I have actually found that optaplanner often gets better results on hard problems than a lot of constraint solvers do. Part of that has to do with a genius aspect of how it was designed: constraint evaluation is built on top of the drools rule engine algorithm (an evolution of the Rete algorithm).
I think the fact that it is written for the JVM is especially helpful because you can write constraints using JVM libraries, which are a massive boon in some of the very domain specific areas that I’ve worked in. Writing geospatial or RF propagation constraints in a DSL like minizinc is a total nonstarter.
I think the fact that it is written for the JVM is especially helpful because you can write constraints using JVM libraries, which are a massive boon in some of the very domain specific areas that I’ve worked in. Writing geospatial or RF propagation constraints in a DSL like minizinc is a total nonstarter.