GitLab

Background

Now that I've become an employee of Eyeo GmbH as of today, that should hopefully give me a bit more clout to influence the codebase to make it more suited for all the listmakers out there.

Essentially, I've noticed that many listmakers out there, somewhere north of 50%, possibly as many as 80%, are completely unaware of the existence of the :-abp-has and :-abp-contains syntaxes. Instead they only know about :has and :has-text, thinking that such pseudo-selectors are non-existent in ABP, and they also often lock them behind !#include file-loadings meant for uBlock Origin).

This makes their lists often work considerably worse in ABP than in other major adblockers. Example of general unawareness among listmakers: The ":has" entries in the extra-file of Frellwit's Swedish Filters

In other cases, they may be unaware of such pseudoselectors at all, which means their lists are still pretty much only using 2015-16-era ABP syntaxes. Example: Slovenian List

What to change

Create a syntax synonym(?) where :has entries are treated as :-abp-has, and :has-text entries are treated as :-abp-contains.

For 100% implementation of this, it may also become necessary to have ABP begin to treat all entries as if they had question marks in the hashtags (e.g. #?#.), since question-mark use is also pretty rare among listmakers who are more familiar with uBO. I plan to make a separate issue report for that, if this particular issue report became a success. In fact I have a lot of change suggestions I could've made in the next couple weeks or so.

It is decided to keep the Element Hiding Emulation separator #?# as a requirement.

• uBO silently converts #?# into ##, so filters' authors can write #?#:has(...) once, and see this working as expected in multiple extensions. AdGuard supports it.
• ##:has(...) will be CSS4 native spec and is not widely available yet. When it is it can be used, still with the condition that the browser supports it without change he.
• we keep separated our own emulation vs native CSS as it may have subtle behaviour difference that could break when the is a native CSS4 implementation. Also there is no :has-text() in any current proposal, so this ensure there is no long term conflict.

Ref: https://github.com/uBlockOrigin/uBlock-issues/issues/1011#issuecomment-884839072

Hints for testers

• Any filter #?# with :-abp-has() or :-abp-has-content() can be duplicated and changed to :has() and :has-text(). :-abp-has() and :-abp-has-content() remain unchanged
• ##:has() and ##:has-text() don't work (behaviour unchanged)
• testpages could be updated, albeit it's not backward compatible

Integration notes

• If any validation was performed, this new allowance in the syntax might need to be taken care of.
• Documentation should be updated.