From 2282d27cf1da04bfdad5173c412e77a51d7c36e6 Mon Sep 17 00:00:00 2001
From: Ankur Dave <ankurdave@gmail.com>
Date: Wed, 16 Oct 2013 23:56:15 -0700
Subject: [PATCH] Cache msgsByPartition

---
 .../src/main/scala/org/apache/spark/graph/impl/GraphImpl.scala | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/graph/src/main/scala/org/apache/spark/graph/impl/GraphImpl.scala b/graph/src/main/scala/org/apache/spark/graph/impl/GraphImpl.scala
index 5398631d26..a3cdb97cea 100644
--- a/graph/src/main/scala/org/apache/spark/graph/impl/GraphImpl.scala
+++ b/graph/src/main/scala/org/apache/spark/graph/impl/GraphImpl.scala
@@ -486,7 +486,8 @@ object GraphImpl {
         .flatMap { case (vid, (vdata, pids)) =>
           pids.iterator.map { pid => MessageToPartition(pid, (vid, vdata)) }
         }
-        .partitionBy(eTable.partitioner.get) //@todo assert edge table has partitioner
+        .partitionBy(eTable.partitioner.get).cache()
+    // @todo assert edge table has partitioner
 
     val vTableReplicationMap: IndexedRDD[Pid, VertexIdToIndexMap] =
       msgsByPartition.mapPartitionsWithIndex( (pid, iter) => {
-- 
GitLab