Drop :root limitation using new selector/rules injections ?

[MoOx]

Like said here I was that dropping :root limitation should be that hard using selector injection like it has been already said in other issues. This could also fix #17

:root {
  --mainColor: red;
}

a {
  color: var(--mainColor);
  other: rule;
}

p {
  --mainColor: blue;
  border: 1px solid var(--mainColor)
}

I would expect

a {
  color: red;
  other: rule;
}

p {
  border: 1px solid blue;
}

p a {
  color: blue;
}

Not sure about what I'm missing. I know we cant guess dom & so update appropriate result, but we could inject new selectors with rules that make use of var() like in the example above.
Obviously that can represent a huge amount of selectors to add if there is a lot of variable & a lot of usage, but that will do the job "as expected".

I was thinking of an incoming issue with #id selectors related to injection but can't remind it right now it's too early in the morning here.

[necolas]

It's impossible to make it work like native css vars would. Non-:root variables are tied to the DOM tree so you can't know what selectors to create ahead of time.

[MoOx]

I know we can't guess the dom, that's why i was mentionning the possibility to inject all possibilities under selectors that are defining non root custom prop. This may sound brutal indeed but I dont understand why this will be not enough.
Could you provide me an example that cant be reproduced with injection please ?

[necolas]

Because there are too many possibilities. What about nested components each with their own local variable overrides? It's not something we can get right outside of the DOM tree. Even if we could, it would be a nightmare to debug this kind of thing.

[tabatkins]

Agreed. It's theoretically possible to do this, and might even be plausible if you only override a given variable in one place, but the number of rules necessary scales at least as fast as (number of places setting the var)*(number of places using the var). I think it actually goes combinatorial on one of those factors, but I don't want to think it through. ^_^