Online igraph Compiler

igraph is a network analysis library with a core written in C, and python-igraph is its Python interface, imported as igraph. The algorithms run in compiled code, so it handles large graphs much faster than pure-Python libraries such as NetworkX. It covers graph generators, paths, centrality, communities and layouts. People use it to analyse social, citation and biological networks. This page is an online igraph compiler: the code runs in your browser, so you can try it without installing anything. Run the example first, then paste any snippet below into a new cell to try it.

What the example does

The ig.config lines make plots use matplotlib, set the default layout to Fruchterman-Reingold, a force-directed layout, and make colour numbers refer to the "rainbow" palette. python-igraph takes its random numbers from Python's random module, so random.seed(1) gives the same graph on every run. ig.Graph.Barabasi(n=100, m=1) grows a graph one vertex at a time. Each new vertex links to one existing vertex and prefers those that already have many links. With m=1 the result is a tree: 100 vertices and 99 edges. g.betweenness() counts, for each vertex, how many shortest paths between other vertices pass through it. The list comprehension scales those counts to colour numbers from 0 to 200, and ig.plot() draws the graph on a matplotlib figure.

Build a graph with names and weights

Graph.TupleList() builds a graph from (source, target, weight) tuples and keeps the names in the name vertex attribute. Vertices get IDs from 0, and g.vs and g.es hold the vertex and edge attributes:

import igraph as ig

g = ig.Graph.TupleList(
    [("Home", "School", 10.0), ("Home", "Park", 2.0), ("Park", "School", 1.5),
     ("Park", "Mall", 3.0), ("Mall", "School", 1.0)],
    weights=True,
)
print(g.vcount(), "vertices,", g.ecount(), "edges, directed:", g.is_directed())
print(g.vs["name"])                      # vertex attribute
print(g.es["weight"])                    # edge attribute
print(dict(zip(g.vs["name"], g.degree())))
print([g.vs[i]["name"] for i in g.neighbors("Park")])
print(g)

Find shortest paths

get_shortest_path() returns vertex IDs, and distances() returns a matrix of path lengths. Both count edges unless you pass weights:

import igraph as ig

g = ig.Graph.TupleList(
    [("Home", "School", 10.0), ("Home", "Park", 2.0), ("Park", "School", 1.5),
     ("Park", "Mall", 3.0), ("Mall", "School", 1.0)],
    weights=True,
)
path = g.get_shortest_path("Home", "School", weights="weight")   # vertex IDs
print([g.vs[i]["name"] for i in path])
print(g.distances("Home", "School", weights="weight"))           # 3.5 km
print(g.distances("Home", "School"))                             # 1 edge
print(g.distances(weights="weight"))                             # every pair

Rank vertices by centrality

Graph.Famous("Zachary") is Zachary's karate club, built into igraph: 34 members and the 78 ties between them. betweenness() returns raw path counts. NetworkX divides by the number of pairs of other vertices, so vertex 0's 231.1 is 0.438 there:

import igraph as ig

g = ig.Graph.Famous("Zachary")
print(g.vcount(), "vertices,", g.ecount(), "edges")

degree = g.degree()
betweenness = g.betweenness()
pagerank = g.pagerank()
top = sorted(range(g.vcount()), key=lambda v: betweenness[v], reverse=True)[:3]
for v in top:
    print(f"vertex {v}: degree {degree[v]}, betweenness {betweenness[v]:.1f}, pagerank {pagerank[v]:.3f}")

Find communities and plot them

community_multilevel() is the Louvain method: it groups vertices so that most edges fall inside a group. It is randomized, so seed Python's random module first. Passing a matplotlib Axes as target draws with matplotlib whatever ig.config says, and mark_groups=True shades each community:

import random
import igraph as ig
import matplotlib.pyplot as plt

random.seed(0)
g = ig.Graph.Famous("Zachary")
communities = g.community_multilevel()
print(len(communities), "communities, sizes", communities.sizes())
print("modularity", round(communities.modularity, 3))

fig, ax = plt.subplots(figsize=(6, 6))
ig.plot(communities, target=ax, layout=g.layout("fr"), mark_groups=True,
        vertex_size=15, vertex_label=range(g.vcount()), vertex_label_size=7)
plt.show()

Good to know

• igraph plots with Cairo by default, and Cairo is not available here: ig.plot(g) on its own raises AttributeError: Plotting not available. Set ig.config["plotting.backend"] = "matplotlib", as the example does, or pass target=ax.
• igraph checks for matplotlib when it is first imported. If you import igraph before any cell has imported matplotlib, plotting fails with a FakeModule error for the rest of the session. Reload the page and import both in the same cell, as the example does.
• Vertex IDs always run from 0 to n-1, so deleting a vertex renumbers the ones after it. Refer to vertices by their name attribute.