This commit is contained in:
gggaaallleee
2025-02-28 19:00:58 +08:00
committed by GitHub
parent cf0aaa1091
commit f7b2a57ca3
29 changed files with 7469 additions and 0 deletions
@@ -0,0 +1,77 @@
import NodeCache from 'node-cache';
import { MongoClient } from 'mongodb';
import crypto from 'crypto';
import dotenv from 'dotenv';
dotenv.config();
const cache = new NodeCache({ stdTTL: parseInt(process.env.STD_TTL || '3600') });
const mongoClient = new MongoClient(process.env.MONGODB_URI || 'mongodb://localhost:27017');
const dbName = 'pageCache';
const collectionName = 'pages';
const connectToMongo = async () => {
await mongoClient.connect();
return mongoClient.db(dbName);
};
const createTTLIndex = async () => {
try {
const db = await connectToMongo();
await db.collection(collectionName).createIndex({ "updatedAt": 1 }, { expireAfterSeconds: parseInt(process.env.EXPIRE_AFTER_SECONDS || '9000') });
console.log("TTL index created successfully");
} catch (error) {
console.error("Error creating TTL index:", error);
}
};
const getPageHash = (content: string) => {
return crypto.createHash('md5').update(content).digest('hex');
};
export const getCachedPage = async (url: string) => {
const cachedPage = cache.get(url);
if (cachedPage) return cachedPage;
try {
const db = await connectToMongo();
const page = await db.collection(collectionName).findOne({ url });
if (page) cache.set(url, page);
return page;
} catch (error) {
console.error('Error getting cached page:', error);
throw error;
}
};
const savePageToCache = async (url: string, content: string) => {
const hash = getPageHash(content);
const page = { url, content, hash, updatedAt: new Date() };
cache.set(url, page); // 更新内存缓存
try {
const db = await connectToMongo();
await db.collection(collectionName).updateOne(
{ url },
{ $set: page },
{ upsert: true }
); // 更新持久化缓存
} catch (error) {
console.error('Error saving page to cache:', error);
throw error;
}
};
export const updateCacheAsync = async (url: string, content: string) => {
await savePageToCache(url, content);
};
process.on('SIGINT', async () => {
await mongoClient.close();
process.exit(0);
});
// 在应用启动时创建 TTL 索引
createTTLIndex();
@@ -0,0 +1,140 @@
import { Cluster } from 'puppeteer-cluster';
import * as cheerio from 'cheerio';
import UserAgent from 'user-agents';
import { setupPage } from './setupPage';
import { getCachedPage, updateCacheAsync } from './cacheUpdater';
import { handleSpecialWebsite } from '../specialHandlers';
import fetch from 'node-fetch';
interface CachedPage {
url: string;
content: string;
hash: string;
updatedAt: Date;
}
export const performDeepSearch = async (clusterInstance: Cluster, resultUrls: string[], results: Map<string, any>, strategies: any[], detectWebsites: string[], pageCount: number) => {
const tasks = [];
await clusterInstance.task(async ({ page, data: { searchUrl } }) => {
try {
const cachedPage = await getCachedPage(searchUrl) as CachedPage | null;
if (cachedPage) {
const result = results.get(searchUrl);
if (result) {
result.content = cachedPage.content;
result.crawlStatus = 'Success';
}
return;
}
} catch (error) {
console.error(`从缓存获取页面 ${searchUrl} 时发生错误:`, error);
results.set(searchUrl, { url: searchUrl, error: (error as Error).message, crawlStatus: 'Failed' });
return;
}
try {
const response = await fetch(searchUrl, {
headers: {
'User-Agent': new UserAgent({ deviceCategory: 'desktop', platform: 'Linux x86_64' }).toString(),
'Referer': 'https://www.google.com/',
'Accept-Language': 'en-US,en;q=0.9',
'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',
'Connection': 'keep-alive',
'Cache-Control': 'no-cache'
}
});
if (response.ok) {
const content = await response.text();
const $ = cheerio.load(content);
const cleanedContent = $('body').html() || '';
const result = results.get(searchUrl);
if (result) {
result.content = cleanedContent;
result.crawlStatus = 'Success';
}
await updateCacheAsync(searchUrl, cleanedContent || '');
return;
} else {
throw new Error(`HTTP error! status: ${response.status}`);
}
} catch (error) {
console.error(`快速抓取页面 ${searchUrl} 时发生错误:`, error);
}
try {
if (detectWebsites.some(website => searchUrl.includes(website))) {
await setupPage(page);
} else {
const userAgent = new UserAgent({ deviceCategory: 'desktop', platform: 'Linux x86_64' });
