Mailing-List: contact dev-help@river.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@river.apache.org
Received-SPF: pass (athena.apache.org: local policy)
From: Peter <jini@zeus.net.au>
Reply-To: Peter <jini@zeus.net.au>
To: dev@river.apache.org
Subject: Soft reference blocking queue
Content-Type: multipart/alternative; boundary="=-rmvwpsmTVvRHBTuuSX+C"
Date: Fri, 23 Sep 2011 19:29:18 +1000
Message-Id: <1316770158.11253.5.camel@Nokia-N900-51-1>
Mime-Version: 1.0

--=-rmvwpsmTVvRHBTuuSX+C
Content-Type: text/plain; charset=utf-8
Content-ID: <1316770157.11253.2.camel@Nokia-N900-51-1>
Content-Transfer-Encoding: 7bit

Just pondering if there's a use case for weak or soft references in a blocking queue.  I was writing a unit test for it today.

One use case might be around cancelled future tasks with an executor, you just null out any strong references, if it hasn't yet executed it probably won't.  Eg future has timed out, null out reference and it is removed from the queue by garbage collection.  This might help to keep a queue small when there are a large number of tasks being cancelled before they execute, rather than occupying memory in the queue between other tasks awaiting execution.

The other case might be an unbounded queue, with a single thread taking, it blocks waiting for data, instead of the queue causing a memory explosion when it can't handle the data given to it , it just starts dropping objects, a bit like a network reciever.  The sender can retry later. 

Cheers,

Peter.

--=-rmvwpsmTVvRHBTuuSX+C--