代码之家 › 专栏 › 技术社区 › J. Taylor

如何在异步中并发运行任务?

python-asyncio task concurrency asynchronous python

J. Taylor · 技术社区 · 7 年前

我正在尝试学习如何使用Python的异步模块并发地运行任务。在下面的代码中,我举了一个模拟的“网络爬虫”。基本上,我尝试在任何给定时间最多发生两个活动fetch()请求的情况下进行,我希望在sleep()期间调用process()。

import asyncio

class Crawler():

    urlq = ['http://www.google.com', 'http://www.yahoo.com', 
            'http://www.cnn.com', 'http://www.gamespot.com', 
            'http://www.facebook.com', 'http://www.evergreen.edu']

    htmlq = []
    MAX_ACTIVE_FETCHES = 2
    active_fetches = 0

    def __init__(self):
        pass

    async def fetch(self, url):
        self.active_fetches += 1
        print("Fetching URL: " + url);
        await(asyncio.sleep(2))
        self.active_fetches -= 1
        self.htmlq.append(url)

    async def crawl(self):
        while self.active_fetches < self.MAX_ACTIVE_FETCHES:
            if self.urlq:
                url = self.urlq.pop()
                task = asyncio.create_task(self.fetch(url))
                await task
            else:
                print("URL queue empty")
                break;

    def process(self, page):
        print("processed page: " + page)

# main loop

c = Crawler()
while(c.urlq):
    asyncio.run(c.crawl())
    while c.htmlq:
        page = c.htmlq.pop()
        c.process(page)

但是,上面的代码一个接一个地下载URL(一次不同时下载两个),并且直到所有URL都被获取之后才进行任何“处理”。如何使fetch()任务并发运行,并使其在sleep()期间在其间调用process()?

2 回复 | 直到 7 年前

dtanabe 7 年前

你的 crawl 方法正在等待每个单独的任务;您应该将其更改为:

async def crawl(self):
    tasks = []
    while self.active_fetches < self.MAX_ACTIVE_FETCHES:
        if self.urlq:
            url = self.urlq.pop()
            tasks.append(asyncio.create_task(self.fetch(url)))
    await asyncio.gather(*tasks)

编辑 :这里有一个更清晰的版本,注释可以同时提取和处理所有内容,同时保留对最大提取数设置上限的基本能力。

import asyncio

class Crawler:

    def __init__(self, urls, max_workers=2):
        self.urls = urls
        # create a queue that only allows a maximum of two items
        self.fetching = asyncio.Queue()
        self.max_workers = max_workers

    async def crawl(self):
        # DON'T await here; start consuming things out of the queue, and
        # meanwhile execution of this function continues. We'll start two
        # coroutines for fetching and two coroutines for processing.
        all_the_coros = asyncio.gather(
            *[self._worker(i) for i in range(self.max_workers)])

        # place all URLs on the queue
        for url in self.urls:
            await self.fetching.put(url)

        # now put a bunch of `None`'s in the queue as signals to the workers
        # that there are no more items in the queue.
        for _ in range(self.max_workers):
            await self.fetching.put(None)

        # now make sure everything is done
        await all_the_coros

    async def _worker(self, i):
        while True:
            url = await self.fetching.get()
            if url is None:
                # this coroutine is done; simply return to exit
                return

            print(f'Fetch worker {i} is fetching a URL: {url}')
            page = await self.fetch(url)
            self.process(page)

    async def fetch(self, url):
        print("Fetching URL: " + url);
        await asyncio.sleep(2)
        return f"the contents of {url}"

    def process(self, page):
        print("processed page: " + page)


# main loop
c = Crawler(['http://www.google.com', 'http://www.yahoo.com', 
             'http://www.cnn.com', 'http://www.gamespot.com', 
             'http://www.facebook.com', 'http://www.evergreen.edu'])
asyncio.run(c.crawl())

user4815162342 7 年前

你可以做 htmlq 安 asyncio.Queue() 和改变 htmlq.append 到 htmlq.push . 那么你的 main 可以是异步的,如下所示:

async def main():
    c = Crawler()
    asyncio.create_task(c.crawl())
    while True:
        page = await c.htmlq.get()
        if page is None:
            break
        c.process(page)

您的顶级代码可以归结为调用 asyncio.run(main()) .

一旦你完成了爬行, crawl() 可以入队 None 通知主协同工作已完成。