aio-pool 0.2.0

Extending Python's process pool to support async functions.

aio-pool

Extending Python's multiporcessing.Pool to support coroutine functions.
Can be useful for when using a server with very high bandwidth or doing both very large IO and CPU tasks at the same time.

All methods of multiprocessing.Pool are supported.
All paramters for multiprocessing.Pool are supported.

examples:

Setting concurrency limit. This means each process can run with up to 8 concurrent tasks at a time.

import asyncio
from aio_pool import AioPool


async def powlong(a):
  await asyncio.sleep(1)
  return a**2

if __name__ == '__main__':
  with AioPool(processes=2, concurrency_limit=8) as pool:
    results = pool.map(powlong, [i for i in range(16)])  # Should take 2 seconds (2*8).
    print(results) 

Async initliazers are also suppported.

import asyncio
from aio_pool import AioPool

async def start(message):
  await asyncio.sleep(1)
  print(message)

async def powlong(a):
  await asyncio.sleep(1)
  return a**2

if __name__ == '__main__':
  with AioPool(processes=2, 
               concurrency_limit=8, 
               initializer=start,
               init_args=("Started with AioPool", )) as pool:
    results = pool.map(powlong, [i for i in range(16)])  # Should take 2 seconds (2*8).
    print(results) 
    

By default, AioPool also set up a default executor for any non-async tasks.
The size can be determined by threadpool_size arguemnt, which defaults to 1.
None default event loops(uvloop for example) are supported as well, using the loop_initializer argument.
Also, non-async functions are supported by default, as the AioPool worker identify if the function is async or not.
If the function is not async, it runs inside the threadpool, to allow the requested concurrency.
This means that order of execution is not guaranteed, even if the function is not async.
However, the order of results is guaranteed through the pool API (map, starmap, apply, etc...).

from aio_pool import AioPool
import uvloop

with AioPool(loop_initializer=uvloop.new_event_loop, threadpool_size=4) pool:
  pool.map(print, [i for i in range(8)])