ProxyRefreshSchedule.py 3.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111
  1. # -*- coding: utf-8 -*-
  2. # !/usr/bin/env python
  3. """
  4. -------------------------------------------------
  5. File Name: ProxyRefreshSchedule.py
  6. Description : 代理定时刷新
  7. Author : JHao
  8. date: 2016/12/4
  9. -------------------------------------------------
  10. Change Activity:
  11. 2016/12/4: 代理定时刷新
  12. 2017/03/06: 使用LogHandler添加日志
  13. 2017/04/26: raw_proxy_queue验证通过但useful_proxy_queue中已经存在的代理不在放入
  14. -------------------------------------------------
  15. """
  16. import sys
  17. import time
  18. import logging
  19. from threading import Thread
  20. from apscheduler.schedulers.background import BackgroundScheduler
  21. sys.path.append('../')
  22. from Util.utilFunction import validUsefulProxy
  23. from Manager.ProxyManager import ProxyManager
  24. from Util.LogHandler import LogHandler
  25. __author__ = 'JHao'
  26. logging.basicConfig()
  27. class ProxyRefreshSchedule(ProxyManager):
  28. """
  29. 代理定时刷新
  30. """
  31. def __init__(self):
  32. ProxyManager.__init__(self)
  33. self.log = LogHandler('refresh_schedule')
  34. def validProxy(self):
  35. """
  36. 验证raw_proxy_queue中的代理, 将可用的代理放入useful_proxy_queue
  37. :return:
  38. """
  39. self.db.changeTable(self.raw_proxy_queue)
  40. raw_proxy_item = self.db.pop()
  41. self.log.info('ProxyRefreshSchedule: %s start validProxy' % time.ctime())
  42. # 计算剩余代理,用来减少重复计算
  43. remaining_proxies = self.getAll()
  44. while raw_proxy_item:
  45. raw_proxy = raw_proxy_item.get('proxy')
  46. if isinstance(raw_proxy, bytes):
  47. # 兼容Py3
  48. raw_proxy = raw_proxy.decode('utf8')
  49. if (raw_proxy not in remaining_proxies) and validUsefulProxy(raw_proxy):
  50. self.db.changeTable(self.useful_proxy_queue)
  51. self.db.put(raw_proxy)
  52. self.log.info('ProxyRefreshSchedule: %s validation pass' % raw_proxy)
  53. else:
  54. self.log.info('ProxyRefreshSchedule: %s validation fail' % raw_proxy)
  55. self.db.changeTable(self.raw_proxy_queue)
  56. raw_proxy_item = self.db.pop()
  57. remaining_proxies = self.getAll()
  58. self.log.info('ProxyRefreshSchedule: %s validProxy complete' % time.ctime())
  59. def refreshPool():
  60. pp = ProxyRefreshSchedule()
  61. pp.validProxy()
  62. def batchRefresh(process_num=30):
  63. # 检验新代理
  64. pl = []
  65. for num in range(process_num):
  66. proc = Thread(target=refreshPool, args=())
  67. pl.append(proc)
  68. for num in range(process_num):
  69. pl[num].daemon = True
  70. pl[num].start()
  71. for num in range(process_num):
  72. pl[num].join()
  73. def fetchAll():
  74. p = ProxyRefreshSchedule()
  75. # 获取新代理
  76. p.refresh()
  77. def run():
  78. scheduler = BackgroundScheduler()
  79. # 不用太快, 网站更新速度比较慢, 太快会加大验证压力, 导致raw_proxy积压
  80. scheduler.add_job(fetchAll, 'interval', minutes=10, id="fetch_proxy")
  81. scheduler.add_job(batchRefresh, "interval", minutes=1) # 每分钟检查一次
  82. scheduler.start()
  83. fetchAll()
  84. while True:
  85. time.sleep(3)
  86. if __name__ == '__main__':
  87. run()