增加免费代理的抓取

package.json

@@ -34,6 +34,8 @@
     "request": "^2.88.0",
     "request-promise": "^4.2.4",
     "socks5-http-client": "^1.0.4",
+    "superagent": "^5.0.5",
+    "superagent-proxy": "^2.0.0",
     "ws": "^6.1.2"
   },
   "devDependencies": {

src/app.js

@@ -8,7 +8,7 @@ import movie from './sites/movie';
 import book from './sites/book';
 import bookChapter from './sites/bookChapter';
 import dandanzan from './sites/dandanzan';
-
+import proxy from './sites/proxy';
 mongoose.Promise = Promise;
 
 
@@ -24,7 +24,8 @@ db.once('open', function () {
   // book.run();
   // movie.run();
   // bookChapter.run();
-  dandanzan.run();
+  // dandanzan.run();
+  proxy.run();
 });
 mongoose.connect(config.db, {promiseLibrary: Promise, useNewUrlParser: true});
 

src/models/spider/ProxyInfo.js

@@ -0,0 +1,46 @@
+import mongoose from 'mongoose';
+import stringUtil from '../../utils/string.util';
+
+let Schema = mongoose.Schema;
+
+
+let ProxyInfoSchema = new Schema({
+  //http, socket4, socket5
+  type: {type: String},
+  link: {type: String},
+  info: {type: String},
+  last_check: {type: Date},
+  //相应时间
+  response_time: {type: Number},
+  // 1: 存活， 0: 未知，-1: 不可用
+  status: {type: Number, default: 0},
+  // 连续尝试次数
+  try_count: {type: Number, default: 0},
+}, {
+  collection: 'proxy_info',
+  timestamps: true
+});
+
+
+let ProxyInfoModel = mongoose.model('ProxyInfo', ProxyInfoSchema);
+ProxyInfoModel.updateOne = async function (link, record) {
+  const query = {link: link};
+  const options = {upsert: true, setDefaultsOnInsert:true};
+  await ProxyInfoModel.update(query, record, options);
+}
+
+// 返回所有待检查的代理
+ProxyInfoModel.needCheckList = function() {
+  return ProxyInfoModel.find({status: {$in: [0 ,1]}});
+}
+
+// 所有可用代理
+ProxyInfoModel.availableList = function() {
+  return ProxyInfoModel.find({status: 1});
+}
+
+ProxyInfoModel.randomProxy = function () {
+  return ProxyInfoModel.find({status: 1}).limit(1);
+}
+
+export default ProxyInfoModel;

src/sites/dandanzan.js

@@ -145,12 +145,9 @@ const parseAllMovie = async (category) => {
 
 export default {
   run: async () => {
-    // await parseAllMovie('movie');
+    await parseAllMovie('movie');
-    // await parseAllMovie('tv');
+    await parseAllMovie('tv');
-    // await parseAllMovie('show');
+    await parseAllMovie('show');
     await parseAllMovie('cartoon');
-    // console.log('all done');
-    // let html = await netUtil.getData('https://wechat-test.kingsome.cn/', {})
-    // console.log(html);
   }
 }

