我新手上路，用正則表達式去根據(jù)你給出的例子的數(shù)據(jù)去匹配的，我是根據(jù)two給出的值去匹配one，所以前提是two.csv文件不能太大，我剛剛測試了一千萬條（大概500M左右,每一百條有一條匹配）的數(shù)據(jù)用了4分鐘，效率不是很高，程序是默認編碼，如果要轉碼效率可能會再差點，數(shù)據(jù)真是超大的話我覺得還是用c語言去寫工作效率會高點

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import re
import os
#用來匹配two.csv列2和列3的值
reg_list2 = re.compile(r'.+?,.+?,\s*(.+?)\s*[;]*\s*,\s*(\d+)\s*')
#用來匹配one.csv列2的值
reg_name = re.compile(r'.+?,.+?,\s*(.+?)\s*[;]*,')
#匹配頭
reg_title = re.compile(r'\s*\S+.+')
def main(argv):
    try:
        fptwo = open("two.csv", "r")
        fpone = open("one.csv", "r")
        fpnew = open("result.csv", "a")
    except UnicodeDecodeError:
        exit(1)
    text = fptwo.read()
    fptwo.close()
    
    data_two = [(re.escape(x),n) for x,n in reg_list2.findall(text)]
    otime = datetime.datetime.now()
    title = None
    new = "列4\n" #新列名字
    tell = fpone.tell()
    nu = 0    
    
    while True:
        line = fpone.readline()
        mate = reg_name.match(line)

        if not title:
            mate = reg_title.search(line)
            if mate:
                title = mate.group(0)
                title = (title + ",").ljust(len(title)+15) + new
                fpnew.write(title)
                continue
        
        if mate:
            name = mate.group(1)
            for item in data_two:
                if re.search(item[0],name):
                    line = line[:-1]
                    line = (line + ",").ljust(len(title))  + item[1] + "\n"
                    print ("Pos L:", nu, "name: ", name)
        
        fpnew.write(line)
        nu += 1
        if tell == fpone.tell():
            break
        else:
            tell = fpone.tell()
    fpone.close()
    fpnew.close()
    return 0

if __name__ == '__main__':
    import sys
    sys.exit(main(sys.argv))

# coding: utf-8

import pandas as pd

lst1 = [
    [1, 'a', 'ww.ok;ww.ant', 'anything'],
    [2, 'e', 'ww.kdi', 'ihy'],
    [3, 'se', 'ww.sdd', 'sld'],
    [4, 'sd', 'ww.akd', 'sdjfa']
]

lst2 = [
    [1, 'sd', 'ww.ant', 1245],
    [2,'2e3', 'ww.kdi', 432],
    [3,'de', 'ww.sdd', 232]
]

df1 = pd.DataFrame(lst1, columns=['A1', 'B1', 'C1', 'D1'])
df2 = pd.DataFrame(lst2, columns=['A2', 'B2', 'C2', 'D2'])

#列拆成多行
df11 = df1.join(df1['C1'].str.split(';', expand=True).stack().reset_index(level=1, drop=True).rename('E'))
df22 = df2.join(df2['C2'].str.split(';', expand=True).stack().reset_index(level=1, drop=True).rename('E'))

#左連接, 并只取需要的字段
df3 = pd.merge(df11, df22, how='left', on='E').loc[:, ['A1', 'B1', 'C1', 'D1', 'D2']]

#去重
print df3.groupby(['A1', 'B1', 'C1', 'D1'], as_index=False).max()