await page.setUserAgent(userAgent.toString());
}
} catch (error) {
console.error(`访问页面 ${searchUrl} 设置用户代理时发生错误:`, error);
}
let pageLoaded = false;
let pageLoadError: Error | null = null;
for (const strategy of strategies) {
try {
await page.goto(searchUrl, { waitUntil: strategy.waitUntil, timeout: strategy.timeout });
pageLoaded = true;
break;
} catch (error: any) {
if (error.name === 'TimeoutError') {
pageLoadError = error;
continue;
} else {
pageLoadError = error;
throw error;
}
}
}
if (!pageLoaded) {
const result = results.get(searchUrl);
if (result) {
result.error = pageLoadError;
result.crawlStatus = 'Failed';
}
return;
}
try {
let cleanedContent = await handleSpecialWebsite(page, searchUrl);
if (!cleanedContent) {
const content = await page.content();
const $ = cheerio.load(content);
cleanedContent = $('body').html() || '';
}
const result = results.get(searchUrl);
if (result) {
result.content = cleanedContent;
result.crawlStatus = 'Success';
}
await updateCacheAsync(searchUrl, cleanedContent || '');
} catch (error) {
results.set(searchUrl, { url: searchUrl, error: (error as Error).message, crawlStatus: 'Failed' });
} finally {
await page.close().catch(() => {});
}
});
for (const url of resultUrls) {
if (tasks.length >= pageCount + 10) {
break;
}
tasks.push(clusterInstance.queue({ searchUrl: url }));
}
await Promise.all(tasks);
await clusterInstance.idle();
await clusterInstance.close();
return Array.from(results.values()).sort((a, b) => b.score - a.score);
};
@@ -0,0 +1,88 @@
import { Page } from 'puppeteer';
import randomUseragent from 'random-useragent';
const getRandomUserAgent = () => {
return randomUseragent.getRandom();
};
const getRandomPlatform = () => {
const platforms = ["Win32", "MacIntel", "Linux x86_64"];
return platforms[Math.floor(Math.random() * platforms.length)];
};
//代理池
const validateproxy = [
{ ip: "39.102.210.222", port: 8080 },
{ ip: "8.130.71.75", port: 8080 },
{ ip: "39.102.214.208", port: 9999 },
{ ip: "39.104.59.56", port: 8080 },
{ ip: "8.130.37.235", port: 3128 },
{ ip: "8.138.131.110", port: 8080 },
{ ip: "8.140.105.75", port: 8009 },
{ ip: "114.80.38.120", port: 3081 },
{ ip: "8.148.23.165", port: 8081 },
{ ip: "119.96.72.199", port: 59394 },
{ ip: "120.55.14.137", port: 80 },
{ ip: "47.116.181.146", port: 5060 },
{ ip: "39.102.214.199", port: 3128 },
{ ip: "47.121.183.107", port: 8080 },
{ ip: "39.104.16.201", port: 8080 },
{ ip: "39.102.209.163", port: 10002 },
{ ip: "101.201.76.157", port: 9090 },
{ ip: "122.224.124.26", port: 12080 },
{ ip: "180.105.244.199", port: 1080 },
{ ip: "119.3.113.150", port: 9094 }
];
const getRandomProxy = () => {
return validateproxy[Math.floor(Math.random() * validateproxy.length)];
};
const getRandomLanguages = () => {
const languages = [
["zh-CN", "zh", "en"],
["en-US", "en", "fr"],
["es-ES", "es", "en"]
];
return languages[Math.floor(Math.random() * languages.length)];
};
export const setupPage = async (page: Page): Promise<void> => {
const proxy = getRandomProxy();
await page.authenticate({
username: proxy.ip,
password: proxy.port.toString()
});
await page.evaluateOnNewDocument(() => {
const newProto = (navigator as any).__proto__;
delete newProto.webdriver;
(navigator as any).__proto__ = newProto;
(window as any).chrome = {};
(window as any).chrome.app = {"InstallState":"testt", "RunningState":"estt", "getDetails":"stte", "getIsInstalled":"ttes"};
(window as any).chrome.csi = function(){};
(window as any).chrome.loadTimes = function(){};
(window as any).chrome.runtime = function(){};
Object.defineProperty(navigator, 'userAgent', {
get: () => getRandomUserAgent(),
});
Object.defineProperty(navigator, 'platform', {
get: () => getRandomPlatform(),
});
Object.defineProperty(navigator, 'plugins', {
get: () => [{"description": "Shockwave Flash",
"filename": "pepflashplayer.dll",
"length": 1,
"name": "Shockwave Flash"}]
});
Object.defineProperty(navigator, 'languages', {
get: () => getRandomLanguages(),
});
const originalQuery = (window.navigator.permissions as any).query;
(window.navigator.permissions as any).query = (parameters: any) => (
parameters.name === 'notifications' ?
Promise.resolve({ state: Notification.permission } as PermissionStatus) :
originalQuery(parameters)
);
});
};