superstring - greedy approximation of the shortest common superstring

Finding the shortest string containing each member of the input string
set as a substring is a problem with applications in data compression,
computer security and bioinformatics. As the problem is NP-complete
(it can be reformulated as ordering the input strings to maximize
their overlap), it is usually solved approximately. This package
implements a simple approximation algorithm described in

Jonathan S. Turner
Approximation algorithms for the shortest common superstring problem
Information and Computation 83, 1-20, 1989

The code does have some shortcuts - most importantly, it uses a
3rd-party suffix tree instead of an implementation tuned to the
algorithm - so Turner's performance analysis isn't wholly applicable,
but it should be enough to play with.

When run without arguments, superstring expects the input set on its
standard input, one string per line (IOW the strings cannot contain a
newline; apart from that, they are manipulated as sequences of bytes,
without any encoding) and prints the resulting superstring (terminated
by a newline) on its standard output. It also prints some debug
messages to standard error. superstring can also be run with a -t
parameter, in which case it runs the algorithm on a few hardcoded test
cases and checks the results.