Writing my first addon for Firefox wasn't too hard or annoying

February 12, 2018

I prefer the non-JavaScript version of Google search results, but they have an annoyance, which is that Google rewrites all the URLs to indirect through themselves (with tracking numbers, but that's a lesser annoyance for me than the loss of knowing what I've already read). The Firefox 56 version of NoScript magically fixed this up, but I've now switched to uMatrix.

(The pre-WebExtensions NoScript does a lot of magic, which has good and bad aspects. uMatrix is a lot more focused and regular.)

To cut a long story short, today I wrote a Firefox WebExtensions-based addon to fix this, which I have imaginatively called gsearch-urlfix. It's a pretty straightforward fix because Google embeds the original URL in their transformed URL as a query parameter, so you just pull it out and rewrite the link to it. Sane people would probably do this as a GreaseMonkey user script, but for various reasons I decided it was simpler and more interesting to write an addon.

The whole process was reasonably easy. Mozilla has very good documentation that will walk you through most of the mechanics of an addon, and it's easy enough to test-load your addon into a suitable Firefox version to work on it. The actual JavaScript to rewrite hrefs was up to me, which made me a bit worried about needing to play around with regular expressions and string manipulation and parsing URLs, but it turns out that modern Firefox-based JavaScript has functions and objects that do all of the hard work; all I had to do was glue them together correctly. I had to do a little bit of debugging because of things that I got wrong, but console.log() worked fine to give me my old standby of print based debugging.

(Credit also goes to various sources of online information, which pointed me to important portions of the MDN JavaScript and DOM documentation, eg 1, 2, 3, and 4. Now you can see why I say that I just had to connect things. Mozilla also provides a number of sample extensions, so I looked at their emoji substitution example to see what I needed to do to transform the web page when it had loaded, which turned out to be a pleasantly simple process.)

There are a couple of things about your addon manifest.json that the MDN site won't tell you directly. The first is that if you want to make your addon into an unsigned XPI and load it permanently into your developer or nightly Firefox, it must have an id attribute (see the example here and the discussion here). The second is that the matches globs for what websites your content scripts are loaded into cannot be used to match something like 'any website with .google. in it'; they're very limited. I assume that this restriction is there because matches feeds into the permissions dialog for your addon.

(It's possible to have Firefox further filter what sites your content scripts will load into, see here, but the design of the whole system insures that your content scripts can only be loaded into fewer websites than the user approved permissions for, not more. If you need to do fancy matching, or even just *.google.*, you'll probably have to ask for permission for all websites.)

This limitation is part of the reason why gsearch-urlfix currently only acts on www.google.com and www.google.ca; those are the two that I need and going further is just annoying enough that I haven't bothered (partly because I want to actually limit it to Google's sites, not have it trigger on anyone who happens to have 'google' as part of their website name). Pull requests are welcome to improve this.

I initially wasn't planning to submitting this to AMO to be officially signed so it can be installed in normal Firefox versions; among other things, doing so feels scary and probably calls for a bunch of cleanup work and polish. I may change my mind about that, if only so I can load it into standard OS-supplied versions of Firefox that I wind up using. Also, I confess that it would be nice to not have my own Firefox nag at me about the addon being unsigned, and the documentation makes the process sound not too annoying.

(This is not an addon that I imagine there's much of an audience for, but perhaps I'm wrong.)

Comments on this page:

By Liam at unc edu at 2018-02-13 08:00:40:

I would use it - the other work arounds to get a real URL off the google results page are complex enough to be irritating, and I look for just such an add-on from time to time in the hope that someone would write one.

By cks at 2018-02-13 11:34:07:

I've put it through the AMO process, which was relatively painless, and people can now get it as Google Search URL Fixup in signed and thus usable-in-anything form.

(Although it's a WebExtension, it doesn't require Firefox 57+; it definitely works in 56. I think it will also work in Firefox ESR, although I'm not sure how good ESR's WebExtensions support is. I'm using only very basic WE stuff, though, just the ability to run regular JS in the page's context and alter bits of the page's DOM.)

By Annonymous Admirer at 2018-02-13 18:17:16:

I love your website style I am very impressed by it.

Written on 12 February 2018.
« Sending emails to your inbox is a dangerous default
Some things about ZFS block allocation and ZFS (file) record sizes »

Page tools: View Source, View Normal, Add Comment.
Login: Password:
Atom Syndication: Recent Comments.

Last modified: Mon Feb 12 23:59:21 2018
This dinky wiki is brought to you by the Insane Hackers Guild, Python sub-branch.