networkx: Draw text on edges
control-flow, directed-graph, graph-theory, networkx, python
Solution
import networkx as nx
# Sample graph
G = nx.Graph()
G.add_edge(0,1)
G.add_edge(1,2)
G.add_edge(2,3)
G.add_edge(1,3)
labels = {(0,1):'foo', (2,3):'bar'}
pos=nx.spring_layout(G)
nx.draw(G, pos)
nx.draw_networkx_edge_labels(G,pos,edge_labels=labels,font_size=30)
import pylab as plt
plt.show()
EDIT: If you need multidigraphs with edge labels, I don't think you can do this completely within `networkx`. However you can do most of it in python and do the rendering and layout with another program:
import networkx as nx
G = nx.MultiDiGraph()
G.add_edge(0,1, label='A')
G.add_edge(1,0, label='B')
G.add_edge(2,3, label='foo')
G.add_edge(3,1, label='bar')
nx.write_dot(G, 'my_graph.dot')
I use graphviz to turn this into an image. On my Unix machine I run this from the command-line
dot my_graph.dot -T png > output.png
Which gives the desired output you are looking for. Note that graphviz has many ways to customize the visual appearance. The above example simply produces:
Problem
for my thesis I need to draw some probabilistic control flow graphs. i.e. control flow graphs with probabilities depicted on the edges. I found graph-tool which seems quite useful, since it can use deep-copies of existing graphs and my graphs are quite similar. So my question is, if there is a possibility to draw edge properties (or some strings) on/next to the edges? If it's not possible or highly complicated, is there a tool which is better to use in this case? Edit: I need directed edges that can even create loops between 2 nodes and have different values. Is there a possibility for this too? So I can see both values? By now I can see the directed graph with a 2-directional edge, but there's only one value on it. So e.g. in networkx (in reference to Hooked) it would look like: ``` G = nx.MultiDiGraph() G.add_edge(0,1) G.add_edge(1,0) labels = {(0,1):'foo', (1,0):'bar'} ``` So that both 'foo' and 'bar' are visible and you can see which direction they are connected to. But as networkx renders it, I get 1 bidirectional edge with 1 of the labels.