TIL - In Python concurrent.futures.as_completed Yields Tasks as Soon as They Finish

When executing multiple tasks in parallel with thread or process pools, as_completed gives you the results in the exact order they finish, not the order they were started.

from concurrent.futures import ThreadPoolExecutor, as_completed
import time

def slow_task(n):
    time.sleep(n)
    return n

with ThreadPoolExecutor() as executor:
    # Submit tasks with different sleep times
    futures = [executor.submit(slow_task, t) for t in [3, 1, 2]]
    
    # 1 prints first, then 2, then 3
    for future in as_completed(futures):
        print(f"Finished sleeping for: {future.result()}s")

This prevents fast tasks from being blocked in a list comprehension waiting for an earlier, slower task.