Contributions to the declarative approach to debugging Prolog programs

• Logic programs have the characteristic that their intended semantics can be expressed declaratively or operationally. Since the two semantics coincide, programmers may find it easier to adopt the declarative view when writing the program. But this causes a problem when the program is to be debugged. The actual semantics of a logic program is dependent on the specific implementation on which the program is run. Although the actual semantics is of operational nature it is usually different from the theoretical operational semantics. Hence debugging may require a comparison of the actual (operational) semantics of a program and its intended declarative semantics.The idea of declarative debugging, first proposed by Shapiro under the term algorithmic debugging, is a methodology which leads to detecting errors in a logic program through knowledge about its intended declarative semantics. Current Prolog systems do not employ declarative diagnosis as an alternative to the basic tracer. This is partly due to the fact that the Shapiro’s declarative debugging system only dealt with pure Prolog programs, and partly due to practical limitations of the suggested methods and algorithms. This thesis consists of three papers. In these papers we point out practical problems with the use of basic declarative debugging systems, and present methods and algorithms which make the framework applicable to a wider range of Prolog programs. We introduce the concept of assertion that can ease communication between the user and the debugging system by reducing the number of necessary interactions, and introduce new debugging algorithms which are adapted to this extended notion. Further, we extend the basic debugging scheme to cover some built-in features of Prolog, and report on practical experience with a prototype declarative debugging system which incorporates the extensions.

Ämnesord

NATURVETENSKAP  -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Computer Sciences (hsv//eng)

Publikations- och innehållstyp

vet (ämneskategori)

lic (ämneskategori)