spark-graphx pagerank

在工作中，在图进行初始化的时候，需要根据边的权重去计算每个点再迭代过程中获得到的能量值。
下面呈现出简单实现：


val rdd = sc.textFile("hdfs://master:9000/graph").map( line =>{
val pair = line.split("\\s+")
(pair(0).toLong,(pair(1).toLong,pair(2).toDouble))
}).partitionBy( new HashPartitioner( 100 )).persist( StorageLevel.MEMORY_ONLY_SER )
val edges = rdd.flatMap{ case(x,y)=> Seq(Edge(x,y._1,y._2))}
val graph = Graph.fromEdges(edges,1L).cache()
val edgeInital = graph.aggregateMessages[Double]( ctx => ctx.sendToSrc( ctx.attr),_+_,TripletFields.EdgeOnly )
var rankGraph: Graph[Double, Double] = graph.outerJoinVertices(edgeInital) {
(vid, vdata, deg) => deg.getOrElse( 0.0 ) }
.mapTriplets( e => 1.0 / e.srcAttr * e.attr, TripletFields.Src )
.mapVertices { (id, attr) => 1.0
}
var iteration = 0
val resetProb = 0.15
var prevRankGraph: Graph[Double, Double] = null
while (iteration < 10) {
rankGraph.cache()
val ranks = rankGraph.aggregateMessages[Double](
ctx => ctx.sendToDst( ctx.srcAttr * ctx.attr ), _ + _, TripletFields.Src)
prevRankGraph = rankGraph
rankGraph = rankGraph.joinVertices(ranks) {
(id, oldRank, msgSum) => resetProb + (1.0 - resetProb) * msgSum
}.cache()
prevRankGraph.vertices.unpersist(false)
prevRankGraph.edges.unpersist(false)

iteration += 1
}
“`还有许多地方待优化，希望有大神指出。