Affiliations: Instituto de Investigaciones en Matemáticas
Aplicadas y en Sistemas, Universidad Nacional Autónoma de México
Apdo. 20-726, 01000 México D. F. México. E-mail:
drosenbl@servidor.unam.mx
Abstract: A program-transformation system is determined by a repertoire of
correctness-preserving rules, such as folding and unfolding. Normally, we would
like the folding rule to be in some sense the inverse of the unfolding rule.
Typically, however, the folding rule of logic program transformation systems is
an inverse of a limited kind of unfolding. In many cases this limited kind of
folding suffices. We argue, nevertheless, that it is both important and
possible to extend such a folding so as to be able to fold the clauses
resulting from any unfolding of a positive literal. This extended folding rule
allows us to derive some programs underivable by the existing version of this
rule alone. In addition, our folding rule has applications to decompilation and
reengineering, where we are interested in obtaining high-level program
constructs from low-level program constructs. Moreover, we establish a
connection between logic program transformation and inductive logic
programming. This connection stems fromviewing our folding rule as a common
extension of the existing multipleclause folding rule, on the one hand, and an
operator devised in inductive logic programming, called
"intra-construction",on the other hand. Hence, our folding rule can be
regarded as a step towards incorporating inductive inference into logic program
transformation. We prove correctness with respect to Dung and
Kanchanasut's semantic kernel.
Keywords: Logic program transformation, folding rule, most specific generalization, semantic kernel, decompilation, inductive logic programming
