A hacky demonstration of how you can, with just a few lines in the header.php of your (non-FSE) WordPress theme,
configure HTMLTidy to run across all of your output files, resulting in:
- Cleaner, tidier, more human-readable HTML code
- Code repaired and formatted to the standard you prefer (e.g. HTML5
<tags>without XHTML-styles self-closing<tags/>where they're not required) - Updating logical emphasis tags e.g.
<b>➡️<strong>,<i>➡️<em> - Hoisting any inline
<style>blocks to the<head>, and re-writing any repeated inline styles as classes - Suppressing comments
- Repairing invalid HTML to save the browser from having to do it every time
Please don't just use this as-is without understanding the implications. Read this entire README and make sure you grok it first. I'm happy to take questions.
To see this working:
- Check out this codebase
- Run
docker compose upto create a new development WordPress site at https://localhost:8890/ - Visit https://localhost:8890/ and go through the install (pick a site name, username, password, etc.)
- Use your new credentials to install and activate the TwentyTwentyOne theme (http://localhost:8890/wp-admin/theme-install.php?search=twentytwentyone)
- Visit http://localhost:8890/ and View Source; see that the code looks like regular old WordPress
- Activate the HTMLTidy21 theme (http://localhost:8890/wp-admin/themes.php), which is a child theme of TwentyTwentyOne.
- Refresh http://localhost:8890/ and View Source again: note that the code has been tidied for consistency, readability, and standards! 🎉
HTMLTidy21 is a child theme of TwentyTwentyOne. src/header.php, in this repo, overrides TwentyTwentyOne's header.php only.
It adds a new output buffer with a callback function: ob_start( 'tidy_entire_page' );. Because it's never explicitly finished, that output buffer is automatically closed when PHP finishes executing i.e. the page content has been produced. Then the function tidy_entire_page is called and passed a string containing all the HTML code of the resulting page.
function tidy_entire_page($buffer) { ... } is where the magic happens. It instantiates the PHP library to HTMLTidy and runs it over the HTML code, producing "tidied" code (following the rules defined in src/header.php).
Note that PHP's bindings to HTMLTidy aren't enabled as-standard, but they can sometimes look as if they are. PHP's a deceptive beast. Try to run this code without the 'tidy' extension properly configured and working and it won't throw an error... and it might even try to re-indent your code! But it won't run the full HTMLTidy suite. This will cause you some frustration, because it just won't seem to be working properly. If you examine Dockerfile you'll see the steps I took to ensure that the PHP-to-HTMLTidy bindings were properly set up in this container, but your real-life hosting may be different (see below).
(Fun fact: PHP has this same strange "Don't have the extension? I'mma try my hardest anyway!" behaviour with ZIP files. If you don't have the 'zip' extension installed, PHP can still make ZIP files, it's just they can't be compressed ZIP files. PHP is a weird language.)
To implement this in your own theme, on your own website:
- Install a recent version of
libtidy. If you're running a relatively up-to-date distro you can probably get this from your favourite package manager with e.g.apt update && apt install -y libtidy-dev. If you're not, then you might have to compile your own. If you get problems like HTMLTidy not believing that certain "modern" elements like<video>exist, check yourlibtidyversion. - Install the PHP extension for HTMLTidy. Depending on how your PHP is built this might be as simple as
apt update && apt install -y php-tidyor uncommenting a line inphp.ini... or it might as complicated as having to recompile PHP but with the--with-tidyflag. - Test that HTMLTidy is installed by running
php -i | grep tidy: if you get any output, that's probably great! - Edit your theme's
header.phpto include the code fromsrc/header.phpin this repository. Note that theob_start(...)command needs to come before any output, even the<!DOCTYPE html>declaration.
- Alternatively, you might be able to get away with hooking into
initfromfunctions.phpor even a plugin, but you run the risk of a higher-priority hook making you have a Very Bad Day. Up to you.
- Reconfigure the HTMLTidy settings according to your preferences. You might find the full list of options helpful.
- Implement any exception rules you want to. E.g. you might opt to allow logged-in-users to not tidy HTML content by sending a GET parameter
?skip-tidy=true, e.g. with some code like this:if( isset( $_GET['skip-tidy'] ) && is_user_logged_in() ) return $buffer; - Make sure you've got some kind of caching set up. Seriously.
HTMLTidy isn't the most expensive operation, but you don't want to be doing it for every single load of every single page. Make sure your webserver's configured to microcache content, or else use a plugin like WP Super Cache with the "advanced" configuration that allows your webserver to send pages to anonymous users without hitting PHP at all.
An alternative approach might be to implement something that runs HTMLTidy on posts as you save them (and then tidy up your header.php/footer.php etc. manually). That's not for me, but maybe that's the direction you want to go in.
I do all kinds of other hacks in my tidy_entire_page function. Some of them... like running preg_replace on a HTML page... I'm not proud of (there's nothing like running an irregular language through a regular expression to make you feel bad about your life choices).
Others are fine, like stripping out <link rel="stylesheet> elements and replacing them with inline <style>...</style> blocks (my HTML + CSS is compact enough that for 95%+ of pages I can fit both into a single <12kb block, at which point it becomes more-efficient to bundle them into a single file and avoid the round-trip!). Your mileage may vary.
But personally, I'm really excited for the upcoming release of PHP 8.4's modern \Dom\Document implementation, which will allow me to dynamically manipulate my output content using XPath rather than the hackiest-of-hacks. Until then, though, HTMLTidy's still a great step-up (if you're the kind of person who cares about how pretty their HTML is!).
Thanks to @edent for encouraging me to share this.