RFR: 8345188: Support tree-structural pseudo-classes [v8]

[John Hendrikx]

On Wed, 8 Jan 2025 20:23:15 GMT, Andy Goryachev <angorya at openjdk.org> wrote:

> Anecdata: I did not see a noticeable performance difference on startup between this PR and the master branch with 48 children with the monkey tester

I wouldn't think you'd see it in those cases as that would primarily be hidden by the cost of adding all the children.  The cases I worry about are containers that have say 1000+ children, and where just a few children (or just one) may be getting added/removed near the start of the container.  Such containers can be some of the View classes that even though virtualized may be displaying 1000+ cells, or some other custom class (a Go/Minesweeper board made out of buttons or something). 

I think that normally a modification of the children list would primarily involve an array copy (but perhaps also peer updates).  Now it will need to change/reset pseudo class state for odd/even states. Now, I'm assuming that if those states are unused in CSS that it will probably be unmeasurable, but a check here might be good, unless we just want to wait for someone to file a performance regression in the future...

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

PR Comment: https://git.openjdk.org/jfx/pull/1652#issuecomment-2578793401