Constraint Programming and a Usability Quest

Abstract

In 2004, Jean-Francois Puget presented [2] an analysis of the “simplicity of Use” of Constraint Programming from which he articulated a series of challenges to make Constraint Programming systems accessible and easier to use. The core of the argument was a contrast between mathematical programming and constraint programming tools. Mathematical programming adopts a model and run paradigm, rely on a simple vocabulary to model problems (i.e., linear constraints), support standard formats for sharing models and benefit from extensive documentation on how to model [5]. Constraint programming features a model and search paradigm, rich modeling languages with combinatorial objects and has a distinctive flavor of programming. While it can be construed as CP’s Achilles’ heel, it is also its most potent strength and is supported by modeling aids [3,4]. The very existence of sophisticated parameter tuning solutions for SAT solvers and Math Programming solvers to determine ideal parameters (e.g., ParamILS [1]) certainly cast a major shadow on the potency of the model and run mantra that is evolving into model and search for the right parameters.