Skip to content

DonKult/dotPageMod

Repository files navigation

Logo of dotPageMod dotPageMod

Firefox WebExtension to load local CSS and JavaScript from your dotfiles into webpages.

Features

  • each hostname has its own directory in your config directory of your own choosing
  • the hostname directories itself can be stored in independent packs for easier sharing and syncing (e.g. private, public, work, home, …)
  • port specific hosts with hostname_port
  • hostnames like mozilla.org match developer.mozilla.org
  • in fact: org matches all hosts in that top-level domain, too
  • ALL for all hosts (on http/https) or ALL_scheme for all hosts on a specific scheme
  • use FRAMEWORK directory to include any/all JavaScript/CSS frameworks (like JQuery) you might want to use in your scripts – they are automatically included before your scripts are sourced.
  • scripts can show desktop notifications without WebAPI (see Cheatsheet)
  • scripts are run in readyState interactive by default, but filenames ending in .start.js run scripts in loading and .idle.js at complete instead (see also RunAt).
  • your JavaScript/CSS works even if the page has them blocked by a Content Security Policy or uMatrix
  • your PageMods apply to the top window (not to frames) and apply to existing pages on (re)load
  • on deactivation CSS sheets are automatically unapplied, JavaScript modifications can be reverted by registering an undo method
  • on Linux with inotifywait (packaged in Debian in inotify-tools) the addon will reload automatically on relevant changes in the config directory
  • a badge on the toolbar button indicates how many files modify the current tab. A list of these files can be accessed in the panel. A red badge & filename indicates a file couldn't be applied, e.g. due to a programming error in it.

(Better?) Alternatives

So, why another one?

A long while ago I was using Greasemonkey, but just hiding a few elements with it felt very daunting with all this metadata and editing in a browser window. I regressed to using Adblock Plus and later uBlock Origin for cosmetic filtering and more and more to block the execution of JavaScript by default. Eventually I introduced uMatrix with a rigid block-policy to the mix which ended up leaving uBlock mostly jobless beside a bunch of cosmetic filters. Management of these isn't perfect through and sometimes you want more than just display: none – especially now that I had all of JavaScript blocked by default and some websites downright refuse to be usable without it (e.g. default collapsed infoboxes). So I moved my filters to ~/.css and started fixing up websites in ~/.js with dotjs. Quickly I ended up hitting issue #27: console.log doesn't work from dotjs which I researched and after commenting researched even more. Set out to write a patch to have this option set automatically I ended up changing other things as well until I realized that the architecture as-is wasn't to my liking (using a single global PageMod reading files dynamically and sending the content to be processed by eval (JS) and by DOM insertion (CSS) – the later failing in the event of a content policy forbidding inline CSS) – and I always wanted to look into developing Firefox extensions…

So, with a "how hard can it be?" I moved on to write my own extension to resolve my real as well as my imaginary problems by introducing new problems – not for me (hopefully), but potentially for anyone (else) wanting to use it…

Cheatsheet

@-moz-document url-prefix(http://www.w3.org/Style/) { }
@-moz-document regex("https:.*") { }
if (window.location.pathname === '/Style/')
if (window.location.pathname.startsWith('/Style/'))

undo changes

browser.runtime.onMessage.addListener(l => {
	if (l.cmd !== 'detach') return;
	// TODO: react here
});

Note: The example nano-framework has some wrappers and examples to undo common changes like event handlers, style toggles and removal of added elements.

Note: If a CSS file is changed and config reloaded the old CSS will be cleanly removed from open tabs and the new one applied. The situation is more complicated for JS: Individual files can't be unapplied (obviously), so in this case the tab gets the detach message described above and all JS files effecting this tab are reapplied. If the scripts effecting this tab have no clean detach behaviour you might be better off reloading the page to get a fresh start.

