Python实现从url中提取域名的几种方法
从url中找到域名,首先想到的是用正则,然后寻找相应的类库。用正则解析有很多不完备的地方,url中有域名,域名后缀一直在不断增加等。通过google查到几种方法,一种是用Python中自带的模块和正则相结合来解析域名,另一种是使第三方用写好的解析模块直接解析出域名。
要解析的url
urls = ["http://meiwen.me/src/index.html",
"http://1000chi.com/game/index.html",
"http://see.xidian.edu.cn/cpp/html/1429.html",
"https://docs.python.org/2/howto/regex.html",
"""https://www.google.com.hk/search?client=aff-cs-360chromium&hs=TSj&q=url%E8%A7%A3%E6%9E%90%E5%9F%9F%E5%90%8Dre&oq=url%E8%A7%A3%E6%9E%90%E5%9F%9F%E5%90%8Dre&gs_l=serp.3...74418.86867.0.87673.28.25.2.0.0.0.541.2454.2-6j0j1j1.8.0....0...1c.1j4.53.serp..26.2.547.IuHTj4uoyHg""",
"file:///D:/code/echarts-2.0.3/doc/example/tooltip.html",
"http://api.mongodb.org/python/current/faq.html#is-pymongo-thread-safe",
"https://pypi.python.org/pypi/publicsuffix/",
"http://127.0.0.1:8000"
]
使用urlparse+正则的方式
import re
from urlparse import urlparse
topHostPostfix = (
'.com','.la','.io','.co','.info','.net','.org','.me','.mobi',
'.us','.biz','.xxx','.ca','.co.jp','.com.cn','.net.cn',
'.org.cn','.mx','.tv','.ws','.ag','.com.ag','.net.ag',
'.org.ag','.am','.asia','.at','.be','.com.br','.net.br',
'.bz','.com.bz','.net.bz','.cc','.com.co','.net.co',
'.nom.co','.de','.es','.com.es','.nom.es','.org.es',
'.eu','.fm','.fr','.gs','.in','.co.in','.firm.in','.gen.in',
'.ind.in','.net.in','.org.in','.it','.jobs','.jp','.ms',
'.com.mx','.nl','.nu','.co.nz','.net.nz','.org.nz',
'.se','.tc','.tk','.tw','.com.tw','.idv.tw','.org.tw',
'.hk','.co.uk','.me.uk','.org.uk','.vg', ".com.hk")
regx = r'[^\.]+('+'|'.join([h.replace('.',r'\.') for h in topHostPostfix])+')$'
pattern = re.compile(regx,re.IGNORECASE)
print "--"*40
for url in urls:
parts = urlparse(url)
host = parts.netloc
m = pattern.search(host)
res = m.group() if m else host
print "unkonw" if not res else res
运行结果如下:
meiwen.me
1000chi.com
see.xidian.edu.cn
python.org
google.com.hk
unkonw
mongodb.org
python.org
127.0.0.1:8000
基本可以接受
urllib来解析域名
import urllib
print "--"*40
for url in urls:
proto, rest = urllib.splittype(url)
res, rest = urllib.splithost(rest)
print "unkonw" if not res else res
运行结果如下:
meiwen.me
1000chi.com
see.xidian.edu.cn
docs.python.org
www.google.com.hk
unkonw
api.mongodb.org
pypi.python.org
127.0.0.1:8000
会把www.也带上,还需要进一步解析才可以
使用第三方模块 tld
from tld import get_tld
print "--"*40
for url in urls:
try:
print get_tld(url)
except Exception as e:
print "unkonw"
运行结果:
meiwen.me
1000chi.com
xidian.edu.cn
python.org
google.com.hk
unkonw
mongodb.org
python.org
unkonw
结果都可以接受
其他可以使用的解析模块:
tld
tldextract
publicsuffix