Hi, yea, we have no simple way to do that in GraphX because the GraphX class has both vertex and edge rdds and we cannot simply implement mapPartitions there to keep vertex/edge semantics inside. Another idea is to generate edge files by using RDD#mapPartitions and write them into HDFS, and then you use GraphLoader#edgeListFile to load them.
// maropu On Thu, Jun 2, 2016 at 11:20 PM, Roman Pastukhov <metaignat...@gmail.com> wrote: > As far as I understand, best way to generate seeded random numbers in > Spark is to use mapPartititons with a seeded Random instance for each > partition. > But graph.pregel in GraphX does not have anything similar to mapPartitions. > > Can something like this be done in GraphX Pregel API? > -- --- Takeshi Yamamuro
