Oscar Dowson
Oscar Dowson
I now have https://github.com/jump-dev/MiniZinc.jl. This writes the JuMP model into a flattened form of MiniZinc, and we then call out to different solvers. The benefit of this is that it...
Bill says > i assume that MOI++ including constraints you're working on will support reified constraints so that in principle we might > support something like > alldiff(x1,x3,x5) or alldiff(x2,x4,x6)...
Initial support for the sets has been merged, and MiniZinc.jl is being registered: https://github.com/JuliaRegistries/General/pull/62010
MILP bridges are largely done. I haven't done Path and Cumulative because they're quite complicated, but Cbc, HiGHS, and Gurobi are all passing tests with the new constraint programming sets!...
I've started looking into reified constraints: https://github.com/jump-dev/MiniZinc.jl/pull/13
Blog post is: https://jump.dev/announcements/2022/07/12/constraint-programming/
This issue is pretty close to completion. One blocker is deciding whether to add the Reified set. I've added Reified to MiniZinc.jl, but adding to MOI would add a lot...
There are some related points in https://github.com/jump-dev/MathOptInterface.jl/issues/525
Main blockers for this are that it's limited to first-order derivatives, so we'd need nested AD to compute second-order. I might be missing something, but it also seems setup to...
I took a look at this, and I don't see a need to integrate this into JuMP/MOI. It's not hard to write in user-space: ```Julia using JuMP import ChainRules import...