[ https://jira.jboss.org/jira/browse/DNA-443?page=com.atlassian.jira.plugin... ] Brian Carothers commented on DNA-443: ------------------------------------- The problem seems to be around line 151 of InMemoryRequestProcessor. This trap is hitting: InMemoryNode parentNode = workspace.getNode(parent); if (parentNode == null) { Path lowestExisting = workspace.getLowestExistingPath(parent); request.setError(new PathNotFoundException(request.under(), lowestExisting, GraphI18n.inMemoryNodeDoesNotExist.text(parent))); return; } The parentNode seems to be null because the sequence of create node requests generated by the sequencer starts off like so: dna:compilationUnit dna:compilationUnit/dna:import/dna:importDeclaration/dna:importOnDemand/dna:typeImportOnDemandDeclaration It seems as though the problem is that the request processor is trying to create dna:typeImportOnDemandDeclaration before dna:compilationUnit/dna:import exists. I see this as a problem with the sequencer. We could make the request processors smart enough to build any missing nodes, but the interpolated nodes wouldn't have useful properties like jcr:primaryType, so I think that approach would just defer the problem. I'll poke around the sequencer and see if I can fix this. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/DNA-443?page=com.atlassian.jira.plugin... ] Randall Hauch commented on DNA-443: ----------------------------------- Actually, the batch 'create(...)' methods on the Graph.Batch class do indeed have the ability to create nodes if they are missing. In fact, both the batch and non-batch works the same way: graph.create(path).ifAbsent()... // or even .orUpdate() or batch.create(path).ifAbsent()... In fact, this style makes available several other methods than "ifAbsent()", such as "orUpdate()", "byAppending()", and "orReplace()". These correspond to the different NodeConflictBehavior options. I'll grant you that the "createIfMissing()" methods are not available in the batch mode. Perhaps we should deprecate them, since the above methods are newer, more complete, and just as useful. I think the only capability that the "createIfMissing()" methods have above the others is that they return an interface that has the "andReturn():Node" method, but I think all the non-batch "create()" methods could do this. I will open an issue to resolve this discrepency in the Graph API, but I'm also going to reopen this issue for 0.6, since sequencers really should be operating in batch (in case there is an error in the middle of processing, and because a large batch with many operations will be significantly faster for most connectors than the same number of operations submitted individually). -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/DNA-443?page=com.atlassian.jira.plugin... ] Randall Hauch reopened DNA-443: ------------------------------- -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/DNA-443?page=com.atlassian.jira.plugin... ] Brian Carothers resolved DNA-443. --------------------------------- Fix Version/s: 0.5 (was: 0.6) Resolution: Done Applied patch that changes all of the node creation to use the destination instead of doing direct writes to the graph. Had to add an accumulator to keep track of the created paths. This should have a beneficial effect on performance over the use of createIfMissing, since it avoids having to continually read in the nodes at the root of the graph. The patch also adds a new test case to verify that the StreamSequencerAdapter is not creating extra nodes. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira