Back in April, Google announced that it will be shipping Headless Chrome in Chrome 59. Since the respective flags are already available on Chrome Canary, the Duo Labs team thought it would be fun to test things out and also provide a brief introduction to driving Chrome using Selenium and Python.
Before we dive into any code, let’s talk about what a headless browser is and why it’s useful. In short, headless browsers are web browsers without a graphical user interface (GUI) and are usually controlled programmatically or via a command-line interface.
Headless Chrome and Python
The Dark Ages
Prior to the release of Headless Chrome, any time that you did any automated driving of Chrome that potentially involved several windows or tabs, you had to worry about the CPU and/or memory usage. Both are associated with having to display the browser with the rendered graphics from the URL that was requested.
When using a headless browser, we don’t have to worry about that. As a result, we can expect lower memory overhead and faster execution for the scripts that we write.
Before we get started, we need to install Chrome Canary and download the latest ChromeDriver (currently 5.29).
Next, let’s make a folder that will contain all of our files:
$ mkdir going_headless
Now we can move the ChromeDriver into the directory that we just made:
$ mv Downloads/chromedriver going_headless/
Since we are using Selenium with Python, it’s a good idea to make a Python virtual environment. I use virtualenv, so if you use another virtual environment manager, the commands may be different.
$ cd going_headless && virtualenv -p python3 env $ source env/bin/activate
$ pip install selenium
Now that we’ve gotten all of that out of the way, let’s get to the fun part. Our goal is to write a script that searches for my name “Olabode” on duo.com, and checks that a recent article I wrote about Android security is listed in the results. If you’ve followed the instructions above, you can use the headless version of Chrome Canary with Selenium like so:
import os from selenium import webdriver from selenium.webdriver.common.keys import Keys from selenium.webdriver.chrome.options import Options chrome_options = Options() chrome_options.add_argument("--headless") chrome_options.binary_location = '/Applications/Google Chrome Canary.app/Contents/MacOS/Google Chrome Canary'` driver = webdriver.Chrome(executable_path=os.path.abspath(“chromedriver"), chrome_options=chrome_options) driver.get("https://duo.com")` magnifying_glass = driver.find_element_by_id("js-open-icon") if magnifying_glass.is_displayed(): magnifying_glass.click() else: menu_button = driver.find_element_by_css_selector(".menu-trigger.local") menu_button.click()` search_field = driver.find_element_by_id("site-search") search_field.clear() search_field.send_keys("Olabode") search_field.send_keys(Keys.RETURN) assert "Looking Back at Android Security in 2016" in driver.page_source driver.close()`
Let’s break down what’s going on in the script. We start by importing the requisite modules.
The Keys provides keys in the keyboard like
RETURN, F1, ALT, etc.
import os from selenium import webdriver from selenium.webdriver.chrome.options import Options from selenium.webdriver.common.keys import Keys
Next, we create a ChromeOptions object which will allow us to set the location of the Chrome binary that we would like to use and also pass the headless argument. If you leave out the headless argument, you will see the browser window pop up and search for my name.
In addition, if you don’t set the binary location to the location of Chrome Canary on your system, the current version of Google Chrome that is installed will be used. I wrote this tutorial on a Mac, but you can find the location of the file on other platforms here. You just need to substitute Chrome for Chrome Canary in the respective file paths.
chrome_options = Options() chrome_options.add_argument("--headless") chrome_options.binary_location = '/Applications/Google Chrome Canary.app/Contents/MacOS/Google Chrome Canary' driver = webdriver.Chrome(executable_path=os.path.abspath(“chromedriver"), chrome_options=chrome_options)
The driver.get function will be used navigate to the specified URL.
The duo.com website is responsive, so we have to handle different conditions. As a result, we check to see if the expected search button is displayed. If it isn’t, we click the menu button to enter our search term.
magnifying_glass = driver.find_element_by_id("js-open-icon") if magnifying_glass.is_displayed(): magnifying_glass.click() else: menu_button = driver.find_element_by_css_selector(".menu-trigger.local") menu_button.click()
Now we clear the search field, search for my name, and send the RETURN key to the drive.
search_field = driver.find_element_by_id("site-search") search_field.clear() search_field.send_keys("Olabode") search_field.send_keys(Keys.RETURN)
We check to make sure that the blog post title from one of my most recent posts is in the page’s source.
assert "Looking Back at Android Security in 2016" in driver.page_source
And finally, we close the browser.
Head to Headless
So, it’s cool that we can now control Chrome using Selenium and Python without having to see a browser window, but we are more interested in the performance benefits we talked about earlier. Using the same script above, we profiled the time it took to complete the tasks, peak memory usage, and CPU percentage. We polled CPU and memory usage with psutil and measured the time for task completion using timeit.
|Headless (60.0.3102.0)||Headed (60.0.3102.0)|
|Median Time||5.29 seconds||5.51 seconds|
|Median Memory Use||25.3 MiB||25.47 MiB|
|Average CPU Percentage||1.92%||2.02%|
For our small script, there were very small differences in the amount of time taken to complete the task (4.3%), memory usage (.5%), and CPU percentage (5.2%). While the gains in our example were very minimal, these gains would prove to be beneficial in a test suite with dozens of tests.
Manual vs. Adhoc
In the script above, we start the ChromeDriver server process when we create the WebDriver object and it is terminated when we call quit(). For a one-off script, that isn’t a problem, but this can waste a nontrivial amount of time for a large test suite that creates a ChromeDriver instance for each test. Luckily, we can manually start and stop the server ourselves, and it only requires a few changes to the script above.
import os from selenium import webdriver from selenium.webdriver.common.keys import Keys from selenium.webdriver.chrome.options import Options service = webdriver.chrome.service.Service(os.path.abspath(“chromedriver")) service.start() chrome_options = Options() chrome_options.add_argument("--headless") # path to the binary of Chrome Canary that we installed earlier chrome_options.binary_location = '/Applications/Google Chrome Canary.app/Contents/MacOS/Google Chrome Canary' driver = webdriver.Remote(service.service_url, desired_capabilities=chrome_options.to_capabilities())
While there are only three lines of code that have changed, let’s talk about what’s going on in them. In order to manually control the ChromeDriver server, we have to use the ChromeDriverService. We do so by creating a service object with a path to the ChromeDriver and then we can start the service.
service = webdriver.chrome.service.Service(os.path.abspath(“chromedriver")) service.start()
The final thing we have to do is create a WebDriver that can connect to a remote server. In order to use Chrome Canary and the headless portion, we have to pass the the dictionary of all the options since the remote WebDriver object doesn’t accept an Option object.
driver = webdriver.Remote(service.service_url, desired_capabilities=chrome_options.to_capabilities())
By adding the manual starting of the service, we saw the expected speed increases. The median time for the headless and headed browser to complete the tasks in the script decreased by 11% (4.72 seconds) and respectively 4% (5.29 seconds).
|Headed Browser||Headless Browser|
|Median Time(% decrease)||4%||11%|
|Median Time (Seconds)||5.29 seconds||4.72 seconds|
The release of headless Chrome has long been awaited. And with the announcement that the creator of PhantomJS is stepping down as a maintainer, we strongly believe that headless Chrome is the future of headless browsers.
While we covered Selenium in this walkthrough, it is worth mentioning that the Chrome DevTools API can be a useful resource if you’re doing any type of profiling or need to create PDFs of pages that you visit. We hope this helps you get started using the headless version of Chrome whether you’re doing any type of QA testing or are automating all your daily web-related tasks.