src/sites/proxy.js

@@ -0,0 +1,144 @@
+import ProxyInfo from '../models/spider/ProxyInfo';
+import netUtil from "../utils/net.util";
+import cheerio from "cheerio";
+import generalQueue from '../utils/general.queue';
+import proxyUtil from '../utils/proxy.util';
+
+/* *
+ * 获取www.xiladaili.com的代理数据
+ * http://www.xiladaili.com/https/1/
+ * */
+const parseOneXiladaili = async (idx) => {
+  let url = `http://www.xiladaili.com/https/${idx + 1}/`;
+  console.log('begin parse page:', url);
+  try {
+    let html = await netUtil.getData(url, {})
+    if (html) {
+      const $ = cheerio.load(html);
+      const trArr = $('.fl-table tbody tr');
+      $(trArr).each(async function (i, tr) {
+        const link = $(this).find('td').first().text();
+        const record = {
+          link: link,
+          type: 'http',
+          info: $(this).find('td').eq(3).text()
+        }
+        await ProxyInfo.updateOne(link, record)
+      });
+    }
+  } catch (err) {
+    console.log(err);
+  }
+}
+/* *
+ * 获取www.xicidaili.com的代理数据
+ * https://www.xicidaili.com/nn/1
+ * */
+const parseOneXicidaili = async (idx) => {
+  let url = `https://www.xicidaili.com/nn/${idx + 1}`;
+  console.log('begin parse page:', url);
+  try {
+    let html = await netUtil.getData(url, {})
+    if (html) {
+      const $ = cheerio.load(html);
+      const trs = $("#ip_list tr");
+      for (let i = 1; i < trs.length; i++) {
+        const tds = trs.eq(i).children("td");
+        const link = `${tds.eq(1).text()}:${tds.eq(2).text()}`
+        const type = tds.eq(5).text();
+        const info = tds.eq(3).text();
+        let speed = tds.eq(6).children("div").attr("title");
+        speed = speed.substring(0, speed.length - 1);
+        let connectTime = tds.eq(7).children("div").attr("title");
+        connectTime = connectTime.substring(0, connectTime.length - 1);
+        if (speed <= 5 && connectTime <= 1 && type === 'HTTPS') { //用速度和连接时间筛选一轮
+          const record = {
+            link: link,
+            type: 'http',
+            info: info
+          }
+          await ProxyInfo.updateOne(link, record)
+        }
+      }
+    }
+  } catch (err) {
+    console.log(err);
+  }
+}
+
+const checkAndUpdate = async (record) => {
+  console.log('begin check:', record.link);
+  try {
+    let time = await proxyUtil.checkProxy(`http://${record.link}`);
+    record.status = 1;
+    record.response_time = time;
+  } catch (err) {
+    console.log('not available:', record.link);
+    record.status = -1;
+  }
+  record.try_count += 1;
+  record.last_check = new Date();
+  try {
+    await record.save();
+  } catch (err) {
+    console.log('error update proxy record', record.link);
+  }
+  return record.status > 0;
+}
+
+const parseXiladaili = async (maxPage) => {
+  for (let i = 0; i < maxPage; i++) {
+    try {
+      await parseOneXiladaili(i);
+    } catch (err) {
+      console.log(err);
+    }
+  }
+  console.log('finish parse all page, Xiladaili');
+}
+const parseXicidaili = async (maxPage) => {
+  for (let i = 0; i < maxPage; i++) {
+    try {
+      await parseOneXicidaili(i);
+    } catch (err) {
+      console.log(err);
+    }
+  }
+  console.log('finish parse all page, Xicidaili');
+}
+const checkAllProxy = async () => {
+  let records;
+  return new Promise(async (resolve, reject) => {
+    console.time('all')
+    generalQueue.setCb(function () {
+      console.log('checkAllProxy finished');
+      resolve()
+    })
+    try {
+      records = await ProxyInfo.needCheckList();
+    } catch (err) {
+      reject(err);
+    }
+    for(let record of records) {
+      try {
+        generalQueue.addQueue({
+          run: async function () {
+            await checkAndUpdate(record);
+          }
+        })
+      } catch (err) {}
+    }
+  })
+
+}
+export default {
+  run: async () => {
+    try {
+      // await parseXiladaili(20);
+      // await parseXicidaili(20);
+      await checkAllProxy();
+    } catch (err) {
+      console.log(err);
+    }
+  }
+}

src/utils/general.queue.js

@@ -1,5 +1,6 @@
 import async from 'async';
 
