RFR 8169808 Stream returning methods should specify if they are late binding

[Doug Lea]

On 11/23/2016 10:05 AM, Martin Buchholz wrote:
> Right now, PBQ's spliterator is in violation of
>
> """A Spliterator that does not report IMMUTABLE or CONCURRENT is
> expected to have a documented policy concerning: when the spliterator
> binds to the element source;"""
>
> so I think we only get a choice between IMMUTABLE and CONCURRENT.  I
> propose:

Well, OK. Hopefully no user program will hardwire which
choice is made, in case the implementation changes.

-Doug


>
> --- src/main/java/util/concurrent/PriorityBlockingQueue.java23 Nov 2016
> 14:51:10 -00001.116
> +++ src/main/java/util/concurrent/PriorityBlockingQueue.java23 Nov 2016
> 15:01:29 -0000
> @@ -962,3 +962,6 @@
>          public int characteristics() {
> -            return Spliterator.NONNULL | Spliterator.SIZED |
> Spliterator.SUBSIZED;
> +            return (Spliterator.IMMUTABLE |
> +                    Spliterator.NONNULL |
> +                    Spliterator.SIZED |
> +                    Spliterator.SUBSIZED);
>          }
>
>
>
> On Wed, Nov 23, 2016 at 6:14 AM, Doug Lea <dl at cs.oswego.edu
> <mailto:dl at cs.oswego.edu>> wrote:
>
>     On 11/22/2016 08:33 PM, Martin Buchholz wrote:
>
>         PriorityBlockingQueue has a late-binding-style, snapshot spliterator
>         that does not report IMMUTABLE.  It *is* immutable ... after the
>         first
>         access!  Should we add IMMUTABLE?
>
>
>     Probably not.
>     This might overly constrain future improvements; for example, someday
>     replacing with some linked structure.
>     Arguably, if so, we could just change the reported characteristics,
>     but that could be surprising to users, and possibly lead to subtle
>     user bugs.
>
>     -Doug
>
>