Beispiel #1
0
  /**
   * Converts runtime edge data to persistent edge data (includes source/target vertex data) and
   * writes it to HBase.
   *
   * @param epgmDatabase EPGM database instance
   * @param edgeDataHandler edge data handler
   * @param persistentEdgeDataFactory persistent edge data factory
   * @param edgeDataTableName HBase edge data table name
   * @param <PED> persistent edge data type
   * @throws IOException
   */
  public <PED extends PersistentEdgeData<VD>> void writeEdgeData(
      final EPGMDatabase<VD, ED, GD> epgmDatabase,
      final EdgeDataHandler<ED, VD> edgeDataHandler,
      final PersistentEdgeDataFactory<ED, VD, PED> persistentEdgeDataFactory,
      final String edgeDataTableName)
      throws IOException {

    Graph<Long, VD, ED> graph = epgmDatabase.getDatabaseGraph().getGellyGraph();

    DataSet<PersistentEdgeData<VD>> persistentEdgeDataSet =
        graph
            .getVertices()
            // join vertex with edges on edge source vertex id
            .join(graph.getEdges())
            .where(0)
            .equalTo(1)
            // join result with vertices on edge target vertex id
            .join(graph.getVertices())
            .where("1.1")
            .equalTo(0)
            // ((source-vertex-data, edge-data), target-vertex-data)
            .with(new PersistentEdgeDataJoinFunction<>(persistentEdgeDataFactory));

    // write (persistent-edge-data) to HBase table
    Job job = Job.getInstance();
    job.getConfiguration().set(TableOutputFormat.OUTPUT_TABLE, edgeDataTableName);

    persistentEdgeDataSet
        .map(new HBaseWriter.EdgeDataToHBaseMapper<>(edgeDataHandler))
        .output(new HadoopOutputFormat<>(new TableOutputFormat<LongWritable>(), job));
  }
Beispiel #2
0
  /**
   * Converts runtime vertex data to persistent vertex data (includes incoming and outgoing edge
   * data) and writes it to HBase.
   *
   * @param epgmDatabase EPGM database instance
   * @param vertexDataHandler vertex data handler
   * @param persistentVertexDataFactory persistent vertex data factory
   * @param vertexDataTableName HBase vertex data table name
   * @param <PVD> persistent vertex data type
   * @throws Exception
   */
  public <PVD extends PersistentVertexData<ED>> void writeVertexData(
      final EPGMDatabase<VD, ED, GD> epgmDatabase,
      final VertexDataHandler<VD, ED> vertexDataHandler,
      final PersistentVertexDataFactory<VD, ED, PVD> persistentVertexDataFactory,
      final String vertexDataTableName)
      throws Exception {

    final Graph<Long, VD, ED> graph = epgmDatabase.getDatabaseGraph().getGellyGraph();

    // group edges by source vertex id (vertex-id, [out-edge-data])
    GroupReduceOperator<Edge<Long, ED>, Tuple2<Long, Set<ED>>> vertexToOutgoingEdges =
        graph
            .getEdges()
            .groupBy(0) // group by source vertex id
            .reduceGroup(
                new GroupReduceFunction<Edge<Long, ED>, Tuple2<Long, Set<ED>>>() {
                  @Override
                  public void reduce(
                      Iterable<Edge<Long, ED>> edgeIterable,
                      Collector<Tuple2<Long, Set<ED>>> collector)
                      throws Exception {
                    Set<ED> outgoingEdgeData = Sets.newHashSet();
                    Long vertexId = null;
                    boolean initialized = false;
                    for (Edge<Long, ED> edgeData : edgeIterable) {
                      if (!initialized) {
                        vertexId = edgeData.getSource();
                        initialized = true;
                      }
                      outgoingEdgeData.add(edgeData.getValue());
                    }
                    collector.collect(new Tuple2<>(vertexId, outgoingEdgeData));
                  }
                });

