Fix #11994: TabView prevent detached DOM elements on TabClose

[melloware]

Fix #11994: TabView prevent detached DOM elements on TabClose

[tandraschko]

why did you refactor like this?

#fireTabCloseEvent should just fire the event, everything else was done before in #remove
doing the cleanseDomElement in #remove just would work fine?

[melloware]

So i will double check but since cleanse removes all elements, events, data off all elements when using <p:ajax event="tabClose" it could affect the lifecycle anything that was on that tab expecting the normal widget lifecycle so I chose to do if using tabClose to wait until the AJAX is done to clean up the DOM else just clean it up immediately if not using tabClose.

It felt like the cleaner / safer way to do it but I could be convinced otherwise?

[tandraschko]

both elements were already destroyed before sending the event in the past and also feels better for me
as there is no delay waiting for the ajax response, its just a notifier and we never had any feature request for something here

[melloware]

Ok I will test and update.

[melloware]

@tandraschko did some testing in my clients real world app.

Test with just changing to cleanseElement instead of panel.remove()

Here is what it looks like with this PR where the cleans runs BEFORE the AJAX runs. Memory recovered back from 16.5MB to 13.3MB with lots of detached elements still in memory (which were in OverlayPanels) in the tab.

Current PR

Here is what it looks like with this PR where the cleans runs AFTER the AJAX runs and all the widget.destroy's properly take place. Memory recovered back to from 16.5MB to 9.5MB and barely any detached elements in memory.

I think I know what is going on but still testing.

[melloware]

@tandraschko ok PR updated I got to the bottom of it. In cleansDomElement we must call jq.remove() before removing all event listeners with jq.off() This is so the widget on("remove) events get fire to register widgets as detached widgets so they clean themselves up.

I have verified with my testing above the memory is now recovered properly in my test scenario.