totaltimeout 1.0.2

Spread one timeout over many operations

Spread one timeout over many operations.

Correctly and efficiently spreads one timeout over many steps by recalculating the time remaining after some amount of waiting has already happened, to pass an adjusted timeout to the next step.

Versioning

This library’s version numbers follow the SemVer 2.0.0 specification.

The current version number is available in the variable __version__, as is normal for Python modules.

Installation

pip install totaltimeout

Usage

Import the Timeout class.

from totaltimeout import Timeout

Waiting in a “timed loop” for an API with retries (useful for unreliable APIs that may either hang or need retries):

for time_left in Timeout(SOME_NUMBER_OF_SECONDS):
     reply = requests.get(some_flaky_api_url, timeout=time_left)
     if reply.status == 200:
         break

Same as above, but with a wait between retries:

timeout = Timeout(SOME_NUMBER_OF_SECONDS)
for time_left in timeout:
     reply = requests.get(some_flaky_api_url, timeout=time_left)
     if reply.status == 200:
         break
     if timeout.time_left() <= RETRY_DELAY:
         break
     time.sleep(RETRY_DELAY)

Waiting for multiple tasks to finish:

timeout = Timeout(10.0)
my_thread_foo.join(timeout.time_left())
my_thread_bar.join(timeout.time_left())
my_thread_qux.join(timeout.time_left())
# Wait only as long as the slowest
# thread to finish, as if they all
# got a 10 second wait in parallel.

Waiting for multiple tasks within each iteration of a “timed loop”:

timeout = Timeout(SOME_NUMBER_OF_SECONDS)
for time_left in timeout:
     foo.some_work(timeout=time_left)
     # The first timeout can be *either* be the for loop value or
     # the ``time_left()`` method. The rest *have to be* the latter.
     foo.some_more_work(timeout=timeout.time_left())
     some_other_work(timeout=timeout.time_left())

Explanation

If you’re confused about what’s going on, run this example program:

from time import sleep

from totaltimeout import Timeout

def demo(timeout_in_seconds):
    timeout = Timeout(timeout_in_seconds)
    for time_left in timeout:
        print(time_left)
        sleep(1)
        print(timeout.time_left())
        sleep(1)

if __name__ == '__main__':
    demo(10)

You should get output kinda like this:

9.99990844912827
8.996184696443379
7.992705063894391
6.990415567532182
5.983945298939943
4.981594786979258
3.979213748127222
2.9768632212653756
1.9745127055794
0.9699955033138394

Advanced Usage Notes

Timeout uses time.monotonic as the default time function, falling back to time.time if time.monotonic is unavailable.

You can override this by passing in a callable as the second argument.

For example, if you’ve installed the monotonic backport:

from monotonic import monotonic

timeout = Timeout(10.0, now=monotonic)

Any callables that return time in seconds as floating point values are supported as part of the interface subject to SemVer backwards compability guarantees.

However, any callables that return time values that can be subtracted from each other to produce duration values which in turn can be subtracted from each other and compared to zero should work, and seconds are expected only because Python’s idiomatic unit for timeouts is seconds. If the timeout, now, and usage are consistent, any choice that fits these criteria is likely to work.