在前一篇博客中,爬了谣言百科中baby类的百科,现在要同时爬所有类别的百科时应该怎么做呢? 无非是添加一个网址list,和一个类别名list,然后进行遍历爬取数据即可! 上代码:
在前一篇博客中,爬了谣言百科中baby类的百科,现在要同时爬所有类别的百科时应该怎么做呢?
无非是添加一个网址list,和一个类别名list,然后进行遍历爬取数据即可!
上代码:
# -*- coding: utf-8 -*-"""
Spyder Editor
This is a temporary script file.
"""
from urllib import request
from bs4 import BeautifulSoup
import re
#import sys
import codecs
if __name__ == "__main__":
urlList = ['http://www.yaoyanbaike.com/category/health','http://www.yaoyanbaike.com/category/food','http://www.yaoyanbaike.com/category/baby','http://www.yaoyanbaike.com/category/science','http://www.yaoyanbaike.com/category/life','http://www.yaoyanbaike.com/category/legend','http://www.yaoyanbaike.com/category/news','http://www.yaoyanbaike.com/category/car','http://www.yaoyanbaike.com/category/love','http://www.yaoyanbaike.com/category/sexual']
fileNameList = ['health','food','baby','science','life','legend','news','car','love','sexual']
categoryNum = 0
while(categoryNum < 9):
text_file_number = 0 # 同一类新闻下的索引数
number = 1 # 同类别新闻不同页面下的索引数
while (number <= 6):
if number==1: # 第一个新闻下地址是baby不是baby_数字所以要区分判断一下
get_url = urlList[categoryNum] + '.html'
else:
get_url = urlList[categoryNum] + '_' + str(number) + '.html' #这个是baby_数字,number就是目录索引数
head = {} #设置头
head['User-Agent'] = 'Mozilla/5.0 (Linux; Android 4.1.1; Nexus 7 Build/JRO03D) AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18.0.1025.166 Safari/535.19'
# 模拟浏览器模式,定制请求头
download_req_get = request.Request(url = get_url, headers = head)# 设置Request
download_response_get = request.urlopen(download_req_get)# 设置urlopen获取页面所有内容
download_html_get = download_response_get.read().decode('UTF-8','ignore') # UTF-8模式读取获取的页面信息标签和内容
soup_texts = BeautifulSoup(download_html_get, 'lxml') # BeautifulSoup读取页面html标签和内容的信息
for link in soup_texts.find_all(["a"]):
print(str(text_file_number)+ " " + str(number) + " "+ link.get('href'))# 打印文件地址用于测试
s = link.get('href')
if s.find("/a/") == -1:
print("无效网址") # 只有包含"/a/"字符的才是有新闻的有效地址
else:
download_url = link.get('href')
head = {}
head['User-Agent'] = 'Mozilla/5.0 (Linux; Android 4.1.1; Nexus 7 Build/JRO03D) AppleWebKit/535.19 (KHTML, like Gecko) Chrome/18.0.1025.166 Safari/535.19'
download_req = request.Request(url = "http://www.yaoyanbaike.com" + download_url, headers = head)
print("http://www.yaoyanbaike.com" + download_url)
download_response = request.urlopen(download_req)
download_html = download_response.read().decode('UTF-8','ignore')
soup_texts = BeautifulSoup(download_html, 'lxml')
texts = soup_texts.find_all('article')
soup_text = BeautifulSoup(str(texts), 'lxml')
p = re.compile("<[^>]+>")
text=p.sub("", str(soup_text))# 去除页面标签
f1 = codecs.open('F:\\test\\' + fileNameList[categoryNum] + '\\' +str(text_file_number)+'.txt','w','UTF-8') # 将信息存储在本地
f1.write(text)
f1.close()
text_file_number = text_file_number + 1
number = number + 1
categoryNum = categoryNum + 1
当然,自己应提前建好类别的文件夹,如图:
然后运行即可得到数据!