MPI-I-94-105

We give an implementation of an incremental construction algorithm for convex hulls in $\IR^d$ using {\em Literate Programming} and {\em LEDA} in C++. We treat convex hulls in arbitrary dimensions without any non-degeneracy assumption. The main goal of this paper is to demonstrate the benefits of the literate programming approach. We find that the time we spent for the documentation parts is well invested. It leads to a much better understanding of the program and to much better code. Besides being easier to understand and thus being much easier to modify, it is first at all much more likely to be correct. In particular, a literate program takes much less time to debug. The difference between traditional straight forward programming and literate programming is somewhat like the difference between having the idea to a proof of some theorem in mind versus actually writing it down accurately (and thereby often recognizing that the proof is not as easy as one thought).