94 lines
3.0 KiB
Plaintext
94 lines
3.0 KiB
Plaintext
|
||
import asyncio
|
||
import multiprocessing
|
||
import os
|
||
import time
|
||
from multiprocessing import Manager
|
||
|
||
|
||
# 业务类
|
||
class BaiJiaHao():
|
||
|
||
async def get_author(self, rec):
|
||
"""
|
||
协程代码
|
||
"""
|
||
print('enter get author,wait for: %d' % rec['num'])
|
||
# 模拟IO操作，耗时根据传进来的num决定
|
||
await asyncio.sleep(rec['num'])
|
||
# 返回协程任务完成后的结果
|
||
return rec
|
||
|
||
|
||
def run(self):
|
||
# 假定我们有11个任务要跑，每个任务耗时为num秒，串行的话需要43秒。
|
||
# 但我们这个demo跑完只需要这些任务中的最大值：8秒
|
||
list = [{'title': 'title1', 'num': 2},
|
||
{'title': 'title2', 'num': 1},
|
||
{'title': 'title3', 'num': 3},
|
||
{'title': 'title4', 'num': 8},
|
||
{'title': 'title5', 'num': 2},
|
||
{'title': 'title6', 'num': 5},
|
||
{'title': 'title7', 'num': 7},
|
||
{'title': 'title8', 'num': 3},
|
||
{'title': 'title9', 'num': 4},
|
||
{'title': 'title10', 'num': 3},
|
||
{'title': 'title11', 'num': 5},
|
||
]
|
||
result = run_get_author_in_multi_process(list)
|
||
print('result', result)
|
||
|
||
|
||
def get_chunks(iterable, chunks=1):
|
||
"""
|
||
此函数用于分割若干任务到不同的进程里去
|
||
"""
|
||
lst = list(iterable)
|
||
return [lst[i::chunks] for i in range(chunks)]
|
||
|
||
|
||
def run_get_author(lists, queue):
|
||
"""
|
||
这个就是子进程运行的函数，接收任务列表和用于进程间通讯的Queue
|
||
"""
|
||
print('exec run_get_author.child process id : %s, parent process id : %s' % (os.getpid(), os.getppid()))
|
||
# 每个子进程分配一个新的loop
|
||
loop = asyncio.new_event_loop()
|
||
# 初始化业务类，转成task或future
|
||
spider = BaiJiaHao()
|
||
tasks = [loop.create_task(spider.get_author(rec)) for rec in lists]
|
||
# 协程走起
|
||
loop.run_until_complete(asyncio.wait(tasks))
|
||
# 往queue写入每个任务的结果
|
||
for task in tasks:
|
||
queue.put(task.result())
|
||
|
||
|
||
def run_get_author_in_multi_process(task_lists):
|
||
"""
|
||
父进程函数，主要是分割任务并初始化进程池，启动进程并返回结果
|
||
"""
|
||
# process_count = len(tasks) % 2
|
||
# 进程数这里我用机器上的核心数，注意：未考虑核心数比任务多的情况
|
||
process_count = multiprocessing.cpu_count()
|
||
print('process_count: %d' % process_count)
|
||
split_lists = get_chunks(task_lists, process_count)
|
||
pool = multiprocessing.Pool(process_count)
|
||
queue = Manager().Queue()
|
||
for lists in split_lists:
|
||
pool.apply_async(run_get_author, args=(lists, queue,))
|
||
pool.close()
|
||
pool.join()
|
||
result = []
|
||
# 从子进程读取结果并返回
|
||
while not queue.empty():
|
||
result.append(queue.get())
|
||
return result
|
||
|
||
now = lambda : time.time()
|
||
|
||
if __name__ == '__main__':
|
||
start = now()
|
||
spider = BaiJiaHao()
|
||
spider.run()
|
||
print('done','TIME: ', now() - start) |