Information-theoretic Complexity of Program Specifications
Neal S. Coulter · The Computer Journal · 1987
Various models and metrics that are based on cognitive theories and assumptions have been proposed in an effort to formalise the concept of programming difficulty. The model and metric proposed here are, instead, based on a problem-space formalisation and an information-theoretic measure (entropy) that provides the complexity of a program specification; this metric can be applied before the specification's implementation is complete. The problem space for a program specification is the set of possible state descriptions that could arise from the execution of the program on that input, as ascertainable from the specification. We show how to define the entropy metric based on the problem space definition, and demonstrate the applicability of this metric to modularity problems.