How to get a specified node 2-hop subgraph with pyspark

[Reid00]

Hi guys,

I am fresh man about grahpx. I read the doc want to get the 2-hop subgraph, find this

from graphframes.examples import Graphs
g = Graphs.friends()  # Get example graph

# Select subgraph based on edges "e" of type "follow"
# pointing from a younger user "a" to an older user "b".
paths = g.find("(a)-[e]->(b)")\
  .filter("e.relationship = 'follow'")\
  .filter("a.age < b.age")
# "paths" contains vertex info. Extract the edges.
e2 = paths.select("e.src", "e.dst", "e.relationship")
# In Spark 1.5+, the user may simplify this call:
#  val e2 = paths.select("e.*")

# Construct the subgraph
g2 = GraphFrame(g.vertices, e2)

but I noticed, this just 1-hop subgraph. I try use bfs, but I found it just get out direction like, get subgraph from a, bfs below code, just get (a)-[]->(b)-[]->(c). In fact, I want to both directions, including (a)<-[]-(b)<-[]-(c) and (a)-[]->(b)<-[]-(c).

    two_hop_graph = graph.bfs(fromExpr="id='f625a5b661058ba5082ca508f99ffe1b'", 
    toExpr="id<>'f625a5b661058ba5082ca508f99ffe1b'", maxPathLength=hop)
    two_hop_graph.show(truncate=False)
    print(f"count: {two_hop_graph.count()}")

is there any method can get a subgraph with specified node id?

[Reid00]

hello, is there anyone take a look? thanks！

[tusharg1993]

@Reid00 Trying to follow your query. Let me know if I misunderstood. If I understand correctly, you are looking for edges in the graph to be treated as undirected?

In GraphFrame, the edges are directed by nature. Thus, if you want the both reverse edges to be taken, you would need to add the reciprocal edges to the edges dataframe.

[Reid00]

thanks for your reply. I sorry confused you. I want to how to get a specified node 2-hop subgraph, I do some tests as above code, but It's all wrong, so I want to know is there any way in GraphFrame?

what is 2-hop subgraph like below:

[coreyabs-db]

Hi @Reid00 - I think you can do this with a motif and appropriate filters, for example something like:

hops = g.find("(a)-[e1]->(b); (b)-[e2]->(c)").filter("a.id == 0 and a.id != c.id")
new_edges = hops.select("e1.*").unionAll(hops.select("e2.*")).distinct()
subgraph = GraphFrame(g.vertices, new_edges).dropIsolatedVertices()

bfs won't work here as you noticed because it doesn't go beyond the first node in the path matching the expression (and the first non target node of course matches your toExpr). as @tusharg1993 mentioned, be sure you have reverse edges for undirected behavior.

Hope this helps.