Create AdGuard's :matches-attr

[piquark6046]

Prerequisites

• I verified that this is not a filter list issue. Report any issues with filter lists or broken website functionality in the uAssets issue tracker.
• This is not a support issue or a question. For any support, questions or help, visit /r/uBlockOrigin.
• I performed a cursory search of the issue tracker to avoid opening a duplicate issue.
• The issue is not present after disabling uBO in the browser.
• I checked the documentation to understand that the issue I am reporting is not normal behavior.

I tried to reproduce the issue when...

• uBO is the only extension.
• uBO uses default lists and settings.
• using a new, unmodified browser profile.

Description

Unlike AdGuard's :matches-attr, uBO's :watch-attr can only observe any changes of specific element attributes.
A filter developer cannot use a regexp in uBO's :watch-attr in order to match a randomized attribute.

Therefore, I request expansion of uBO's :watch-attr.

Related: AdguardTeam/AdguardFilters#132468

A specific URL where the issue occurs.

https://www.clien.net/service/board/news

Steps to Reproduce

1. Visit the provided URL.
2. Apply the following filter:

   clien.net#?#div[class*="_"]:watch-attr("class"="/[a-z]+_[a-z]+_[a-z]+/")

3. Reload the website.

Expected behavior

Only one element is blocked correctly as AdGuard Browser Extension performs.

Actual behavior

All element having class attribute are blocked.

uBO version

1.44.4

Browser name and version

Mozilla Firefox Developer Edition 106.0b9

Operating System and version

Ubuntu 22.04.1 LTS

[krystian3w]

So AdGuard forgot fix use filter without fallback: AdguardTeam/ExtendedCss#122 (broken unhide node action)

His implementation works fine in add-on 3.3.2 so don't added watch-attr, I don't checked in any version of apps for Windows/macOS/Android (maybe these need "pseudoclass").

---

In short I mean this is two different mechanisms.

---

To avoid break pages in outdated uBo better copy AdGuard name if worth for only few Korean pages (WP.pl in uBo no need implementation matches-attr/property - but hide ads is moved to uAssets team instead waiting to MajkiIT fixes on almost every day).

[gorhill]

What you want is a new operator which matches text on attribute name and value. It's best to add this new operator :matches-attr() rather than trying to extend :watch-attr() which purpose is completely different than :matches-attr().

[piquark6046]

@gorhill Great. Thanks a lot.

[gwarser]

In the meantime, https://stackoverflow.com/questions/8308328/xpath-to-find-attributes-where-the-name-starts-with-a-given-value

---

Ah, and you don't need to specify attribute in :watch-attr(), this was a bug apparently: gorhill/uBlock@8aa379e

[gorhill]

I can find only two instances of matches-attr in all of AdGuard lists, and even the issue you link to is an instance of a matches-attr filter being removed. Just trying to assess priority of implementing this, which from what I see should be quite low.

[peace2000]

It seems this got implemented. Great!

Though as I'm testing it, it seems that if I put quotes around the attribute, it doesn't match. If I remove them, it matches.

Example:

https://www.k-rauta.fi/tuote/valmistasoite-hole-in-1-250ml/6408070100905

Putting this rule to the picker

##button:matches-attr(class="/[\w]{7}/")

Shows matches, but when I add the quotes around the attribute, matches are gone:

##button:matches-attr("class"="/[\w]{7}/")

I'm trying to match this element:

Did I understand something wrong or am I just being stupid now?

What I understood is that having quotes is better so that it's AG-compatible rule as well.

[gorhill]

##button:matches-attr("class"="/[\w]{7}/")

I would expect this to work the same as ##button:matches-attr(class="/[\w]{7}/").

I will investigate.

[peace2000]

Thanks gorhill, the fix works.

Though I noticed another strange thing: uBO seems to accept a rule if it has a single square bracket before the last round bracket:

k-rauta.fi##button:matches-attr("class"="product-info-anchor"])

Shouldn't that be invalid?

It hides this element:

[gorhill]

Shouldn't that be invalid?

Ok, it will be considered invalid in next release -- currently the [ is discarded.

[gorhill]

Is it all good now? Can it be closed?

[piquark6046]

Yup. It can be closed.