Simplest async/await example possible in Python

Question

To answer your questions I will provide 3 different solutions to the same problem.

case 1: just normal python

import time

def sleep():
    print(f'Time: {time.time() - start:.2f}')
    time.sleep(1)

def sum(name, numbers):
    total = 0
    for number in numbers:
        print(f'Task {name}: Computing {total}+{number}')
        sleep()
        total += number
    print(f'Task {name}: Sum = {total}\n')

start = time.time()
tasks = [
    sum("A", [1, 2]),
    sum("B", [1, 2, 3]),
]
end = time.time()
print(f'Time: {end-start:.2f} sec')

output:

Task A: Computing 0+1
Time: 0.00
Task A: Computing 1+2
Time: 1.00
Task A: Sum = 3

Task B: Computing 0+1
Time: 2.01
Task B: Computing 1+2
Time: 3.01
Task B: Computing 3+3
Time: 4.01
Task B: Sum = 6

Time: 5.02 sec

case 2: async/await done wrong

import asyncio
import time

async def sleep():
    print(f'Time: {time.time() - start:.2f}')
    time.sleep(1)

async def sum(name, numbers):
    total = 0
    for number in numbers:
        print(f'Task {name}: Computing {total}+{number}')
        await sleep()
        total += number
    print(f'Task {name}: Sum = {total}\n')

start = time.time()

loop = asyncio.get_event_loop()
tasks = [
    loop.create_task(sum("A", [1, 2])),
    loop.create_task(sum("B", [1, 2, 3])),
]
loop.run_until_complete(asyncio.wait(tasks))
loop.close()

end = time.time()
print(f'Time: {end-start:.2f} sec')

output:

Task A: Computing 0+1
Time: 0.00
Task A: Computing 1+2
Time: 1.00
Task A: Sum = 3

Task B: Computing 0+1
Time: 2.01
Task B: Computing 1+2
Time: 3.01
Task B: Computing 3+3
Time: 4.01
Task B: Sum = 6

Time: 5.01 sec

case 3: async/await done right (same as case 2 except the sleep function)

import asyncio
import time

async def sleep():
    print(f'Time: {time.time() - start:.2f}')
    await asyncio.sleep(1)

async def sum(name, numbers):
    total = 0
    for number in numbers:
        print(f'Task {name}: Computing {total}+{number}')
        await sleep()
        total += number
    print(f'Task {name}: Sum = {total}\n')

start = time.time()

loop = asyncio.get_event_loop()
tasks = [
    loop.create_task(sum("A", [1, 2])),
    loop.create_task(sum("B", [1, 2, 3])),
]
loop.run_until_complete(asyncio.wait(tasks))
loop.close()

end = time.time()
print(f'Time: {end-start:.2f} sec')

output:

Task A: Computing 0+1
Time: 0.00
Task B: Computing 0+1
Time: 0.00
Task A: Computing 1+2
Time: 1.00
Task B: Computing 1+2
Time: 1.00
Task A: Sum = 3

Task B: Computing 3+3
Time: 2.00
Task B: Sum = 6

Time: 3.01 sec

case 1 with case 2 give the same 5 seconds, whereas case 3 just 3 seconds. So the async/await done right is faster.

The reason for the difference is within the implementation of sleep function.

# case 1
def sleep():
    print(f'Time: {time.time() - start:.2f}')
    time.sleep(1)

# case 2
async def sleep():
    print(f'Time: {time.time() - start:.2f}')
    time.sleep(1)

# case 3
async def sleep():
    print(f'Time: {time.time() - start:.2f}')
    await asyncio.sleep(1)

sleep function in case 1 and case 2 are the "same". They "sleep" without allowing others to use the resources. Whereas case 3 allows access to the resources when it is asleep.

In case 2 we added async to the normal function. However the event loop will run it without interruption. Why? Because we didn't tell where the loop is allowed to interrupt your function to run another task.

In case 3 we told the event loop exactly where to interrupt the function to run another task. Where exactly?

# case 3
async def sleep():
    print(f'Time: {time.time() - start:.2f}')
    await asyncio.sleep(1) # <-- Right here!

More on this read here

Update 02/May/2020

Consider reading

• A Hitchhikers Guide to Asynchronous Programming
• Asyncio Futures and Coroutines

is it possible to give a simple example showing how async / await works, by using only these two keywords + asyncio.get_event_loop() + run_until_complete + other Python code but no other asyncio functions?

This way it's possible to write code that works:

import asyncio


async def main():
    print('done!')


if __name__ ==  '__main__':
    loop = asyncio.get_event_loop()
    loop.run_until_complete(main())

But this way it's impossible to demonstrate why you need asyncio.

By the way, why do you need asyncio, not just plain code? Answer is - asyncio allows you to get performance benefit when you parallelize I/O blocking operations (like reading/writing to network). And to write useful example you need to use async implementation of those operations.

Please read this answer for more detailed explanation.

Upd:

ok, here's example that uses asyncio.sleep to imitate I/O blocking operation and asyncio.gather that shows how you can run multiple blocking operations concurrently:

import asyncio


async def io_related(name):
    print(f'{name} started')
    await asyncio.sleep(1)
    print(f'{name} finished')


async def main():
    await asyncio.gather(
        io_related('first'),
        io_related('second'),
    )  # 1s + 1s = over 1s


if __name__ ==  '__main__':
    loop = asyncio.get_event_loop()
    loop.run_until_complete(main())

Output:

first started
second started
first finished
second finished
[Finished in 1.2s]

Note how both io_related started then, after only one second, both done.

Python 3.7+ now has a simpler API (in my opinion) with a simpler wording (easier to remember than "ensure_future"): you can use create_task which returns a Task object (that can be useful later to cancel the task if needed).

