def calculateSimilarity()

in cassovary-core/src/main/scala/com/twitter/cassovary/algorithms/similarity/Similarity.scala [45:75]

• 21 lines of code
• 7 McCabe index (conditional complexity)

  def calculateSimilarity(u: Int, v: Int, dir: GraphDir = GraphDir.OutDir): Double

  /**
   * Iterate over graph nodes and calculate similarity scores for each node. If the graph stores edges in
   * both in and out direction, then iterate over the neighbors in reverse `dir` of neighbors of node `u`.
   * @param u    current node ID
   * @param k    limit for similar nodes
   * @param dir  direction of edges in Directed Graph
   * @return     Seq of top `k` similar node ids and their similarity score with node `u`.
   *             Nodes with non-zero score are added. So, the length of the Seq can be less than `k`
   */
  def getTopKSimilarNodes(u: Int, k: Int, dir: GraphDir = GraphDir.OutDir): Seq[(Int, Double)] = {
    val similarNodesQueue = new SmallBoundedPriorityQueue[SimilarNodes](k)
    val graphNodes = {
      if (graph.storedGraphDir == StoredGraphDir.BothInOut) {
        getNeighbors(u, dir) match {
          case Some(neighbors) => neighbors.flatMap(getNeighbors(_, GraphDir.reverse(dir)).getOrElse(Seq.empty))
          case None => Seq.empty[Int]
        }
      } else {
        graph map (node => node.id)
      }
    }
    graphNodes foreach { node =>
      if (node != u) {
        val similarityScore = calculateSimilarity(u, node, dir)
        if (similarityScore > 0.0) similarNodesQueue += SimilarNodes(node, similarityScore)
      }
    }
    similarNodesQueue.top(k).map(node => (node.nodeId, node.score))
  }