[selectors‑5] Making :: a (pseudo‑)combinator

[ExE-Boss]

Due to backwards compatibility issues, :: can’t be a proper combinator (see #2284 (comment)).

There’s however nothing preventing us from making :: a combinator‑like thing that would preserve the current semantics, while allowing for #2284 (comment) and #4565.

---

The goal is to support:

* :: *:is(before, after, marker) { … }
/* being roughly equivalent to: */
*::before, *::before::marker,
*::after, *::after::marker,
*::marker,

*::slotted(*)::before, *::slotted(*)::before::marker,
*::slotted(*)::after, *::slotted(*)::after::marker,
*::slotted(*)::marker,

*::part(*)::before, *::part(*)::before::marker,
*::part(*)::after, *::part(*)::after::marker,
*::part(*)::marker { … }

But for backwards compatibility:

*::*:is(before, after, marker) { … }
/* meaning: */
*::before, *::after, *::marker { … }

and

* ::*:is(before, after) { … }
/* meaning: */
* ::before, * ::after, * ::marker { … }

Related issues

• [selectors-4]Why Pseudo-elements cannot be represented by the :is()/:matches()? #2284 (comment)
• [selectors] New universal selector including tree-abiding pseudo-elements #4565
• [css-shadow-parts] Unifying ::part() and ::--foo #4900
• [selectors] Child & descendant pseudo element combinators #7346

[emilio]

*::*:is(before, after, marker) { … }

I don't think such a selector would make sense. :is(before, after, marker) is already a valid selector, and means a totally different thing.

[ExE-Boss]

It’s on the selector that’s after the :: bit.

[valtlai]

What about adding new pseudo-elements? Like ::is(before, after, marker), ::where(before, after, marker) and maybe also ::not(before, after, marker).