    // group edges by target vertex id (vertex-id, [in-edge-data])
    GroupReduceOperator<Edge<Long, ED>, Tuple2<Long, Set<ED>>> vertexToIncomingEdges =
        graph
            .getEdges()
            .groupBy(1) // group by target vertex id
            .reduceGroup(
                new GroupReduceFunction<Edge<Long, ED>, Tuple2<Long, Set<ED>>>() {
                  @Override
                  public void reduce(
                      Iterable<Edge<Long, ED>> edgeIterable,
                      Collector<Tuple2<Long, Set<ED>>> collector)
                      throws Exception {
                    Set<ED> outgoingEdgeData = Sets.newHashSet();
                    Long vertexId = null;
                    boolean initialized = false;
                    for (Edge<Long, ED> edgeData : edgeIterable) {
                      if (!initialized) {
                        vertexId = edgeData.getTarget();
                        initialized = true;
                      }
                      outgoingEdgeData.add(edgeData.getValue());
                    }
                    collector.collect(new Tuple2<>(vertexId, outgoingEdgeData));
                  }
                });

    // co-group (vertex-data) with (vertex-id, [out-edge-data]) to simulate left
    // outer join
    DataSet<Tuple2<Vertex<Long, VD>, Set<ED>>> vertexDataWithOutgoingEdges =
        graph
            .getVertices()
            .coGroup(vertexToOutgoingEdges)
            .where(0)
            .equalTo(0)
            .with(
                new CoGroupFunction<
                    Vertex<Long, VD>, Tuple2<Long, Set<ED>>, Tuple2<Vertex<Long, VD>, Set<ED>>>() {
                  @Override
                  public void coGroup(
                      Iterable<Vertex<Long, VD>> vertexIterable,
                      Iterable<Tuple2<Long, Set<ED>>> outEdgesIterable,
                      Collector<Tuple2<Vertex<Long, VD>, Set<ED>>> collector)
                      throws Exception {
                    Vertex<Long, VD> vertex = null;
                    Set<ED> outgoingEdgeData = null;
                    // read vertex data from left group
                    for (Vertex<Long, VD> v : vertexIterable) {
                      vertex = v;
                    }
                    // read outgoing edge from right group (may be empty)
                    for (Tuple2<Long, Set<ED>> oEdges : outEdgesIterable) {
                      outgoingEdgeData = oEdges.f1;
                    }
                    collector.collect(new Tuple2<>(vertex, outgoingEdgeData));
                  }
                });

    // co-group (vertex-data, (vertex-id, [out-edge-data])) with (vertex-id,
    // [in-edge-data]) to simulate left outer join
    DataSet<PersistentVertexData<ED>> persistentVertexDataSet =
        vertexDataWithOutgoingEdges
            .coGroup(vertexToIncomingEdges)
            .where("0.0")
            .equalTo(0)
            .with(new PersistentVertexDataCoGroupFunction<>(persistentVertexDataFactory));

    // write (persistent-vertex-data) to HBase table
    Job job = Job.getInstance();
    job.getConfiguration().set(TableOutputFormat.OUTPUT_TABLE, vertexDataTableName);

    persistentVertexDataSet
        .map(new HBaseWriter.VertexDataToHBaseMapper<>(vertexDataHandler))
        .output(new HadoopOutputFormat<>(new TableOutputFormat<LongWritable>(), job));
  }
Beispiel #3
0
  /**
   * Converts runtime graph data to persistent graph data (including vertex and edge identifiers)
   * and writes it to HBase.
   *
   * @param epgmDatabase EPGM database instance
   * @param graphDataHandler graph data handler
   * @param persistentGraphDataFactory persistent graph data factory
   * @param graphDataTableName HBase graph data table name
   * @param <PGD> persistent graph data type
   * @throws IOException
   */
  public <PGD extends PersistentGraphData> void writeGraphData(
      final EPGMDatabase<VD, ED, GD> epgmDatabase,
      final GraphDataHandler<GD> graphDataHandler,
      final PersistentGraphDataFactory<GD, PGD> persistentGraphDataFactory,
      final String graphDataTableName)
      throws IOException {
    final Graph<Long, VD, ED> graph = epgmDatabase.getDatabaseGraph().getGellyGraph();

