NuCS vs Choco: A Pure-Python Constraint Solver Meets a JVM Veteran

NuCS, a new pure-Python constraint solver, goes head-to-head in a performance benchmark against Choco, a veteran solver built on the Java Virtual Machine. The core architectural difference is that NuCS is limited to bound consistency using simple intervals, while Choco can perform stronger arc consistency by representing domains with holes. When running identical models, the Numba-accelerated Python code in NuCS surprisingly matches and can even surpass the speed of the heavily optimized JVM solver on larger problems. However, when models differ, the best solver becomes a strategic trade-off, with Choco's powerful filtering winning on some problems and NuCS's cheaper consistency and modeling flexibility proving superior on others.