mirror of
https://github.com/Brandon-Rozek/website.git
synced 2024-11-09 10:40:34 -05:00
1.2 KiB
1.2 KiB
title | date | draft | tags | medium_enabled | |
---|---|---|---|---|---|
Quick Python: Concurrent Futures | 2020-04-11T20:40:28-04:00 | false |
|
true |
Another way to perform concurrency in python is to use the concurrent.futures
module.
from concurrent.futures import ThreadPoolExecutor, ProcessPoolExecutor
def add(x, y):
return x + y
with ThreadPoolExecutor(max_workers=4) as executor:
future = executor.submit(add, 1, 2)
result = future.result(timeout=30) # unit: seconds
If max_workers=None
then it will default to the number of processors on the machine multiplied by 5.
If timeout=None
then there is no time limit applied.
You can also apply a function to a list or iterables
def double(x):
return 2 * x
with ThreadPoolExecutor() as executor:
future = executor.map(function_handle, [1, 2, 3])
result = future.result()
Instead of threads, it is also possible to spawn processes to side-step the global interpreter lock. The documentation warns that only picklable objects can be executed and returned though.
def add(x, y):
return x + y
with ProcessPoolExecutor() as executor:
future = executor.submit(add, 1, 2)
result = future.result()