添加:匿名性校验

cwjokaka · Sep 28, 2019 · dd9f22f · dd9f22f
1 parent b4e5c06
commit dd9f22f
Show file tree

Hide file tree

Showing 9 changed files with 185 additions and 50 deletions.
diff --git a/main.py b/main.py
@@ -1,44 +1,4 @@
-import asyncio
-import typing
-
-from apscheduler.schedulers.background import BackgroundScheduler
-
-from src.database.sqlite_opt import sqlite_opt
-from src.entity.proxy_entity import ProxyEntity
-from src.log.logger import logger
-from src.spider.spiders import spider_collection
-from setting import WEB_SERVER, VALIDATOR, SPIDER
-from src.validator.validator import validator
-from src.web.web_flask import app
-
-
-def crawl():
-    proxies = []
-    tasks = []
-    for spider_name in SPIDER['list']:
-        tasks.append(spider_collection[spider_name].crawl())
-        # proxies.extend(spider_collection[spider_name].crawl())
-    loop = asyncio.new_event_loop()
-    asyncio.set_event_loop(loop)
-    results = loop.run_until_complete(asyncio.gather(*tasks))
-    loop.close()
-    for proxies_list in results:
-        proxies.extend(proxies_list)
-    # proxies = loop.run_until_complete(asyncio.gather(*tasks))
-    # 持久化
-    save(proxies)
-
-
-def save(proxies: typing.List[ProxyEntity]):
-    for proxy in proxies:
-        sqlite_opt.add_proxy(proxy)
-
+from src.runner import run
 
 if __name__ == '__main__':
-    logger.info('初始化sqlite数据库...')
-    sqlite_opt.init_db()
-    scheduler = BackgroundScheduler()
-    scheduler.add_job(crawl, 'interval', seconds=SPIDER['crawl_interval'])
-    scheduler.add_job(validator.run, 'interval', seconds=VALIDATOR['validate_interval'])
-    scheduler.start()
-    app.run(host=WEB_SERVER['host'], port=WEB_SERVER['port'])
+    run()
diff --git a/setting.py b/setting.py
@@ -1,6 +1,6 @@
 # 代理爬虫配置
 SPIDER = {
-    'crawl_interval': 60,       # 爬取IP代理的间隔(秒)
+    'crawl_interval': 120,       # 爬取IP代理的间隔(秒)
     'list': [                   # 使用的代理爬虫(类名)
         'Spider66Ip',
         'SpiderQuanWangIp',
@@ -14,9 +14,17 @@
 
 # 校验器配置
 VALIDATOR = {
-    'test_url': 'http://www.baidu.com',
+    'test_url': 'http://www.baidu.com',     # 可用校验url
     'request_timeout': 4,           # 校验超时时间
-    'validate_interval': 30
+    'validate_interval': 60         # 校验间隔(秒)
+}
+
+# 匿名性校验配置
+ANONYMITY_VALIDATOR = {
+    'http_test_url': 'http://httpbin.org/get',      # 匿名校验url
+    'https_test_url': 'https://httpbin.org/get',
+    'request_timeout': 4,                           # 校验最大超时时间
+    'interval': 180                                 # 校验间隔(秒)
 }
 
 # 数据库配置

diff --git a/src/database/abs_database.py b/src/database/abs_database.py
@@ -6,12 +6,18 @@ def add_proxy(self, proxy):
     def get_all_proxies(self):
         raise NotImplementedError
 
+    def get_unknown_anonymity_proxies(self):
+        raise NotImplementedError
+
     def increase_reliability(self, url):
         raise NotImplementedError
 
     def reduce_reliability(self, url):
         raise NotImplementedError
 
+    def update_anonymity(self, url, value):
+        raise NotImplementedError
+
     def remove(self, key):
         raise NotImplementedError
 

diff --git a/src/database/sqlite_opt.py b/src/database/sqlite_opt.py
@@ -5,6 +5,7 @@
 from sqlalchemy import create_engine, desc
 from sqlalchemy.orm import sessionmaker
 from src.entity.proxy_entity import ProxyEntity
+from src.enum.common import ProxyCoverEnum
 from src.log.logger import logger
 import sqlite3
 
