[
https://issues.apache.org/jira/browse/TINKERPOP-2187?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16834470#comment-16834470
]
Florian Hockmann commented on TINKERPOP-2187:
---------------------------------------------
{quote}The message scope should be {{Global}} only, SPVP never sends local
messages.
{quote}
Good to know, it was the first time I worked on a graph computer and it looked
to me like {{Local}} scope was also needed, but then I'm probably wrong on that
one.
In other words, a {{GraphComputer}} should work in any case, whether the
message combiner is provided or not.
That's definitely good to know as it probably means that the
FulgoraGraphComputer has a different problem. While debugging a failing test, I
ran into a situation where a message was received that couldn't be processed as
there was already a different message that needed to be processed first and it
looked to me like FulgoraGraphComputer expected a message combiner in that case
to combine the two messages. (It has been a few weeks since I've worked on that
so my description could be a bit off.) If the message combiner is optional,
then FulgoraGraphComputer needs to be able to handle such a situation.
> Two issues with the ShortestPathVertexProgram
> ---------------------------------------------
>
> Key: TINKERPOP-2187
> URL: https://issues.apache.org/jira/browse/TINKERPOP-2187
> Project: TinkerPop
> Issue Type: Bug
> Components: process
> Affects Versions: 3.4.1
> Reporter: Florian Hockmann
> Assignee: Daniel Kuppitz
> Priority: Minor
>
> While trying to update JanusGraph to TinkerPop 3.4.0, I encountered two
> issues with the {{ShortestPathVertexProgramm}}:
> # {{ShortestPathVertexProgram.getMessageScopes()}} returns an empty
> collection which doesn't work for a graph computer that uses these scopes to
> execute the edge traversal. I think the scopes should be {{Global}} and
> {{Local}} with the edge traversal.
> # The {{ShortestPathVertexProgram}} defines no message combiner, but
> {{sendMessages()}} can apparently still send multiple messages for the same
> vertex.
> For some reason, this doesn't lead to any problems for the
> {{TinkerGraphComputer}}. (Maybe it receives each message before another one
> can be sent for the same vertex?)
> For {{FulgoraGraphComputer}} however some tests result in a situation where
> multiple messages are sent to the same vertex which fails because no combiner
> is defined.
> So, I'd say that the {{ShortestPathVertexProgram}} should have a message
> combiner.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
