asyncio的另一个解决方案:

如果你想取消后台任务，而不仅仅是在运行的主代码上超时，那么你需要一个来自主线程的显式通信，要求任务的代码取消，比如threading.Event()

import asyncio
import functools
import multiprocessing
from concurrent.futures.thread import ThreadPoolExecutor


class SingletonTimeOut:
    pool = None

    @classmethod
    def run(cls, to_run: functools.partial, timeout: float):
        pool = cls.get_pool()
        loop = cls.get_loop()
        try:
            task = loop.run_in_executor(pool, to_run)
            return loop.run_until_complete(asyncio.wait_for(task, timeout=timeout))
        except asyncio.TimeoutError as e:
            error_type = type(e).__name__ #TODO
            raise e

    @classmethod
    def get_pool(cls):
        if cls.pool is None:
            cls.pool = ThreadPoolExecutor(multiprocessing.cpu_count())
        return cls.pool

    @classmethod
    def get_loop(cls):
        try:
            return asyncio.get_event_loop()
        except RuntimeError:
            asyncio.set_event_loop(asyncio.new_event_loop())
            # print("NEW LOOP" + str(threading.current_thread().ident))
            return asyncio.get_event_loop()

# ---------------

TIME_OUT = float('0.2')  # seconds

def toto(input_items,nb_predictions):
    return 1

to_run = functools.partial(toto,
                           input_items=1,
                           nb_predictions="a")

results = SingletonTimeOut.run(to_run, TIME_OUT)

有很多建议，但没有一个是使用并发的。期货，我认为这是最清晰的处理方式。

from concurrent.futures import ProcessPoolExecutor

# Warning: this does not terminate function if timeout
def timeout_five(fnc, *args, **kwargs):
    with ProcessPoolExecutor() as p:
        f = p.submit(fnc, *args, **kwargs)
        return f.result(timeout=5)

超级简单的阅读和维护。

我们创建一个池，提交一个进程，然后等待5秒，然后引发一个TimeoutError，你可以根据需要捕获和处理它。

本机为python 3.2+，并反向移植到2.7 (pip install futures)。

线程和进程之间的切换非常简单，只需将ProcessPoolExecutor替换为ThreadPoolExecutor。

如果您想在超时时终止进程，我建议您查看Pebble。

我们也可以用信号来表示。我认为下面的例子会对你有用。与线程相比，它非常简单。

import signal

def timeout(signum, frame):
    raise myException

#this is an infinite loop, never ending under normal circumstances
def main():
    print 'Starting Main ',
    while 1:
        print 'in main ',

#SIGALRM is only usable on a unix platform
signal.signal(signal.SIGALRM, timeout)

#change 5 to however many seconds you need
signal.alarm(5)

try:
    main()
except myException:
    print "whoops"

如果工作没有完成，我打算杀死进程，使用线程和进程来实现这一点。

from concurrent.futures import ThreadPoolExecutor

from time import sleep
import multiprocessing


# test case 1
def worker_1(a,b,c):
    for _ in range(2):
        print('very time consuming sleep')
        sleep(1)

    return a+b+c

# test case 2
def worker_2(in_name):
    for _ in range(10):
        print('very time consuming sleep')
        sleep(1)

    return 'hello '+in_name

作为上下文管理器的实际类

class FuncTimer():
    def __init__(self,fn,args,runtime):
        self.fn = fn
        self.args = args
        self.queue = multiprocessing.Queue()
        self.runtime = runtime
        self.process = multiprocessing.Process(target=self.thread_caller)

    def thread_caller(self):
        with ThreadPoolExecutor() as executor:
            future = executor.submit(self.fn, *self.args)
            self.queue.put(future.result())

    def  __enter__(self):
        return self

    def start_run(self):
        self.process.start()
        self.process.join(timeout=self.runtime)
        if self.process.exitcode is None:
            self.process.kill()
        if self.process.exitcode is None:
            out_res = None
            print('killed premature')
        else:
            out_res = self.queue.get()
        return out_res


    def __exit__(self, exc_type, exc_value, exc_traceback):
        self.process.kill()

如何使用

print('testing case 1') 
with FuncTimer(fn=worker_1,args=(1,2,3),runtime = 5) as fp: 
    res = fp.start_run()
    print(res)

print('testing case 2')
with FuncTimer(fn=worker_2,args=('ram',),runtime = 5) as fp: 
    res = fp.start_run()
    print(res)

我也遇到过同样的问题，但我的情况是需要在子线程上工作，信号不适合我，所以我写了一个python包:timeout-timer来解决这个问题，支持用作上下文或装饰器，使用信号或子线程模块来触发超时中断:

from timeout_timer import timeout, TimeoutInterrupt

class TimeoutInterruptNested(TimeoutInterrupt):
    pass

def test_timeout_nested_loop_both_timeout(timer="thread"):
    cnt = 0
    try:
        with timeout(5, timer=timer):
            try:
                with timeout(2, timer=timer, exception=TimeoutInterruptNested):
                    sleep(2)
            except TimeoutInterruptNested:
                cnt += 1
            time.sleep(10)
    except TimeoutInterrupt:
        cnt += 1
    assert cnt == 2

查看更多信息:https://github.com/dozysun/timeout-timer

我是wrapt_timeout_decorator的作者。

这里介绍的大多数解决方案乍一看在Linux下都工作得很好——因为我们有fork()和signals()——但在windows上看起来有点不同。 当涉及到Linux上的子线程时，你不能再使用信号了。

为了在Windows下生成一个进程，它需要是可pickle的——许多装饰函数或Class方法都不是。

所以你需要使用一个更好的pickler像莳萝和multiprocess(不是pickle和multiprocessing) -这就是为什么你不能使用ProcessPoolExecutor(或只有有限的功能)。

For the timeout itself - You need to define what timeout means - because on Windows it will take considerable (and not determinable) time to spawn the process. This can be tricky on short timeouts. Lets assume, spawning the process takes about 0.5 seconds (easily !!!). If You give a timeout of 0.2 seconds what should happen? Should the function time out after 0.5 + 0.2 seconds (so let the method run for 0.2 seconds)? Or should the called process time out after 0.2 seconds (in that case, the decorated function will ALWAYS timeout, because in that time it is not even spawned)?

嵌套的装饰器也很讨厌，你不能在子线程中使用信号。如果你想要创建一个真正通用的、跨平台的装饰器，所有这些都需要考虑(并测试)。

其他问题是将异常传递回调用者，以及记录问题(如果在装饰函数中使用-不支持记录到另一个进程中的文件)

我试图涵盖所有的边缘情况，您可以查看包wrapt_timeout_decorator，或者至少测试您自己的解决方案，受到那里使用的单元测试的启发。

@Alexis Eggermont -不幸的是，我没有足够的分数来评论-也许其他人可以通知你-我认为我解决了你的进口问题。