知乎用户地域分析

 1 # -*- coding: utf-8 -*-
 2 
 3 import re
 4 import sqlite3
 5 import sys
 6 reload(sys)
 7 sys.setdefaultencoding(‘utf-8‘)
 8 
 9 class CountCollect(object):
10     """docstring for CountCollect"""
11     
12     pro_city = {}
13     city_pro = {}
14     prolist = []
15     citylist = []
16     conn = sqlite3.connect(‘zhihuuser.db‘)
17     conn.text_factory = str
18     cur = conn.cursor()
19 
20     def getss(self):
21         shengshi = open(‘shengshi.txt‘,‘r‘)
22         for line in shengshi.readlines():
23             province = line.split(":")[0]
24             # print province
25             self.prolist.append(province)
26             self.pro_city.setdefault(province,{})
27             self.pro_city[province].setdefault(‘number‘,0)            
28             self.pro_city[province].setdefault(‘cities‘,{})
29             cities = line.split(‘:‘)[-1].split(‘\n‘)[0].split(‘ ‘)
30             # print cities
31             for city in cities:
32                 # print city,type(city)
33                 self.citylist.append(city)
34                 self.pro_city[province][‘cities‘].setdefault(city,0)
35                 self.city_pro[city] = province
36         # print self.prolist
37         # for i in self.citylist:
38         #     print i
39         
40 
41     def getlocation(self):
42         # 从数据库中检索出数据
43         result = self.cur.execute(‘select location from userstb 
44             where location is not NULL‘).fetchall()
45         string = []
46         for line in result:
47             try:
48                 string.append(line[0].encode("gbk"))
49             except Exception, e:
50                 string.append(line[0])
51         ww = open(‘location.txt‘,‘a‘)
52         for x in string:
53             ww.write(x)
54             ww.write(‘\n‘)
55         # 符合省
56         for pro in self.prolist:
57             for x in string:
58                 if re.search(pro,x):
59                     print "%s----%s" %(pro,x)
60                     self.pro_city[pro][‘number‘] += 1
61                 # 符合市
62         for cit in self.citylist:
63             for x in string:
64                 if re.search(cit,x):
65                     print "%s----%s" %(cit,x)
66                     pro = self.city_pro[cit]
67                     self.pro_city[pro][‘number‘] += 1
68                     self.pro_city[pro][‘cities‘][cit] += 1
69         
70         
71         keys = self.pro_city.keys()
72         r = open(‘result2.txt‘,‘a‘)
73         for key in keys:
74             # print key,type(key)
75             # print cmp(‘直辖市‘,key.decode(‘gbk‘))
76             if cmp(key.decode(‘gbk‘),‘直辖市‘)==0 or 
77                         cmp(key.decode(‘gbk‘),‘港澳台‘)==0:
78                 for city in self.pro_city[key][‘cities‘]:
79                     print city, self.pro_city[key][‘cities‘][city]
80                     r.write("{name:‘")
81                     r.write(city)
82                     r.write("‘,value: ")
83                     r.write(str(self.pro_city[key][‘cities‘][city]))
84                     r.write("},")
85                     r.write(‘\n‘)
86             else:
87                 print key, self.pro_city[key][‘number‘]
88                 r.write("{name:‘")
89                 r.write(key)
90                 r.write("‘,value: ")
91                 r.write(str(self.pro_city[key][‘number‘]))
92                 r.write("},")
93                 r.write(‘\n‘)
94 
95 
96 if __name__ == ‘__main__‘:
97     cc = CountCollect()
98     cc.getss() #生成省市地域字典
99     cc.getlocation() #分类统计