# graph-canon
[](./LICENSE.md)

[](https://codecov.io/gh/noamteyssier/graph-canon)
[](https://docs.rs/graph-canon/latest/graph-canon/)
Super fast and barebones graph canonicalization using `nauty` C-lib
and built on `petgraph`.
## Usage
### Hashable Labels
If you are just looking to create a hashable object to determine isomorphism
then it is simples to use the `CanonLabeling` struct.
This can be created from a `Graph` object directly.
#### Directed Graphs
```rust
use petgraph::{Directed, Graph};
use graph_canon::CanonLabeling;
let e1 = vec![(0, 1), (0, 2), (1, 2)]; // Isomorphic
let e2 = vec![(1, 0), (1, 2), (0, 2)]; // Isomorphic
let e3 = vec![(1, 0), (1, 2), (2, 1)]; // Non-Isomorphic
let g1 = Graph::<(), (), Directed>::from_edges(&e1);
let g2 = Graph::<(), (), Directed>::from_edges(&e2);
let g3 = Graph::<(), (), Directed>::from_edges(&e3);
let l1 = CanonLabeling::new(&g1);
let l2 = CanonLabeling::new(&g2);
let l3 = CanonLabeling::new(&g3);
assert_eq!(l1, l2);
assert_ne!(l1, l3);
```
#### Undirected Graphs
```rust
use petgraph::{Undirected, Graph};
use graph_canon::CanonLabeling;
let e1 = vec![(0, 1), (0, 2), (1, 2)]; // Isomorphic
let e2 = vec![(1, 0), (1, 2), (0, 2)]; // Isomorphic
let e3 = vec![(1, 0), (1, 2)]; // Non-Isomorphic
let g1 = Graph::<(), (), Undirected>::from_edges(&e1);
let g2 = Graph::<(), (), Undirected>::from_edges(&e2);
let g3 = Graph::<(), (), Undirected>::from_edges(&e3);
let l1 = CanonLabeling::new(&g1);
let l2 = CanonLabeling::new(&g2);
let l3 = CanonLabeling::new(&g3);
assert_eq!(l1, l2);
assert_ne!(l1, l3);
```
### Recovering the Canonical `Graph`
If instead you are interested in working with the graph itself,
you can use the `canonize` function to return a new `Graph` object
```rust
use petgraph::{Directed, Graph};
use graph_canon::canonize;
let edges = vec![(0, 1), (0, 2), (1, 2)];
let graph = Graph::<(), (), Directed>::from_edges(&edges);
let canon = canonize(&graph);
assert_eq!(canon.edge_count(), 3);
```
## Timing Comparison
This crate is inspired by [`nauty-pet`](https://crates.io/crates/nauty-pet)
but is much faster as it is much simpler.
(tests measured with [`criterion`](https://docs.rs/criterion/latest/criterion/))
This test is using a randomly generated graph of `10` nodes and `0.5` probability
of edge connection using [`random_gpn_graph`](https://docs.rs/petgraph-gen/0.1.3/petgraph_gen/fn.random_gnp_graph.html)
```text
graph-canon time: [1.3272 µs 1.3276 µs 1.3285 µs]
Found 14 outliers among 100 measurements (14.00%)
3 (3.00%) high mild
11 (11.00%) high severe
nauty-pet time: [6.2591 µs 6.2738 µs 6.2956 µs]
Found 10 outliers among 100 measurements (10.00%)
2 (2.00%) low mild
4 (4.00%) high mild
4 (4.00%) high severe
```