Partition edges of a graph with NetworkX and Python

Question

I would like to partition edges of a graph g based on edge attributes, using Python and NetworkX. In this snippet:

import networkx as nx

g = nx.Graph()
g.add_node(1, pos=[0, 0])
g.add_node(2, pos=[0, 2])
g.add_node(3, pos=[1, 1])
g.add_node(4, pos=[2, 1])
g.add_edge(1, 3, cat='a')
g.add_edge(4, 3, cat='a')
g.add_edge(1, 2, cat='b')
g.add_edge(4, 2, cat='b')

I would like to partition graph g into graphs ga and gb based on cat attribute and retain node attribute pos . Note that nodes 1 and 4 will belong to both ga and gb . Is there a library support for this operation?

Answer 1

you could use a combination of list comprehension and sets:

ga = g.subgraph(set(*[[u,v] for (u,v,data) in g.edges(data=True) if data.get('cat')=='a']))
gb = g.subgraph(set(*[[u,v] for (u,v,data) in g.edges(data=True) if data.get('cat')=='b']))

Answer 2

This seems to be a simple solution:

ga = g.edge_subgraph(e for e in g.edges() if g.edges[e]['cat']=='a')
gb = g.edge_subgraph(e for e in g.edges() if g.edges[e]['cat']=='b')