Diagnosis of CHIP Programs Using Type Information

The paper presents an approach to diagnosis of CLP programs and a diagnoser for CHIP based on this approach. The objective is to achieve partial correctness of the program w.r.t. a specification of type terms which describe sets of calls and successes of program predicates. Parametric regular term grammars with constraints are used to define the meaning of type terms as sets. The diagnoser compares the type specification given by the user with the type information inferred from the program and localizes all fragments of the program responsible for the discrepancy between the two. No information about error symptoms is used. The type specification is not requested a priori. It is developed “by need” querying the user about the types inferred. A type error is often localized by providing only a small portion of type specification.