Beware: This doesn't work on addon update/removal as there is no API for it. The closest might be onSuspend but it is neither implemented in Firefox nor a good drop-in as it can still be cancelled and comes with no runtime guarantees whatsoever. I really hope browser vendors will make up their mind on this as this doesn't feel like a good user experience. On the upside, you hopefully don't need to update the extension itself all too often – and of course never need to remove it. ;)

showing desktop notifications

browser.runtime.sendMessage({'cmd': 'notify', 'title': title, 'message': msg});

Note: A notification via WebAPI requires the website to have permission for it, but our scripts are written by the user, so permission is implicitly given – and independent from the permission status of the website.

modifying current tab (pin, active, mute, reload, close, …)

browser.runtime.sendMessage('tab/activate');
browser.runtime.sendMessage('tab/pin');
browser.runtime.sendMessage('tab/unpin');
browser.runtime.sendMessage('tab/mute');
browser.runtime.sendMessage('tab/unmute');
browser.runtime.sendMessage('tab/close');
browser.runtime.sendMessage('tab/reload');
browser.runtime.sendMessage('tab/force-reload');
window.location = 'https://example.org';
browser.runtime.sendMessage({ cmd: 'tab/url', url: 'https://example.org'});

open a new tab

window.open(url, '_blank');
browser.runtime.sendMessage({ cmd: 'tab/open', active: true, pinned: false, window: 'tab', url: 'https://example.org' });

Note: The first option requires the website to have popup permissions. The second is less simple, but can open tabs in the 'current' window or in the window the tab belongs to the script runs in (default).

embedding images and co in CSS

SVG

URL encoding is enough here and has the benefit of allowing modifications still, the encoded string can be used as url(data:image/svg+xml,ENCODED).

perl -MURI::Escape -e 'print uri_escape($ARGV[0]);' "to encode"
echo 'to decode' | awk -niord '{printf RT?$0chr("0x"substr(RT,2)):$0}' RS=%..

(Source for encode and decode commands)

binary image formats and other things

The URI needs to be url(data:image/TYPE;base64,ENCODED). Encoding can be done with base64 -w 0 < image.file.

URL to directory resolution

  • http://dotpagemod.example.com:8080/ => ALL, ALL_http, com, com_8080, example.com, example.com_8080, dotpagemod.example.com, dotpagemod.example.com_8080
  • http://dotpagemod.example.com/ => ALL, ALL_http, com, com_80, example.com, example.com_80, dotpagemod.example.com, dotpagemod.example.com_80
  • https://dotpagemod.example.com/ => ALL, ALL_https, com, com_443, example.com, example.com_443, dotpagemod.example.com, dotpagemod.example.com_443
  • ftp://dotpagemod.example.com/ => ALL_ftp, com, com_21, example.com, example.com_21, dotpagemod.example.com, dotpagemod.example.com_21

Examples

Websites like OpenUserJS, Greasy Fork and UserStyles can be an inspiration of what you could possibly do with your new JS/CSS powers given enough dedication. More (humble) examples can be easily found e.g. on github by searching for dotjs.

If you are more interested in seeing an example of how this extension could be used in practice by a user you can have a look at the examples folder.

Note: The examples are provided as-is, I am neither endorsing nor recommending using any of the examples or even the websites they might apply to even if they might have been used by me at some point in the past. No guarantees are made that they work as intended and/or (not) eat your babies (instead). You have been warned.

Website owners who find examples applying to their sites should remember that obviously a user has cared deeply enough about the site to modify it to cater even better to its own needs instead of moving away to a competitor before starting a "you are experiencing it all wrong – professionals have designed that for you!" frenzy. You might even recognize the chance to incorporate the changes into your design for the benefit of all users (if applicable).

If you want to have a look you should start exploring with FRAMEWORK/david.js as this nano framework is heavily used in the example configuration – nothing prevents you from using heavier guns like JQuery in your own set of course.

Feel free to propose more examples in patches and pull-requests, but note that I reserve the right to deny any proposal if I deem the example unworthy by my sole reasoning. This explicitly includes examples whose sole purpose it is to hide elements on a page as nobody needs to see an example of display: none¹ even if that is a valid usecase in practice and my own config contains plenty of those.

¹ if you really want to see one: ALL/antisocial.css was made for you.

Installation

At the moment you have to build the extension yourself to install it. Given that I approximate the userbase to be only me, I have no plans to shuffle the addon into the review queue as it would just waste valuable reviewer time. That also means you have to run a Developer/Nightly edition of Firefox as it isn't signed and only those allow disabling the signature requirement with the about:config option xpinstall.signatures.required you have to set to false (see also Mozillas FAQ on extension signing).

