今天我们使用python来搭建简易的搜索引擎。
搜索引擎的本质其实就是对数据的预处理,分词构建索引和查询。
(这边我们默认所有的数据都是utf-8的数据类型)
我们在一个网站上去获取所有的URL:
def crawl(pages,depth=2):
for i in range(depth):
newpages = set()
for page in pages:
try:
c = urllib.request.urlopen(page)
except:
print('Invaild page:',page)
continue
soup = bs4.BeautifulSoup(c.read())
links = soup('a')
for link in links:
if('href' in dict(link.attrs)):
url = urllib.urljoin(page,link['href'])
if url.find("'")!=-1:continue
url = url.split('#')[0]
if url[0:3]=='http':
newpages.add(url)
pages = newpages
通过一个循环抓取当前页面上所有的链接,我们尽可能多的去抓取链接,之所以选择set而不使用list是防止重复的现象,我们可以将爬取的的网站存放到文件或者MySQL或者是MongoDB里。
output = sys.stdout
outputfile = open('lujing.txt', 'w')
sys.stdout = outputfile
list = GetFileList(lujing, [])
将生成的路径文件lujing.txt读取,并按照路径文件对文本处理
# 将生成的路径文件lujing.txt读取,并按照路径文件对文本处理,去标签
for line in open("lujing.txt"):
print(line)
# line=line[0:-2]
line1 = line[0:12]
line2 = line[13:16]
line3 = line[17:-1]
line4 = line[17:-6]
line = line1 + '\\' + line2 + '\\' + line3
print(line4)
path = line
fb = open(path, "rb")
data = fb.read()
bianma = chardet.detect(data)['encoding'] # 获取当前文件的编码方式,并按照此编码类型处理文档
page = open(line, 'r', encoding=bianma, errors='ignore').read()
dr = re.compile(r'<[^>]+>', re.S) # 去HTML标签
dd = dr.sub('', page)
print(dd)
fname = 'TXT' + "\\" + line4 + ".txt"
# print(fname)
f = open(fname, "w+", encoding=bianma) # 将去标签的文件写到文件夹内,并按照原命名以txt文档方式保存
# fo=open(fname,"w+")
f.write(dd)
下面我们进行分词索引:
因为大家都比较熟悉sql语句那我在这里就写成MySQL的版本了,如果需要mongodb的可以私信公众号。
import jieba
import chardet
import pymysql
import importlib, sys
importlib.reload(sys)
# 如果使用MongoDB
# from pymongo import MongoClient
# #data processing
# client = MongoClient('localhost',27017)
# apiDB = client['urlDB'] #serverDB_name:test_nodedata
# questionnaires = apiDB['weburl']
# data = list(questionnaires.find())
conn = pymysql .connect(host="localhost",user="root",
password="123456",db="suoyin",port=3307)
conn.text_factory = str
c = conn.cursor()
c.execute('drop table doc')
c.execute('create table doc (id int primary key,link text)')
c.execute('drop table word')
c.execute('create table word (term varchar(25) primary key,list text)')
conn.commit()
conn.close()
def Fenci():
num = 0
for line in open("url.txt"):
lujing = line
print(lujing)
num += 1
print(line)
line = line[17:-5]
print(line)
line = 'TXT' + '\\' + line + 'Txt' # line为文件位置
print(line) # 文件名称
path = line
fb = open(path, "rb")
data = fb.read()
bianma = chardet.detect(data)['encoding'] # 获取文件编码 print(bianma)
# page = open(line, 'r', encoding=bianma, errors='ignore').read()
# page1=page.decode('UTF-8')
if bianma == 'UTF-16':
data = data.decode('UTF-16')
data = data.encode('utf-8')
word = jieba.cut_for_search(data)
seglist = list(word)
print(seglist)
# 创建数据库
c = conn.cursor() # 创建游标
c.execute('insert into doc values(?,?)', (num, lujing))
# 对每个分出的词语建立词表
for word in seglist:
# print(word)
# 检验看看这个词语是否已存在于数据库
c.execute('select list from word where term=?', (word,))
result = c.fetchall()
# 如果不存在
if len(result) == 0:
docliststr = str(num)
c.execute('insert into word values(?,?)', (word, docliststr))
# 如果已存在
else:
docliststr = result[0][0] # 得到字符串
docliststr += ' ' + str(num)
c.execute('update word set list=? where term=?', (docliststr, word))
conn.commit()
conn.close()
Fenci()
最后一步,查询:
import pymsql
import jieba
import math
conn = pymysql .connect(host="localhost",user="root",
password="123456",db="suoyin",port=3307)
c = conn.cursor()
c.execute('select count(*) from doc')
N = 1 + c.fetchall()[0][0] # 文档总数
target = input('请输入搜索词:')
seggen = jieba.cut_for_search(target)
score = {} # 文档号:匹配度
for word in seggen:
print('得到查询词:', word)
# 计算score
tf = {} # 文档号:文档数
c.execute('select list from word where term=?', (word,))
result = c.fetchall()
if len(result) > 0:
doclist = result[0][0]
doclist = doclist.split(' ')
# 把字符串转换为元素为int的list
doclist = [int(x) for x in doclist]
# 当前word对应的df数
df = len(set(doclist))
idf = math.log(N / df)
print('idf:', idf)
for num in doclist:
if num in tf:
tf[num] = tf[num] + 1
else:
tf[num] = 1
# tf统计结束,现在开始计算score
for num in tf:
if num in score:
# 如果该num文档已经有分数了,则累加
score[num] = score[num] + tf[num] * idf
else:
score[num] = tf[num] * idf
sortedlist = sorted(score.items(), key=lambda d: d[1], reverse=True)
cnt = 0
for num, docscore in sortedlist:
cnt = cnt + 1
c.execute('select link from doc where id=?', (num,))
url = c.fetchall()[0][0]
print("Result Ranking:", cnt)
print('url:', url, 'match degree:', docscore)
if cnt > 20:
break
if cnt == 0:
print('No result')
搞定。