Python中的循环多线程

Question

我正在尝试用Python的多线程技术解决Project Euler的第8题。

题目是找出在一个1000位数字中，五个连续数字的最大乘积。这个数字可以在这里找到。

我的思路是从原始列表中每5个数字为一组来计算乘积，并且这个过程要重复5次，每次的起始位置向右移动一个。

这是我的线程类：

class pThread(threading.Thread):
    def __init__(self, l):
        threading.Thread.__init__(self)
        self.l = l
        self.p = 0

    def run(self):

        def greatest_product(l):
        """
        Divide the list into chunks of 5 and find the greatest product
        """
            def product(seq):
                return reduce(lambda x,y : x*y, seq)

            def chunk_product(l, n=5):
                for i in range(0, len(l), n):
                    yield product(l[i:i+n])

            result = 0
            for p in chunk_product(num):
                result = result > p and result or p 

            return result

        self.p = greatest_product(self.l)

当我尝试创建5个线程来处理原始列表中的所有5位数字组合时，下面这种手动的方法可以得到正确的答案，其中num是我从文本中解析出来的单个数字列表：

thread1 = pThread(num)
del num[0]
thread2 = pThread(num)
del num[0]
thread3 = pThread(num)
del num[0]
thread4 = pThread(num)
del num[0]
thread5 = pThread(num)

thread1.start()
thread2.start()
thread3.start()
thread4.start()
thread5.start()

thread1.join()
thread2.join()
thread3.join()
thread4.join()
thread5.join()

def max(*args):
    result = 0
    for i in args:
        result = i > result and i or result
    return result

print max(thread1.p, thread2.p, thread3.p, thread4.p, thread5.p)

但是这个方法却没有得到正确的结果：

threads = []
for i in range(0, 4):
    tmp = num[:]
    del tmp[0:i+1]
    thread = pThread(tmp)
    thread.start()
    threads.append(thread)

for i in range(0, 4):
    threads[i].join()

我在这里做错了什么呢？我对多线程非常陌生，所以请多多指教。

Answer 1

我尝试做这个主要是为了练习多进程编程，并学习如何使用argparse这个工具。

这个程序大约占用了4到5个GB的内存，所以如果你的电脑内存不多的话，要注意哦。

python euler.py -l 50000000 -n 100 -p 8

Took 5.836833333969116 minutes
The largest product of 100 consecutive numbers is: a very large number

如果你在命令行输入python euler.py -h，你会看到：

usage: euler.py [-h] -l L [L ...] -n N [-p P]

Calculates the product of consecutive numbers and return the largest product.

optional arguments:
  -h, --help    show this help message and exit
  -l L [L ...]  A single number or list of numbers, where each # is seperated
                by a space
  -n N          A number that specifies how many consecutive numbers should be
                multiplied together.
  -p P          Number of processes to create. Optional, defaults to the # of
                cores on the pc.        

还有代码如下：

"""A multiprocess iplementation for calculation the maximum product of N consecutive
numbers in a given range (list of numbers)."""

import multiprocessing
import math
import time
import operator
from functools import reduce
import argparse

def euler8(alist,lenNums):
    """Returns the largest product of N consecutive numbers in a given range"""
    return max(reduce(operator.mul, alist[i:i+lenNums]) for i in range(len(alist)))

def split_list_multi(listOfNumbers,numLength,threads):
    """Split a list into N parts where N is the # of processes."""
    fullLength = len(listOfNumbers)
    single = math.floor(fullLength/threads)
    results = {}
    counter = 0
    while counter < threads:
        if counter == (threads-1):
            temp = listOfNumbers[single*counter::]
            if counter == 0:
                results[str(counter)] = listOfNumbers[single*counter::]
            else:
                prevListIndex = results[str(counter-1)][-int('{}'.format(numLength-1))::]
                newlist = prevListIndex + temp
                results[str(counter)] = newlist
        else:
            temp = listOfNumbers[single*counter:single*(counter+1)]
            if counter == 0:
                newlist = temp
            else:
                prevListIndex = results[str(counter-1)][-int('{}'.format(numLength-1))::]
                newlist = prevListIndex + temp
            results[str(counter)] = newlist
        counter += 1
    return results,threads

