Mixins Better for Performance

Written by Harry Roberts on CSS Wizardry.

N.B. All code can now be licensed under the permissive MIT license. Read more about licensing CSS Wizardry code samples…

When it comes to preprocessors, one of the most frequent questions I’m asked is Mixins or @extend? I’ve always been quite vocal about this topic, and I firmly believe you should avoid @extend for a number of reasons:

1. It alters your source order, which is always risky in CSS.
2. It creates awkward groupings in your code, putting unrelated selectors together.
3. It is very greedy, @extending every instance of a given subject, not just the one you actually wanted..
4. It can get really out of control, really fast.

@extend is now widely considered an anti-pattern, so its usage is thankfully fading out, but we’re not quite there yet.

I was workshopping with a client yesterday and was asked about the mixin vs. @extend situation, to which I gave my usual answer of Don’t use @extend, ever!, and in return I was asked But isn’t @extend better for performance? It generates less code.

It is true that @extend (when used correctly) will produce less CSS, but my answer was a firm no: mixins are better for performance.

I answered the question with quite some confidence, despite having never actually done any tests. The reason for my confidence was a pretty solid theory that I had:

Because gzip favours repetition, surely we’ll get a better compression ratio if we share the exact same declarations, say, 1000 times, than if we shared 1000 unique classes twice.

Y’see, when people talk about the performance of mixins, they’re usually thinking about filesize on the filesystem. But because we have gzip enabled (you do have gzip enabled, right?), we should be thinking about filesize over the network.

My thinking was that once we’ve gzipped our CSS, files with greater repetition of identical strings will end up being smaller than files whose repetition is less frequent, regardless of the size of the files on the filesystem. I posited that a bigger file would end up smaller after gzip if that extra filesize was comprised of repeated string.

I got back to my hotel room and decided to put my theory to the test.

The Experiment

Here’s what I did.

1. I created two CSS files.

2. Each file had 1000 unique classes, generated using Sass:

   @for $i from 1 through 1000 {
     .#{unique-id()}-#{$i} {
       ...
     }
   }
   

3. I gave each class a unique declaration, simply reusing the same random string that formed the name itself by using the parent selector, and I put some nonsense strings either side of that:

   @for $i from 1 through 1000 {
     .#{unique-id()}-#{$i} {
       content: "ibf#{&}jaslbw";
     }
   }
   

4. I then chose three simple declarations that would remain the same across all 1000 classes:

   color: red;
   font-weight: bold;
   line-height: 2;
   

5. In one file, I shared these declarations via a mixin:

   @mixin foo {
     color: red;
     font-weight: bold;
     line-height: 2;
   }
   
   .#{unique-id()}-#{$i} {
     @include foo;
     content: "ibf#{&}jaslbw";
   }
   

6. And in the other I shared them via @extend:

   %foo {
     color: red;
     font-weight: bold;
     line-height: 2;
   }
   
   .#{unique-id()}-#{$i} {
     @extend %foo;
     content: "ibf#{&}jaslbw";
   }
   

All of these test files (and more) are available on GitHub.

This left me with two files made up of totally unique classes and 1000 unique declarations, and with three identical declarations shared in two different ways.

The filesizes of these should not surprise you in the slightest:

• mixin.css came in at 108K.
• extend.css came in at 72K.
• This gives a difference in filesize of 36K.
• Using mixins was 150% larger than using @extend.

This is exactly what I expected—mixins do produce more CSS than @extend does.

But! We have to remember that we should not be worried about filesize on the filesystem—we only care about the sizes of our gzipped files.

I minified and gzipped the two files and got the results I expected:

• mixin.css came in at 12K.
• extend.css came in at 18K.
• This gives a difference in filesize of 6K.
• Using mixins was 33.333% smaller than using @extend.

Amazing! We’ve gone from mixins being 1.5× larger than using @extend, to mixins being 0.3× smaller than using @extend. My theory seems correct!

Making Things More Realistic

I do feel that the test files were pretty fair—creating unique strings for class names was designed to hinder compression, so that we could more accurately test the effects of gzip on our actual subject: the shared declarations.

That said, the test files were pretty unrealistic, so I decided to make things a little more reasonable.

I grabbed the compiled CSS from an existing project, made two copies of it, and I @imported each of my test files into each project respectively. This meant that my test files were surrounded by some 1794 lines of actual, realistic CSS.

I compiled each new test file and these were the results:

• mixin.css came in at 16K.
• extend.css came in at 22K.
• This gives a difference in filesize of 6K.
• Using mixins was 27% smaller than using @extend.

The absolute numbers seem trivial (a mere 6K), but in relative terms, we can achieve a 27% saving over the wire simply by opting to use mixins to repeat declarations over and over, as opposed to using @extend to repeat a handful of selectors.

Mixins Are Better for Performance

My tests showed pretty conclusively that mixins end up being better for network performance than using @extend. The way gzip works means that we get better savings even when our uncompresssed files are substantially larger.

This means that the performance argument for @extend is non-existent. As well as being bad for your CSS, @extend is bad for performance. Please stop using it.

N.B. All code can now be licensed under the permissive MIT license. Read more about licensing CSS Wizardry code samples…

By Harry Roberts

Harry Roberts is an independent consultant web performance engineer. He helps companies of all shapes and sizes find and fix site speed issues.