Python并发编程:从多线程到异步IO的演进与实践

• IO密集型 → 优先异步IO（效率最高）  
• CPU密集型 → 多进程（绕过GIL）  
• 简单IO任务 → 多线程（快速上手）

import timeimport requestsurls = ["http://httpbin.org/delay/2"] * 10start = time.time()for url in urls:    requests.get(url)print(f"顺序执行: {time.time() - start:.2f}s")  # 输出: 20.01s

import timeimport threadingimport requestsurls = ["http://httpbin.org/delay/2"] * 10deffetch(url):    requests.get(url)start = time.time()threads = [threading.Thread(target=fetch, args=(url,)) for url in urls]for t in threads:    t.start()for t in threads:    t.join()print(f"多线程: {time.time() - start:.2f}s")  # 输出: 2.55s

import timefrom multiprocessing import Poolimport requestsurls = ["http://httpbin.org/delay/2"] * 10deffetch(url):    requests.get(url)start = time.time()with Pool(10) as p:    p.map(fetch, urls)print(f"多进程: {time.time() - start:.2f}s")  # 输出: 2.23s

import timeimport asyncioimport aiohttpurls = ["http://httpbin.org/delay/2"] * 10asyncdeffetch(session, url):asyncwith session.get(url) as response:returnawait response.text()asyncdefmain():asyncwith aiohttp.ClientSession() as session:        tasks = [fetch(session, url) for url in urls]await asyncio.gather(*tasks)start = time.time()asyncio.run(main())print(f"异步IO: {time.time() - start:.2f}s")  # 输出: 2.11s

import asyncioasyncdefsay_after(delay, word):await asyncio.sleep(delay)  # 模拟IO等待（非阻塞！）    print(word)asyncdefmain():# 创建两个任务（并行执行）    task1 = asyncio.create_task(say_after(1, "Hello"))    task2 = asyncio.create_task(say_after(2, "World"))# 等待任务完成await task1await task2# 启动事件循环asyncio.run(main())

HelloWorld

import asyncioimport aiohttpimport timeasyncdeffetch_url(session, url):"""异步HTTP请求"""asyncwith session.get(url) as response:returnawait response.text()  # 非阻塞等待响应asyncdefmain():    urls = ["http://httpbin.org/delay/2"] * 5# 5个延迟请求asyncwith aiohttp.ClientSession() as session:        tasks = [fetch_url(session, url) for url in urls]# 并行执行所有任务        results = await asyncio.gather(*tasks)        print(f"成功获取 {len(results)} 个页面")if __name__ == "__main__":    start = time.time()    asyncio.run(main())    print(f"异步爬虫耗时: {time.time() - start:.2f}s")

成功获取 5 个页面异步爬虫耗时: 2.10s

1. 先用多线程：快速验证需求，避免过早优化。
2. 性能瓶颈时：若任务IO密集，转战异步IO；若CPU密集，用多进程。
3. 永远警惕GIL：CPU密集型任务绝不用多线程。