By now Drupal is a fairly known content management framework, which is why some companies choose to build their site using it given its history and how robust it is.
I've been working for a client for about 2 years now, and they have a rather large Drupal site with multiple dependencies. For this client in particular I'm in charge of front-end development and some back-end integrations, and currently we're using a Gulp-based workflow to manage static assets. This process involves Sass compilation, image compression, and JS minification/concatenation, among other stuff.
In a routine check, a member of the client's team decided to run the site through Google's PageSpeed Insights and to my dismay our initial score was pretty low, ranging between 20 and 30. After this report I decided to dig in a bit deeper and see how we could improve our PSI score which led to some interesting takeaways. Not only we were using a bunch of third party scripts for different tasks (some scripts weren't even necessary anymore) but we also realised that Drupal tends to position render-blocking content at the top of the page, inside the head tag, that could be either deferred, preloaded or moved to the bottom of the DOM right before the closing body tag.
But moving our render-blocking content to the bottom of the document wasn't enough, since we were now getting pretty awful performance on metrics like our First Meaningful Paint. Based on this we decided to see if there was a proper way to create critical CSS and include said declarations inline in the head of the DOM, this would help us improve our FMP and perceived loading times keeping the benefits of moving the rest of our render blocking resources to the end.
Approach No. 1: Hand-picked critical CSS
Our first thought when moving forward to create critical CSS rules to include on the site was to generate a hand-crafted separate file. This process was running smoothly until we tried to import some partial Sass that depended on Bootstrap mixins and variables, which eventually led to dependency hell defeating the purpose of critical CSS. We were unable to create a critical CSS file since we were including a bunch of unneeded declarations because of dependencies.
Approach No. 2: Fetch the homepage critical CSS using a tool like Chrome/Chromium DevTools' Code Coverage
After learning about Chrome/Chromium DevTools' Code Coverage we thought "What if we could run a headless browser when the build process runs and use the DevTools to fetch our homepage's actually used CSS which also includes stuff like navbar, menu, text sizing and color, etcetera?"
Enter Puppeteer: Puppeteer is a Node library which provides a high-level API to control Chrome or Chromium over the DevTools Protocol. Puppeteer runs headless by default, but can be configured to run full (non-headless) Chrome or Chromium.
The first step to include Puppeteer in our workflow was to add it as a dependency:
npm install --save-dev puppeteer
And then we include the dependency in our gulpfile.js
const puppeteer = require('puppeteer');
After Puppeteer is available to work within our Gulpfile we proceed to create a new task (named css-critical) in charge of generating the critical CSS file and declare a variable to hold the URL from which Puppeteer will fetch our critical CSS:
gulp.task('css-critical', async function() {
const URL = 'https://exampleurl.com';
});
With that in place, we now need to declare a new empty string variable to hold whatever we gather as critical CSS, and launch a headless browser with a viewport of 1440x900 pixels:
gulp.task('css-critical', async function() {
const URL = 'https://exampleurl.com';
let criticalCSS = '';
const browser = await puppeteer.launch({
headless: true,
args: [`--window-size=1440,900`],
defaultViewport: null
});
});
Our next step is to open a new page, start the CSS Coverage tool, load our site, store the results in a variable called cssCoverage and finally stop the CSS coverage tool.
gulp.task('css-critical', async function() {
const URL = 'https://exampleurl.com';
let criticalCSS = '';
const browser = await puppeteer.launch({
headless: true,
args: [`--window-size=1440,900`],
defaultViewport: null
});
const page = await browser.newPage();
await page.coverage.startCSSCoverage();
await page.goto(URL, {waitUntil: 'load'})
const cssCoverage = await page.coverage.stopCSSCoverage();
});
Next we'll need to select the used data ranges returned by the Coverage tool in order to compose our final CSS file.
gulp.task('css-critical', async function() {
const URL = 'https://exampleurl.com';
let criticalCSS = '';
const browser = await puppeteer.launch({
headless: true,
args: [`--window-size=1440,900`],
defaultViewport: null
});
const page = await browser.newPage();
await page.coverage.startCSSCoverage();
await page.goto(URL, {waitUntil: 'load'})
const cssCoverage = await page.coverage.stopCSSCoverage();
for (const entry of cssCoverage) {
for (const range of entry.ranges) {
criticalCSS += entry.text.slice(range.start, range.end) + "\n"
}
}
});
After this is done and ready, we'll proceed to close the page, close the browser and dump the content of our criticalCSS into an actual file, that will later be inlined into our Drupal html.html.twig template.
gulp.task('css-critical', async function() {
const URL = 'https://exampleurl.com';
let criticalCSS = '';
const browser = await puppeteer.launch({
headless: true,
args: [`--window-size=1440,900`],
defaultViewport: null
});
const page = await browser.newPage();
await page.coverage.startCSSCoverage();
await page.goto(URL, {waitUntil: 'load'})
const cssCoverage = await page.coverage.stopCSSCoverage();
for (const entry of cssCoverage) {
for (const range of entry.ranges) {
criticalCSS += entry.text.slice(range.start, range.end) + "\n"
}
}
await page.close();
await browser.close();
require('fs').writeFileSync('css/critical.css', criticalCSS);
});
With everything in place, all that's left to do is to inject our critical CSS file into our template and move all render-blocking CSS and JS to the bottom of our DOM. html.html.twig should end up looking something like this:
{%
set body_classes = [
logged_in ? 'user-logged-in',
not root_path ? 'path-frontpage' : 'path-' ~ root_path|clean_class,
node_type ? 'node--type-' ~ node_type|clean_class,
db_offline ? 'db-offline',
]
%}
<!DOCTYPE html>
<html{{ html_attributes }}>
<head>
<head-placeholder token="{{ placeholder_token|raw }}">
<title>{{ head_title|safe_join(' | ') }}</title>
<style media="screen">
{% include directory ~ '/css/critical.css' ignore missing %}
</style>
<js-placeholder token="{{ placeholder_token|raw }}">
</head>
<body{{ attributes.addClass(body_classes) }}>
<a href="#main-content" class="visually-hidden focusable skip-link">
{{ 'Skip to main content'|t }}
</a>
{{ page_top }}
{{ page }}
{{ page_bottom }}
<css-placeholder token="{{ placeholder_token|raw }}">
<js-bottom-placeholder token="{{ placeholder_token|raw }}">
</body>
</html>
And that's it! This approach helped us move our PageSpeed Insights score between 50 and 60 points up from the initial 20-30 we were getting.
Some improvements that can potentially be done:
- Remove duplicate declarations by comparing critical CSS generated and regular CSS
- Remove unwanted elements that might not be considered critical for the site, i.e.: a slider, video decoration, animations
- Create a page-by-page approach to serve critical CSS that's adjusted for each page rather than just one page used in general
Top comments (0)