How to Create and Visualize a Directed Weighted Graph with Parallel Edges and Different Weights Using Python?

I'm working on a project where I need to create a directed weighted graph in Python that allows parallel edges with different weights between nodes. I am using the networkx library and Matplotlib for visualization.

My goal is to:

• Create a directed graph with parallel edges (multiple edges between the same nodes).
• Assign random weights to these edges.
• Visualize the graph with edge labels showing the weights.

import random
import networkx as nx
import matplotlib.pyplot as plt

def create_graph(n_nodes, alpha = 0.5):
    G = nx.MultiDiGraph()
    G.add_nodes_from(range(n_nodes))
    for i in range(n_nodes):
        for j in range(i+1,n_nodes):
            if random.random() < alpha:
                weight=random.randint(1,10)
                G.add_edge(i, j, weight=weight)

            if random.random() < alpha:
                weight=random.randint(1,10)
                G.add_edge(j, i, weight=weight)
    return G
  

def display_graph(G):
    pos = nx.spring_layout(G)  
    weight_labels = nx.get_edge_attributes(G, 'weight')  
    nx.draw(G, pos, with_labels=True, node_color='skyblue', edge_color='gray', node_size=700)
    nx.draw_networkx_edge_labels(G, pos, edge_labels=weight_labels)  
    plt.show()


n_nodes = 5
G = create_graph(n_nodes, alpha = 0.5)
display_graph(G)

However, when I try to visualize the graph with edge labels, I got this error message:

networkx.exception.NetworkXError: draw_networkx_edge_labels does not support multiedges.

This error occurs when I try to display edge labels for a MultiDiGraph, and it seems like the draw_networkx_edge_labels function doesn't support parallel edges.

1. How can I resolve this error and properly visualize a directed weighted graph with parallel edges?
2. Is there a better way to create and visualize directed graphs with parallel edges in networkx?

I'd appreciate any guidance or examples to help me achieve this. Thank you in advance!

A core issue of Networkx's drawing utilities is that they are separated into different functions. For labeling multi-graph edges, the function that determines the location of the edge label (draw_networkx_edge_labels) requires knowledge of the edge path; however, the edge path is computed in draw_networkx_edges and there is no cross-talk between the two.

To work around this and other design issues of the Networkx drawing utilities, I wrote a drop-in replacement called Netgraph. Multi-graph support, however, is pretty new so you have to install from the dev branch (stable releases are distributed via pip and conda-forge):

pip install https://github.com/paulbrodersen/netgraph/archive/dev.zip

import random
import networkx as nx
import matplotlib.pyplot as plt

from netgraph import MultiGraph

def create_graph(n_nodes, alpha = 0.5):
    G = nx.MultiDiGraph()
    G.add_nodes_from(range(n_nodes))
    for i in range(n_nodes):
        for j in range(i+1,n_nodes):
            if random.random() < alpha:
                weight=random.randint(1,10)
                G.add_edge(i, j, weight=weight)

            if random.random() < alpha:
                weight=random.randint(1,10)
                G.add_edge(j, i, weight=weight)
    return G

G = create_graph(5)
MultiGraph(
    G,
    node_labels=True,
    node_color="skyblue",
    edge_color="gray",
    edge_labels=nx.get_edge_attributes(G, 'weight'),
    edge_label_fontdict=dict(fontsize=8),
    arrows=True,
)
plt.show()