networkx - can't remove_node from graph - python-3.x

Question

A G = nx.DiGraph() whose nodes and edges are the following:
G.nodes() = ['10.2.110.1', '10.2.25.65', '10.2.94.87', '10.2.20.209', '10.2.6.206', '10.2.94.55', '10.2.182.10', '10.2.94.86', '10.2.20.2', '10.2.20.1', '10.2.94.94']
G.edges() = [('10.2.110.1', '10.2.20.2'), ('10.2.110.1', '10.2.20.1'), ('10.2.25.65', '10.2.6.206'), ('10.2.94.87', '10.2.94.55'), ('10.2.20.209', '10.2.110.1'), ('10.2.94.55', '10.2.20.209'), ('10.2.182.10', '10.2.182.10'), ('10.2.94.86', '10.2.94.87'), ('10.2.20.2', '10.2.25.65'), ('10.2.20.1', '10.2.182.10'), ('10.2.94.94', '10.2.94.86')]
That above, produces the following topology.
As you can see, node_94 is green, because is the starting node. Both node_10 and node_206 are the farEnds.
I want to remove nodes from the graph depending on the number of hops away from the farEnds for node_94.
I have this function which tries to remove nodes depending on how far a node is from a given farEnd.
def getHopToNH(G):
labelList = {}
nodes = G
for startNode in nodes.nodes():
try:
farInt = nx.get_node_attributes(nodes,'farInt')[startNode]
except:
farInt = 'NA'
try:
p = min([len(nx.shortest_path(nodes,source=startNode,target=end)) for end in farInt])
except:
p = 0
if p < 7:
labelList = {**labelList,**{str(startNode):'node_'+str(startNode).split(".")[3]}}
else:
nodes.remove_node(startNode)
return labelList,nodes
However, when running that function, I get the following error:
File "trace_1_7.py", line 87, in getHopToNH
for startNode in nodes.nodes():
RuntimeError: dictionary changed size during iteration
The problem arises with the nodes.remove_node(startNode). If I remove that line, the code works nice and produces the plot that you can see above.
How can I accomplish the removal based on the number of hops towards a farEnd?
Thanks!
Lucas

Answer 1

Since networkx internally represents graphs using dicts, when iterating over a graph's nodes, we are iterating over the keys of a dictionary (this dictionary maps each node to its attributes). Using remove_node will change the size of this dictionary, which is not allowed when we are iterating over its key, hence the RuntimeError.
To remove nodes, we maintain a list containing the nodes we want to remove, then remove the nodes in this list after the for loop.
def getHopToNH(G):
labelList = {}
nodes = G
nodes_to_remove = []
for startNode in nodes.nodes():
try:
farInt = nx.get_node_attributes(nodes,'farInt')[startNode]
except:
farInt = 'NA'
try:
p = min([len(nx.shortest_path(nodes,source=startNode,target=end)) for end in farInt])
except:
p = 0
if p < 7:
labelList = {**labelList,**{str(startNode):'node_'+str(startNode).split(".")[3]}}
else:
nodes_to_remove.append(startNode)
nodes.remove_nodes_from(nodes_to_remove)
return labelList,nodes