python 性能测试（1）-- % vs + vs str.join vs +=

54yuri

浏览: 28342 次
性别:
来自: 北京

最近访客更多访客>>

kliyufei

colin115

xieyulin19

博主相关

博客

微博

相册

留言

关于我

文章分类

社区版块

存档分类

博客分类：

python

python

网上广为流传着"不要使用+" 连接字符串的“经验”，特别是这篇文章当中提到了
http://www.oschina.net/question/129540_61768

引用

要使用 out = "<html>%s%s%s%s</html>" % (head, prologue, query, tail)
而避免 out = "<html>" + head + prologue + query + tail + "</html>"

-------

实际上果真如此吗？连接字符串主要有四种用法：

%
+
str.join
+=

我的理解是这样的
% 和 + 都是应付参数个数已知的、固定的，比如已经知道了a,b,c 三个参数了，那么a+b+c
str.join 和 += 主要是应付参数个数未知的、不固定的，比如参数的一个列表当中所有的字符串，这个列表长度未知；当然他们也可以应付参数个数固定的情况

我的开发环境是：
CPU: Intel(R) Xeon(R) E5520 @ 2.27GHz
Memory： 2G
OS: Ubuntu 12.04

第1回合使用
解释器：python 2.7.3
'use % to join' run 1000000 times, spent 2.045571 seconds
'use + ' run 1000000 times, spent 2.238076 seconds

'use str.join ' run 1000000 times, spent 1.816502 seconds
'use += ' run 1000000 times, spent 3.134726 seconds
-----
long items test:
'use % to join' run 1000000 times, spent 3.631465 seconds
'use + ' run 1000000 times, spent 3.203789 seconds

'use str.join ' run 1000000 times, spent 2.474077 seconds
'use += ' run 1000000 times, spent 3.927131 seconds

第2回合使用 pypy, 鉴于使用pypy的人多了起来，这回是使用ubuntu12 自带的 pypy 1.8
解释器：pypy 1.8 (== python 2.7.2)
short items test:
'use % to join' run 1000000 times, spent 1.637780 seconds
'use + ' run 1000000 times, spent 1.498548 seconds

'use str.join ' run 1000000 times, spent 2.204772 seconds
'use += ' run 1000000 times, spent 2.955786 seconds
-----
long items test:
'use % to join' run 1000000 times, spent 3.570450 seconds
'use + ' run 1000000 times, spent 1.730548 seconds

'use str.join ' run 1000000 times, spent 2.914669 seconds
'use += ' run 1000000 times, spent 5.924469 seconds

第3回合使用最新的PyPy 2.0 beta1
解释器：pypy 2.0 beta1 (== python 2.7.3)
short items test:
'use % to join' run 1000000 times, spent 0.749635 seconds
'use + ' run 1000000 times, spent 0.423104 seconds

'use str.join ' run 1000000 times, spent 1.319862 seconds
'use += ' run 1000000 times, spent 1.985512 seconds
-----
long items test:
'use % to join' run 1000000 times, spent 2.512883 seconds
'use + ' run 1000000 times, spent 0.414724 seconds

'use str.join ' run 1000000 times, spent 1.934615 seconds
'use += ' run 1000000 times, spent 4.973543 seconds

可以下个结论了：
在纯python上，无论长字符串还是短字符串: 参数个数已知 or 参数个数未知, 都用join吧
如果是pypy, 无论长字符串还是短字符串: 参数个数已知用+ ，未知用join吧； BTW， pypy 2.0beta 的速度确实挺快的

无论哪种情况+哪种解释器，的确是不要用 "+=" ！

我的测试代码如下(特别感谢提醒“暗能量有点甜”http://www.weibo.com/kyuseishu 的提示，代码更简洁了)：

#!/usr/bin/env python
import time


def show_run_time(name, count, func, args):
    start = time.time()
    for i in xrange(count):
        apply(func, [],  args)
 
    print("'%s' run %d times, spent %.6f seconds"%(name, count, time.time()-start))


if __name__ == '__main__':

    count = 1000 * 1000
    args1 = { #short items
        "a": "1"*3,
        "b": "2"*4, 
        "c": "3"*5,
        "d": "4"*6,
        "e": "5"*7,
        "f": "6"*8,
        "g": "7"*9,
        "h": "8"*10,
    }
    

    args2 = { #long items
        "a": "1"*101,
        "b": "2"*131, 
        "c": "3"*161,
        "d": "4"*191,
        "e": "5"*221,
        "f": "6"*251,
        "g": "7"*281,
        "h": "8"*311,
    }


    def test_str_format(a, b, c, d, e, f, g, h):
        out1 = "<html>%s%s%s%s%s%s%s%s</html>" % (a, b, c, d, e, f, g, h)
        out2 = "<table>%s%s%s%s%s%s%s%s</table>" % (h, b, c, d, e, f, g, a)
        return  ("%s | %s"%(out1, out2))

    def test_plus(a, b, c, d, e, f, g, h):
        out1 = "<html>" + a + b  + c + d + e + f + g + b + "</html>"
        out2 = "<table>" + h + b  + c + d + e + f + g + a + "</table>"
        return  out1 + " | " + out2

    def test_plus_and_equal(a, b, c, d, e, f, g, h):
        out = "<html>"
        for i in [a, b, c, d, e, f, g, h]:
             out += i
        out += "</html> | <table>"

        for i in [h, b, c, d, e, f, g, a]:
             out += i
        out += "</table>"

        return out


    def test_join(a, b, c, d, e, f, g, h):
        out1 = "".join(["<html>", a, b, c, d, e, f, g, h, "</html>"])
        out2 = "".join(["<table>", h, b, c, d, e, f, g, a, "</table>"])
        return  " | ".join((out1, out2))

    print "short items test:"
    show_run_time("use % to join", count, test_str_format, args1)
    show_run_time("use + ", count, test_plus, args1)

    show_run_time("use str.join ", count, test_join, args1)
    show_run_time("use += ", count, test_plus_and_equal, args1)

    print "-----"

    print "long items test:"
    show_run_time("use % to join", count, test_str_format, args2)
    show_run_time("use + ", count, test_plus, args2)

    show_run_time("use str.join ", count, test_join, args2)
    show_run_time("use += ", count, test_plus_and_equal, args2)

分享到：

MIT、GPL、LPGL、BSD 的区别 | python 当中类的实例也能当函数使？靠__ca ...

2013-01-15 09:48
浏览 1448
评论(0)
分类:编程语言
查看更多

发表评论

您还没有登录,请您登录后再发表评论

最近访客更多访客>>

博主相关

文章分类

社区版块

存档分类

最新评论