If the intro wasn't discouraging enough for you want to first choose a permanent location on your disk to store this repository in for simplicity as you are about to create a softlink to the native application configuration file (at least that is what you need to do on Linux, see Native messaging documentation for details). make install tries to do the right thing for your current user.

With the native app done, you need to install the webextension itself. You can either choose to use one of the releases published on Github or built it yourself from source; that isn't hard, don't worry! You even have two options:

Building with Docker

You can use Docker (packaged as docker.io in Debian) to build the add-on if you so choose. The included Dockerfile has the details if you are unsure how to do it. If you follow it by the letter you will end up with a checkout of the git repository in /mnt/dotPageMod as your permanent location which will including a built dotpagemod.xpi.

Building manually

Apart from a typical Linux system you already have, you will likely need to install Pandoc and zip. Checkout the repository to the permanent location you have chosen, then it is as simple as running make which will produce a dotpagemod.xpi file for you.

Configuring the extension

After acquiring an xpi file including the webextension you can install it e.g. via about:addonsInstall Add-on from File….

Now that you have the extension installed you will need to configure it. Clicking on the new button and in the opening panel on the addon name in the included title bar brings you the fastest to the option panel, but you can also go the usual way as for any other extensions of course.

The options want you to specify a directory containing your scripts and stylesheets. Where you want to store them is up to you and depends a bit on how much you want to share the collections (across profiles, users, computers, …), but you can change the path anytime you like so don't worry too much now and just pick a place. In this directory copy/link the examples/ directory in (not the content, the entire directory). After you have set and saved the option you should be able to visit a website modified by the examples and see the extension working. If not something is probably wrong with the app: See burger menuDeveloperBrowser console. From there you can either keep the examples (or not), modify them and start your own collections. Have fun fixing the web!

Contributing aka Where are all the testcases?

Looking at the source it doesn't take long to figure out that the addon code comes with no tests attached. Why is that you might ask – and you are right! I pondered hard about this myself, but ultimately decided that the way this addon works is too hard to test for me at this point in time given no previous experience with addon development as a whole and that nearly all interesting functions deal heavily with files which isn't exactly a strong suite of the SDK. Until this will inescapably bite me (and it surely will), I will kid you and me alike that this addon is sufficiently small to not need automatic tests…

You can treat the examples folder as well as your personal collection as a testcase – it is what I do. So, if you happen to want to provide a patch, feel free to implement an example for its use as well and fling it my way.

What is in the name?

Back in early 2016 the best way to create a Firefox extension was to use the Add-on SDK. In the SDK an API called page-mod was used to modify webpages by inserting JS and CSS files. The SDK is no more and the API used by WebExtensions is entirely different, but the name stayed as the underlying idea remains true: Having a firefox extension which reads dotfiles and modifies pages with it.

Why isn't it working on about: pages or addons.mozilla.org?

The reason is security. WebExtensions aren't allowed to insert code into privileged tabs like about: pages and a few other domains to avoid people being tricked by evil extensions and co. The older extension types didn't have such restrictions. Perhaps one day there is a way to get them working again for all tabs if a user can manage the risk.

At least for "restricted domains" in Firefox you can use the hidden about:config option extensions.webextensions.restrictedDomains to modify the set – including making it empty, so that e.g. the included example for addons.mozilla.org works again. This option does NOT affect about: pages or the PDF viewer through.

Logo

The logo is derived from Technology Class CSS3-Styling Icon which accompanies the official HTML5 Logo HTML5 Logo by W3C under the CC-BY-3.

The original is black – I am coloring the 'J' in this stylised 3 in yellow and the remaining 2 strokes in blue as JavaScript and CSS tend to be shown with those shield colors accompanying the HTML5 icon. They are also the colors of the SelfHTML logo which was a happy accident through.

Not very creative, I know, but it seemed better than using a gear or wrench…

License

Apart from the logo, which is (also) licensed under the Creative Commons Attribution 3.0 as mentioned in the previous paragraph, the extension is MIT (Expat) licensed.