#小练习 SGMLParser 解析 HTML 分类： HTMLParser 2013-11-12 15:52 472人阅读 评论(0) 收藏
#coding:utf-8
from sgmllib import SGMLParser
class GetIdList(SGMLParser):
def reset(self):
self.IDlist=[]
self.flag=False
self.getdata=False
self.verbatim = 0
SGMLParser.reset(self)
def start_div(self,attrs):
if self.flag==True:
self.verbatim+=1 #进入子层div了，层数+1
for k,v in attrs:
if k=="class" and v=="entry-content":#确定进入了<div class='entry-content'>
self.flag=True
def end_div(self):#遇到</div>
if self.verbatim==0:
self.flag=False
if self.flag==True:#退出子层div了，层数减 1
self.verbatim-=1
def start_p(self,attrs):
if self.flag==False:
return
self.getdata=True
def end_p(self):#遇到</p>
if self.getdata:
self.getdata=False
def handle_data(self,text):#处理文本
if self.getdata:
self.IDlist.append(text)
def printID(self):
for i in self.IDlist:
print i.decode("utf-8")
if __name__ == '__main__':
the_page ='''<html>
<head>
<title>test</title>
</head>
<body>
<h1>title</h1>
<div class='entry-content'>
<p>感兴趣内容1</p>
<p>感兴趣内容2</p>
……
<p>感兴趣内容n</p>
<div class='entry-content'>我是来捣乱的2<div class= 'ooxx'>我是来捣乱的3</div></div>
</div>
<div class='content'>
<p>内容1</p>
<p>内容2</p>
……
<p>内容n</p>
</div>
</body>
</html>
'''
gid=GetIdList()
gid.feed(the_page)
gid.printID()
gid.close()
结果：
感兴趣内容1
感兴趣内容2
感兴趣内容n
