Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Add support for Jest in stylelint-rule-tester #815
(opening issue here since stylelint-rule-tester doesn't have them enabled)
It would be great to add support for Jest in stylelint-rule-tester.
I think you guys might have guessed but we at Facebook are in the process of moving our old crappy linter to use wonderful Stylelint and PostCSS instead :) We're using Jest here for testing. What we've done for now is create a separate jest-stylelint-rule-tester which is essentially a stylelint-rule-tester but with API replaced to work with Jest (
But of course, ideally, we'd love to add support for this "natively" so that other people can use it too and to have 1 source of truth instead of maintaining our own version.
So I wanted to throw this out there and get your thoughts. Would you be open to replacing Tape with Jest entirely? Or would we add support for Jest on top (and if so, is there anything Tape can do that Jest can't)?
@cpojer tells me that Jest 0.9.0 is like super fast so it might be even faster than what you currently have ;)
I am definitely interested in figuring out a better system for rule testing. I will probably have some time later today to think about this a little more and maybe write down some of my thoughts about alternatives we might consider.
Thanks for raising the issue. And I'm really glad to hear that you're adopting stylelint at Facebook! That's awesome.
Here's what I'd really like --- I'd like to have an abstraction of some kind (for running a CSS string against a stylelint rule and checking the warnings) that it itself test-engine-agnostic. Ideally, you could use that abstraction with whatever test engine you preferred -- Jest, tape, Mocha, Jasmine, AVA, etc. -- either directly or with a very light wrapper that wired up the particulars of the test engine.
So instead of just switching from tape to Jest, or AVA, or something else, I'd really like to build the abstraction that would make testing rules easy for everyone. If that's doable. I'm trying to brainstorm possible ways of approaching this problem. Input, feedback, help is all very appreciated.
We could use Node's
We could make a test-describing schema that is just arrays of test-case descriptions, which should or should not trigger warnings. But then we still need an abstraction that interprets the schema and applies some assertions to it. And so you change the syntax of the tests without solving the fundamental problems.
One big problem standing in the way is asynchronicity. We run the string through PostCSS (using stylelint as a plugin) then look at the generated result for warnings. That process is async. The trouble is that it seems each testing framework deals with async in a different way, so I'm having a hard time deciding how to approach that and come out with a syntax that's succinct enough to be valuable. (Being able to write tests as succinctly as we can currently is essential. That's the only way I've stayed sane writing so many tests.)
Another problem is getting good reports when something fails --- helping you to understand the failure and track down the specific test that failed. Creating an abstraction over any test framework seems to cause a problem here: they're all calibrated to rely on the location of their assertions, or something like that.
The last problem is that there are over 10000 unit tests already in this repo, and if we were to change the way the tests themselves are written, we'd really really really want to be able to update the tests automatically, with some regexp find-replace or something, rather than manually ...
Again, any ideas welcome. Before considering switching from tape to another test engine I'd really like to see if it's feasible to create something like the abstraction I'm getting at.