Are there any plans to generalize the type of VertexId in GraphX? Our keys are particularly long. We could use the hashCode() trick, but the chance of collisions is not acceptable. Given our data volume, we have encountered hashCode() collisions more than once.
I see this Jira, but it is specific to UUIDs: https://issues.apache.org/jira/browse/SPARK-1153 Looking through the code, it seems doable, but I'm not aware of the consequences. Is it entirely a performance issue? Support for an arbitrary type would be ideal, but arbitrarily long byte arrays are a reasonable compromise, if that helps. ----- -- Madhu https://www.linkedin.com/in/msiddalingaiah -- View this message in context: http://apache-spark-developers-list.1001551.n3.nabble.com/VertexId-type-in-GraphX-tp10104.html Sent from the Apache Spark Developers List mailing list archive at Nabble.com. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org For additional commands, e-mail: dev-h...@spark.apache.org
