Python3并发写文件

Posted 2020-11-22 川川籽

使用python2在进行并发写的时候，发现文件会乱掉，就是某一行中间会插入其他行的内容。
但是在使用python3进行并发写的时候，无论是多进程，还是多线程，都没有出现这个问题，难道是python3的特性吗？

import time
import os
import multiprocessing
from multiprocessing.dummy import Pool as ThreadPool


def write(val, file):
    w = open(file, "a")
    for i in range(100):
        w.write("%s
" % val)
        time.sleep(0.001)

def thread_write(file):
    res, pools = [], ThreadPool(10)
    for i in range(10):
        val = str(i) * 1000
        res.append(pools.apply_async(func=write, args=(val, file, )))

    while res:
        for ret in res:
            if ret.ready():
                res.remove(ret)
        time.sleep(0.01)

def mutil_write(file):
    pools = multiprocessing.Pool(processes=10)
    res = []
    for i in range(100):
        res.append(pools.apply_async(thread_write, args=(file, )))

    while res:
        for ret in res:
            if ret.ready():
                res.remove(ret)
        time.sleep(0.01)

if __name__ == '__main__':
    file = "./write_test"
    mutil_write(file)

    with open(file) as fb:
        lines = 0
        line_len = []
        for line in fb:
            lines += 1
            line = line.strip()
            line_len.append(len(line))
            if len(line) != 1000:
                raise(Exception("error line: %s, len: %d" % (line, len(line))))

        print("lines:%d, max len:%d, min:%d, avg:%.2f" % (lines, max(line_len), min(line_len), sum(line_len)/len(line_len)))
    os.remove(file)

上面代码，多进程并发写结束后，校验每一行的长度是否是设置好的长度。用python3反复运行，均通过测试没有异常。

$ python3 --version
Python 3.7.4

$ python3 t.py
lines:10000, max len:1000, min:1000, avg:1000.00

如果用python2，则会出现异常：

$ python2 --version
Python 2.7.15

$ python2 t.py
Traceback (most recent call last):
  File "t.py", line 49, in <module>
    raise(Exception("error line: %s, len: %d" % (line, len(line))))
Exception: error line: 333333333333333333333333333333333333333333333333333333333333333333333333333333333333333333330000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000, len: 1092