我有一列有1000万个字符串。字符串中的字符需要以某种方式重新排列。
原始字符串:AAA01188P001
随机播放的字符串:188A1A0AP001
现在我有一个for循环运行,该循环接受每个字符串并重新定位每个字母,但这需要几个小时才能完成。有没有更快的方法来达到这个结果?
这是for循环。
for i in range(0, len(OrderProduct)):
s = list(OrderProduct['OrderProductId'][i])
a = s[1]
s[1] = s[7]
s[7] = a
a = s[3]
s[3] = s[6]
s[6] = a
a = s[2]
s[2] = s[3]
s[3] = a
a = s[5]
s[5] = s[0]
s[0] = a
OrderProduct['OrderProductId'][i] = ''.join(s)
答案 0 :(得分:5)
我使用不同的方法进行了一些性能测试:
这是我获得1000000次随机播放的结果:
188A1AA0P001 usefString 0.518183742
188A1AA0P001 useMap 1.415851829
188A1AA0P001 useConcat 0.5654986979999999
188A1AA0P001 useFormat 0.800639699
188A1AA0P001 useJoin 0.5488918539999998
基于此,带有硬编码子字符串的格式字符串似乎是最快的。
这是我用来测试的代码:
def usefString(s): return f"{s[5:8]}{s[0]}{s[4]}{s[1:4]}{s[8:]}"
posMap = [5,6,7,0,4,1,2,3,8,9,10,11]
def useMap(s): return "".join(map(lambda i:s[i], posMap))
def useConcat(s): return s[5:8]+s[0]+s[4]+s[1:4]+s[8:]
def useFormat(s): return '{}{}{}{}{}'.format(s[5:8],s[0],s[4],s[1:4],s[8:])
def useJoin(s): return "".join([s[5:8],s[0],s[4],s[1:4],s[8:]])
from timeit import timeit
count = 1000000
s = "AAA01188P001"
t = timeit(lambda:usefString(s),number=count)
print(usefString(s),"usefString",t)
t = timeit(lambda:useMap(s),number=count)
print(useMap(s),"useMap",t)
t = timeit(lambda:useConcat(s),number=count)
print(useConcat(s),"useConcat",t)
t = timeit(lambda:useFormat(s),number=count)
print(useFormat(s),"useFormat",t)
t = timeit(lambda:useJoin(s),number=count)
print(useJoin(s),"useJoin",t)
性能:(由@jezrael添加)
N = 1000000
OrderProduct = pd.DataFrame({'OrderProductId':['AAA01188P001'] * N})
In [331]: %timeit [f'{s[5:8]}{s[0]}{s[4]}{s[1:4]}{s[8:]}' for s in OrderProduct['OrderProductId']]
527 ms ± 16.7 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
In [332]: %timeit [s[5:8]+s[0]+s[4]+s[1:4]+s[8:] for s in OrderProduct['OrderProductId']]
610 ms ± 18.8 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
In [333]: %timeit ['{}{}{}{}{}'.format(s[5:8],s[0],s[4],s[1:4],s[8:]) for s in OrderProduct['OrderProductId']]
954 ms ± 76.6 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
In [334]: %timeit ["".join([s[5:8],s[0],s[4],s[1:4],s[8:]]) for s in OrderProduct['OrderProductId']]
594 ms ± 10.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
答案 1 :(得分:2)
如果逻辑一致,您能否只用切片来重建字符串?
s = OrderProduct['OrderProductId'][i]
new_s = s[5]+s[7]+s[1:2]+s[6]+s[4]+s[0]+s[3]+s[1]
或作为格式字符串:
new_s = '{}{}{}{}{}{}{}'.format(s[5],s[7]...)
编辑:+1为戴夫建议''.join()列表与串联。
答案 2 :(得分:1)
如果您只想对字符串进行混洗(没有特殊的逻辑),则可以通过以下几种方法进行:
使用string_utils:
import string_utils
print string_utils.shuffle("random_string")
使用内置方法:
import random
str_var = list("shuffle_this_string")
random.shuffle(str_var)
print ''.join(str_var)
使用numpy:
import numpy
str_var = list("shuffle_this_string")
numpy.random.shuffle(str_var)
print ''.join(str_var)
但是,如果您需要采用某种逻辑(例如,将每个元素放在特定位置),则可以这样做:
s = 'some_string'
s = ''.join([list(s)[i] for i in [1,6,2,7,9,4,0,8,5,10,3]])
print(s)
输出:
otmrn_sisge
如果仍然花费太长时间,则可以使用多处理。像这样:
from multiprocessing import Pool
p = Pool(4) # 4 is the number of workers. usually is set to the number of CPU cores
def shuffle_str(s):
# do shuffling here, and return
list_of_strings = [...]
list_of_results = p.map(shuffle_str, list_of_strings)