西宁公安网站建设四川成都最新消息

程序跑的很慢，使用bottleneck一分析发现问题出在deepcopy上。
因为程序是需要对列表进行操作，而我的程序逻辑是要求必须对列表进行深拷贝，于是没有办法绕过。
只能想办法加速这个深层拷贝。
我想到四种方法
第一种：deepcopy，这没得啥说的。是baseline.

##直接deepcopy
def method1(origin_list, step):for each in range(step):l = copy.deepcopy(origin_list)return l

第二种：使用numpy，先转为numpy对象，然后再tolist

##转换为numpy, 然后再tolist()
def method2(origin_list, step):for each in range(step):l = np.array(origin_list).tolist()assert type(l) == type(origin_list)return l

第三种：使用pickle,先pickle.dump，再pickle.load

##使用pickle
def method3(origin_list, step):for each in range(step):l = pickle.loads(pickle.dumps(origin_list))assert type(l)== type(origin_list)return l

第四种，使用ujson,也是先usjon.dump，在ujson.load。

##使用ujson
def method4(origin_list, step):for each in range(step):l = ujson.loads(ujson.dumps(origin_list))assert type(l)== type(origin_list)return l

分别拷贝10000次，测试如下：

__author__ = 'dk'
import numpy as np
import copy
import time
import pickle
import ujson##直接deepcopy
def method1(origin_list, step):for each in range(step):l = copy.deepcopy(origin_list)return l##转换为numpy, 然后再tolist()
def method2(origin_list, step):for each in range(step):l = np.array(origin_list).tolist()assert type(l) == type(origin_list)return l##使用pickle
def method3(origin_list, step):for each in range(step):l = pickle.loads(pickle.dumps(origin_list))assert type(l)== type(origin_list)return l##使用ujson
def method4(origin_list, step):for each in range(step):l = ujson.loads(ujson.dumps(origin_list))assert type(l)== type(origin_list)return l
if __name__ == '__main__':origin_list = [[i for  i in range(30)] for i in range(30)]step = 10000t = time.time()method1(origin_list,step)e = time.time()print(e-t)t = time.time()method2(origin_list,step)e = time.time()print(e-t)t = time.time()method3(origin_list,step)e = time.time()print(e-t)t = time.time()method4(origin_list,step)e = time.time()print(e-t)

输出结果：

所以，如果list深层次拷贝绕不过去，就用pickle吧！
哎~·