Search code examples
pythonnetworkx

How to simplify a networkx graph?

I am looking to "simplify" a NetworkX Directed Graph on a chosen set nodes. For e.g. I have this starting directed graph

enter image description here

And I want to see relationships in subgraph of only nodes [1,2,4 and 6]. i.e. as follows: enter image description here

Basically I want to show the "resulting dependencies" on just my chosen nodes.

So far, I have tried the subgraph approach as follows.

Creating the original graph:

G = nx.DiGraph() 
G.add_edges_from([(1,3),(3,4),(1,5),(5,6),(2,6),(4,6)])

..and then creating a subgraph:

chosen_nodes = [1,2,4,6]
SG = G.subgraph(chosen_nodes)

But the resulting subgraph doesn't retain any dependency relationship in edges other than the 2-6 and 4-6 relationship.

print(list(SG.edges))
>> [(4, 6), (2, 6)]
Solution

  • I think one clean way to get at this is to generate a subgraph of the transitive closure.

    from networkx.algorithms.dag import transitive_closure

def dependency_graph(G, nodes):
    return transitive_closure(G).subgraph(nodes)

print(dependency_graph(G, [1, 2, 4, 6]).edges)
>> [(1, 4), (1, 6), (4, 6), (2, 6)]