Hacker News Reader

Comment by pseudoramble | Hacker News Reader

When I try and check out their new page linked from here, I'm getting a certificate error. It looks like it's only issued for Netlify subdomains. Minor, but kinda annoying.

One question I have about Rome is why it tries to take on the testing framework too. I really like the idea of it having all the build tooling together, but testing seems like a sort of separate thing. Other environments I've used let libraries fill the gaps for testing, including JS. Is there a big advantage to pulling it into Rome?

View on HN

If you look at configuring a tool like Jest, there's a ton of "build tool" overlap, typically involving configuring Babel/TypeScript, parsing, path aliases, and so on. In particular, you usually have duplicate configuration for Jest and Webpack for all those things, but with completely different config file formats. Test runners also need to be able to iterate over files, run checks, and print diagnostics.

Those all sound like things that Rome has the ability to do already, _and_ it's another tool that Rome could eliminate from a typical install.

I'm not planning to switch to Rome any time soon, but I definitely understand the mindset involved in that approach.

This is a weird peculiarity of the JS ecosystem. The rest of programming ecosystems do not have dozens of tools all doing a build for you. When you run tests in C you do not use a special cli testing tool that also compiles your project... You just run cc on your test files and run it like everything else. Why JS prefers hiding the build I'll never understand.

Not sure I agree with that statement. For example, I know that nUnit and jUnit are test runners for the .NET and Java ecosystems, and both of those are also runnable via the CLI.

As for the JS ecosystem, it's definitely developed differently than other languages, because it deals with a very different set of constraints. See this article for background explanation (which also links to further reading)

https://www.swyx.io/jobs-of-js-build-tools/

ayy tx for shoutout

Not really: dynamic languages (Python, Ruby, PHP) also have special cli testing tools just like JS, and that's where JS got inspiration from. What makes JS different is that it started having a compilation phase along the way, so both approaches hat to be combined.

You absolutely have this in the Java ecosystem also. Maven or Gradle + IDE plugins that need to be configured for most projects.

Python too, it's lessened by the fact that you can usually skips the build test, but there's often tox, pytest, setup.py, and coverage.py configuration in even simple projects.

That's a good point, and one that I've forgotten since I've been out of the JS world for a while now! Often you're trying to run the code in different environments depending on the context, which does require a lot of extra build tool support. Consolidating that is a reasonable goal, I'd say.

It will be neat to see where they draw the line. For example, will Rome include an assertion library? What about mock and fake support? That feels a bit far from what Rome is trying to do, but if they've implemented a test runner, maybe defaults provided would be good for the DX?

_That_ is an excellent question, and I honestly have no idea what their plans are in that respect :) I've seen a number of Sebastian's tweets over the last couple years as he was working on Rome, and have glanced at the repo README, but haven't poked at the actual work they're doing.

Part of the sales pitch for Jest is that it includes a set of APIs for mocking and assertions built in. Wouldn't surprise me if Rome did the same, but curious how they'd design things.

It will probably include module-mocking in it (aka jest.mock), since it happens during "bundling"/compilation, but the assertion and function mocking can certainly be handled by a third-party library.

I guess the argument is that testing is very often another piece of the "developer experience" of a project, akin to linting or compilation. It makes some amount of sense to me, although I see where you're coming from.

Right. I'm genuinely not trying to make a judgement on the choice - it's just not the choice I'm used to seeing, including in JS environments. So it seems like there must be some reasoning behind it.

I suppose that's true! Practically speaking, it would be nice to have one testing framework to learn and use over-and-over again, but that's also kinda counter to what I'm used to in the JS world.

Reply on news.ycombinator.com