Add polyfill for `:has()` CSS4 pseudo-selector to element hiding filters

[greiner]

Background

(taken from original description in #2360 about hiding elements by selecting their child elements)

Currently, there's no way to hide an element on a page dependent on its children. This proposed solution is based on existing CSS functionality, so it will be possible to bind an observer to the document to check for new parent nodes and bind another observer object to them to check for changes within only these particular nodes in case specific child node will be added. Over time it will be possible to completely disable such functionality due to native support for CSS4 selectors in newer browser versions.

It may be possible to look at the jQuery code for this functionality since they supported :has() pseudo-selector in their selectors long before it was even added into the CSS4 drafts.

Generally, this proposal has similar issues as CSS property filters such as requiring to hide elements manually and to observe the page for changes at runtime.

:has() and CSS property filters have to work in conjunction, therefore they need to be stored in the same filter class (see #4394). In addition, we will probably have to change the way CSS property filters work: Currently we eventually generate standard CSS rules for them, which are then being applied by the browser. For :has(), we have to apply the rules manually.

What to change

• A new special pseudo-classe :-abp-has(...) should be added. Here, the parameter is a selector - this pseudo-class should only match elements where the inner selector produces any hits.
• It should be possible to nest pseudo-classes, in particular :-abp-has(:-abp-properties(...)) has to be supported.
• ElemHideEmulation class should get a new removeElements callback, to be called for any elements selected by filters with :-abp-has() pseudo-class. Filters only using :-abp-properties() should use addSelectors callback as before.

Optional changes

• Allow :-abp-has() pseudo-classes to be nested.
• Observe DOM changes to apply filters to dynamically added/modified elements

References

• ​The draft CSS spec for :has
• ​uBlock's documentation of their :has filters

[fhd]

I'll have a go at this one.

[Lain_13]

I think it's worth mentioning that :has() selector must work in combination with -abp-properties. So, filter like site.name##.block:has([-abp-properties="background: yellow"]) must hide all elements of class "block" with anything with word "yellow" in a "background" property inside of them.

[mapx]

"waiting for #2388 to be closed probably makes the most sense to get the necessary architecture in place first"

#2388 ==> still opened #2398 ==> still opened #2402 (unit tests)

(fhd comment: Given how CSS property filters support has already landed in Firefox - how about closing this one (and #2388) and removing #2402 as a dependency? )

[kzar]

Yea I think you're right, I don't think this is blocked by #2388 any more. I'll remove that as blocking now.

[mapx]

Replying to Lain_13:

I think it's worth mentioning that :has() selector must work in combination with -abp-properties. So, filter like site.name##.block:has([-abp-properties="background: yellow"]) must hide all elements of class "block" with anything with word "yellow" in a "background" property inside of them.

it's exactly the last trick FB used to defeat the various userscripts (using css ::after)

[kzar]

(I have had nearly no chance to look at this so far and am on vacation next week. Will reassign myself again when I have time if no one else has picked it up.)

[ethaniel]

Can't wait for this ticket to be resolved. Yandex is using every possibility to evade the block. Still seeing ads.

[fhd]

With Dave gone for a while, I'll pick this up again, this time for real :)

I guess I still won't find enough time to finish it, but there's some ground work to be done I can get out of the way at the least, I'll create a blocking issue for that soon.

[fhd]

I've created #4394 for the ground work I was talking about. I'll take care of that first.

I realised that in order for :has() and CSS property filters to work in conjunction (which I also see as a must), we'll have to change the way CSS property filters are being applied somewhat fundamentally. I've updated the description a bit to reflect that.

[trev]

I updated the description, the syntax modification is now #5287.

[Lain_13]

After recent changes in the description I'd like to clarify one thing:
Will ABP support :has() as originally were requested or only :-abp-has()? uBO already supports :has() and it would be nice to have compatibility between ABP and uBO from the get-go.

[hfiguiere]

Right now it is settled on :-abp-has() and the new filter syntax specified in #5287. :has() is not be supported.

[abpbot]

A commit referencing this issue has landed:
​Issue 3143 - Filter elements with :-abp-has()

[abpbot]

A commit referencing this issue has landed:
​Issue 3143 - Remove internal tests