Building a constraint programming solver in Julia

Can you give me an example of how this would look like? This is an ongoing series so at the moment my solver can't solve much :D

I think the idea of a constraint solver is you begin with a constraint; Test(X), where X is in some large space S and Test is some expression yielding true or false.

Then the solver uses some magic to parse Test so that it can search S more efficiently than just loop over it. So if the solver is small enough, it seems like 3d animation could be done but the might need to be somewhat aware of that use.

See:

https://en.wikipedia.org/wiki/Constraint_programming

From a quick glance, this and most software doing “constraint programming” deal with discrete problems. Not saying those are useless for mechanical modeling or 3D animation, but those areas typically deal with continuous constraints which require different solvers. My experience in this is limited to a single class in CP 3 years ago so take this with a grain of salt!!

As part of my PhD dissertation, I wrote an "interval valued" constraint solver and a Python internal DSL based originally on miniKanren, that ran it. The intervals handle the floating point in interval chunks, allowing for set-like reasoning and relational constraint programming atop floating point.

I used it to operate on a design space as a single entity, or sliced into finite chunks, and to cut out sub-spaces of infeasible designs by constraint propagation.

I think interval valued relational constraint programming could be extended (and probably already has) for mechanical modeling, and 3D animation, yes. But the utility of a purely discrete constraint solver, such as the one posted here, is maybe less for these types of cases. Interval systems methods (for solving nonlinear equations) have a long history in optimization and motion planning. No reason a constraint solver couldn't be applied there. Similarly, if you have a specific use case, I could try and offer more about constraint solving for it.

I should also note that constraint solving with real or and especially interval data goes back to at least the 80s. I just made use of some low hanging fruit!

You probably want to develop on the Constraint solver instead of just installing, right? Then you can use `] dev https://github.com/Wikunia/ConstraintSolver.jl` You currently try to develop a package which already exists and that doesn't work. Yes the error message is unhelpful.

This is not able to solve linear problems as it's built only for discrete variables in mind. The approach is more removing values from the search space to find a feasible solution whereas LP focuses on optimization. There are problems were both approaches can be used. Currently my solver is quite limited. Stay tuned