Inline styles.css in <head> #145

KyleAMathews · 2016-02-25T06:50:18Z

No description provided.

KyleAMathews · 2016-02-26T00:22:02Z

Actually... I have no idea if this is possible to do automatically. If someone knows, please do share.

KyleAMathews · 2016-03-01T17:24:06Z

Possibly could use the val-loader. Output the styles.css file then sync load the file when generating the and inline the css if < 10 kb or just add a link to the external file.

https://github.com/webpack/val-loader

KyleAMathews · 2016-03-01T19:12:21Z

Actually easier than that. Just use the raw-loader to load the styles.css file. If it's over 10kb, include a link, if it's below 10kb, inline it. This could actually be turned into a react component pretty easily something like <CSSInliner path={"styles.css"} />

gesposito · 2016-03-01T19:49:06Z

Loaders usually accept the limit parameter
https://github.com/gatsbyjs/gatsby/blob/master/lib/utils/webpack.config.js#L226

KyleAMathews · 2016-03-01T20:25:15Z

Interestingly Google AMP's project says all css styles should be inlined in the and not be external files. Regardless of the size (though they say limit styles to 50kb)... https://www.smashingmagazine.com/2016/02/everything-about-google-accelerated-mobile-pages/

So perhaps this should be turned on by default. Webpack extracts and minifies CSS and then we inline it in the <head>

KyleAMathews · 2016-03-12T20:34:56Z

So played around with this a bit today. To make this work, we'll need to have a special styles build step it seems so that the extracted/minified styles.css exists for the html.js to require it.

KyleAMathews · 2016-03-12T20:43:07Z

Which is a bit weird perhaps + extra complexity but I think it's totally worth it. Inlining CSS in head in my testing improves time to first paint by ~300ms on even a very simple page by avoiding the extra request.

kylegach · 2016-05-04T03:48:42Z

I was able to get purifycss-webpack-plugin working with this gatsby-node.js:

var ExtractTextPlugin = require("extract-text-webpack-plugin");
var Purify = require("purifycss-webpack-plugin");

exports.modifyWebpackConfig = function(config, env) {
  if(env === 'production') {
    config.removeLoader('css');
    config.loader('css', function(cfg) {
      cfg.test = /\.css$/;
      cfg.loader = ExtractTextPlugin.extract(['css', 'postcss']);
      return cfg
    })
    config.plugin('extract-css',
                  ExtractTextPlugin,
                  ["styles.css", { allChunks: true }]);
    config.plugin('purify-css',
                  Purify,
                  [{
                    basePath: __dirname,
                    resolveExtensions: ['.js, .html'],
                    paths: [
                      'html.js',
                      'components/**/*.js',
                      'pages/**/{*.md,*.js,*.html}',
                      'wrappers/*.js'
                    ],
                    purifyOptions: {
                      info: true,
                      rejected: true,
                      whitelist: [
                        '*markdown*',
                        '*footnotes*',
                        '*webfonts-loaded*',
                        '*is-reverse*'
                      ]
                    }
                  }]);
  }

  return config;
}

(Explanation of Purify’s options, here.)

As you can see, that implementation is quite fragile, requiring both specifically-defined file globs for the paths to scan and carefully selected whitelisted selectors (the webfonts-loaded and is-reverse classes are specific to my project, but everyone is likely to need at least markdown).

The tool, PurifyCSS, also doesn't seem to work all that well. I have quite a few selectors in my styles (cssstats), and I expected at least of third of those to be un-used. But it only reduced the un-minified filesize by 9.5%, and here’s the full list of removed selectors:

menu
canvas
progress
progress
strong
dfn
mark
sub
sup
sub
sup
kbd
samp
hr
select
textarea
optgroup
select
fieldset
legend
textarea
ol
ol
ol>li
ol>li:before
ol
ol>li:before
.fontFamily-inherit
.textDecoration-none
.underline
.textLeft
.textJustify
.nowrap
.breakWord
.truncate
.listStyle-none
.tableCell
.overflow-scroll
.overflowx-hidden
.overflowx-scroll
.overflowx-auto
.overflowy-hidden
.overflowy-scroll
.overflowy-auto
.fit
.flex-columnReverse
.items-stretch
.self-start
.self-end
.self-center
.self-baseline
.self-stretch
.justify-around
.content-around
.content-stretch
.rounded
.rounded-top
.rounded-right
.rounded-bottom
.rounded-left
.not-rounded
.bgLighter-1
.bgLighter-2
.bgLighter-3
.bgLighter-4

