Graphx处理janusGraph数据实现

最新推荐文章于 2023-12-29 01:23:56 发布

weixin_33882452

最新推荐文章于 2023-12-29 01:23:56 发布

阅读量1.8k

点赞数

CC 4.0 BY-SA版权

文章标签：大数据人工智能 java

原文链接：https://my.oschina.net/zhouwang93/blog/3052345

该博客提出在spark直接执行gremlinSQL方案受阻时的备选方案。将gremlinSql查询结果转化为startGraph并转写到HDFS，spark读取其构建graphx可用的图，调用graphx图计算算法，实现graphX操作janusgraph，还介绍了结果转json及读取转换的方法。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

2019独角兽企业重金招聘Python工程师标准>>>

声明： 此方案是在spark直接执行gremlinSQL方案实现受阻的情况下的备选方案，不涉及工作机密，不存在泄密可能，纯属个人思考，希望抛砖引玉

方案： 将gremlinSql的查询结果转化为startGraph，然后转写到HDFS，spark读取hdfs的starGraphJSon构建graphx可用的图，然后就可以调用graphx丰富的图计算算法；从而将实现graphX操作janusgraph的目的

1. gremlinSql的查询结果转换成starGraphJson

由于org.apache.tinkerpop.gremlin.structure.io.graphson.GraphSONWriter保存的graphSon格式无法满足需求，所以将查询出的带path的点边数据自己转换成单点图的json结构，转化方法如下，然后存到hdfs，存储方法不再赘述。

public StringBuilder generatorStarGraphJson(Vertex vertex, Edge edge, StringBuilder starGraphJson){
	String inVId;
	String outVId;
	String VId;
	starGraphJson.append("{").append("\"id\":"+vertex.id()+","+"\"label\":\""+vertex.label()+"\",");
	//这种情况有outE和inE，outE中会有inV信息,inE会有一个inV
	inVId = edge.inVertex().id().toString();
	outVId = edge.outVertex().id().toString();
	VId = vertex.id().toString();
	if(inVId.equalsIgnoreCase(VId)){
		starGraphJson.append("\"outE\":{").append("\""+edge.label()+"\":[{").append("\"id\":\""+edge.id()+"\",")
				.append("\"inV\":"+edge.inVertex().id()+",").append("\"properties\":{"+concatEdgeProperties(edge)+"}}]},");
	}else if(outVId.equalsIgnoreCase(VId)){
		starGraphJson.append("\"inE\":{").append("\""+edge.label()+"\":[{").append("\"id\":\""+edge.id()+"\",")
				.append("\"outV\":"+edge.inVertex().id()+",").append("\"properties\":{"+concatEdgeProperties(edge)+"}}]},");
	}else{
		throw new Exception("点边不对应数据错误！！！");
	}
	//拼接点的properties
	starGraphJson.append("\"properties\":{").append(concatVertexProperties(vertex)).append("}}");
	return  starGraphJson;
}

2. spark读取指定路径的starGraph转成graph

class GraphSon2GraphXRDD() extends Serializable {

def getGraphConf(HDFSFilePath : String): BaseConfiguration ={ val inputGraphConf = new BaseConfiguration inputGraphConf.setProperty("gremlin.graph", classOf\[HadoopGraph\].getName) inputGraphConf.setProperty(Constants.GREMLIN\_HADOOP\_GRAPH\_READER, classOf\[GraphSONInputFormat\].getName) inputGraphConf.setProperty(Constants.GREMLIN\_HADOOP\_INPUT\_LOCATION, HDFSFilePath) inputGraphConf.setProperty(Constants.MAPREDUCE\_INPUT\_FILEINPUTFORMAT_INPUTDIR, HDFSFilePath) inputGraphConf }

def getSc(sparkHost:String ,isRemote:Boolean): SparkContext ={ var sparkConf = new SparkConf() if(isRemote){ //待完善 }else{ sparkConf.setMaster("local\[*\]").setAppName("GraphSon2GraphX") } val sc = new SparkContext(sparkConf) sc }

def getJavaRDD(conf : BaseConfiguration, sc : SparkContext): JavaPairRDD\[AnyRef, VertexWritable\] ={ val jsc = JavaSparkContext.fromSparkContext(sc) val graphRDDInput = new InputFormatRDD val vertexWritableJavaPairRDD = graphRDDInput.readGraphRDD(conf, jsc) vertexWritableJavaPairRDD }

def getVertexRDD(vertexWritableJavaPairRDD : JavaPairRDD\[AnyRef, VertexWritable\]): RDD\[(Long,util.HashMap\[String,java.io.Serializable\])\] ={ vertexWritableJavaPairRDD.rdd.map((tuple2: Tuple2\[AnyRef, VertexWritable\]) => { // Get the center vertex val v = tuple2._2.get val g = StarGraph.of(v) // In case the vertex id in TinkerGraph is not long type // val vid = convertStringIDToLongID([v.id](http://v.id)().toString) val vid = [v.id](http://v.id)().toString.toLong // Pass the vertex properties to GraphX vertex value map and remain the original vertex id var graphxValueMap : util.HashMap\[String,java.io.Serializable\] = new util.HashMapString,java.io.Serializable graphxValueMap.put("originalID",[v.id](http://v.id)().toString) graphxValueMap.putAll(g.traversal.V([v.id](http://v.id)).valueMap().next(1).get(0)) (vid,graphxValueMap) }) }

def getEdgeRDD(vertexWritableJavaPairRDD : JavaPairRDD\[AnyRef, VertexWritable\]): RDD\[graphx.Edge\[util.HashMap\[String, java.io.Serializable\]\]\] ={ val edge = vertexWritableJavaPairRDD.rdd.flatMap((tuple2: Tuple2\[AnyRef, VertexWritable\]) => { val v = tuple2._2.get val g = StarGraph.of(v) val edgelist:util.List\[Edge\] = g.traversal.V([v.id](http://v.id)).outE().toList

