[css-scoping-1] add :has-slotted() pseudo class

[keithamus]

Based on the discussion in #6867, and the conclusions within, I thought I'd have a go at specifying a :has-slotted( <<compound-selector>> ) pseudo class.

I put this in css-scoping-1; the resolution was to add it in css-selectors-5 but I thought it might be useful living next to :host, :host(), :host-context() and ::slotted(). I am of course happy to move it if folks disagree with this choice.

My intent for this PR is largely to present something worthy of discussion. I hope I've achieved that, apologies if this PR amounts to just noise.

[Westbrook]

Exciting start! Points to the need for :host:has(...) support so that we aren't required to ship arbitrary wrapping elements, but that's a argument for another day.

If you haven't already, I think I might take a pass some tentative WPTs based on this outline during the week of the 12th.

[Westbrook]

"after flattening" is interesting here because it give "non-slotted" content while being required to gather all slotted content. E.G.:

<div>
  <template shadowrootmode="open">
    <style>
      div {
        color: green;
      }
      :has-slotted(div) {
        color: red;
      }
    </style>
    <slot><div>Test</div></slot>
  </template>
</div>

In the example above "Test" will display red in the case that flattened elements are leveraged.

However, if you don't use flattened elements:

<div>
  <template shadowrootmode="open">
    <div>
      <template shadowrootmode="open">
        <style>
          div {
            color: green;
          }
          :has-slotted(div) {
            color: red;
          }
        </style>
        <slot></slot>
      </template>
      <slot><div>Test</div></slot>
    </div>
  </template>
</div>

This "Test" will also be red. And, without flattened elements,

<div>
  <template shadowrootmode="open">
    <div>
      <template shadowrootmode="open">
        <style>
          div {
            color: green;
          }
          :has-slotted(div) {
            color: red;
          }
        </style>
        <slot></slot>
      </template>
      <div>Test</div>
    </div>
  </template>
</div>

This "Test" will also be red.

I'm not sure it's a real thing, but is there a place between ::slotted(*) and .assignedElements({flatten: true}) that would resolve the external assigned to a slot and not the default ones internal to the shadow host?

[sorvell]

For consistency, I think it's important that :has-slotted matches using the same rules for matchable elements as ::slotted.

Concretely, since ::slotted() never matches either <slot> or fallback content, :has-slotted also should not match when it would by selecting either <slot> or fallback content.

See #5482. IMO, the spec is still very confusing and possibly indicates that this should match, but the CSS working group resolved that the current behavior is intended in #5482.

I think matching against fallback content both for ::slotted and :has-slotted is valuable, but I suggest it should be tackled as a follow-on to this feature.