This is a function to merge several nodes into one in a Networkx graph

nx_merge_nodes.py

# This is a function to merge several nodes into one in a Networkx graph

def merge_nodes(G,nodes, new_node, attr_dict=None, **attr):
    """
    Merges the selected `nodes` of the graph G into one `new_node`,
    meaning that all the edges that pointed to or from one of these
    `nodes` will point to or from the `new_node`.
    attr_dict and **attr are defined as in `G.add_node`.
    """
    
    G.add_node(new_node, attr_dict, **attr) # Add the 'merged' node
    
    for n1,n2,data in G.edges(data=True):
        # For all edges related to one of the nodes to merge,
        # make an edge going to or coming from the `new gene`.
        if n1 in nodes:
            G.add_edge(new_node,n2,data)
        elif n2 in nodes:
            G.add_edge(n1,new_node,data)
    
    for n in nodes: # remove the merged nodes
        G.remove_node(n)

Hi, thanks for sharing this function.

FYI -- I noticed a problem if new_node is in nodes: all new_node related edges get deleted at the end. If you always check if the 'current working node' is not new_node it seems to fix it.

Cheers!

Hi, Nice function.
But if the graph is a weighted graph then u should also add one more line to your function that new_node.weight = nodes.weight (nodes sent for merging)

There is an issue with this code. For example if we merge two nodes and both have a connection with same node then we have to add the edge weights of both connections when we make new edge.

@bilalzahid make a pull request!!