The "Burning Man App" (I'm settling on the working title BMapp) that I'm building has stalled for a minute. Before I dig in, and move from proof of concept to actual build, I want to make sure that the architecture and pipeline are solid. That is to say, I'd like a good framework and a build task that keeps the coding itself relatively simple. For the framework, I've settled on Inferno, because it is light and fast. While its key developer has left (?) the project to move onto React at Facebook, and I'm not sure what that means for Inferno's future, for now, the relative advantages yet go to Inferno.
As to a build task, the Inferno team itself appears to recommend Brunch, an open-source, community-driven "build tool" rather than a "generic task runner". Its main selling point is that it reduces some of the monstrous complexity that some build task runners can encompass. A single brunch-config.js file pointing to input files, plugins and some output directories, and boom, you have a minified, uglified, compressed, tightened, tested web app. No fuss, no muss. Sounds great! Also, during setup I ran into a few errors (that were ultimately caused by a flawed npm release, not Brunch): with one issue ticket the Brunch team was incredibly responsive. They were on it! Even committing to closing out the issue over the weekend! That is dedication. "Windows support is very important to us."
So, when I say that Brunch isn't working for me, it is by no means a slight of Brunch itself or the hardworking team that created it. If you're looking for a build pipeline, do check it out! This post is more a documentation that I tried it, and why I'm moving on.
I generally like to code javascript using Typescript, for a number of reasons, but mostly because it is javascript, only with strict typing. Coding with it prevents me from easily relying on type ambiguity (and the advantages that ambiguity can sometimes confer), and in return, I'm made aware in the editor, even before compilation, of potential problems. Other advantages: I can use the latest and greatest coding practices from ES6 and later (does what Babel does), and my code will automagically be compiled into whatever version necessary.
Brunch kinda has typescript support, but see, that's the thing: it suffers from the same issue that all popular task runners do: you need to rely on some intermediary plugin that may, or may not, be up-to-date. Brunch uses 'skeletons' which are a kind of template or starter pack for different configurations. Their official 'brunch with typescript' skeleton, out of the box, gives you a `brunch-typescript is unsupported, use typescript-brunch instead` deprecation message. Okay, no big deal, I'll just swap out the brunch-typescript plugin for the typescript-brunch plugin. But typescript-brunch has this notation: "From version 1.8.2 up to current version, this plugin may report TypeScript errors that you are not expecting" and "We are hoping to support the full language service, at least for brunch build at some point, but until then...[etc]". Not exactly inspiring confidence. Okay, no big deal. I'll just use Typescript itself to compile to javascript, and once there, let Brunch takeover to do what it does best.
On to uglify js, the minimizer. I followed all instructions as I understood them to get it working and it just plain did not work, meaning, to minimize and 'mangle' the compiled javascript; or really, to have any effect at all. Simply installing the plugin and referencing it in the config file appeared to have zero effect on the code itself. No doubt, and I mean this, it's my bad. I didn't understand the instructions, or didn't fiddle with it enough. Perhaps if I had changed the order of the plugins, or if I had only spent another 10 minutes or half-an-hour, then I'm certain I could get it working as expected.
And, it's community driven! If it's not working, get on the stick! It's not like you can demand your money back. Write up a ticket describing the issue. Ask questions on the forum! Write the bug fix, if that's what's necessary. I get it! But that's the thing: I had already spent a lot of time with it already. A task runner / builder / DevOp tool / what-have-you is supposed to make things easier, minimize errors and time in the setup and manipulation of code. If you're spending as much time on the tool as you would just rolling your own, then it's maybe best to roll your own. My spare coding time, right now, is dedicated to getting this BMapp up and going.
So, for now, it's back to using npm as my build tool. It's simple, it works, and if something isn't there, it's super easy to code up a module that does what you want.