Define :user-invalid / :user-valid for form-associated custom elements

[nt1m]

#9047 omitted defining the pseudos for FACE. Perhaps a new ElementInternals state is needed here?

[annevk]

See also #8938.

[annevk]

I think we pretty much have to expose a way to set "user interacted". But that also means we need some kind of opt-in for setting that unless we're okay with existing form-associated custom elements never matching :user-valid and :user-invalid. Maybe we are?

cc @mfreed7 @smaug---- @emilio

[nt1m]

I would propose ElementInternals.setUserInteracted() along with a getter. Then the selector matching would function like other form controls by checking both the validity and that flag.

[smaug----]

Not sure. Could it possibly happen automatically as part of focusing? It feels rather error prone if one needs to remember to explicitly call setUserInteracted() - FACE would always need to have a listener for focus which calls setUserInteracted().

[nt1m]

Focus isn't representative of the user interacted flag. I would say the change event is more accurate. Potentially the flag could be set on that event, but not sure the developer would expect it.

[annevk]

I think that's why it needs a separate opt-in. We need to know if you want to participate in this mechanism. I suppose we could also make it a tri-state:

• internals.userInteracted = "auto" | "true" | "false"

[nt1m]

auto is a good idea in theory, however, I'm concerned it's really a fake "auto" if the behavior can't be updated later on for compat reasons.

[annevk]

If you control a custom element presumably you can update it to use the other values (which would then influence matching)? I suppose your consumers might not expect the change, but that's not really something we can concern ourselves with.

[vospascal]

@annevk is there any progress on this i just walked into this issue with implementing FACE I'm seeing it working with :valid and :invalid but not yet with :user-valid and :user-invalid I was already thinking of implementing pseudo-states but since that is also not supported everywere that doesn't help much

one lingering question i have is these combined states are very tailored towards a combination of states (inspired by good old angularjs): dirty/pristine, touched/untouched, and valid/invalid If one would be able to read these one could also determine the state user-valid / user-invalid. Wouldn't that be a more flexible solution in the long run?

[annevk]

I think that's what https://github.com/whatwg/html/issues/9639#issuecomment-1723223481 suggests, no?

WebKit is in principle okay with doing something here, but other browsers haven't weighed in much.

[josepharhar]

It feels rather error prone if one needs to remember to explicitly call setUserInteracted() - FACE would always need to have a listener for focus which calls setUserInteracted().

I agree, I don't really see the value in supporting :user-valid/:user-invalid if you have to do this. At that point, you could create your own css custom state selector for your custom element and use that instead. Maybe :user-valid/:user-invalid is better than custom state selectors, but it seems like the same thing.

If there is a good way for the user agent to know if there was any interaction inside the custom element that sounds preferable, but I'm not sure what that would be.

[annevk]

I think the value in supporting :user-valid / :user-invalid is the same as with supporting :disabled and others. In that the web developers that end up using the form-associated custom elements get a consistent API surface across them and built-in elements.

[josepharhar]

Good point. I suppose that if we just have to add something to ElementInternals, then it should be pretty easy to implement