Bug in LongDoubleFloatDoubleVertex.sendMsgToAllEdges()
------------------------------------------------------
Key: GIRAPH-125
URL: https://issues.apache.org/jira/browse/GIRAPH-125
Project: Giraph
Issue Type: Bug
Components: graph
Affects Versions: 0.1.0
Reporter: Yuanyuan Tian
I just found a bug in the sendMsgToAllEdges() function of the
LongDoubleFloatDoubleVertex class. The segment of the code that contains the
bug is:
final LongWritable destVertex = new LongWritable();
final MutableVertex<LongWritable, DoubleWritable, FloatWritable,
DoubleWritable> vertex = this;
verticesWithEdgeValues.forEachKey(new LongProcedure() {
@Override
public boolean apply(long destVertexId) {
destVertex.set(destVertexId);
vertex.sendMsg(destVertex, msg);
return true;
}
});
Here destVertex is a final object, but this single object is reused in the
forEachKey function many times. Each time its actual value is changed but the
same object is put to the underlying message list (a hashmap) through
vertex.sendMsg. Because the single destVertex object has been put into the
underlying hashmap again and again, destVertex.set(destVertexId) will change
the existing keys in the hashmap. Eventually, every keys added to the hash map
will have the same value as the last key.
A simple fix is as follows:
final MutableVertex<LongWritable, DoubleWritable, FloatWritable,
DoubleWritable> vertex = this;
verticesWithEdgeValues.forEachKey(new LongProcedure() {
@Override
public boolean apply(long destVertexId) {
vertex.sendMsg(new LongWritable(destVertexId), msg);
return true;
}
});
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
