Networks

(NOTE: AS OF 20170729, THIS PACKAGE IS NO LONGER BEING MAINTAINED. Users of this package are encouraged to migrate to MetaGraphs.jl.

Networks.jl is a Julia package that extends the basic graph types of LightGraphs.jl to also store graph/vertex/edge properties.

Installation

Pkg.clone("https://github.com/JuliaGraphs/Networks.jl")

Usage

The types Net and DiNet are the two ready-to-use network types, for undirected and directed networks respectively. They are able to store any number of graph/vertex/edge properties of any type. It is based on a dictionary of dictionaries structures.

using Networks
using LightGraph

n = Net(10)
nv(n) == 10 # number of vertices
ne(n) == 0 # number of edges

#vertex properties
add_vertex!(n, label="pino", weight=2.1)
getprop(n, 11, :label) == "pino"
setprop!(n, 2, label="rose")

#edge properties
add_edge!(n, 1, 2, width=10)
setprop!(n, Edge(1,2), label="violet", a = "b")
getprop(n, 1, 2, :width) == 10
getprop(n, 1, 2, :a) == "b"

nv(n) == 11 # number of vertices
ne(n) == 1 # number of edges

# graph properties
setprop!(n, name="mynetwork", id=1)
getprop(n, :name) == "mynetwork"

# construct from LightGraphs graphs
n = Net(CompleteGraph(10, 10))
n = DiNet(DiGraph(10, 20)) # random graph with 10 vertices and 20 edges

If you are looking for a more specialized and eventually more performant data structure, take a look to the (Di)Network type.

Internals

Networks.jl is a Julia package for supporting Graphs with vertex and edge properties.

This includes weighted graphs, labeled vertices, categories edges. The idea is to support graphs through the Adjacency List storage format provided in the LightGraphs.Graph (undirected) and LightGraphs.DiGraph (directed) graph types. This is achieved by encapsulation of the LightGraphs.Graph type. Goals of this project are to be simple, performant, and flexible in that order. This is similar to the LightGraphs.jl philosophy but with the understanding that applications of networks are much more diverse than a simple graph datastructure.

The idea is that any Network can be converted into a graph by forgetting the properties. Thus convert(Graph, net::Network) are defined for for both Graph and DiGraph. If combined with default definitions of the graph functions such as fadj that do conversion before calling the "real" definition then one can use Network as a graph transparently.

For example: fadj(g::Any) = fadj(convert(SimpleGraph, g)) allows one to call fadj on an arbitrary network and get the forward adjacency list.