plotEmails.py

'''
@date Nov 27,  2009
@author: Michael Bommarito
@contact michael.bommarito@gmail.com
This is provided for purely academic purposes.
'''
import igraph,cairo

# Set of nodes
nodes = set()

# List of edges - edges can have multiplicity > 1
edges = []

# Read in the data from emailList.txt, which stores the output of parseEmails.py
emailData = [line.split(',') for line in open('emailList.txt').read().strip().splitlines()]

# Iterate over all emails
for email in emailData:
	# Update the node set
	nodes.update(email[2:])

	# Now update the edge list
	for b in email[3:]:
		edges.append((email[2],b))

'''
Convert the node set into a sorted list so we can index the nodes to integer.
This is necessary for igraph.

Then build the node map which goes from (node->index) for fast indexing next step.
'''
nodes = sorted(list(nodes))
nodeMap = dict([(v,i) for i,v in enumerate(nodes)])

'''
This line is tricky but fast - we want a dictionary with key = integer-indexed edge
and value = edge multiplicity.  This line does that.

At this point, our edgelist is edgeDict.keys() and our edgeweights are edgeDict.values().
'''
edgeDict = dict([((nodeMap[e[0]],nodeMap[e[1]]),edges.count(e)) for e in set(edges)])
edges = edgeDict.keys()
weights = edgeDict.values()

# Create the graph!
g = igraph.Graph(edges, directed = True)

'''
Now do some cosmetic stuff.
Label the nodes and then resize them and the edges.
'''
degree = g.degree()
for i,v in enumerate(g.vs):
	v['label'] = nodes[i].split('@')[1]
	v['label_color'] = 'red'
	v['label_size'] = 8
	v['label_dist'] = 0
	v['size'] = degree[i] / 10.0
	v['color'] = 'blue'

for e in g.es:
	e['width'] = weights[i]/5.0
	e['color'] = 'grey'
	e['arrow_width'] = 0.4
	e['arrow_size'] = 0.4

'''
Drop the pieces out of the largest weakly connected component.
We go from:
Directed graph (|V| = 2239, |E| = 4513)
Directed graph (|V| = 2261, |E| = 4529)
'''
g = g.clusters(mode = igraph.WEAK).subgraph(0)
weights = [e['width'] for e in g.es]

'''
Layout and plot the graph.
The layout is 3-step: randomize, apply grid-FR to get close to home, then apply real FR with weighting.
Deal with bbox and margin size stuff too...
Then output to SVG.
I manually backlit this in Inkscape.
'''
layout = g.layout_random()
layout = g.layout_grid_fruchterman_reingold(maxiter = 100, seed = layout)
layout = g.layout_fruchterman_reingold(maxiter=250, weights = weights, seed = layout)
pixels = 8000
margin = 50
svg = cairo.SVGSurface("emails.svg", pixels, pixels)
p = igraph.drawing.plot(g, target = svg, layout = layout, bbox  = (0,0,pixels,pixels), margin = (0,0,margin,margin))
p.save()