I think this is largely due to my use of escaped characters in my classes, e.g. flex@md (though there’s evidence that the maintainer addressed this already) and my use of quite a few classes in total. You can see that I have many, many classes using escaped characters, yet none were removed. I know for a fact I haven’t yet used any of the ...@lg classes, so at the very least, all of those should have been removed.

Worse, there are false positives in that list. My .md content uses ordered lists, but the tool doesn’t pick up on it because it scans the files before conversion to html.

It’s that very last point that makes me think using a tool like PurifyCSS isn’t workable before the pages are built.

And, the plugin can only output a file, so it doesn’t actually address the key aspect of this issue: inline the styles when they’re small.

KyleAMathews · 2016-05-08T04:58:38Z

With #253 this is about ready to go once the new release is done.

Checklist for finishing this up:

Add new helper in gatsby-helpers for html.js that inlines css using the Webpack raw-loader.
Convert starters to inline css
Document

And some testing I just did with #253 with the blog starter. Switching to inlined css dropped the Speed Index score by 1/2(!!!) on both 3G and cable. Which is even a bigger drop than I expected.

3G:
not-inlined: http://www.webpagetest.org/result/160508_FF_8WW/
inlined: http://www.webpagetest.org/result/160508_95_8YQ/

Cable:
not-inlined: http://www.webpagetest.org/result/160508_T2_8ZK/
inlined: http://www.webpagetest.org/result/160508_C5_8ZQ/

It's crazy but with inlining CSS, the site loads just as fast on 3G as not-inlined does on cable. Nuts. Also on cable the inlined css site is 89% visually complete in 0.5 secs. That's a fast website.

This also scores us 100/100 on PageSpeed https://developers.google.com/speed/pagespeed/insights/?url=http%3A%2F%2F572ec371d6865d789f20e371.peddler-tapir-41575.netlify.com%2F&tab=mobile

/cc @scottnonnenberg @kylegach

KyleAMathews · 2016-05-11T03:49:43Z

Tested this on my website relaterocket.co tonight for a more real-world test as this site uses https + custom fonts from Typekit. Redeploying with inlined css dropped the first-load PageSpeed from ~1600 to ~1000. Re-running the test multiple times and from multiple locations I saw significant variation from ~700-1400 but the majority at around 1000. Also re-running the same test seemed to speed up the tests. Perhaps the browser at the location does cache some things.

Before: http://www.webpagetest.org/result/160511_SG_A5T/
After: http://www.webpagetest.org/result/160511_5G_Q2J/

KyleAMathews · 2016-05-11T18:44:59Z

So I ended up not adding the helper as well... I couldn't make it work. Webpack can still be incredibly confusing sometimes :-) I put in the helper <style dangerouslySetInnerHTML={{ __html: require('!raw!public/styles.css') }} /> much like what I ended up adding to the starters directly but it'd just complain that it couldn't find the module. I thought with our modulesDirectories as they are that'd work... anyone else want to take a crack at making a simple CSSInline component which inserts public/styles.css in the head?

KyleAMathews · 2016-05-11T18:45:42Z

But ignoring that, things are looking great! This is a fairly significant performance win. Next big win will be making it simpler to inline images in css.

KyleAMathews · 2016-05-11T19:27:12Z

Oh and service workers of course 💥

ben-hamelin · 2020-04-17T14:00:18Z

Just wanted to drop a note here, just needed some quick css in the head. Used:

let css = <style dangerouslySetInnerHTML={{ __html: require('!raw-loader!./inline-styles.css') }} />

  return (
    <html {...props.htmlAttributes}>
      <head>
        <meta charSet="utf-8" />
        <meta httpEquiv="x-ua-compatible" content="ie=edge" />
        <meta
          name="viewport"
          content="width=device-width, initial-scale=1, shrink-to-fit=no"
        />
        {props.headComponents}
        **{css}**

where inline-styles.css is in /src/ next to html.js

scottnonnenberg · 2020-04-17T14:43:22Z

Thanks. Seems like the build process still spends a bunch of time building all that client-side logic? I was actually more concerned with that, with unnecessary build time, when I made those changes back then!