  // Put all edges of the center vertex into the list
  val list = new collection.mutable.ArrayBuffer[graphx.Edge[util.HashMap[String,java.io.Serializable]]]()
  var x = 0
  for(x &lt;- 0 until edgelist.size()){
    var srcId = edgelist.get(x).inVertex.id().toString
    var dstId = edgelist.get(x).outVertex.id().toString
    //        val md1 = convertStringIDToLongID(srcId)
    //        val md2 = convertStringIDToLongID(dstId)
    val md1 = srcId.toLong
    val md2 = dstId.toLong
    // Get the properties of the edge
    var edgeAttr = new util.HashMap[String,java.io.Serializable]()
    var perporties : util.Iterator[Property[Nothing]] = edgelist.get(x).properties()
    while(perporties.hasNext){
      val property = perporties.next()
      edgeAttr.put(property.key(),property.value().toString)
    }
    list.append(graphx.Edge(md1,md2,edgeAttr))
  }
  list
})
val edgeRDD = edge.distinct()
edgeRDD

}

def doLAP(vertexWritableJavaPairRDD : JavaPairRDD\[AnyRef, VertexWritable\], iterationNum : Int): Array\[Array\[String\]\] = { val vertexRDD = getVertexRDD(vertexWritableJavaPairRDD)

val edgeRDD = getEdgeRDD(vertexWritableJavaPairRDD)

val graph = graphx.Graph[util.HashMap[String,java.io.Serializable],
  util.HashMap[String,java.io.Serializable]](vertexRDD,edgeRDD,new util.HashMap[String,java.io.Serializable]())

val LVMRsult = lib.LabelPropagation.run(graph , iterationNum).vertices.collect.sortWith (_._1 &lt; _._1).map(f =&gt; {
  println(f.toString())
  f})
getFinalCommunit(LVMRsult)

}

def getFinalCommunit(LVMRsult:Array\[(Long,Long)\]): Array\[Array\[String\]\] ={ var result = new Array[Array\[String\]](LVMRsult.length) var tmp = new ArrayBufferString for(i <- 0 until LVMRsult.length){ var k = 0 val array = new ArrayBufferString
  //社区中包含多个值
  for(j &lt;- (i+1) until LVMRsult.length) {
    if(LVMRsult(i)._2.equals(LVMRsult(j)._2)){
      if(!tmp.contains(LVMRsult(i)._1.toString)){
        array += LVMRsult(i)._1.toString
        tmp += LVMRsult(i)._1.toString
      }
      if(!tmp.contains(LVMRsult(j)._1.toString)){
        array += LVMRsult(j)._1.toString
        tmp += LVMRsult(j)._1.toString
      }
      k = k+1
    }
  }

  //自己为一个社区
  if(k.equals(0)){
    if(!tmp.contains(LVMRsult(i)._1.toString)){
      array += LVMRsult(i)._1.toString
      tmp += LVMRsult(i)._1.toString
    }

  }
  if(array.length &gt; 0){
    result.update(i,array.toArray.distinct)
  }
}
result.filter(f =&gt; {
  println(if (f.length &gt;0) f.mkString("(",",",")"))
  f != null
})

}

def doPageRank(vertexWritableJavaPairRDD : JavaPairRDD\[AnyRef, VertexWritable\], stopThreshold : Double): Array\[Array\[Any\]\] = { val vertexRDD:RDD\[(Long,util.HashMap\[String,java.io.Serializable\])\] = getVertexRDD(vertexWritableJavaPairRDD)


val edgeRDD = getEdgeRDD(vertexWritableJavaPairRDD)

val graph = graphx.Graph[util.HashMap[String,java.io.Serializable],
  util.HashMap[String,java.io.Serializable]](vertexRDD,edgeRDD,new util.HashMap[String,java.io.Serializable]())
val gpgraph = graph.pageRank(stopThreshold).cache()

val titleAndPrGraph = graph.outerJoinVertices(gpgraph.vertices) {
  (v, title, rank) =&gt; (rank.getOrElse(0.0), title)
}

//倒序 false  正序 true

// titleAndPrGraph.vertices.sortBy((entry: (VertexId, (Double, Object))) => entry.\_2.\_1, false).foreach(f => println(f.\_1+":"+f.\_2._1))

val pageRank = titleAndPrGraph.vertices.sortBy((entry: (VertexId, (Double, Object))) =&gt; entry._2._1, false).map(f =&gt; {
  println(f._1+":"+f._2._1)
  Array(f._1.toString,f._2._1)
})
pageRank.collect()

}

}

这样就贯通了janusgraph和graphx，调用graphx的丰富的图计算功能就畅通无阻，就是实现有点挫，希望抛砖引玉

转载于:https://my.oschina.net/zhouwang93/blog/3052345