All Posts

5m

How To Scrape A Job Board With Crawlee

Why

Having known many restauranters, their most pressing and consistent pain point has been sourcing willing and competent labor. Especially when it comes to small businesses, it can be a struggle to find candidates since most flock to big name brands for perks that most nascent businesses can't compete with. To compound, mom-and-pop shops tend to be overlooked since they reside off the busy boulevard corner-lots the larger corporations inhabit. Altogether, this can make it quite difficult to find new people to work with.

I decided to take it into my own hands to level the playing field for small businesses in the arena of finding work. Which is why I'll be building a job board specifically for small business restaurants in Burbank, CA. Think that's niche enough?

My first problem will be solving the "chicken or the egg" problem. A job board is a two-sided marketplace: candidates and employers. Which do I attempt to serve first? If I go for businesses first, I will need to have a large pool of candidates actively searching for a job. If I go for candidates, I will need to have a large pool of employers with unfilled positions. Luckily there are already several open datasets of job postings I can pull from and serve under my own site. This will enable me to serve candidates first with existing job postings, build up the candidate traffic, then market to businesses in search of candidates.

So yes, I will be scraping Indeed. I'm just- I'm just scraping Indeed.

How

After much investigation and internal debate, I've landed on crawlee and playwright as my tool of choice. Crawlee is purpose-built as a crawler. It has some really neat architectural decisions that keep the code real clean (I'll touch on that in a bit). As for Playwright, it's the latest iteration of Puppeteer and I've always enjoyed Puppeteer's SDK shape.

First thing's first, go ahead and run this command to get a starter project going:

npx crawlee create my-crawler

Once that's done, go ahead and replace the URL where the crawler begins in main.ts:

// For more information, see https://crawlee.dev/
import { PlaywrightCrawler, ProxyConfiguration } from 'crawlee';

import { router } from './routes.js';

- const startUrls = ['https://crawlee.dev/'];
+ const startUrls = ['https://www.indeed.com/jobs?q=[ENTER YOUR DESIRED JOB HERE]&l=Burbank%2C+CA&radius=35'];

const crawler = new PlaywrightCrawler({
    // proxyConfiguration: new ProxyConfiguration({ proxyUrls: ['...'] }),
    requestHandler: router,
    // Comment this option to scrape the full website.
    maxRequestsPerCrawl: 20,
});

await crawler.run(startUrls);

Next, we'll want to set up our scraping logic in routes.ts:

import { createPlaywrightRouter } from "crawlee";
import { scrapeButton, scrapeMultiButton } from "./buttonScapers.js";

export const router = createPlaywrightRouter();

// the default handler runs before everything else
router.addDefaultHandler(async ({ enqueueLinks, log }) => {
  log.info(`enqueueing new URLs`);

  // queue up all the job links found in the search results
  await enqueueLinks({
    globs: ["https://www.indeed.com/rc/clk**"],
    label: "job",
  });

  // queue up all the page links found in the pagination buttons
  await enqueueLinks({
    globs: ["https://www.indeed.com/jobs**&start=**"],
    regexps: [new RegExp(/https:\/\/www\.indeed\.com\/jobs.*&start=.*/)],
    label: "new-page",
  });
});

// for every new result page, queue other result pages and job pages
router.addHandler("new-page", async ({ request, page, log, enqueueLinks }) => {
  const urlParams = new URLSearchParams(page.url().split("?")[1]);
  const skip = Number(urlParams.get("start") || 0);

  log.info(`On page ${skip / 10}`);

  // queue up all the job links found in the search results
  await enqueueLinks({
    globs: ["https://www.indeed.com/rc/clk**"],
    label: "job",
  });

  // queue up all the page links found in the pagination buttons
  await enqueueLinks({
    globs: ["https://www.indeed.com/jobs**&start=**"],
    regexps: [new RegExp(/https:\/\/www\.indeed\.com\/jobs.*&start=.*/)],
    label: "new-page",
  });
});

router.addHandler("job", async ({ request, page, log, pushData }) => {
  const url = request.loadedUrl;

  const rawPageTitle = await page.title();
  const pageTitle = rawPageTitle.slice(0, rawPageTitle.length - 13);

  log.info(`${pageTitle}`, { url: request.loadedUrl });

  const urlParams = new URLSearchParams(page.url().split("?")[1]);
  const indeedJobId = urlParams.get("jk");

  const jobTitle = await page.locator("h1 span").textContent();

  const pay = await scrapeButton(page, `[aria-label='Pay'] li div div`);
  const match = pay.match(/\$(\d+(?:\.\d+)?)/); // grab the first string to match `$[NUMBER]`
  const payNumber = match ? parseFloat(match[1]) : null;

  const commitment = await scrapeMultiButton(
    page,
    `[aria-label='Job type'] li div div`
  );
  const schedule = await scrapeMultiButton(
    page,
    `[aria-label='Shift and schedule'] li div div`
  );

  const location = await page.locator("#jobLocationText div").textContent();
  const benefits = await page.locator("#benefits ul li").allInnerTexts();
  const description = await page.locator("#jobDescriptionText").innerHTML();

  await pushData({
    indeedJobId,
    url,
    pageTitle,
    jobTitle,
    pay,
    payNumber,
    commitment,
    schedule,
    location,
    benefits,
    description,
  });
});

Crawlee has this incredibly clever routing technique called labels. It allows you to queue up other pages for crawling and send those tasks to different "channels" so you can easily compartmentalize your logic. I always ended up with spaghetti code when writing my own crawlers from scratch because I'd have a mess of for loops and a ton of deeply nested statements. With this new paradigm, it's easy to keep things clean.

You might be wondering what scrapeMultiButton and scrapeButton are: they are two helper functions I built to handle this funky UI: Indeed's user interface with complex button representing the various properties of a job posting

If you want to scrape these as well, here's the implementation in buttonScrapers.ts:

import { Page } from "playwright";

export const scrapeButton = async (page: Page, selector: string) => {
const elementFound = await page.locator(selector).first().isVisible();

  if (!elementFound) {
    return "";
  }

  return await page.locator(selector).first().innerText();
};

export const scrapeMultiButton = async (page: Page, selector: string) => {
  const elementFound = await page.locator(selector).first().isVisible();

  if (!elementFound) {
    return [];
  }

  const buttons = await page.locator(selector).all();
  const buttonTexts = [];


  for (let i = 0; i < buttons.length; i += 3) {
    buttonTexts.push(await buttons[i].innerText());
  }

  return buttonTexts
};

Lastly, we'll want to dump this data into a CSV so we can pipe it into our database of choice:

// For more information, see https://crawlee.dev/
- import { PlaywrightCrawler, ProxyConfiguration } from 'crawlee';
+ import { PlaywrightCrawler, ProxyConfiguration, Dataset } from 'crawlee';

import { router } from './routes.js';

const startUrls = ['https://www.indeed.com/jobs?q=[ENTER YOUR DESIRED JOB HERE]&l=Burbank%2C+CA&radius=35'];

const crawler = new PlaywrightCrawler({
    // proxyConfiguration: new ProxyConfiguration({ proxyUrls: ['...'] }),
    requestHandler: router,
    // Comment this option to scrape the full website.
-    maxRequestsPerCrawl: 20,
+    maxRequestsPerCrawl: 500, // let's also get more results
});

await crawler.run(startUrls);

+await Dataset.exportToCSV('jobs')

That's it! Now you just have to run npm run start:dev!