from bs4 import BeautifulSoup
import codecs
import os
import sys
###########################
## 输入地址path，输出地址pathTxt，输出地址里面有个_log.txt，记录结果
## 重新运行前先清空输出地址的所有文件
path='E:\\BaiduYunDownload\\www.cworksafety.com'
pathTxt='E:\\BaiduYunDownload\\www.cworksafety.com.Txt'
##
##########################
if not os.path.exists(pathTxt):
os.mkdir(pathTxt)
print('目录已新建'+pathTxt)
else:
print('目录已存在'+pathTxt)
def write_lable(soup,lable,file_w,file_log):
soup_list=soup.find_all(lable)
if soup_list and any(soup_list):
n=0
for p in soup_list:
ps=p.stripped_strings
if ps is not None:
#print(ps)
for string in ps:
n=n+1
file_w.write(string)
if n:
file_log.write(' has been extracted with '+str(n)+' <'+lable+'> \r\n')
else:
file_log.write(' has no <'+lable+'>\r\n')
else:
file_log.write(' has no <'+lable+'>\r\n')
for root,dirs,files in os.walk(path):
file_log=codecs.open(pathTxt+'\\'+'_log.txt','a',encoding='utf8')
for file in files:
soup=BeautifulSoup(codecs.open(root+'\\'+file,encoding='utf8'))
file_w=codecs.open(pathTxt+'\\'+file.split('.')[0],'a',encoding='utf8')
file_log.write(file.split('.')[0]+':'+'\r\n')
######################################
## 这部分根据具体的网页，自己修改，查看效果，下面这三个效果还不错
write_lable(soup,'title',file_w,file_log)
write_lable(soup,'p',file_w,file_log)
write_lable(soup,'font',file_w,file_log)
##
#####################################
file_w.close()
file_log.close()
