SELENIUM&PHANTOMJS 完成爬取网络代理

源码

#!/usr/bin/env python
#coding:utf-8
"""用selenium&PhantomJS 完成的网络爬虫，最适合使用的情形是爬取有JS的网站，但是用来爬取其他网站同样给力"""

from selenium import webdriver
from myLog import MyLog as mylog
import sys
reload(sys)
sys.setdefaultencoding('utf-8')

class Item(object):           #用于存储代理的属性
    ip = None
    port = None
    anonymouns = None
    type = None
    support = None
    local = None
    speed = None

class GetProxy(object):

    def __init__(self):
        self.starturl = "http://www.kuaidaili.com/free/inha/"  #目标网址
        self.log = mylog()                                       #二次封装的日志模块
        self.urls = self.getUrls()
        self.proxyList = self.getProxyList(self.urls)
        self.fileName = 'proxy.txt'
        self.saveFile(self.fileName,self.proxyList)

    def getUrls(self):                      ##生成目标页面代理的网页页面URL
        urls = []
        for i in xrange(1,3):
            url = self.starturl + str(i)
            urls.append(url)
            self.log.info('get URL %s to urls' %url)
        return urls

    def getProxyList(self,urls):           ##分析网页源代码
        browser = webdriver.PhantomJS()     ## 使用selenium&PhantomJS
        proxyList = []
        for url in urls:
            browser.get(url)                #打开网页
            browser.implicitly_wait(5)      #智能等待加载JS模块
            elements = browser.find_elements_by_xpath('//tbody/tr')    #使用xpath来解析网页源码
            for element in elements:
                item = Item()
                item.ip = element.find_element_by_xpath('./td[1]').text
                item.port = element.find_element_by_xpath('./td[2]').text
                item.anonymouns = element.find_element_by_xpath('./td[3]').text
                item.type = element.find_element_by_xpath('./td[4]').text
                item.support = element.find_element_by_xpath('./td[5]').text
                item.local = element.find_element_by_xpath('./td[6]').text
                item.speed = element.find_element_by_xpath('./td[7]').text
                proxyList.append(item)
                self.log.info('add proxy %s: %s to list'%(item.ip,item.port))
        browser.quit()                                #浏览器的退出
        return proxyList

    def saveFile(self,fileName,proxyList):            #生成文件
        self.log.info('add all proxy to %s'%self.fileName)
        with open(fileName,'w') as fp:
            for item in proxyList:
                # print item.ip
                fp.write(item.ip+'\t')
                fp.write(item.port + '\t')
                fp.write(item.anonymouns + '\t')
                fp.write(item.type + '\t')
                fp.write(item.support + '\t')
                fp.write(item.local + '\t')
                fp.write(item.speed + '\n')
                self.log.info('write %s:%s to file successfuly......'%(item.ip,item.port))

if __name__ == "__main__":
    USE = GetProxy()

 

日志输出

 

#!/usr/bin/env python
#coding:utf-8
#date 2017_11_30
#author chenjisheng

import logging
import getpass

class MyLog(object):
    '''此类用来封装已有logging类，方便自己使用'''

    def __init__(self):
        self.user = getpass.getuser()                       #获取使用的用户
        self.logger = logging.getLogger(self.user)          #初始化一个logger 对象
        self.logger.setLevel(logging.DEBUG)                 #设置logger 的日志级别
        self.logfile = "progress_log.log"                 #设置日志保存的文件
        self.formates = logging.Formatter(
            "[%(asctime)s] - USERNAME:[%(name)s] - [%(levelname)s] - %(message)s"
        )                                                   #设置日志的格式
        self.Hand = logging.FileHandler(self.logfile)       #设置日志的输出为文件
        self.Hand.setFormatter(self.formates)               #使用日志的格式
        self.Hand.setLevel(logging.ERROR)                   #设置输出为文件的日志级别

        self.HandStream = logging.StreamHandler()           #设置日志的输出格式为console
        self.HandStream.setFormatter(self.formates)         #使用日志的格式
        self.HandStream.setLevel(logging.DEBUG)             #设置输出为窗口的日志级别

        self.logger.addHandler(self.HandStream)             #增加日志的输出对象给logger(负责发送消息)
        self.logger.addHandler(self.Hand)                   #增加日志的输出对象给logger


    def debug(self,messages):
        self.logger.debug(messages)

    def info(self,messages):
        self.logger.info(messages)

    def warning(self,messages):
        self.logger.warning(messages)

    def error(self,messages):
        self.logger.error(messages)

    def critical(self,messages):
        self.logger.critical(messages)

if __name__ == "__main__":
    mylog = MyLog()
    mylog.debug("I'm debug")
    mylog.info("I'm info")
    mylog.warning("I'm warning")
    mylog.error("I'm error ")
    mylog.critical("I'm critical")