power-confluent drawing

A package for drawing confluent drawings by leveraging power graph decomposition [https://arxiv.org/abs/1810.09948]

The package may be installed using pip install pconfluent

and an example use case looks like

import pconfluent
I = [0,0,0,1,1,1,2,2,2]
J = [3,4,5,3,4,5,3,4,5]
Ir,Jr,Ip,Jp = pconfluent.find_power_graph(I,J)
pconfluent.draw_confluent(Ir,Jr,Ip,Jp, filepath='K33.svg')

where the function

find_power_graph(I, J, w_intersect=10, w_difference=1)

takes I and J as edge indices, and w_intersect and w_difference as weights for the greedy power graph construction algorithm. Returns a 4-tuple (Ir,Ir,Ip,Jp) where Ir and Jr are routing edges, and Ip and Jp are power edges.

draw_confluent(Ir, Jr, Ip, Jp, nodesplit=True, split_length=.5, filepath=None)

takes the routing and power edges from the find_power_graph function, and draws the corresponding power-confluent drawing. nodesplit and split_length can be used to choose if/how nodes should be split to remove crossings, and filepath as the destination for the output .svg file.

The power graph construction algorithm is written in C++ for speed, while the power-to-routing graph conversion and drawing is in Python. Splines are rendered in .svg format by using the Böhm algorithm to glue together quadratic Bézier curves (a good explanation can be found here).