@@ -35,7 +36,19 @@ def get_all_proxies(self):
             return session.query(ProxyEntity).all()
         except Exception as e:
             logger.exception(e)
-            pass
+        finally:
+            session.close()
+        return []
+
+    def get_unknown_anonymity_proxies(self):
+        session = self._DBSession()
+        try:
+            return (session.query(ProxyEntity)
+                    .filter(ProxyEntity.reliability > 0)
+                    .filter(ProxyEntity.proxy_cover == ProxyCoverEnum.UNKNOWN.value)
+                    .all())
+        except Exception as e:
+            logger.exception(e)
         finally:
             session.close()
         return []
@@ -79,6 +92,21 @@ def reduce_reliability(self, url):
     def remove(self, key):
         return super().remove(key)
 
+    def update_anonymity(self, url, value):
+        conn = self._get_connect()
+        cursor = conn.cursor()
+        try:
+            cursor.execute(f"""
+            UPDATE {DB["table_name"]} SET proxy_cover = {value}
+            WHERE url='{url}'
+            """)
+            conn.commit()
+        except Exception as e:
+            logger.exception(e)
+        finally:
+            cursor.close()
+            conn.close()
+
     def init_db(self):
         conn = self._get_connect()
         cursor = conn.cursor()

diff --git a/src/entity/proxy_entity.py b/src/entity/proxy_entity.py
@@ -1,11 +1,12 @@
 from src.enum.common import ProxyTypeEnum, ProxyCoverEnum
 from sqlalchemy.ext.declarative import declarative_base
 from sqlalchemy import Column, Integer, String
+from setting import DB
 Base = declarative_base()
 
 
 class ProxyEntity(Base):
-    __tablename__ = 'proxy'
+    __tablename__ = DB['table_name']
     url = Column(String(36), primary_key=True)
     # ip = Column(String(20))
     # port = Column(String(5))

diff --git a/src/runner.py b/src/runner.py
@@ -0,0 +1,46 @@
+import asyncio
+import typing
+
+from apscheduler.schedulers.background import BackgroundScheduler
+
+from src.database.sqlite_opt import sqlite_opt
+from src.entity.proxy_entity import ProxyEntity
+from src.log.logger import logger
+from src.spider.spiders import spider_collection
+from setting import WEB_SERVER, VALIDATOR, SPIDER, ANONYMITY_VALIDATOR
+from src.validator.validator import validator
+from src.validator.anonymity_validator import anonymity_validator
+from src.web.web_flask import app
+
+
+def crawl():
+    proxies = []
+    tasks = []
+    for spider_name in SPIDER['list']:
+        tasks.append(spider_collection[spider_name].crawl())
+        # proxies.extend(spider_collection[spider_name].crawl())
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    results = loop.run_until_complete(asyncio.gather(*tasks))
+    loop.close()
+    for proxies_list in results:
+        proxies.extend(proxies_list)
+    # proxies = loop.run_until_complete(asyncio.gather(*tasks))
+    # 持久化
+    save(proxies)
+
+
+def save(proxies: typing.List[ProxyEntity]):
+    for proxy in proxies:
+        sqlite_opt.add_proxy(proxy)
+
+
+def run():
+    logger.info('初始化sqlite数据库...')
+    sqlite_opt.init_db()
+    scheduler = BackgroundScheduler()
+    scheduler.add_job(crawl, 'interval', seconds=SPIDER['crawl_interval'])
+    scheduler.add_job(validator.run, 'interval', seconds=VALIDATOR['validate_interval'])
+    scheduler.add_job(anonymity_validator.run, 'interval', seconds=ANONYMITY_VALIDATOR['interval'])
+    scheduler.start()
+    app.run(host=WEB_SERVER['host'], port=WEB_SERVER['port'])
diff --git a/src/spider/abs_spider.py b/src/spider/abs_spider.py
@@ -10,13 +10,12 @@ def __init__(self, name='unknown') -> None:
         self._name = name
 
     async def crawl(self):
-        res = []
         logger.info(f'{self._name}开始爬取...')
         try:
-            res.extend(await self.do_crawl())
+            return await self.do_crawl()
         except Exception as e:
             logger.exception(f'{self._name}爬取失败:e:{e}')
-        return res
+        return []
 
