Spark GraphX图计算【代码实现，源码分析】

一.简介

　　参考：https://www.cnblogs.com/yszd/p/10186556.html

二.代码实现　　

 1 package big.data.analyse.graphx
 2
 3 import org.apache.log4j.{Level, Logger}
 4 import org.apache.spark.graphx._
 5 import org.apache.spark.rdd.RDD
 6 import org.apache.spark.sql.SparkSession
 7
 8 class VertexProperty()
 9 case class UserProperty(val name: String) extends VertexProperty
10 case class ProductProperty(val name: String, val price: Double) extends VertexProperty
11
12 /*class Graph[VD, ED]{
13   val vertices: VertexRDD[VD]
14   val edges: EdgeRDD[ED]
15 }*/
16
17 /**
18   * Created by zhen on 2019/10/4.
19   */
20 object GraphXTest {
21   /**
22     * 设置日志级别
23     */
24   Logger.getLogger("org").setLevel(Level.WARN)
25   def main(args: Array[String]) {
26     val spark = SparkSession.builder().appName("GraphXTest").master("local[2]").getOrCreate()
27     val sc = spark.sparkContext
28     /**
29       * 创建vertices的RDD
30       */
31     val users : RDD[(VertexId, (String, String))] = sc.parallelize(
32       Array((3L, ("Spark", "GraphX")), (7L, ("Hadoop", "Java")),
33             (5L, ("HBase", "Mysql")), (2L, ("Hive", "Mysql"))))
34
35     /**
36       * 创建edges的RDD
37       */
38     val relationships: RDD[Edge[String]] = sc.parallelize(
39       Array(Edge(3L, 7L, "Fast"), Edge(5L, 3L, "Relation"),
40       Edge(2L, 5L, "colleague"), Edge(5L, 7L, "colleague")))
41
42     /**
43       * 定义默认用户
44       */
45     val defualtUser = ("Machical", "Missing")
46
47     /**
48       * 构建初始化图
49       */
50     val graph = Graph(users, relationships, defualtUser)
51
52     /**
53       * 使用三元组视图呈现顶点之间关系
54       */
55     val facts : RDD[String] = graph.triplets.map(triplet =>
56       triplet.srcAttr._1 + " is the " + triplet.attr + " with " + triplet.dstAttr._1)
57     facts.collect().foreach(println)
58
59     graph.vertices.foreach(println) //顶点
60     graph.edges.foreach(println) //边
61     graph.ops.degrees.foreach(println) // 各顶点的度
62     graph.triplets.foreach(println) // 顶点，边，关系
63     println(graph.ops.numEdges) // 边的数量
64     println(graph.ops.numVertices) // 顶点的数量
65   }
66 }

三.结果

　　1.三元组视图

　　2.顶点

　　3.边

　　4.各顶点的度

　　5.三元组视图

　　6.边/顶点数量

四.源码分析

 1 class Graph[VD, ED] {
 2    // Information about the Graph
 3 　　val numEdges: Long
 4 　　val numVertices:Long
 5 　　val inDegrees: VertexRDD[Int]
 6 　　val outDegrees: VertexRDD[Int]
 7 　　val degrees: VertexRDD[Int]
 8 　　
 9    // Views of the graph as collections
10 　　val vertices: VertexRDD[VD]
11 　　val edges: EdgeRDD[ED]
12 　　val triplets: RDD[EdgeTriplet[VD,ED]]
13 　
14  　//Functions for caching graphs
15 　　def persist(newLevel1:StorageLevel = StorageLevel.MEMORY_ONLY): Graph[VD, ED]//默认存储级别为MEMORY_ONLY
16 　　def cache(): Graph[VD, ED]
17 　　def unpersistVertices(blocking: Boolean = true): Graph[VD, ED]
18
19 　　// Change the partitioning heuristic
20 　　def partitionBy(partitionStrategy: PartitionStrategy)
21
22 　　// Transform vertex and edge attributes
23 　　def mapVertices[VD2](map: (VertexId, VD) => VD2): Graph[VD2, ED]
24 　　def mapEdges[ED2](map: Edge[ED] => ED2): Graph[VD, ED2]
25 　　def mapEdges[ED2](map: (PartitionID, Iterator[Edge[ED]]) => Iterator[ED2]): Graph[VD, ED2]
26 　　def mapTriplets[ED2](map: EdgeTriplet[VD, ED] => ED2): Graph[VD, ED2]
27 　　def mapTriplets[ED2](map: (PartitionID, Iterator[EdgeTriplet[VD, ED]]) => Iterator[ED2]): Graph[VD, ED2]
28
29 　　// Modify the graph structure
30 　　def reverse: Graph[VD, ED]
31 　　def subgraph(epred: EdgeTriplet[VD,ED] => Boolean,vpred: (VertexId, VD) => Boolean): Graph[VD, ED]
32 　　def mask[VD2, ED2](other: Graph[VD2, ED2]): Graph[VD, ED] // 返回当前图和其它图的公共子图
33 　　def groupEdges(merge: (ED, ED) => ED): Graph[VD,ED]
34
35 　　// Join RDDs with the graph　　
36 　　def joinVertices[U](table: RDD[(VertexId, U)])(mapFunc: (VertexId, VD, U) => VD): Graph[VD, ED]
37 　　def outerJoinVertices[U, VD2](other: RDD[(VertexId, U)])(mapFunc: (VertexId, VD, Option[U]))
38 　　
39 　　// Aggregate information about adjacent triplets
40 　　def collectNeighborIds(edgeDirection: EdgeDirection): VertexRDD[Array[VertexId]]
41 　　def collectNeighbors(edgeDirection: EdgeDirection): VertexRDD[Array[(VertexId, VD)]]
42 　　def aggregateMessages[Msg: ClassTag](sendMsg: EdgeContext[VD, ED, Msg] => Unit, merageMsg: (Msg, Msg) => Msg, tripletFields: TripletFields: TripletFields = TripletFields.All): VertexRDD[A]
43 　　
44 　　//Iterative graph-parallel computation
45 　　def pregel[A](initialMsg: A, maxIterations: Int, activeDirection: EdgeDiection)(vprog: (VertexId, VD, A) => VD, sendMsg: EdgeTriplet[VD, ED] => Iterator[(VertexId, A)], mergeMsg: (A, A) => A): Graph[VD, ED]
46 　　
47 　　// Basic graph algorithms
48 　　def pageRank(tol: Double, resetProb: Double = 0.15): Graph[Double, Double]
49 　　def connectedComponents(): Graph[VertexId, ED]
50 　　def triangleCount(): Graph[Int, ED]
51 　　def stronglyConnectedComponents(numIter: Int): Graph[VertexId, ED]
52 }

原文地址：https://www.cnblogs.com/yszd/p/11621791.html