…

On Fri, Apr 17, 2020, 7:00 AM ben-hamelin ***@***.***> wrote: Just wanted to drop a note here, just needed some quick css in the head. Used: `let css = <style dangerouslySetInnerHTML={{ __html: require('!raw-loader!./inline-styles.css') }} /> return ( <html {...props.htmlAttributes}> {props.headComponents} *{css}*` where inline-styles.css is in /src/ next to html.js — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#145 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADME7L3AKIKGT5MSYALNDLRNBOIHANCNFSM4B4J4LVQ> .

KyleAMathews changed the title ~~Inline smaller CSS files~~ FEATURE: Inline smaller CSS files in <head> if lower than x KBs Feb 26, 2016

KyleAMathews changed the title ~~FEATURE: Inline smaller CSS files in <head> if lower than x KBs~~ FEATURE: Inline styles.css in <head> if lower than x KBs Feb 26, 2016

KyleAMathews mentioned this issue Mar 14, 2016

build is not working #192

Closed

KyleAMathews changed the title ~~FEATURE: Inline styles.css in <head> if lower than x KBs~~ FEATURE: Inline styles.css in <head> Apr 4, 2016

KyleAMathews changed the title ~~FEATURE: Inline styles.css in <head>~~ Inline styles.css in <head> Apr 4, 2016

KyleAMathews added the enhancement label Apr 4, 2016

KyleAMathews self-assigned this Apr 4, 2016

KyleAMathews mentioned this issue Apr 4, 2016

Test performance of starters #234

Closed

This was referenced Apr 27, 2016

Webpack config refactor; CSS now generated with HTML #253

Merged

Add kylegach.com to list of sites #270

Merged

MoOx mentioned this issue May 8, 2016

Ability to inline CSS in HTML MoOx/phenomic#459

Closed

KyleAMathews closed this as completed May 11, 2016

KyleAMathews mentioned this issue May 4, 2017

Upgrade to webpack 2 in 1.0 branch #461

Closed

andrewagain mentioned this issue Jul 9, 2019

inline styles in <head> sveltejs/sapper#796

Closed

LekoArts mentioned this issue Jan 20, 2020

Add option to generate external css file instead of inlining all styles in a <style> tag #20710

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Inline styles.css in <head> #145

Inline styles.css in <head> #145

KyleAMathews commented Feb 25, 2016

KyleAMathews commented Feb 26, 2016

KyleAMathews commented Mar 1, 2016

KyleAMathews commented Mar 1, 2016

gesposito commented Mar 1, 2016

KyleAMathews commented Mar 1, 2016

KyleAMathews commented Mar 12, 2016

KyleAMathews commented Mar 12, 2016

kylegach commented May 4, 2016 •

edited

Loading

KyleAMathews commented May 8, 2016 •

edited by xavivars

Loading

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

ben-hamelin commented Apr 17, 2020 •

edited

Loading

scottnonnenberg commented Apr 17, 2020 via email

Inline styles.css in <head> #145

Inline styles.css in <head> #145

Comments

KyleAMathews commented Feb 25, 2016

KyleAMathews commented Feb 26, 2016

KyleAMathews commented Mar 1, 2016

KyleAMathews commented Mar 1, 2016

gesposito commented Mar 1, 2016

KyleAMathews commented Mar 1, 2016

KyleAMathews commented Mar 12, 2016

KyleAMathews commented Mar 12, 2016

kylegach commented May 4, 2016 • edited Loading

KyleAMathews commented May 8, 2016 • edited by xavivars Loading

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

KyleAMathews commented May 11, 2016

ben-hamelin commented Apr 17, 2020 • edited Loading

scottnonnenberg commented Apr 17, 2020 via email

kylegach commented May 4, 2016 •

edited

Loading

KyleAMathews commented May 8, 2016 •

edited by xavivars

Loading

ben-hamelin commented Apr 17, 2020 •

edited

Loading