基于python生成器封装的协程类

来源:互联网 发布:垂衣女土冬天淘宝网 编辑:程序博客网 时间:2024/06/14 06:00

自从python2.2提供了yield关键字之后,python的生成器的很大一部分用途就是可以用来构建协同程序,能够将函数挂起返回中间值并能从上次离开的地方继续执行。python2.5的时候,这种生成器更加接近完全的协程,因为提供了将值和异常传递回到一个继续执行的函数中,当等待生成器的时候,生成器能返回控制。

python提供的生成器设施:

  • yield:能够将自己挂起,并提供一个返回值给等待方
  • send:唤起一个被挂起的生成器,并能够传递一个参数,可以在生成器中抛出异常
  • next:本质上相当于send(None),对每个生成器的第一次调用必须不能传递参数
  • close:主动退出一个生成器

python封装

虽然python3提供了asyncio这样的异步IO库,而且也有greenlet等其他协程库,但目前的需求并不是实际的网络IO并发操作,而是需要模拟状态机的运行,因此使用协程可以很方便的模拟,并加入认为的控制,下面是封装的一个python类。

class Coroutine(object):    """ Base class of the general coroutine object """    STATE_RUNNING = 0    STATE_WAITING = 1    STATE_CLOSING = 2    def __init__(self):        self.state = Coroutine.STATE_WAITING        self.started = False        self.args = None        self.routine = self._co()    def _co(self):        self.ret = None        while True:            self.args = yield self.ret            if not self.started:                self.started = True                continue            else:                self.state = Coroutine.STATE_RUNNING                self.ret = self.run(self.args)            if self.state == Coroutine.STATE_CLOSING:                break            self.state = Coroutine.STATE_WAITING    def start(self):        """ Start the generator """        if self.routine is None:            raise RuntimeError('NO task to start running!')        self.started = True        self.routine.next()    def finish(self):        """ Finish the execution of this routine """        self.state = Coroutine.STATE_CLOSING        self.routine.close()    def run(self, args):        """ The runing method to be executed every once time"""        raise NotImplementedError    def execute(self, arg_obj):        """ Awake this routine to execute once time """        return self.routine.send(arg_obj)

基于上述封装,下面实现了一个协同的生产者消费者示例:

class ProducerCoroutine(Coroutine):    """ The Producer concrete coroutine """    def __init__(self, cnsmr):        if not isinstance(cnsmr, Coroutine):            raise RuntimeError('Consumer is not a Coroutine object')        self.consumer = cnsmr        self.consumer.start()        super(ProducerCoroutine, self).__init__()    def run(self, args):        print 'produce ', args        ret = self.consumer.execute(args)        print 'consumer return:', ret    def __call__(self, args):        """ Custom method for the specific logic """        self.start()        while len(args) > 0:            p = args.pop()            self.execute(p)        self.finish()class ConsumerCoroutine(Coroutine):    """ The Consumer concrete coroutine """    def __init__(self):        super(ConsumerCoroutine, self).__init__()    def run(self, args):        print 'consumer get args: ', args        return 'hahaha' + repr(args)

运行结果如下:

produce  4consumer get args:  4consumer return: hahaha4produce  3consumer get args:  3consumer return: hahaha3produce  2consumer get args:  2consumer return: hahaha2produce  1consumer get args:  1consumer return: hahaha1produce  0consumer get args:  0consumer return: hahaha0
0 0