package com.bigdata.service.ndx.pipeline;
/**
* Extends {@link KVOList} to provide handshaking with a {@link KVOLatch}.
* <p>
* Note: {@link IDuplicateRemover}s MUST create a list from the identified
* duplicates so that the {@link KVOLatch} of each duplicate as well as the
* original are decremented after a successful write. Failure to do this will
* cause a {@link KVOLatch#dec()} to be "lost" and generally results in
* non-termination of {@link KVOLatch#await()} since the count for the "lost"
* latch would never reach zero.
*
* @author <a href="mailto:thompsonbry@users.sourceforge.net">Bryan Thompson</a>
* @version $Id$
*
* @param <O>
* The generic type of the unserialized value object.
*/
public class KVOC<O> extends KVOList<O> {
/** The latch. */
private final KVOLatch latch;
/**
*
* @param key
* The unsigned byte[] key (required).
* @param val
* The byte[] value (optional).
* @param obj
* The paired application object (optional).
* @param latch
* The object that maintains the counter.
*/
public KVOC(final byte[] key, final byte[] val, final O obj,
final KVOLatch latch) {
super(key, val, obj);
if (latch == null)
throw new IllegalArgumentException();
latch.inc();
this.latch = latch;
}
/**
* Extended to decrement the {@link KVOLatch}. This inherits from
* {@link KVOList}, which maps {@link KVOList#done()} over the list of
* duplicates. This ensures that all latches are decremented once the
* original value has been successfully written onto an index partition.
*/
@Override
/*
* @todo This is now synchronized on the same monitor as the super class
* (KVOList) which wants mutex semantics for add(o) and done(). I am not
* sure if this is necessary. It is being done in an attempt to track
* down a client stall related to a hang waiting on the inner Lock for
* the tidsLatch in AsynchronousStatementBufferFactory.
*/
synchronized//
public void done() {
super.done();
latch.dec();
}
}