def worker(listNumbers,number,output):
    """A worker. Used to run seperate processes and put the results in the queue"""
    result = euler8(listNumbers,number)
    output.put(result)

def main(listOfNums,lengthNumbers,numCores=multiprocessing.cpu_count()):
    """Runs the module.
    listOfNums must be a list of ints, or single int
    lengthNumbers is N (an int) where N is the # of consecutive numbers to multiply together
    numCores (an int) defaults to however many the cpu has, can specify a number if you choose."""

    if isinstance(listOfNums,list):
        if len(listOfNums) == 1:
            valuesToSplit = [i for i in range(int(listOfNums[0]))]
        else:
            valuesToSplit = [int(i) for i in listOfNums]
    elif isinstance(listOfNums,int):
        valuesToSplit = [i for i in range(listOfNums)]
    else:
        print('First arg must be a number or a list of numbers')

    split = split_list_multi(valuesToSplit,lengthNumbers,numCores)
    done_queue = multiprocessing.Queue()
    jobs = []
    startTime = time.time()

    for num in range(split[1]):
        numChunks = split[0][str(num)]
        thread = multiprocessing.Process(target=worker, args=(numChunks,lengthNumbers,done_queue))
        jobs.append(thread)
        thread.start()

    resultlist = []
    for i in range(split[1]):
        resultlist.append(done_queue.get())

    for j in jobs:
        j.join()

    resultlist = max(resultlist)
    endTime = time.time()
    totalTime = (endTime-startTime)/60
    print("Took {} minutes".format(totalTime))

    return print("The largest product of {} consecutive numbers is: {}".format(lengthNumbers, resultlist))            

if __name__ == '__main__':
    #To call the module from the commandline with arguments
    parser = argparse.ArgumentParser(description="""Calculates the product of consecutive numbers \
    and return the largest product.""")
    parser.add_argument('-l', nargs='+', required=True,
                       help='A single number or list of numbers, where each # is seperated by a space')
    parser.add_argument('-n', required=True, type=int,
                        help = 'A number that specifies how many consecutive numbers should be \
                        multiplied together.')
    parser.add_argument('-p', default=multiprocessing.cpu_count(), type=int,
                       help='Number of processes to create. Optional, defaults to the # of cores on the pc.')
    args = parser.parse_args()
    main(args.l, args.n, args.p)

Answer 2

这里有三个问题：

1. 第一个问题是，你的“手动”方法并没有给出正确的答案。恰好问题的正确答案在你列表的第4个位置。你可以通过以下方式看到这一点：

   import operator as op
   print max(reduce(op.mul, num[i:i+5]) for i in range(1000))
   for k in range(5):
       print max(reduce(op.mul, num[i:i+5]) for i in range(k, 1000, 5))
   

   你“手动”方法的一个问题是，多个线程共享了 num 这个变量，它们都使用同一个列表。所以当你执行 del num[0] 时，所有的 threadX.l 都会受到影响。你总是得到相同答案的原因是第二个问题。

2. 这一行

   for p in chunk_product(num):
   

   应该改成：

   for p in chunk_product(l):
   

   因为你想使用函数 greatest_product(l) 的参数，而不是全局变量 num。

3. 在第二种方法中，你只启动了4个线程，因为循环的范围是 [0, 1, 2, 3]。另外，你想删除的值是 tmp[0:i]，而不是 tmp[0:i+1]。这里是代码：

   threads = []
   for i in range(5):
       tmp = num[:]
       del tmp[0:i]
       thread = pThread(tmp)
       thread.start()
       threads.append(thread)
   
   for i in range(5):
       threads[i].join()
   
   print len(threads), map(lambda th: th.p, threads)
   print max(map(lambda th: th.p, threads))