Pseudo classes for the `interestfor` API

[mfreed7]

{edited to rename interesttarget --> interestfor}

The explainer for interestfor proposes four new pseudo classes:

• :has-interest - matches on the invoker when the user is currently showing interest in that invoker.
• :has-partial-interest - matches on the invoker when the user is currently showing partial interest in that invoker. (They have merely focused the invoker, but haven't fully-activated it via the hot-key.)
• :target-of-interest - matches on the target of the invoker when the user is showing interest in the invoker for this target.
• :target-of-partial-interest - matches on the target of the invoker when the user is showing partial interest in the invoker for this target.

To hopefully help with the terminology:

<button interestfor=foo>The *invoker*</button>
<div popover id=foo>The *target* of the invoker</div>

A few notes:

• It felt slightly easier to understand if :has-interest means "has any kind of interest" (including partial), and :has-partial-interest means "specifically only has partial interest". But suggestions appreciated.
• @dbaron pointed out that these names are kind of ambiguous, and also that there is actually a third permutation (match on the invoker but describe the target) but I’m not sure there are use cases for that third set, so I'm not proposing it at this point.
• As always, the names need bikeshedding. E.g. :has-interest or just :interest? And target-of-interest is somewhat descriptive but still leaves something to be desired. Maybe it’s ok.

Thoughts?

[tabatkins]

Yeah, these naming patterns aren't quite consistent with what we use elsewhere.

What about:

• :interest-invoker (matching any interest) and :interest-invoker(partial | total)? (total keyword just a guess)
• :interest-target and :interest-target(partial | total), same meanings

Do we want to expose a state keyword for things that are potentially interest invokers/targets, but that are currently without any interest? Or is that ambiguous when interest isn't being shown?

[mfreed7]

Yeah, these naming patterns aren't quite consistent with what we use elsewhere.

That doesn't surprise me! Thanks for the help with suggestions.

• :interest-invoker (matching any interest) and :interest-invoker(partial | total)? (total keyword just a guess)

I like both. I assume the second suggestion was :interest-invoker-partial ? If so, I think that works.

• :interest-target and :interest-target(partial | total), same meanings

Those were my initial stabs at names too, but the confusing thing is that the attribute is called interesttarget. So it might be confusing to have :interest-target match not on the element with interesttarget, but instead on the target of that element. Having said that, I'm also ok with this naming. Because I really don't like :interest-target-target.

Do we want to expose a state keyword for things that are potentially interest invokers/targets, but that are currently without any interest? Or is that ambiguous when interest isn't being shown?

Hmm. I think that can be handled with [interesttarget]:not(:interest-invoker), no? Perhaps you just mean it might be convenient to have a shorthand for that?

[tabatkins]

I assume the second suggestion was :interest-invoker-partial ?

No, it was a functional pseudoclass that took a few keywords letting you indicate exactly what variety of interest you wanted to match on (while the non-functional one defaulted to some form of interest that we think is likely the most useful). So :interest-invoker(partial).

So it might be confusing to have :interest-target match not on the element with interesttarget, but instead on the target of that element.

I think that's fine - the interesttarget attribute indicates which element is the interest target. It doesn't indicate that the element with the attribute is the interest target.

Hmm. I think that can be handled with [interesttarget]:not(:interest-invoker), no? Perhaps you just mean it might be convenient to have a shorthand for that?

I wasn't sure there was a way to express that; not all negative selectors actually can be written explicitly in a simple manner. I think a shorthand is convenient, yes, since you're likely writing your actual-interest selectors as just :interest-invoker, and having to add the [interesttarget] (to avoid :not(:interest-invoker) over-matching on the rest of the page) isn't super obvious.

[mfreed7]

No, it was a functional pseudoclass that took a few keywords letting you indicate exactly what variety of interest you wanted to match on (while the non-functional one defaulted to some form of interest that we think is likely the most useful). So :interest-invoker(partial).

Oh! Duh, reading it again, I can see this clearly. I like this idea! Are there other functional pseudo classes that also work without the ()? Either way I like the idea, and my vote would be for :interest-invoker (without ()) to be sugar for :interest-invoker(total) - that's usually the most useful selector, since it's a superset of :interest-invoker(partial).

I think that's fine - the interesttarget attribute indicates which element is the interest target. It doesn't indicate that the element with the attribute is the interest target.

Fair enough.

Do we want to expose a state keyword for things that are potentially interest invokers/targets, but that are currently without any interest? Or is that ambiguous when interest isn't being shown?

Hmm. I think that can be handled with [interesttarget]:not(:interest-invoker), no? Perhaps you just mean it might be convenient to have a shorthand for that?

I wasn't sure there was a way to express that; not all negative selectors actually can be written explicitly in a simple manner. I think a shorthand is convenient, yes, since you're likely writing your actual-interest selectors as just :interest-invoker, and having to add the [interesttarget] (to avoid :not(:interest-invoker) over-matching on the rest of the page) isn't super obvious.

Ok - I guess maybe the shorthand would be :interest-invoker(none) or something?

[tabatkins]

Are there other functional pseudo classes that also work without the ()?

Yes, :host and :host(), for example. We've used it in a few other designs that ended up being dropped, too.

I guess maybe the shorthand would be :interest-invoker(none) or something?

Something like that. possible, maybe.

[bkardell]

my vote would be for :interest-invoker (without ()) to be sugar for :interest-invoker(total) - that's usually the most useful selector, since it's a superset of :interest-invoker(partial).

Same.

total feels a little weird to me as a name here but 🤷...

[Psychpsyo]

I think some of these use-cases are handled by reference combinators in Selectors 5.

For example :has-interest /interesttarget/ [popover] would be the same as :target-of-interest, just a bit more verbose.

[mfreed7]

I think some of these use-cases are handled by reference combinators in Selectors 5.

Cool feature! I hadn't seen that one. Is that shipping in any browser?

I still think it sounds like there's interest in adding pseudo classes for convenience, even when there are other ways to express it.

my vote would be for :interest-invoker (without ()) to be sugar for :interest-invoker(total) - that's usually the most useful selector, since it's a superset of :interest-invoker(partial).

Same.

total feels a little weird to me as a name here but 🤷...

Suggestions are welcome! The total value was suggested in #12154 (comment).

[keithamus]

I think :interest-target & :interest-target(partial | full) make the most sense to me.

[mfreed7]

Quick summary of the rough direction of the conversation so far:

• :interest-invoker(possible | partial | total):
  • possible matches when an element has the interestfor attribute. (Perhaps it must also be pointing to a valid idref?)
  • partial matches when an interest invoker has "partial interest", but not "full interest". (This is a special case.)
  • total matches when an interest invoker has "partial interest" or "full interest". (This is the general/common case.)
  • :interest-invoker (without parens) is equivalent to :interest-invoker(total)
• :interest-target(possible | partial | total):
  • possible matches when an element is the target (matches the idref value) of an element with the interestfor attribute.
  • partial matches the target of an interest invoker that has "partial interest", but not "full interest".
  • total matches the target of an interest invoker that has "partial interest" or "full interest".
  • :interest-target (without parens) is equivalent to :interest-target(total)

Open questions:

1. Confirmation of the above shape generally.
2. Functionality of possible: does the idref need to be valid? Should :interest-target() support possible?
3. Naming:
   a. :interest-invoker and :interest-target
   b. possible, partial, and total