This directory contains the following files:
multicoloring.py, plot.p, performance-greedy-line.eps, 4nodes-plot.txt,
10nodes-plot.txt, 20nodes-plot.txt, 40nodes-plot.txt, 50nodes-plot.txt.

multicoloring.py:
=================
Python program running the algorithms Greedy and FourBucket on the line 
on uniformly distributed input sequences for request densities
10*2^i, 12.5*2^i, 15*2^i, and 17.5*2^i for i=1,2,...,10

The random seed, the number of repetitions, for each request density and the
length of the path can be defined:

seed = 30
number_of_tests = 100
number_of_nodes = [4,10,20,40,50]

For each path length, three text files are produced:
"number_of_nodes"nodes-results.txt logs the results after each testrun
"number_of_nodes"nodes-nodes.txt logs the nodes and their assigned colors after each testrun
"number_of_nodes"nodes-plot.txt logs the average performance after one round of testruns

plot.p
======
This is a gnuplot plotting file, plotting Greedy's performance on the line, 
in the style of the figure in the paper.

Xnodes-plot.txt
===============
These files are the results of running multicoloring.py with the parameters
set as in the program.

performance-greedy-line.eps
===========================
This file is a black & white graphic illustration of the results from the
files above. To see the same in color on the screen, start up gnuplot and
load the gnuplot file as described below.

Performing the actual test
==========================
To redo the test, first run the python program:
> python multicoloring.py
Then start gnuplot, and load the plotting file:
> gnuplot
> load "./plot.p"

A first test
============
The actual test made takes weeks of computation time. Before starting a
longer test, one can make quick test to see that everything works. To
do that, make the following adjustments:

In multicoloring.py, use:
   number_of_tests = 10
   range_exponent = 3
In plot.p, use:
   set xrange [1:100]

Running as described above should complete in less than one minute. Note
that the results only start becomming interesting for long request sequences,
corresponding to large values of the range_exponent, so this one minute test
is just to establish that it works in principle.