Is there any performance difference between #?#:-abp-has() and #?#:has() on browsers that support the has selector?

[partingscientist]

With the release of Firefox 121 in less that two weeks (December 19th) along with the has() selector, is there any performance difference between #?#:-abp-has() and #?#:has() in Firefox for uBO? I have previously read that uBO discards the ? if I'm not mistaken, which I would assume mean that #?#:has() is handled just like native has selector if available. Is this correct? Is #?#:-abp-has() also converted to native has if the browser supports it? Thank you.

[krystian3w]

It should be same like on Chrome 105+ => can be used as CSS 4.

It looks CSSTree implementation made simple :has() filtering non-procedural (by uBo implementation).

You can check on Beta/Nightly in how method was forced (JS with random attributes or CSS 4).

---

As for the hiding speed, the blinking is very minimal (the only limitation is CSS and browser logic now).

[garry-ut99]

In theory a native :has is supposed to be faster than a procedural :has, in practice I didn't test speed so far, so I dunno.

Whether some of uBO procedural filters are handled as native selectors is another story, some info is in this thread : uBlockOrigin/uBlock-issues#2475, I know that #?#:-abp-has() is an alias of ##:has() which is converted to native on browsers with native support, and adding ? ==> #?#:has() aborts the conversion, and forces to remain as a procedural, I'm not sure whether anything changed since then.

[krystian3w]

Now #?# can't allow back to JS implementation of hiding.

[partingscientist]

#?#:has() aborts the conversion, and forces to remain as a procedural

@garry-ut99 So is it true that all #?#:has() filters in Adguard list do not use the native selector when used in uBO? That does not sound correct.

[krystian3w]

No longer aborts.

Now needed is Invalid CSS4 selector or procedular modifier (e.g. has-text).

Example something Looks as CSS4 but is not:

##:not(.item *):not(:has(.item *))

! uBo parse same to:
! ##.item:others()

[gorhill]

#?#:-abp-has() is equivalent to ##:has(). #?#:has() will force the filter to be procedural, this is to ensure there is a fallback syntax should the result be unexpected when using the native form. Never mind, #?#:has() is also compiled to native if supported by the browser.

[partingscientist]

I originally asked this because I thought migrating from #?#:-abp-has() to #?#:has() will use the native selector instead of procedural in uBO (considering ABP supports #?#:has() as an alias to #?#:-abp-has()).

Now I'm curious, does #?#:has() filters in built-in Adguard filters are treated as being procedural instead of native?

[krystian3w]

They also start mostly as CSS 4, check with DevTools that most of working ads/annoy­ances them do not have 2 attributes added with a random name.

In general, it looks like Microsoft implemented a shadowban on repositories or comments sections.

[gorhill]

does #?#:has() filters in built-in Adguard filters are treated as being procedural

Yes, this forces procedural in uBO. You would have to ask them why they use #?# instead of ##. I am wrong, see below.

[krystian3w]

So why this start in CSS4:

example.*#?#div:has(a)