     async def do_crawl(self) -> List[ProxyEntity]:
         raise NotImplementedError

diff --git a/src/validator/anonymity_validator.py b/src/validator/anonymity_validator.py
@@ -0,0 +1,69 @@
+import asyncio
+import json
+
+import aiohttp
+
+from setting import ANONYMITY_VALIDATOR, HEADERS
+from src.database.sqlite_opt import sqlite_opt
+from src.enum.common import ProxyCoverEnum, ProxyTypeEnum
+from src.log.logger import logger
+
+
+class AnonymityValidator(object):
+
+    urls = {
+        ProxyTypeEnum.UNKNOWN.value: ANONYMITY_VALIDATOR['http_test_url'],
+        ProxyTypeEnum.HTTP.value: ANONYMITY_VALIDATOR['http_test_url'],
+        ProxyTypeEnum.HTTPS.value: ANONYMITY_VALIDATOR['https_test_url'],
+        ProxyTypeEnum.HTTP_AND_HTTPS.value: ANONYMITY_VALIDATOR['https_test_url'],
+    }
+
+    def run(self):
+        # 获取proxy列表
+        proxy_list = sqlite_opt.get_unknown_anonymity_proxies()
+        if len(proxy_list) > 0:
+            tasks = [self.valid_proxy(proxy.url, proxy.proxy_type) for proxy in proxy_list]
+            asyncio.run(asyncio.wait(tasks))
+
+    async def valid_proxy(self, proxy_url, proxy_type):
+        async with aiohttp.ClientSession() as session:
+            try:
+                async with session.get(self.urls[proxy_type],
+                                       proxy=proxy_url,
+                                       headers=HEADERS,
+                                       timeout=ANONYMITY_VALIDATOR['request_timeout']) as resp:
+                    if resp.status == 200:
+                        # 检验其匿名性
+                        r_dict = json.loads(await resp.text())
+                        headers = r_dict.get('headers', '')
+                        ip = r_dict.get('origen')
+                        proxy_connection = headers.get('Proxy-Connection', None)
+                        flag = True
+                        if ',' in ip:
+                            ips = str.split(ip, ',')
+                            first = ips[0]
+                            for p in ips:
+                                if first != p.lstrip():
+                                    proxy_cover = ProxyCoverEnum.TRANSPARENT.value   # 透明
+                                    flag = False
+                                    break
+                        if flag:
+                            if proxy_connection:
+                                proxy_cover = ProxyCoverEnum.NORMAL_COVER.value  # 普匿
+                            else:
+                                proxy_cover = ProxyCoverEnum.HIGH_COVER.value    # 高匿
+                        # 更新匿名性
+                        sqlite_opt.update_anonymity(proxy_url, proxy_cover)
+                        logger.info(f'验证匿名性成功: url:{proxy_url}, coverValue:{proxy_cover}')
+                    else:
+                        logger.warn(f'验证匿名性失败, proxy_url:{proxy_url}, 返回码:{resp.status}')
+            except asyncio.TimeoutError:
+                logger.warn(f'验证匿名性请求超时, proxy_url:{proxy_url}')
+            except ConnectionRefusedError:
+                logger.warn(f'验证匿名性请求被拒绝, proxy_url:{proxy_url}')
+            except Exception as e:
+                # logger.exception(e)
+                logger.warn(f'验证匿名性失败, proxy_url:{proxy_url}, e:{e}')
+
+
+anonymity_validator = AnonymityValidator()
diff --git a/test/validator/test_anonymity_validator.py b/test/validator/test_anonymity_validator.py
@@ -0,0 +1,18 @@
+import unittest
+
+from src.database.sqlite_opt import sqlite_opt
+from src.validator.anonymity_validator import anonymity_validator
+
+
+class TestAnonymityValidator(unittest.TestCase):
+
+    def setUp(self) -> None:
+        self._opt = sqlite_opt
+        self._validator = anonymity_validator
+
+        # self._opt.clean()
+
+    def test_valid_proxy(self):
+        self._validator.run()
+        pass
+