并发模型 - python将进程池放在装饰器里为什么不生效也没报错

Question

我想把进程池封装在装饰器里，但是它既没生效也没报错

# coding:utf-8
import multiprocessing
import tornado
from tornado.httpclient import AsyncHTTPClient

process_num = 20  # 进程数
url = "https://www.baidu.com"

def handle_request(response):
    print str(response)

def run_in_process(process_num):
    def _run_in_process(f):
        def __run_in_process(*args, **kwargs):
            pool = multiprocessing.Pool(processes=process_num)
            for i in range(process_num):
                pool.apply_async(f, args=args, kwds=kwargs, callback=kwargs.get("callback"))
            pool.close()
            pool.join()

        return __run_in_process

    return _run_in_process


@run_in_process(process_num)
def main():
    http_client = AsyncHTTPClient()
    http_client.fetch(url, callback=handle_request)
    global loop
    loop = tornado.ioloop.IOLoop.instance()
    if loop._running is False:
        loop.start()


if __name__ == '__main__':
    main()

结果如下

/usr/bin/python2.7 /home/xxx/workspace/py_project/crawler/center/sample.py

Process finished with exit code 0

但是奇怪的是，我用多进程的方式重写一次，发现是可以生效的

# coding:utf-8
import multiprocessing
import tornado
from tornado.httpclient import AsyncHTTPClient

process_num = 20  # 进程数
url = "https://www.baidu.com"

def handle_request(response):
    print str(response)

def run_in_process(process_num):
    def _run_in_process(f):
        def __run_in_process(*args, **kwargs):
            _processes = []
            for i in xrange(process_num):
                p = multiprocessing.Process(target=f, args=args, kwargs=kwargs)
                p.start()
                _processes.append(p)

            for p in _processes:
                p.join()

        return __run_in_process
    return _run_in_process


@run_in_process(process_num)
def main():
    http_client = AsyncHTTPClient()
    http_client.fetch(url, callback=handle_request)
    global loop
    loop = tornado.ioloop.IOLoop.instance()
    if loop._running is False:
        loop.start()


if __name__ == '__main__':
    main()

日志如下

/usr/bin/python2.7 /home/shufeng/workspace/private_project/jobscrawler/center/sample.py
HTTPResponse(_body=None,buffer=<_io.BytesIO object at 0x7f2fdaa21ef0>,code=200,effective_url='http://www.baidu.com',error=None,headers=<tornado.httputil.HTTPHeaders object at 0x7f2fdaa425d0>,reason='OK',request=<tornado.httpclient.HTTPRequest object at 0x7f2fdaa42250>,request_time=0.014312028884887695,time_info={})
HTTPResponse(_body=None,buffer=<_io.BytesIO object at 0x7f2fdaa21ef0>,code=200,effective_url='http://www.baidu.com',error=None,headers=<tornado.httputil.HTTPHeaders object at 0x7f2fdaa43450>,reason='OK',request=<tornado.httpclient.HTTPRequest object at 0x7f2fdaa430d0>,request_time=0.02327895164489746,time_info={})
HTTPResponse(_body=None,buffer=<_io.BytesIO object at 0x7f2fdaa21ef0>,code=200,effective_url='http://www.baidu.com',error=None,headers=<tornado.httputil.HTTPHeaders object at 0x7f2fdaa43510>,reason='OK',request=<tornado.httpclient.HTTPRequest object at 0x7f2fdaa43190>,request_time=0.026951074600219727,time_info={})
HTTPResponse(_body=None,buffer=<_io.BytesIO object at 0x7f2fdaa21ef0>,code=200,effective_url='http://www.baidu.com',error=None,headers=<tornado.httputil.HTTPHeaders object at 0x7f2fdaa42690>,reason='OK',request=<tornado.httpclient.HTTPRequest object at 0x7f2fdaa42310>,request_time=0.0552978515625,time_info={})
HTTPResponse(_body=None,buffer=<_io.BytesIO object at 0x7f2fdaa24ef0>,code=200,effective_url='http://www.baidu.com',error=None,headers=<tornado.httputil.HTTPHeaders object at 0x7f2fdaa39e10>,reason='OK',request=<tornado.httpclient.HTTPRequest object at 0x7f2fdaa39a90>,request_time=0.05612993240356445,time_info={})

同样的情况也会出现在线程池跟协程的使用上，有谁知道这是怎么回事吗？

Answer 1

知乎灵剑大神已回答此问题：https://www.zhihu.com/questio...

Answer 2

在linux下运行, 会得到下面的报错:

PicklingError: Can't pickle <type 'function'>: attribute lookup __builtin__.function failed

而这个报错是因为, 传入不可序列化的对象进进程池时, 报错导致的, 而这个对象就是实例方法, 可以试下用py3运行下, 因为3的实例方法已经可以支持序列化

参考资料: https://virusdefender.net/ind...