Design and Implementation of a Concurrent Logic Programming Language with Linear Logic Constraints

Prolog is originally rooted in logic with the elegant mapping: “programs = formulas, execution = proof search”. Constraint logic programming extends Prolog to program in a richer structure than mere Herbrand terms. The underlying constraint solving engine requires either to be built-in or to add coroutines mechanisms that are not in the scope of logical reading. Implementations add other non-logical features, like assert/retract, and mutable variables, to mimic imperative programming style. The concurrent constraint programming language enjoys logical semantics and is expressive enough to describe constraint propagators [Sar93]. Agents tell constraints as messages and are synchronised by asking whether the messages entail some constraints. The execution pursues once the guard is entailed. The suspension is therefore a transient state, captured by using linear-logic implication ( [Fag01]. Furthermore, reading constraints as resources in linear logic allows semantics to capture the non-monotonous traits of imperative programming like mutability. LCC enjoys the mapping “programs = linear-logic formulas, execution = logical deduction”: observables are the logical consequences of a program, by opposition to the logical resolution in the Prolog settings. My thesis aims at designing a practical language as close as possible to the linear concurrent constraint (LCC) theory. The main contribution is a new operational semantics which behaves as an angelic scheduler with a tractable algorithmic complexity. This operational semantics is sound and complete with respect to the logical semantics and allows the construction of a rich language over a very simple kernel. The second section presents the kernel, the third describes the operational semantics and the last section describes the state of the implementation work and its perspectives.