@rantanen/igraph

igraph

An embeddable webGL graph visualization library. http://patrickfuller.github.io/igraph/

Examples

• IPython notebook
• les misérables
• github connections

IPython

The IPython notebook is an open-source tool poised to replace MATLAB in many applications. As a scientist of sorts, I'm all about it. Therefore, I made handles to use igraph with the notebook. Install through pip:

pip install igraph

Open a new notebook and test the setup by typing:

import igraph
igraph.draw([(1, 2), (2, 3), (3, 4), (4, 1), (4, 5), (5, 2)])

into a notebook cell. You should get a paddlewheel graph as an output. You can use this in conjunction with other code for educational purposes (try generating a red-black tree!). There are three commands and some optional parameters to check out. Read the docstrings and check out the associated example for more.

Javascript

Start by downloading the minified javascript file:

wget https://raw.githubusercontent.com/patrickfuller/igraph/master/js/build/igraph.min.js

Include this file alongside jQuery in your project, and then use with:

igraph.create('my-selector');
igraph.draw(myGraph);

where 'my-selector' is where you want to place igraph, and myGraph is a plain ol' object. See below for more on the object structure, or just check out the included example. The igraph.create() method takes a few optional parameters, specifying the sizes and colors of nodes, as well as force-directed optimization.

options = {
    directed: true, // Toggles edge arrows
    nodeSize: 2.0, // Default node size
    edgeSize: 0.25, // Edge connection diameter
    arrowSize: 1.0, // If drawn, edge arrow size
    defaultNodeColor: 0xaaaaaa, // Color for nodes without a "color" property
    defaultEdgeColor: 0x777777, // Color for edges without a "color" property
    shader: "toon", // three.js shader to use, can be "toon", "basic", "phong", or "lambert"
    runOptimization: true // Runs a force-directed-layout algorithm on the graph
};

Graph Data Format

igraph takes input graph data structures as plain objects. Here's the most boring graph in the world:

{
    nodes: {
        jane: { },
        bob: { },
        mike: { },
        sally: { }
    },
    edges: [
        { source: "jane", target: "bob" },
        { source: "bob", target: "mike" },
        { source: "mike", target: "sally" }
    ]
}

Nodes require no information outside of their keys. However, there are useful optional parameters that can be specified.

{
    color: 0xffffff, // Color for this node
    size: 1.0, // Scaling factor for this node's size
    location: [0.0, 0.0, 0.0] // Starting location of node. Useful for pre-rendering.
}

By default, the algorithm runs a force-directed layout on the graph. When enabled, the "location" field is optional. However, for larger graphs, you will want to disable this feature and pre-render the locations. Use the associated Python library (igraph.generate) to do so.