Développement d'applications avec Objective CAML by E. Chailloux, P. Manoury and B. Pagano, O'Reilley, 2003

Abstract

This book describes theoretical results about AnsProlog * that have been obtained over the past decade.AnsProlog * or Prolog with Answer Sets 1 is a variation of the Prolog programming language, and extends the language by allowing clauses of the form:in the program.The L i 's are the literals (or atoms) of the Prolog language and may be supplied with a prefix ¬ sign, indicating the negation of a literal, while the prefix of not indicates negation as failure.Hence the semantics of the clause (C) may be read as follows: if all the literals L1, . . ., L m are true and all the literals L m+1 , . . ., L n can be safely assumed false then at least one of the literals L 1 , . . ., L k is true.(The actual semantics of each AnsProlog * program will be defined in terms of the Herbrand Universe of ground terms and the Herbrand Base of ground atoms.)The book takes the approach that the clause (C) is the most general form of a clause in the AnsProlog * language and so various subclasses of AnsProlog * can be defined by restricting this clause.For example: an AnsProlog -not program is when none of the clauses of a program contain the prefix not.In this respect, the book discusses the tractability, the complexity, the expressibility of the various subclasses of AnsProlog * based on the premise that AnsProlog * is both an excellent knowledge representation language and that it has a number of advantages over the Prolog language implementations based on SLDNF.For example, the ordering of goals within a clause and the ordering of clauses within a Prolog program affects whether a solution can or cannot be found (i.e. the program might get into an infinite loop); but not this is not the case within an AnsProlog * program.The reason being is that the semantics of an implementation of the AnsProlog * language can be thought of as allowing all models of the program to exist and then by using the clauses within the program, to impose restrictions on these models.The actual model(s) produced can then be interpreted in either a bi-valent (where a ground atom is either true or false) fashion or a tri-valent (where a ground atom is either true, false or unknown) fashion.The implementation algorithms describing how to restrict the models are described in Chapter 7 and two systems implementing the AnsProlog * language (and various subclasses), viz: (i) lparse+smodels and (ii) dlv are discussed in Chapter 8.The lparse+smodels program produces the stable models (or bi-valent) implementation, while the dlv produces the well-founded models (or tri-valent) implementation.Baral does note that both systems are under development and so implying that Chapter 8 may be out of date within a few years.However this aspect is compensated by Baral having a website www.baral.us/bookonewhere hypertext links to both the two systems and an errata/additional notes for the book are presented.On the application side, the book is peppered with many examples and simple programs illustrating the current point being made in the text.For example: how various forms of the 1 AnsProlog * is sometimes called A-Prolog in the literature.