#2471: Enable jQuery selectors everywhere

[fregante]

• Closes Partially support non-native CSS selectors for trigger "watch" mode #2471

Demo with the selector: h2:lt(1000)

Screen.Recording.1.mov

[fregante]

looks like there's infinite loop issue

Found it:

https://github.com/jquery/sizzle/blob/20390f05731af380833b5aa805db97de0b91268a/src/sizzle.js#L344

In some occasions Sizzle will add an id to limit the scope of the selector. I triggered it with h2:not([id]) ~ h2

This is the suggested workaround and the browser no longer hangs.

[bezborodow]

It looks like you found the issue where the browser was hanging using jQuery selectors? It was a bug with Sizzle? Is this something that has been fixed upstream? If so, do you think it might be safe to include this in jQuery.initialize?

[bezborodow]

Possibly related jquery/jquery/issues/4332?

[twschiller]

Thanks! - good find on the selectors.

Reading the PR, I realized that we had already implemented polling for initialize. So we could have just removed the isNativeSelector checks from triggerExtension b/c they should have worked?

This new PR has different performance characteristics than the old method, as it now has to watch the document for all mutations. I need to dig in a bit more here to see what the tradeoffs are

[fregante]

Unless I made a mistake, the method/scope of the watcher hasn't changed. The only thing that changed was jQuery selectors are no longer polled but observed.

It would not have just worked because initialize uses querySelectorAll. Maybe I'm misunderstanding what you're saying. It’s also possible that you’re misreading the original code like I did in my now-hidden comment 😅

[twschiller]

The only thing that changed was jQuery selectors are no longer polled but observed.

The way that the vendored initialize method works is that it observes the root (document) for mutations, which it then runs the selectors over. The question is what the performance impact will be if we're running JQuery instead of native selectors. (Although, my understanding is JQuery will use native selector evaluation under the hood for native selectors"?)

For 1.5.2 (which needs to go out ASAP), I think I will just remove the assertion on "watch" mode, because our "initialize" method does actually support polling via await element. That will then give us more time to merge in this PR for 1.5.3

It’s also possible that you’re misreading the original code like I did in my now-hidden comment 😅

The Global.ts file's "initialize" refers to the vendored $.initialize method which does not support JQuery. I think it's right that we could use _initialize, but would have to be careful to re-attach it because it only fires once

I might also still be misreading the code though 🤷

[twschiller]

I'm a little confused how the code loop check works as the mutation callback is synchronous. Not fully following how another mutation event could fire while one is being processed. But have tested it and it seems to work OK

[bezborodow]

Interesting.

Possibly related jquery/jquery/issues/4332? See my comments above.