13.2. AsyncIO About

• asyncio in Python standard library

• async and await builtin keywords

• Running asynchronously: 3s + 1s + 1s = bit over 3s [execution time]

• Async is the future of programming

13.2.1. Advantages

• Maximize the usage of a single thread

• Handling I/O asynchronously

• Enabling concurrent code using coroutines

• Async will fill the gaps, otherwise wasted on waiting for I/O

• You control when tasks switches occur, so locks and other synchronization are no longer needed

• Async is the cheapest way to task switch

• Cost task switches is incredibly low; calling a pure Python function has more overhead than restarting a generator or awaitable

• Function builds stack each time it's called, whereas async uses generators underneath, which already has stack created

• In terms of speed async servers blows threaded servers in means of thousands

• Async is very cheap in means of resources

• Async world has a huge ecosystem of support tools

• Coding is easier to get right, than threads

13.2.2. Disadvantages

• Async switches cooperatively, so you do need to add explicit code yield or await to cause a task to switch

• Everything you do need a non-blocking version (for example open())

• Increased learning curve

• Create event loop, acquire, crate non-blocking versions of your code

• You think you know Python, there is a second half to learn (async)

13.2.3. Sync vs Async

Figure 13.8. Source: Michael Kennedy [1]

Figure 13.9. Source: Michael Kennedy [1]

Figure 13.10. Source: Langa, Ł. import asyncio: Learn Python's AsyncIO [4]

Figure 13.11. Source: Langa, Ł. import asyncio: Learn Python's AsyncIO [4]

13.2.4. Execution

Figure 13.12. Source: Michael Kennedy [1]

Figure 13.13. Source: Michael Kennedy [1]

13.2.5. Example

>>> import asyncio
>>>
>>>
>>> async def a():
...     print('a: started')
...     await asyncio.sleep(0.2)
...     print('a: finished')
...     return 'a'
>>>
>>> async def b():
...     print('b: started')
...     await asyncio.sleep(0.1)
...     print('b: finished')
...     return 'b'
>>>
>>> async def c():
...     print('c: started')
...     await asyncio.sleep(0.3)
...     print('c: finished')
...     return 'c'
>>>
>>> async def main():
...     result = await asyncio.gather(a(), b(), c())
...     print(f'Result: {result}')
>>>
>>>
>>> asyncio.run(main())
a: started
b: started
c: started
b: finished
a: finished
c: finished
Result: ['a', 'b', 'c']

13.2.6. Further Reading

• Kennedy, M. Demystifying Python's Async and Await Keywords [1]

• Kennedy, M. Async Techniques and Examples in Python [2]

• Abdalla, A. Creating a Bittorrent Client using Asyncio [3]

• Langa, Ł. import asyncio: Learn Python's AsyncIO [4]

13.2.7. References

[1] (1,2,3,4,5)

Kennedy, M. Demystifying Python's Async and Await Keywords. Publisher: JetBrainsTV. Year: 2019. Retrieved: 2022-03-10. URL: https://www.youtube.com/watch?v=F19R_M4Nay4

[2]

Kennedy, M. Async Techniques and Examples in Python Course. Publisher: TalkPython. Year: 2022. Retrieved: 2022-03-10. URL: https://talkpython.fm/async

[3]

Abdalla, A. Creating a Bittorrent Client using Asyncio. Year: 2017. Retrieved: 2022-03-10. URL: https://www.youtube.com/watch?v=Pe3b9bdRtiE

[4] (1,2,3)

Langa, Ł. import asyncio: Learn Python's AsyncIO. Year: 2020. Retrieved: 2022-03-10. URL: https://www.youtube.com/playlist?list=PLhNSoGM2ik6SIkVGXWBwerucXjgP1rHmB

13.2.8. Assignments

Code 13.32. Solution

"""
* Assignment: OOP Async Coroutine
* Complexity: easy
* Lines of code: 2 lines
* Time: 2 min

English:
    1. Define:
        a. coroutine function `a()`
    2. After running coroutine should:
        b. return 'a'
    3. Run doctests - all must succeed

Polish:
    1. Zdefiniuj:
        a. coroutine function `a()`
    2. Po uruchomieniu coroutine powinna:
        a. zwracać 'a'
    3. Uruchom doctesty - wszystkie muszą się powieść

