Oscar Dowson

We now have bridges from ScalarAffine and ScalarQuadratic to ScalarNonlinearFunction, so solvers who don't exploit linear/quadratic functions are much simpler. If solvers do exploit, then they'll need specialized support anyway....

I propose closing this in favor of #2180. I don't see us changing this anytime soon, and the complexity of adding a new function type to MOI makes me not...

Sure, but making them a first-class part of MOI and JuMP is a lot more work. And I don't really want more functions that solvers can opt-into with a bridge...

This is an item in https://github.com/jump-dev/MathOptInterface.jl/issues/2180, so I think we can close this issue. I don't see us making progress anytime soon. It would be far too disruptive (even though...

Sorry, @matbesancon, yet another issue closed without resolution. At least this one doesn't have >100 comments...

A good example for this would be local solver's `list` and `set` variable types: https://www.localsolver.com/docs/last/modelingfeatures/collectionvariables.html I haven't worked through the details, but perhaps `x in List(10)` an `x in Partition()`...

I wonder if we should close this issue as out-of-scope. Supporting anything other than `VariableIndex` is a huge change, and we have the potential to either develop things as a...

To clarify, the closing of this: I talked to @dourouc05 yesterday, and we agreed that adding something other than `VariableIndex` is a big change. The current suggestion for "interval" variables...

ScalarNonlinearFunction is merged. I'll hold off closing this until we add VectorNonlinearFunction.

Closing as done. We know have `ScalarNonlinearFunction` and `VectorNonlinearFunction`.