JaCoP VS Choco

I used MiniZinc to model a scheduling problem for a customer which used constraints to block off already scheduled time, holiday time, non-work hours, etc. and it also had soft constraints for ensuring that blocks of time would be prioritized first to ensure that preferred representatives would get to a work with a given customer first based on previous relationship and skills. Once we translated the modelled constraints from MiniZinc to JaCoP in Java, the solution was able to generate hundreds of solutions per second. And note that it was able to do this minute level resolution such that no blocks of time were wasted because of scheduler limitations.

There are quite a few out there. I played with gecode as well as https://choco-solver.org/ which I found intuitive and fun to use.

I personally doesn't get surprised by POSIX to consider it's violating this "least surprise" amount.

Instead, for example Java has given me much more surprises, and at catastrophic levels. I was using Choco Solver [0] back in the day, and I created two instances of it, attached to different classes. Which is perfectly normal, right?

Somehow they've cross linked between these two instances, affected the results they have computed, and created persistent memory leaks which needed system reboots to claim back. Java should be immune to that, but no.

Preventing that needed to run only one instance of Choco, which limited my performance greatly. Luckily, the system had a queue/consumer structure, so running only one didn't need extensive changes.

[0]: https://choco-solver.org/

Sudoku is one of these types of problems, but it has very simple rules, so it’s a nice playground to try out different ways to solve graph colouring problems. This post outlines a solution using constraint programming with choco solver.

- Choco (https://choco-solver.org/): state of the art trailing solver, heavily tested, easy to get started with and actively maintained. Written in Java.