Logic meta-programming facilities in 'LOG

Abstract

Abstract.A meta-level extension of a logic programming language is presented. The resulting language, called 'LOG (read quote-log), provides meta-programming facilities similar to those of Prolog while preserving a declarative logical semantics. It also offers new meta-programming opportunities as compared with Prolog due to its ability to treat whole programs, i.e. sequences of clauses, as data objects. The extension basically consists in defining a suitable naming scheme. It associates two different but related meta-representations with every syntactic object of the language, from characters to programs. The choice of the double metarepresentation is motivated by both the user and the implementation viewpoints. All Prolog built-in meta-predicates can be redefined as 'LOG programs by exploiting the new naming scheme. Then some syntactic sugar is added to make the language more concrete. Some examples are given, in particular to show the ability of the language to deal with programs as data.

On the declarative and procedural semantics of definite metalogic programs

“…The semantics of Prolog meta-programs has received some attention, the work of Hill and Lloyd [9] being particularly influential. The design of metalogic programming languages has been a slow process but there have now been several proposals including Gödel [10], Reflective Prolog [6] and Quote-Prolog [4].…”

“…The declarative semantics of Gödel [10] are based on the ground representation of [9]. The proposal of Cervesato and Rossi [4] also uses the ground representation but includes primitive as well as structured names. In fact, every string has both a primitive and a structured name.…”

“…The primitive name is in effect simply the quotation of the string itself. The semantics are not discussed in detail in [4] and the language does not seem to use reflection.…”

In Praise of Impredicativity: A Contribution to the Formalization of Meta-Programming

“…The article van Harmelen (1992) argues that naming relations can encode, together with an object formula, pragmatic and semantic information resulting in a more efficient (meta-language) version of the original formula. Naming relations have indeed been defined for achieving such "compilations" what explains their large number and, as a consequence, the large number of formalizations of meta-programming in first-order logic: metaProlog (Bowen 1985;Bowen and Weinberg 1985), MOL (Eshghi 1986), the language proposed by Barklund in the article Barklund (1989), Reflective Prolog (Costantini and Lanzarone 1989;Costantini and Lanzarone 1994), R-Prolog * (Sugano 1989;Sugano 1990), 'LOG (spoken "quotelog") (Cervesato and Rossi 1992), Gödel (Hill and Lloyd 1994), the language proposed by Higgins in the article Higgins (1996), and the generalization of Reflective Prolog proposed in the article Barklund et al (2000).…”

“…'LOG (Cervesato and Rossi 1992) has two naming relations associating "two different but related meta-representations with every syntactic object of the language, from characters to programs," "a constant name and a structured ground term, called the structural representation." This double naming relation generalizes those of Martelli and Rossi (1988) and Rossi (1989; that apply only to programs.…”