Software developers are an ideal channel for the distribution of discrete optimization (DO) technology. Unfortunately, including even basic optimisation functionality in an application currently requires the use of an entirely separate paradigm with which most software developers are not familiar. We suggest an alternative interface to DO designed to overcome this barrier. The interface allows an optimisation problem to be defined in terms of procedures rather than decision variables and constraints. Optimisation is seamlessly integrated into a wider application through automatic conversion between this definition and a conventional model solved by an external solver.

The core of the method is translating procedural object oriented code into constraints. This translation is also useful for other applications such as symbolic execution of the code, e.g. in Congolic testing. We discuss how to create the best translation of procedural code, better than existing DO and SMT approaches, and how we can translate bounded loops more effectively. Finally we briefly describe a global constraint for reaching definitions, which encapsulates the essential problem of converting procedural code to constraints.

This work is joint with Kathryn Francis.