Tests:
    >>> import sys; sys.tracebacklimit = 0
    >>> from inspect import iscoroutine, iscoroutinefunction
    >>> import asyncio

    >>> assert iscoroutinefunction(a)
    >>> assert iscoroutine(a())

    >>> asyncio.run(a())
    'a'
"""


# Coroutine function `a()`
# Function should return 'a'
# type: Coroutine
def a():
    ...

Code 13.33. Solution

"""
* Assignment: OOP Async Sleep
* Complexity: easy
* Lines of code: 3 lines
* Time: 2 min

English:
    1. Define:
        a. coroutine function `a()`
    2. After running coroutine should:
        a. wait for 1.0 seconds
        b. return 'a'
    3. Run doctests - all must succeed

Polish:
    1. Zdefiniuj:
        a. coroutine function `a()`
    2. Po uruchomieniu coroutine powinna:
        a. czekać 1.0 sekundę
        b. zwracać 'a'
    3. Uruchom doctesty - wszystkie muszą się powieść

Hint:
    * asyncio.sleep()

Tests:
    >>> import sys; sys.tracebacklimit = 0
    >>> from inspect import iscoroutine, iscoroutinefunction
    >>> import asyncio

    >>> assert iscoroutinefunction(a)
    >>> assert iscoroutine(a())

    >>> asyncio.run(a())
    'a'
"""

import asyncio

# coroutine function `a()`
# wait for 1.0 seconds, return 'a'
# type: Coroutine
def a():
    ...

Code 13.34. Solution

"""
* Assignment: OOP Async Concurrent
* Complexity: easy
* Lines of code: 15 lines
* Time: 5 min

English:
    1. Define:
        a. coroutine function `a()`
        b. coroutine function `b()`
        c. coroutine function `c()`
    2. After running coroutine should:
        a. print 'NAME: before'
        b. wait for X seconds
        c. print 'NAME: after'
        d. return NAME, where
    3. Definition of NAME:
        a. for coroutine `a()`, NAME is 'a'
        a. for coroutine `b()`, NAME is 'b'
        a. for coroutine `c()`, NAME is 'c'
    4. Definition of X:
        a. for coroutine `a()`, X is 1.0
        a. for coroutine `b()`, X is 0.5
        a. for coroutine `c()`, X is 1.5
    3. Run doctests - all must succeed

Polish:
    1. Zdefiniuj:
        a. coroutine function `a()`
        b. coroutine function `b()`
        c. coroutine function `c()`
    2. Po uruchomieniu coroutine powinna:
        a. wyświetlić 'NAME: before'
        b. czekać X sekund
        c. wyświetlić 'NAME: after'
        d. zwrócić NAME
    3. Definicja NAME:
        a. dla coroutine `a()`, NAME to 'a'
        a. dla coroutine `b()`, NAME to 'b'
        a. dla coroutine `c()`, NAME to 'c'
    3. Definicja X:
        a. dla coroutine `a()`, X to 1.0
        a. dla coroutine `b()`, X to 0.5
        a. dla coroutine `c()`, X to 1.5
    3. Uruchom doctesty - wszystkie muszą się powieść

Tests:
    >>> import sys; sys.tracebacklimit = 0
    >>> from inspect import iscoroutine, iscoroutinefunction
    >>> import asyncio

    >>> assert iscoroutinefunction(a)
    >>> assert iscoroutinefunction(b)
    >>> assert iscoroutinefunction(c)
    >>> assert iscoroutine(a())
    >>> assert iscoroutine(b())
    >>> assert iscoroutine(c())

    >>> async def main():
    ...     return await asyncio.gather(a(), b(), c())
    >>>
    >>> asyncio.run(main())
    a: before
    b: before
    c: before
    b: after
    a: after
    c: after
    ['a', 'b', 'c']
"""

import asyncio

# coroutine function `a()`
# print 'a: before', wait 1.0 second, print 'a: after', return 'a'
# type: Coroutine
def a():
    ...

# coroutine function `b()`
# print 'b: before', wait 0.5 second, print 'b: after', return 'b'
# type: Coroutine
def b():
    ...

# coroutine function `c()`
# print 'c: before', wait 1.5 second, print 'c: after', return 'c'
# type: Coroutine
def c():
    ...