    // build (graph-id, vertex-id) tuples from vertices
    FlatMapOperator<Vertex<Long, VD>, Tuple2<Long, Long>> graphIdToVertexId =
        graph
            .getVertices()
            .flatMap(
                new FlatMapFunction<Vertex<Long, VD>, Tuple2<Long, Long>>() {
                  @Override
                  public void flatMap(
                      Vertex<Long, VD> vertex, Collector<Tuple2<Long, Long>> collector)
                      throws Exception {
                    if (vertex.getValue().getGraphCount() > 0) {
                      for (Long graphID : vertex.getValue().getGraphs()) {
                        collector.collect(new Tuple2<>(graphID, vertex.f0));
                      }
                    }
                  }
                });

    // build (graph-id, edge-id) tuples from vertices
    FlatMapOperator<Edge<Long, ED>, Tuple2<Long, Long>> graphIdToEdgeId =
        graph
            .getEdges()
            .flatMap(
                new FlatMapFunction<Edge<Long, ED>, Tuple2<Long, Long>>() {
                  @Override
                  public void flatMap(Edge<Long, ED> edge, Collector<Tuple2<Long, Long>> collector)
                      throws Exception {
                    if (edge.getValue().getGraphCount() > 0) {
                      for (Long graphId : edge.getValue().getGraphs()) {
                        collector.collect(new Tuple2<>(graphId, edge.getValue().getId()));
                      }
                    }
                  }
                });

    // co-group (graph-id, vertex-id) and (graph-id, edge-id) tuples to
    // (graph-id, {vertex-id}, {edge-id}) triples
    CoGroupOperator<Tuple2<Long, Long>, Tuple2<Long, Long>, Tuple3<Long, Set<Long>, Set<Long>>>
        graphToVertexIdsAndEdgeIds =
            graphIdToVertexId
                .coGroup(graphIdToEdgeId)
                .where(0)
                .equalTo(0)
                .with(
                    new CoGroupFunction<
                        Tuple2<Long, Long>,
                        Tuple2<Long, Long>,
                        Tuple3<Long, Set<Long>, Set<Long>>>() {

                      @Override
                      public void coGroup(
                          Iterable<Tuple2<Long, Long>> graphToVertexIds,
                          Iterable<Tuple2<Long, Long>> graphToEdgeIds,
                          Collector<Tuple3<Long, Set<Long>, Set<Long>>> collector)
                          throws Exception {
                        Set<Long> vertexIds = Sets.newHashSet();
                        Set<Long> edgeIds = Sets.newHashSet();
                        Long graphId = null;
                        boolean initialized = false;
                        for (Tuple2<Long, Long> graphToVertexTuple : graphToVertexIds) {
                          if (!initialized) {
                            graphId = graphToVertexTuple.f0;
                            initialized = true;
                          }
                          vertexIds.add(graphToVertexTuple.f1);
                        }
                        for (Tuple2<Long, Long> graphToEdgeTuple : graphToEdgeIds) {
                          edgeIds.add(graphToEdgeTuple.f1);
                        }
                        collector.collect(new Tuple3<>(graphId, vertexIds, edgeIds));
                      }
                    });

    // join (graph-id, {vertex-id}, {edge-id}) triples with
    // (graph-id, graph-data) and build (persistent-graph-data)
    JoinOperator.EquiJoin<
            Tuple3<Long, Set<Long>, Set<Long>>, Subgraph<Long, GD>, PersistentGraphData>
        persistentGraphDataSet =
            graphToVertexIdsAndEdgeIds
                .join(epgmDatabase.getCollection().getSubgraphs())
                .where(0)
                .equalTo(0)
                .with(new PersistentGraphDataJoinFunction<>(persistentGraphDataFactory));

    // write (persistent-graph-data) to HBase table
    Job job = Job.getInstance();
    job.getConfiguration().set(TableOutputFormat.OUTPUT_TABLE, graphDataTableName);

    persistentGraphDataSet
        .map(new HBaseWriter.GraphDataToHBaseMapper<>(graphDataHandler))
        .output(new HadoopOutputFormat<>(new TableOutputFormat<LongWritable>(), job));
  }