RFR: 8345188: Support tree-structural pseudo-classes [v6]

[Andy Goryachev]

On Wed, 8 Jan 2025 22:08:38 GMT, John Hendrikx <jhendrikx at openjdk.org> wrote:

>> I think this concern is out of proportion. Generally, nodes will not have such a huge number of children that simply flipping a few pseudo-classes will be prohibitively expensive. CSS recalculation will, in most cases, only happen once per pulse anyway.
>> 
>> I don't know why we would be optimizing for pathological applications that register two pseudo-class listeners for each of their tens of thousands of children in a container, _and_ do heavy work in those listeners.
>
> Perhaps it is not as dire, but it may be worth a short check how this works with say 1000 nodes with and without this new code (no need to register listeners, I agree that that is a different problem, I was more worried about how CSS handles this many changes).  Perhaps also a case where there is also some style change going on linked to the odd or even pseudo class (although if that is much slower, then I guess that's what the user wanted...)

A thousand of nodes scenario is possible and I think it needs to be tested, just to estimate the impact.  For example, a rich text type of control: TextFlow with a thousand small Text instances all styled differently (or tied to odd/even pseudo class).

Although I am struggling to see the use case for such an arrangement.

-------------

PR Review Comment: https://git.openjdk.org/jfx/pull/1652#discussion_r1907928338