Does uBO de-dupe cosmetic-HTML filters if no-script?

[Rudxain]

If no-scripting: example.com true, and we have an HTML-filter such as:

example.com##^.ad-unit-wrapper

Then the following cosmetic-filters are useless for example.com:

! generic: has to stay in memory, but
! `example.com` can be whitelisted
##.ad-unit-wrapper
! specific: can be deduped
example.com##.ad-unit-wrapper
! very specific: not identical, but can be discarded anyways
example.com##.ad-unit-wrapper > .example-ad

This optimization is valid for all host-names, as long as the HTML-filter's CSS-selector is identical (or "broader") than the cosmetic counterpart.

I understand that static-filters are parsed+de-duped independently of rules (dynamic or switch). But what if the user blocks all scripts using a static-filter?

[krystian3w]

#3181 but can be fixed by recommend CSP.

[gorhill]

I don't understand what you are asking. Cosmetic filters are unrelated to no-scripting switch.

[krystian3w]

It seems that he would suggest checking if JS is inactive and if we have HTML filtering match, then we should not load CSS filters because the page will not dynamically replace the content due to lack of access to JS.

[Rudxain]

Cosmetic filters are unrelated to no-scripting

That's what I meant by "independent". So I mentioned static-filters as an alternative, because I assume it's easier to implement the optimization when the script-blocking is specified by a filter rather than a rule (switch and/or dynamic).

if JS is inactive and if we have HTML filtering match, then we should not load CSS filters

Spot-on!

[stephenhawk8054]

Does uBO de-dupe cosmetic-HTML filters if no-script?

Then no.

[gorhill]

if JS is inactive and if we have HTML filtering match, then we should not load CSS filters

Spot-on!

Please provide profiling data showing that the "gain" -- if measurable at all -- is worth the amount of work to invest in making this happen, the added pile of JS code paths, and the added overhead of having to verify that all the conditions are fulfilled to avoid injecting the CSS filters.

[krystian3w]

CSS can be useful if HTML Filtering fails on the first visit #3222 (I assume that disabling JS by any method does not help to reported error 3222).