Allow & Dynamic<T>

[Simn]

Allows this:

typedef KnownDynamic<T> = {known:T} & Dynamic<T>;

function main() {
	var d:KnownDynamic<Int> = { known: 12 };
	d.also = 12;
}

We have this Extend thing for anon_status which I'm using to implement this. In that context, I've noticed that apply_params doesn't recurse into it, which is likely a separate bug.

@RealyUniqueName Could you check this? I might have missed something.

CI failure is unrelated.

closes #9825

[kLabz]

Could this also [be made to] work with {known:T} & Dynamic / {known:T} & Dynamic<Any>?

This is something that would be very useful for react externs, as components there often accept a well defined list of attributes (props) and also pretty much anything else, that would be transferred to child components or something (that's what you get for asking javascript developers to define some typing rules I guess :P)

[Simn]

Both of these work already.

[kLabz]

Thanks! Can't wait for 4.3 :P

[R32]

What if there is a typo error? The compiler doesn't seem to remind you

var d:KnownDynamic<Int> = { knowm: 12 };  // typo

[Simn]

That errors with Object requires field known.

[RealyUniqueName]

Maybe check if current T is not compatible with the one stored in extends_dynamic?
Theoretically multiple Dynamic<T> could get into tl from different sources.

[RealyUniqueName]

It seems this kind of signatures would have issues when both A and B extend Dynamic<T> depending on the case if Ts are different or not.

//@:generic
function unite<A,B,C:A&B>(a:A, b:B):C;

[Simn]

Hmm, I expected unification to handle that case already, but it looks like it doesn't do anything with Extend anons at all. Which makes me wonder why the thing even exists...

[Simn]

I found this: #3228

This means that the Extend status is purely informative, which makes my approach here insufficient.

[Simn]

This approach isn't good so I'll close this for now.