com.apple.concurrent.Dispatch in the modular JDK

[Andrew Thompson]

A few years ago I had a brief email exchange with Doug Lea I the idea of providing an SPI so platforms could plug in implementations for a "default" ExecutorService. The idea was this could be a regular fixed thread pool or something more exotic like libdispatch, which is, after all OS managed and supposedly plays nicely with power management etc. 

I didn't take the idea any further, but something similar cropped up when lamdas were added in Java 8, the idea of a default mechanism for parallel lambda execution. 

I still think there's a good idea in there. The notion of a default thread pool that could be backed by more native code on various platforms. But that whole infrastructure would need to be created. It's more like a JEP. If that were to exist then the Dispatch class would be a natural way to implement on OS X. 

Andrew

> On May 19, 2015, at 8:01 PM, Phil Race <Philip.race at oracle.com> wrote:
> 
> Apple's JDK 6 provided the non-standard "com.apple.concurrent.Dispatch" class
> which was a means for Java code to access Apple's "libdispatch".
> To quote the documentation at
> http://www.coderanch.com/how-to/javadoc/appledoc/api/com/apple/concurrent/Dispatch.html
> 
> " Factory for |Executor|s and |ExecutorService|s backed by libdispatch.
> Access is controlled through the Dispatch.getInstance() method,
> because performed tasks occur on threads owned by libdispatch"
> 
> The class carried over into Oracle JDK for OS X and is there all the way up
> to current JDk 9 dev builds.
> 
> The question is what to do with this class in the 'modular JDK' where
> we need to decide on the 'supportedness' of many APIs and what
> to do with them as a result.
> 
> There is one Eclipse bug from 2010 where one of the methods was
> added to support embedding SWT in AWT applets in the browser :
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=303869
> 
> but that was for JDK 6 and SWT can't be embedded in the current JDK 7/8/9
> Java Plugin for OS X which runs out of process .. so that usage is moot.
> 
> So the current thinking is that since we do not know of anyone who
> uses this class that it should be unsupported in JDK 9 and not
> visible in the modular JDK image.
> 
> So .. any comments ?
> 
> -phil.
> 
> 
> 
> 
>