RFR: 8090110: Very bad TextArea of performance in a language environment that uses the Chinese character (KANJI)

[yosbits]

On Thu, 1 Oct 2020 11:23:03 GMT, Jeanette Winzenburg <fastegal at openjdk.org> wrote:

>> I understand that some people want support for the **new features** of Skin Swiching.
>
>> 
>> 
>> I understand that some people want support for the **new features** of Skin Swiching.
> 
> Assuming that you are joking ;) Just in case you are not and for the record:
> 
>     skinnable.setSkin(skin)
>     skin.dispose()
> 
> are public api without any notion of being optional - making the implementation of TextAreaSkin misbehaving big style.
> 
> That said: verified that the fix improves performance of the example immensely. So I think it should be integrated
> asap, ignoring the big heap of mess already there.
> Nevertheless, it should
> 
> - be implemented in a way to not add to the pile of dirt: instead of adding the changeListeners manually, add them via
>   skin api (which will remove them automatically in dispose). That's possible if they are added once in the life-time of
>   the skin, which seems to be the case, or not?
> - add tests:
>       - unit tests to ensure that changing properties that effect the clip actually are adjusting it
>       - system? test to verify the performance gain

I think testing of the internal structure of Control should be minimal; I think testing for context, such as CSS
classes, is necessary. In particular, I don't think we are at a stage where we are dealing with the internal structure
of TextArea in a fixed way, so we will need to determine the Next Step.

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

PR: https://git.openjdk.java.net/jfx/pull/307