Watch out, HAML will strip an attribute with boolean false, so we need to use a string. Or reconsider using false as a default value..
I wish Jest had the rspec concept of `let`.
It's not usually valid, but can still be entered. I wasn't able to fix the positioning of the :after psuedo element without having a child text node. Maybe it's possible to add an empty child text node, but I didn't think it worth getting down to that level..
Using browser validation. I didn't use model validation because the on_hand pseudo-attribute doesn't support it.
But.. it turned out to not be so simple. Browser validation can't work if the field is hidden, and breaks the javascript. So now I made the javascript smarter, and the end result is more helpful I think.
I couldn't think of a simpler way to hardcode it, so now we have a clever generic method :)
We can assume that hidden elements will stay hidden, but we need to check each time if an element is disabled or not.
I'm starting to think that these stimulus tests are worthless. The environment is not the same as a browser, which creates extra work to deal with the inconsistencies. And it means we're not testing real world behaviour.
So these are just unit tests, but they take extra effort to put together due to the inter-relatedness with the DOM. Hmm.
