[concurrency-interest] jsr166y.forkjoin API comments

Mark Thornton mthornton at optrak.co.uk
Fri Jan 25 17:30:09 EST 2008


Neal Gafter wrote:
> On Jan 25, 2008 12:58 PM, Mark Thornton <mthornton at optrak.co.uk 
> <mailto:mthornton at optrak.co.uk>> wrote:
>
>     During the first public review of JSR 14 (Generics) I proposed an
>     extension that would have supported primitives. ... I think the
>     mechanism
>     I proposed then would still be technically feasible today (absent
>     curious interactions with auto-boxing)...
>
>
> I've never heard a proposal that I thought was technically feasible 
> from a language perspective, other than the obvious 
> performance-problematic scheme of boxing on the way in and unboxing on 
> the way out.  Do you have a writeup of your proposed subtyping rules?
I'll see what I can dredge up from the archives when I'm at work. It 
never progressed very far as the primary reason for rejection wasn't 
technical. Essentially though a type argument was either a regular one 
(as now) or constrained to be a numeric primitive. A type argument 
couldn't be replaced by both object types and primitives. Then instead 
of having Collections.sort methods for Object[] and every numeric 
primitive, just two methods are required, one for Objects (as now) and 
the other for all the numeric primitives.

Mark Thornton


More information about the Concurrency-interest mailing list