An Extended Response

This article, part of the writing collection, was published on .

If you've dabbled in Sass before, you're likely familiar with @extend and @mixin, but is there any use-case for @extend? Given the benchmark performance of the two, can we still be confident in the cascade when @extending in Sass?

A couple of weeks ago, CSS Wizard, Harry Roberts, wrote about some performance benchmarking he did with Sass’s @extend and @mixin, specifically in the context of how the Sass is compiled into CSS in both cases and how that output affects the pipeline of data to the end-user.

N.B. This assumes that you are gzipping your assets on the front end.

I’ve been using both @extend and @mixin in my Sass for a number of years, but finding the balance between the two techniques, to find what works for me, took some time. I think it was Harry’s article that prompted me to re-evaluate how I approach the subject.

The gist of Harry’s article is that using @extend to repeat styles is not worth the fragile nature of how @extend works—declaring the styles once and pulling the selectors into a single declaration (somewhere in the compiled CSS output) can spell disaster for the cascade you work so hard to make performant and succinct. You should always feel in complete control of your Sass’ compiled output.

While using @mixin means repeated code in the compiled CSS output (remember: Don’t Repeat Yourself), gzipping our CSS will actually make the impact of this repeated code almost negligible; however, it’s important to ensure that any and all static output from a @mixin is in the same order in each instance it is used. This is because gzip casts repeated strings to a variable and substitutes in the variable (instead of the full string) wherever the string is repeated in the CSS. By ensuring the same (or as close to the same) order of selectors, properties, values, etc. in the output of a @mixin, you set the maximum-possible-length-string of characters to a variable and thereby reduce the size of your gzipped output CSS.

If you want a clearer explanation or more information on the above, check out Harry Roberts’ article.

Harry ran some benchmarking tests on both @extend and @mixin, and determined that @mixin were as performant as, if not more than, @extend. This spurred me on to refactor my Sass and make sure I was using @extend properly…

While this may be a contentious topic, I believe there are some predictable use-cases for Sass’ @extend, and the key in these use-cases is what exactly is being @extended. Typically, developers use @extend to apply one or more properties to multiple selectors, but as we know, this can cause unpredictable CSS output as it’s difficult to remember every selector for which you’ve extended the properties.

The method which is far more predictable in its use of @extend does something different. Instead of @extending properties to multiple selectors, @extend multiple selectors to a placeholder selector (also called a silent class) and reference the placeholder selector wherever applying styles to those multiple selectors.

  • %content-elements
  • %heading-elements
  • %monospace-elements
  • %button-elements

I have created the above placeholder selectors to represent different groups of multiple selectors, and the terse naming helps understanding what they represent (at least I think so). How these are used is best explained with an example, so let’s look at how I target all heading elements (h1, h2, h3, h4, h5, h6) throughout my Sass:

h6 {
	@extend %heading-elements;

If you recall how placeholder selectors and @extend work, you can understand that what we’re doing here is telling our Sass to do a simple replacement during compilation: Wherever %heading-elements appears in the Sass, replace it with h1, h2, h3, h4, h5, h6. This comes in handy when you want to target heading elements outside the scope of where your heading’s styles are defined.

For example, so that users can easily link to a specific part of the content, attached to most headings is a fragment anchor, .fragment-anchor. These anchors are direct children of heading elements, and should only appear when the user interacts with the heading element (by hovering with the mouse or tapping on a touch-device):

.fragment-anchor {
	display: none;

	h1:hover &,
	h2:hover &,
	h3:hover &,
	h4:hover &,
	h5:hover &,
	h6:hover &,
	h1:focus &,
	h2:focus &,
	h3:focus &,
	h4:focus &,
	h5:focus &,
	h6:focus & {
		display: inline-block;

This makes sense, but is just too long-winded. We can simplify this declaration by referring to our placeholder selector, %heading-elements, like so:

.fragment-anchor {
	display: none;

	%heading-elements:hover &,
	%heading-elements:focus & {
		display: inline-block;

The CSS output of the two techniques is exactly the same, but the Sass source files are now much easier to understand. The advantages here are that we aren’t making any negative impact on the cascade of the compiled CSS because we’re not using @extend in such a way that the source order of declarations doesn’t matter—we’re using @extend to represent multiple selectors. This technique also makes the Sass much easier to scan through, and saves us from having to remember multiple selectors whilst traversing our Sass partials.

I wholeheartedly believe in a very simplistic approach to Sass. Simplicity is king when it comes to writing CSS, and going overboard on Sass features may give you a false sense of power.

The purpose of Sass is not to write less CSS but to make the CSS that we write more understandable.

tl;dr If you find yourself referring to the same group of selectors over and over, try @extending them to a placeholder selector and refer to the placeholder selector outside of the scope of the selectors themselves.