In French, the spectrum of a matrix is called "spectre".
This micro-library contains a small toolkit helpful in analysing network topologies. Namely:
- Degree, adjacency and Laplacian matrices for a graph.
- Various centrality measures for a graph.
- The algebraic connectivity of a graph (Fiedler).
The library is centered around the Graph
structure which can be constructed
from one or more Edge
instances. Once constructed, various measurements and
matrix representations of the graph can be computed.
use std::net::SocketAddr;
use spectre::edge::Edge;
use spectre::graph::Graph;
// Construct the graph instance.
let mut graph = Graph::new();
// Create some addresses to be part of a network topology.
let addrs: Vec<SocketAddr> = (0..3)
.map(|i| format!("127.0.0.1:{i}").parse().unwrap())
.collect();
let (a, b, c) = (addrs[0], addrs[1], addrs[2]);
// Insert some edges, note the IDs can be any type that is `Copy + Eq + Hash + Ord`.
graph.insert(Edge::new(a, b));
graph.insert(Edge::new(a, c));
// Compute some metrics on that state of the graph.
let density = graph.density();
let degree_centrality_delta = graph.degree_centrality_delta();
// Matrices can be pretty printed...
println!("{}", graph.laplacian_matrix());
// ...outputs:
// ┌ ┐
// │ 2 -1 -1 │
// │ -1 1 0 │
// │ -1 0 1 │
// └ ┘