Make Selenium work on Github Actions

Scraping with BeautifulSoup on GitHub Actions is easy-peasy. But what about Selenium?? After you jump through some hoops it's just as simple.

How to use this

I think you can just fork this repository and get to work on scraper.py.

Otherwise, just steal scraper.py and .github/workflows/scrape.yml and you'll be good to go.

Saving CSV files

If you want every scrape to update a CSV or something like this, you'll need to edit your scrape.yml to commit to the repo after the scraper is done running.

Take a look at the bottom of autoscraper-history to see how to do that. It should be one more step, something like this:

      - name: Commit and push if content changed
        run: |-
          git config user.name "Automated"
          git config user.email "[email protected]"
          git add -A
          timestamp=$(date -u)
          git commit -m "Latest data: ${timestamp}" || exit 0
          git push

I'm pretty sure I lifted that code 100% from Simmon Willison.

Scheduling

Right now the yaml is set to only scrape when you click the "Run workflow" button in the Actions tab, but you can add something like

  schedule:
    - cron: '0 * * * *'

To make it run the first minute of every hour.

How this all works

To make Selenium + GitHub Actions work, this repo does a few magic things. In a normal world, you start Chrome like this:

from selenium import webdriver

driver = webdriver.Chrome()

But we.... do things a little differently. Let me walk you through the changes!

webdriver-manager to manage the webdriver

You can add a special action to set up Chromedriver but I feel it's honestly easier to use webdriver-manager. It magically picks out the right version of chromedriver for you.

from selenium import webdriver
from webdriver_manager.chrome import ChromeDriverManager

driver = webdriver.Chrome(ChromeDriverManager().install())

It works on your local machine, too!

Chromium, not Chrome

When you're running GitHub Actions, it's probably on a nice little Ubuntu Linux machine. In those situations, you install software using apt. Since you can't install Chrome with apt, you'll install Chromium instead, the open-source version of Chrome. Works the same, just opens a little differently.

As a result, our chromedriver install code changed a little more:

from selenium import webdriver
from webdriver_manager.chrome import ChromeDriverManager
from webdriver_manager.utils import ChromeType

driver_path = ChromeDriverManager(chrome_type=ChromeType.CHROMIUM).install()
driver = webdriver.Chrome(driver_path)

This also means we need to add a line that does apt-get install -y chromium-browser to our scrape.yml)

Headless Chromium

Since we don't have a monitor plugged into GitHub Actions, we can't actually see what's going on in the browser. In the olden days you had to construct some odd technical fake screen, but these days you just run in headless mode!

from selenium import webdriver 
from selenium.webdriver.chrome.options import Options

chrome_options = Options()
chrome_options.add_argument("--headless")
driver = webdriver.Chrome(options=chrome_options)

Other options and more management of things

To combine the webdriver-manager driver path and the headless Chrome option, there are a lot of hoops to jump through. During the research process a lot of extra chrome options poked up, so I thought hey, let's just add all those, too.

from selenium import webdriver
from webdriver_manager.chrome import ChromeDriverManager
from webdriver_manager.utils import ChromeType
from selenium.webdriver.chrome.options import Options
from selenium.webdriver.chrome.service import Service

chrome_service = Service(ChromeDriverManager(chrome_type=ChromeType.CHROMIUM).install())

chrome_options = Options()
options = [
    "--headless",
    "--disable-gpu",
    "--window-size=1920,1200",
    "--ignore-certificate-errors",
    "--disable-extensions",
    "--no-sandbox",
    "--disable-dev-shm-usage"
]
for option in options:
    chrome_options.add_argument(option)

driver = webdriver.Chrome(service=chrome_service, options=chrome_options)

The biggest change here beyond all those options is the Service thing. Apparently just giving it the path to chromdriver isn't good enough? Who knows, I just do what works.