+let finishCb;
 /**
  * 操作队列
  * */
@@ -14,6 +15,7 @@ let q = async.queue( async (reqObj, cb) => {
 q.drain = function(){
   console.info('all queue done');
   console.timeEnd('all');
+  (finishCb) && (finishCb());
 };
 module.exports = {
   addQueue(obj) {
@@ -22,5 +24,8 @@ module.exports = {
         console.log('error parse: ', obj, err);
       }
     });
+  },
+  setCb(cb) {
+    finishCb = cb;
   }
 }

src/utils/net.util.js

@@ -1,8 +1,6 @@
 import request from 'request';
 import Promise from 'bluebird';
-import proxy from './proxys';
 import random_useragent from 'random-useragent';
-let agent = require('socks5-http-client/lib/Agent')
 
 
 const iconv = require('iconv-lite');
@@ -55,6 +53,7 @@ export default {
     (gzip) && (options.gzip = true);
     return requestData(options, encoding);
   },
+  // 通过代理来get数据
   getDataProxy(url, header, encoding, gzip) {
     header = header || {
       'Cache-Control': 'no-cache',
@@ -65,16 +64,55 @@ export default {
       url: url,
       headers: header,
       proxy: proxy.randomProxy(),
-      // agentClass: agent,
-      // agentOptions: {
-      //   socksHost: '101.71.41.169',
-      //   socksPort: 43,
-      // }
     };
     if (encoding) {
       options.encoding = null;
     }
     (gzip) && (options.gzip = true);
     return requestData(options, encoding);
+  },
+  // 检查proxy是否可用
+  async checkProxy(link){
+    let header = {
+      'Cache-Control': 'no-cache',
+      'User-Agent': random_useragent.getRandom(),
+      'Connection':'close',
+    }
+    //https://wechat-test.kingsome.cn/api/stat
+    //https://www.baidu.com/favicon.ico
+    const options = {
+      url: 'https://wechat-test.kingsome.cn/api/stat',
+      proxy: link,
+      method: 'GET',
+      headers: header,
+      time : true,
+      followRedirect: false,
+      // pool: false,
+    }
+    return new Promise((resolve, reject) => {
+      let req = request(options, (err, response, body) => {
+        if (err) {
+          return reject(err);
+        }
+        console.log('status code:', response.statusCode, 'body:', body);
+        if (response.statusCode === 200 && body === '1') {
+          console.log('Request time in ms', response.elapsedTime);
+          resolve(response.elapsedTime);
+        } else {
+          reject(new Error(' server response code: ' + response.statusCode + ' with url: ' + options.url));
+        }
+      });
+      setTimeout(() => {
+        console.log('check proxy timeout', link);
+        try {
+          req.abort(new Error('timeout'));
+        } catch (err) {
+          console.log(err);
+          reject()
+        }
+        reject()
+      }, 5000);
+    })
+
   }
 }

src/utils/proxy.util.js

@@ -0,0 +1,30 @@
+import stringUtil from './string.util';
+import ProxyInfo from '../models/spider/ProxyInfo';
+
+let request = require('superagent');
+require('superagent-proxy')(request);
+
+let proxys = [];
+
+export default {
+  async randomProxy() {
+    if (proxys.length === 0) {
+      proxys = await ProxyInfo.availableList();
+    }
+    return proxys[stringUtil.randomNum(0, proxys.length - 1)];
+  },
+  async checkProxy(proxy){
+    return new Promise(async (resolve, reject) => {
+      try {
+        let response = await request.get('https://wechat-test.kingsome.cn/api/stat').proxy(proxy).timeout(5000);
+        if(response.statusCode === 200 && response.text === '1' ){
+          resolve();
+        } else {
+          reject(new Error(' server response code: ' + response.statusCode));
+        }
+      } catch (err) {
+        reject(err);
+      }
+    })
+  }
+}

src/utils/proxys.js

@@ -1,14 +0,0 @@
-import stringUtil from './string.util';
-const proxys = [
-     'http://113.200.56.13:8010',
-     'http://65.52.174.40:80',
-     'http://165.22.254.199:8080',
-     'http://88.255.101.241:8080',
-     'http://117.197.117.50:8080'
-    ];
-export default {
-
-  randomProxy() {
-    return proxys[stringUtil.randomNum(0, proxys.length - 1)];
-  }
-}

stat

@@ -0,0 +1 @@
+1