Differing triplet and vertex data

[lbollar <lb...@gmail.com>]

Hello all,

Found this previous post but didn't see answer that appears related.

http://apache-spark-user-list.1001560.n3.nabble.com/template/NamlServlet.jtp?macro=search_page&node=1&query=triplet+vertex


I have a graph algorithm in spark 1.6.2 where I am implementing Louvain
Modularity. The implementation has the same premise that
PageRank.runWithOptions does, in that it has a graph as a variable,
iterates, and overwrites this var from time to time and I have followed some
of their techniques there for materializing the graph and unpersisting. 

One peculiar thing though is that after having iterated and constructing the
new graph, if I traverse the triplets and print each one, my vertices match
the original unaltered graph the algo started with. If I map over the
vertices only, they have been changed to their new updated values. Logging
inserted into the method shows the same behavior. 

In order to actually have the new triplets materialize, I have to overwrite
the graph with a call to a new constructor using graph =
Graph(graph.vertices, graph.edges) which seems a bit hacky. 

Is this normal or expected behavior where triplet vertex data would differ
from vertex data in some circumstances?

Thanks,



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Differing-triplet-and-vertex-data-tp28330.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe e-mail: user-unsubscribe@spark.apache.org

Re: Differing triplet and vertex data

[lbollar <lb...@gmail.com>]

Sorry, wrong link above.

http://apache-spark-user-list.1001560.n3.nabble.com/Differing-triplet-and-vertex-data-td28330.html



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Differing-triplet-and-vertex-data-tp28330p28331.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe e-mail: user-unsubscribe@spark.apache.org