In mid 2019 I got the urge to work on a browser extension. It was pretty fun!
Everyone has different workflows when it comes to their browser tabs. Some people open so many that you can no longer see a tab’s title, and the favicon is all that remains of the content.
I personally like to have a few pinned tabs. For me, if it’s not worthy enough to be pinned then it’s not worthy enough to stick around. And if it IS worthy enough to be pinned, then I want it open by default when the browser starts. This heuristic helps me focus on one task at a time, which I find makes me more effective.
It turns out that my preferences are not shared by most people. This makes a default behavior of most browsers not exactly what I want. And since I’m always looking for a cool new programming project, GrayTabby happened.
Unsurprisingly, the UX for most web extensions is written in HTML and JS – languages that the browser is already good at working with. This fact decides a lot of the development toolchain.
NodeJS / NPM / Webpack
At the end of the day, I need to be in an ecosystem with access to good libraries that can compile down to HTML/JS/CSS. NPM is for downloading and managing dependencies. Webpack is for bundling the UI from its constituent ingredients. Node is the runtime for both of these technologies.
VSCode / Typescript
I’m extremely optimistic on VSCode. While it has the downside of being in a hyperspeed ecosystem, it is also completely open and backed by a major tech player (Microsoft). The TS and JS integrations are the best I’ve seen in the industry.
I’ve gotten a ton of mileage out of the TS type system. Lots of stupid bugs avoided. And the compiler is very flexible in the kind of JS it targets.
DefinitelyTyped is a great repository of people who open-source their quality type declarations.
Web extensions have a long history. I started paying attention when they became a large part of Chrome. Eventually firefox supported them too, mostly reverse engineering the Chrome API since so many extensions had already been written for Chrome. Firefox is pushing a new standardization, but browser makers aren’t in a hurry to get on the same page here.
Testing is a big pain and I almost wish I hadn’t even bothered. I found multiple bugs in the toolchain:
Also, so much of the execution environment is too tightly integrated with the browser. This is reasonable (it IS an browser extension, after all) but the mocking logic is so complicated compared to the system under test.
This is the least fun part of the process. Bundling your code and submitting it to a web store required a lot of extra steps. The process was even outright broken for firefox. It really ought to be a one-click solution for open source projects. If I have a travis-CI on an open source project that produces the code, I should have a simple flow to deploy it on a web store via an OAuth integration.