跳到主要内容

concurrent.futures

问题​

concurrent.futures 模块怎么用？ThreadPoolExecutor 和 ProcessPoolExecutor 有什么区别？

答案​

统一的接口​

concurrent.futures 提供了线程池和进程池的统一高级接口：

from concurrent.futures import ThreadPoolExecutor, ProcessPoolExecutor, as_completed

def task(n: int) -> int:
    return n * n

# 线程池（IO 密集型）
with ThreadPoolExecutor(max_workers=4) as executor:
    futures = [executor.submit(task, i) for i in range(10)]
    
    # as_completed：按完成顺序获取结果
    for future in as_completed(futures):
        print(future.result())
    
    # map：按提交顺序返回结果
    results = list(executor.map(task, range(10)))

# 进程池（CPU 密集型）——接口完全相同
with ProcessPoolExecutor(max_workers=4) as executor:
    results = list(executor.map(task, range(10)))

Future 对象​

from concurrent.futures import ThreadPoolExecutor, Future

def slow_task() -> str:
    import time
    time.sleep(2)
    return "完成"

with ThreadPoolExecutor() as executor:
    future: Future = executor.submit(slow_task)
    
    print(future.done())       # False — 还没完成
    print(future.running())    # True
    result = future.result(timeout=5)  # 阻塞等待结果
    print(future.done())       # True
    
    # 添加回调
    future2 = executor.submit(slow_task)
    future2.add_done_callback(lambda f: print(f"回调: {f.result()}"))

常见面试问题​

Q1: 线程池的 max_workers 设置多少合适？​

答案：

类型推荐值原因
IO 密集型CPU 核心数 × 5IO 等待时线程不占 CPU
CPU 密集型CPU 核心数更多线程反而增加切换开销
import os
cpu_count = os.cpu_count()  # 如 8
# IO 密集型：ThreadPoolExecutor(max_workers=cpu_count * 5)
# CPU 密集型：ProcessPoolExecutor(max_workers=cpu_count)

Q2: submit 和 map 的区别？​

答案：

  • submit：提交单个任务，返回 Future，更灵活
  • map：批量提交，按顺序返回结果，更简洁

submit + as_completed 适合想按完成顺序处理的场景；map 适合想保持输入顺序的场景。

相关链接​