Basic example 1

import asyncio

async def hello(i):
    print(f"hello {i} started")
    await asyncio.sleep(4)
    print(f"hello {i} done")

async def main():
    task1 = asyncio.create_task(hello(1))  # returns immediately, the task is created
    await asyncio.sleep(3)
    task2 = asyncio.create_task(hello(2))
    await task1
    await task2

asyncio.run(main())  # main loop

Result:

hello 1 started
hello 2 started
hello 1 done
hello 2 done

---

Basic example 2

If you need to get the return value of these async functions, then gather is useful. The following example is inspired from the documentation, but unfortunately the doc doesn't show what gather is really useful for: getting the return values!

import asyncio

async def factorial(n):
    f = 1
    for i in range(2, n + 1):
        print(f"Computing factorial({n}), currently i={i}...")
        await asyncio.sleep(1)
        f *= i
    return f

async def main():
    L = await asyncio.gather(factorial(2), factorial(3), factorial(4))
    print(L)  # [2, 6, 24]

asyncio.run(main())

Expected output:

Computing factorial(2), currently i=2...
Computing factorial(3), currently i=2...
Computing factorial(4), currently i=2...
Computing factorial(3), currently i=3...
Computing factorial(4), currently i=3...
Computing factorial(4), currently i=4...
[2, 6, 24]

---

PS: even if you use asyncio, and not trio, the tutorial of the latter was helpful for me to grok Python asynchronous programming.

Since everything is nicely explained, then let's run some examples with event loops compare synchronous code to asynchronous code.

synchronous code:

import time

def count():
    time.sleep(1)
    print('1')
    time.sleep(1)
    print('2')
    time.sleep(1)
    print('3')

def main():
    for i in range(3):
        count()

if __name__ == "__main__":
    t = time.perf_counter()
    main()
    t2 = time.perf_counter()
    
    print(f'Total time elapsed: {t2:0.2f} seconds')

output:

1
2
3
1
2
3
1
2
3
Total time elapsed: 9.00 seconds

We can see that each cycle of count running to completion before the next cycle begins.

asynchronous code:

import asyncio
import time

async def count():
    await asyncio.sleep(1)
    print('1')
    await asyncio.sleep(1)
    print('2')
    await asyncio.sleep(1)
    print('3')

async def main():
    await asyncio.gather(count(), count(), count())

if __name__ == "__main__":
    t = time.perf_counter()
    asyncio.run(main())
    t2 = time.perf_counter()

    print(f'Total time elapsed: {t2:0.2f} seconds')

output:

1
1
1
2
2
2
3
3
3
Total time elapsed: 3.00 seconds

The asynshonous equivalent on the other hand looks somting like this took three seconds to run as opposed to nine secounds. The first count cycle was started and as soon as it hit the awaits sleep one Python was free to do other work, for instance starting the secound and subsequently the third count cycles. This is why we have all the ones than all tubes then all three. In the output programing concurrently can be a very valuable tool. Multiprocessing has the operating do all of the multitasking work and in Python it's the only option for multi-core concurrency that is having your program executed on multiple cores of CPU. If use threads then the operating system is still doing all of the multitasking work and in cpython the global intrepeter lock prevents multi-core concurrency in asynshonous programming. There is no operating system intervention there's one process there's one thread so what's going on well tasks can release the CPU when there are waiting periods, so that other task can use it.

import asyncio

loop = asyncio.get_event_loop()


async def greeter(name):
    print(f"Hi, {name} you're in a coroutine.")

try:
    print('starting coroutine')
    coro = greeter('LP')
    print('entering event loop')
    loop.run_until_complete(coro)
finally:
    print('closing event loop')
    loop.close()

output:

starting coroutine
entering event loop
Hi, LP you're in a coroutine.
closing event loop

Asynchronous frameworks need a scheduler usually called an event loop. This event loop keeps track of all the running tasks and when a function suspended it returns control to the event loop which then will find another function to start or resume and this is called cooperative multitasking. Async IO provides a framework an asynchronous framework that's centered on this event loop and it efficiently handles input/output events an application interacts with the event loop explicitly it registers code to be run and then it lets the event loop the scheduler make the necessary calls into application code when the resources are available. So, if a network server open sockets and then registers them to be told when input events occur on them the event loop will alert the server code when there's a new incoming connection or when there's data to be read. If there's no more data to be read from a socket than the server then yields control back to the event loop.

The mechanism from yielding control back to the event loop depends on co-routines co-routines are a language construct designed for concurrent operation. The co-routine can pause execution using the awake keyword with another co-routine and while it's paused the co-routine state is maintained allowing it to resume where it left off one co-routine can start another and then wait for the results and this makes it easier to decompose a task into reusable parts.

import asyncio

loop = asyncio.get_event_loop()

async def outer():
    print('in outer')
    print('waiting for result 1')
    result1 = await phase1()
    print('waiting for result 2')
    result2 = await phase2(result1)
    return result1, result2


async def phase1():
    print('in phase1')
    return 'phase1 result'

async def phase2(arg):
    print('in phase2')
    return 'result2 derived from {}'.format(arg)

asyncio.run(outer())

output:

in outer
waiting for result 1
in phase1
waiting for result 2
in phase2

This example asks two phases that must be executed in order but that can run concurrently with other operations. The awake keyword is used instead of addingbthe new co-routines to the loop because control flow is already inside of a co-routine being managed by the loop. It isn't necessary to tell the loop to manage the new co-routines.

I don't know why but all of explanations on this topic are too complex or they are using examples with useless asyncio.sleep()... So far the best code sample that I found is this: https://codeflex.co/python3-async-await-example/