Edge attribute filter on GraphFrames motif search not working

Question

I've got some sample data on a family graph I want to query on.

I'd like to use the find method on the GraphFrames object in order to query the motif A->B where the edge is of type "Mother".

Since GraphFrames uses a subset of the cypher language of Neo4J I was wondering if the following would be the correct query?

graph.find("(A)-[edge:Mother]->(B)").show

Or what would be the best way to implement this in GraphFrames?

GraphFrame(vertex, graph.edges.filter("attr=='Mother'")).vertices.show

This doesn't work since I cannot filter on the direction, so I only want to get the mothers :)

Any idea?

Answer 1

Suppose this is your test data:

import org.graphframes.GraphFrame

val edgesDf = spark.sqlContext.createDataFrame(Seq(
  ("a", "b", "Mother"),
  ("b", "c", "Father"),  
  ("d", "c", "Father"),
  ("e", "b", "Mother")    
)).toDF("src", "dst", "relationship")

val graph = GraphFrame.fromEdges(edgesDf)
graph.edges.show()

+---+---+------------+
|src|dst|relationship|
+---+---+------------+
|  a|  b|      Mother|
|  b|  c|      Father|
|  d|  c|      Father|
|  e|  b|      Mother|
+---+---+------------+

You can use a motif query and apply a filter to it:

graph.find("()-[e]->()").filter("e.relationship = 'Mother'").show()

+------------+
|           e|
+------------+
|[a,b,Mother]|
|[e,b,Mother]|
+------------+

Or, since your case is relatively simple, you can apply a filter to the edges of the graph:

graph.edges.filter("relationship = 'Mother'").show()

+---+---+------------+
|src|dst|relationship|
+---+---+------------+
|  a|  b|      Mother|
|  e|  b|      Mother|
+---+---+------------+

Here's some alternative syntax (each gets the same result as immediately above):

graph.edges.filter($"relationship" === "Mother").show()
graph.edges.filter('relationship === "Mother").show()

You mention filtering on direction, but the direction of each relationship is encoded in the graph itself (i.e. from source to destination).