Files
crawlab/spiders/segmentfault/segmentfault_spider.js
2019-03-13 21:33:47 +08:00

74 lines
1.8 KiB
JavaScript

const puppeteer = require('puppeteer');
const MongoClient = require('mongodb').MongoClient;
(async () => {
// browser
const browser = await (puppeteer.launch({
timeout: 10000
}));
// define start url
const url = 'https://segmentfault.com/newest';
// start a new page
const page = await browser.newPage();
// navigate to url
try {
await page.goto(url);
await page.waitFor(2000);
} catch (e) {
console.error(e);
// close browser
browser.close();
// exit code 1 indicating an error happened
code = 1;
process.emit("exit ");
process.reallyExit(code);
return
}
// scroll down to fetch more data
for (let i = 0; i < 10; i++) {
console.log('Pressing PageDown...');
await page.keyboard.press('PageDown', 200);
await page.waitFor(500);
}
// scrape data
const results = await page.evaluate(() => {
let results = [];
document.querySelectorAll('.news-list .news-item').forEach(el => {
results.push({
url: 'https://segmentfault.com' + el.querySelector('.news__item-info > a').getAttribute('href'),
title: el.querySelector('.news__item-title').innerText
})
});
return results;
});
// open database connection
const client = await MongoClient.connect('mongodb://192.168.99.100:27017');
let db = await client.db('crawlab_test');
const colName = process.env.CRAWLAB_COLLECTION || 'results_segmentfault';
const taskId = process.env.CRAWLAB_TASK_ID;
const col = db.collection(colName);
// save to database
await results.forEach(d => {
d.task_id = taskId;
console.log(d);
col.insertOne(d);
});
console.log(`results.length: ${results.length}`);
// close database connection
client.close();
// shutdown browser
browser.close();
})();