A novel approach to string constraint solving

Abstract

String processing is ubiquitous across computer science, and arguably more so in web programming. In order to reason about programs manipulating strings we need to solve constraints over strings. In Constraint Programming, the only approaches we are aware for representing string variables—having bounded yet possibly unknown size—degrade when the maximum possible string length becomes too large. In this paper, we introduce a novel approach that decouples the size of the string representation from its maximum length. The domain of a string variable is dynamically represented by a simplified regular expression that we called a dashed string, and the constraint solving relies on propagation of information based on equations between dashed strings. We implemented this approach in G-Strings, a new string solver—built on top of Gecode solver—that already shows some promising results.