    curr_date = now.date()
    
    # Make 23h30 string into datetime, adding the same date as current time above
    bkp_time = datetime.datetime.strptime("02:00:00","%H:%M:%S").time()
    bkp_datetime = datetime.datetime.combine(curr_date, bkp_time)
    
    # extract the difference from both dates and a day in seconds
    bkp_minus_curr_seconds = (bkp_datetime - now).total_seconds()
    a_day_in_seconds = 60 * 60 * 24
    
    # if the difference is a negative value, we will subtract (- with + = -)
    # it from a day in seconds, otherwise it's just the difference
    # this means that if the time is the next day, it will adjust accordingly
    wait_time = a_day_in_seconds + bkp_minus_curr_seconds if bkp_minus_curr_seconds < 0 else bkp_minus_curr_seconds
    
    return wait_time

# Here will be the function we will call at threading
def function_we_will_thread():
    # this will make it infinite during the threading
    while True:
        seconds = get_waiting_time_till_two(TIMEZONE)
        time.sleep(seconds)
        # Do your routine

# Now this is the part where it will be threading
thread_auto_update = Thread(target=function_we_will_thread)
thread_auto_update.start()

It takes only one of the very basic libraries.

import time

sleep_until = 'Mon Dec 25 06:00:00 2020' # String format might be locale dependent.
print("Sleeping until {}...".format(sleep_until))
time.sleep(time.mktime(time.strptime(sleep_until)) - time.time())

• time.strptime() parses the time from string -> struct_time tuple. The string can be in different format, if you give strptime() parse-format string as a second argument. E.g. time.strptime("12/25/2020 02:00AM", "%m/%d/%Y %I:%M%p")
• time.mktime() turns the struct_time -> epoch time in seconds.
• time.time() gives current epoch time in seconds.
• Substract the latter from the former and you get the wanted sleep time in seconds.
• sleep() the amount.

If you just want to sleep until whatever happens to be the next 2AM, (might be today or tomorrow), you need an if-statement to check if the time has already passed today. And if it has, set the wake up for the next day instead.

import time

sleep_until = "02:00AM" # Sets the time to sleep until.

sleep_until = time.strftime("%m/%d/%Y " + sleep_until, time.localtime()) # Adds todays date to the string sleep_until.
now_epoch = time.time() #Current time in seconds from the epoch time.
alarm_epoch = time.mktime(time.strptime(sleep_until, "%m/%d/%Y %I:%M%p")) # Sleep_until time in seconds from the epoch time.

if now_epoch > alarm_epoch: #If we are already past the alarm time today.
    alarm_epoch = alarm_epoch + 86400 # Adds a day worth of seconds to the alarm_epoch, hence setting it to next day instead.

time.sleep(alarm_epoch - now_epoch) # Sleeps until the next time the time is the set time, whether it's today or tomorrow.

⏱ Asynchronous version of Omrii's solution

import datetime
import asyncio

async def sleep_until(hour: int, minute: int, second: int):
    """Asynchronous wait until specific hour, minute and second

    Args:
        hour (int): Hour
        minute (int): Minute
        second (int): Second

    """
    t = datetime.datetime.today()
    future = datetime.datetime(t.year, t.month, t.day, hour, minute, second)
    if t.timestamp() > future.timestamp():
        future += datetime.timedelta(days=1)
    await asyncio.sleep((future - t).total_seconds())

from datetime import datetime
import time, operator
time.sleep([i[0]*3600 + i[1]*60 for i in [[H, M]]][0] - [i[0]*3600 + i[1]*60 for i in [map(int, datetime.now().strftime("%H:%M").split(':'))]][0])

Instead of using the wait() function, you can use a while-loop checking if the specified date has been reached yet:

if datetime.datetime.utcnow() > next_friday_10am:
    # run thread or whatever action
    next_friday_10am = next_friday_10am()
    time.sleep(30)

def next_friday_10am():
    for i in range(7):
        for j in range(24):
            for k in range(60):
                if (datetime.datetime.utcnow() + datetime.timedelta(days=i)).weekday() == 4:
                    if (datetime.datetime.utcnow() + datetime.timedelta(days=i, hours=j)).hour == 8:
                        if (datetime.datetime.utcnow() + datetime.timedelta(days=i, hours=j, minutes=k)).minute == 0:
                            return datetime.datetime.utcnow() + datetime.timedelta(days=i, hours=j, minutes=k)

Still has the time-checking thread check the condition every after 30 seconds so there is more computing required than in waiting, but it's a way to make it work.