An execution scheme for interactive problem-solving in concurrent constraint logic programming languages

Van Emden's incremental queries address the inadequacy of current Prolog-style querying mechanism in most logic programming systems for interactive problem-solving. In the context of constraint logic programming, incremental queries involve adding new constraints or deleting old ones from a query after a solution is found. This paper presents an implementation scheme IQ of incremental queries in Constraint Pandora, which defines a class of non-deterministic concurrent constraint logic programming languages. We use Van Hentenryck and Le Provost's scheme (VHLP-scheme hereafter), a re-execution approach, as a starting point. Re-execution is costly in concurrent languages, in which process creation and inter-process communications are common operations. The main idea of IQ is that the basic trail unwinding operation used in backtracking is more efficient than re-execution in reaching an execution context along a recorded execution path. We modify the conventional trail-unwinding operation in such a way that constraints are used actively to prune the search space. Analysis shows that the IQ-scheme is at least as efficient as the VHLP-scheme in sequential systems and is much more efficient in concurrent systems. We show the feasibility of our proposal by incorporating the IQ-scheme into IFD-Constraint Pandora, an instance of Constraint Pandora supporting interval and finite domain constraint solving. Our preliminary results agree with that of theoretical analysis. (C) 2000 Elsevier Science Ltd. All rights reserved.