Ran into another problem this time -
I'm trying to populate a graph with about 3300 links. I've a CSV file
as input which has approx 3300 records - of the form (A, B, n) -
specifying a link from A to B, with a weight n. I am reading each line
of the file, and creating each node (A and B). If the nodes are new, a
new link is created. If both A and B exist, the link between them is
updated with the new weight. Now here is the problem. Say I have a
dataset as follows -
A, B, 3
C, D, 5
D, E, 2
E, A, 7
My algorithm is something like this -
check whether node1 already exists or not, create if it doesn't exist;
check whether node2 already exists or not, create if it doesn't exist;
if (either node1 or node2 or both are new) {
create a new relationship between them;
}
if (both node1 and node2 exist) {
if (! node1.hasRelationship(RelTypes.KNOWS, Direction.OUTGOING) {
// there is no link between node1 and node2 because node1 has
no outgoing links
create a new relationship;
} else {
// node1 has outgoing links, but none of them may link with node2
check whether there is any link from node1 to node2;
update if found, else create new;
}
}
The trouble seems to occur because of the "hasRelationship" statement.
The data I have has many cases where node1 and node2 have been created
before separately - which means, in many cases, this "hasRelationship"
check is done and a new link is created. It's working ok with small
dataset, but for a dataset with many such cases (like the last line in
the sample data: E, A, 7) I am getting an OutOfMemory error with "GC
overhead limit exceeded".
This is the code -
public CallInfo createOrUpdateCallInfo(final Subscriber callingParty,
final Subscriber calledParty,
final long count,
final long duration) {
CallInfo callInfo = null;
if (callingParty == null) {
throw new IllegalArgumentException("Null CallingParty");
}
if (calledParty == null) {
throw new IllegalArgumentException("Null CalledParty");
}
final Node callingPartyNode = ((SubscriberImpl)
callingParty).getUnderlyingNode();
final Node calledPartyNode = ((SubscriberImpl)
calledParty).getUnderlyingNode();
if (!callingPartyNode.hasRelationship(RelTypes.CALLS,
Direction.OUTGOING)) {
//no outgoing relationships - bound to be a new one
final Relationship rel =
callingPartyNode.createRelationshipTo(calledPartyNode,
RelTypes.CALLS);
callInfo = new CallInfoImpl(rel);
callInfo.setCount(count);
callInfo.setDuration(duration);
} else {
// could be an update or a create
// get all the outgoing relationships and check if there
is one which
// ends with the calledParty node
final Iterable<Relationship> currentRels =
callingPartyNode.getRelationships(RelTypes.CALLS, Direction.OUTGOING);
for (Relationship rel : currentRels) {
final boolean found = rel.getEndNode().equals(calledPartyNode);
if (found) {
// update the existing relationship
callInfo = new CallInfoImpl(rel);
callInfo.updateCount(count);
callInfo.updateDuration(duration);
break;
} else {
final Relationship newRel =
callingPartyNode.createRelationshipTo(calledPartyNode,
RelTypes.CALLS);
callInfo = new CallInfoImpl(newRel);
callInfo.setCount(count);
callInfo.setDuration(duration);
}
}
}
return callInfo;
}
I've tried to allocate 3GB to this application, but that didn't help.
It crawls for 10-15 mins, then throws the OutOfMemory error...
Can anyone give me some pointers please?
Regards
Arijit
--
"And when the night is cloudy,
There is still a light that shines on me,
Shine on until tomorrow, let it be."
_______________________________________________
Neo4j mailing list
[email protected]
https://lists.neo4j.org/mailman/listinfo/user
