Jesse Perla

Trying to track references in issues, the guess is that this is the solution to https://github.com/TuringLang/Turing.jl/issues/1754 or am I missing something? If so, is this PR sufficiently solid that it...

> and then we don’t need extra configuration. Edit: I maybe was remembering some reluctance related to adding more Pkg-related config to the Project.toml as opposed to using it for...

@mlubin I really hope that NLopt gets up with the MOI soon so that I can swap out Ipopt when using juMP Maybe I am misunderstanding, but if you are...

Great @arnavs will prepare a full PR for this with the necessary tests, documentation, etc. At this point, we think that this will require no direct changes to the NLopt...

@arnavs: A few notes: - If we try to make new methods for the `min_objective!`, etc. , we will need to deal with the fact that we cannot construct a...

> but if you're already in the business of designing APIs for nonlinear optimization (as discussed in this thread), That is close to the last thing that I want to...

OK, after sketching out a knitro one that I am hacking together, I propose a few additions if you are in the middle of this stuff: - `verbosity` or `print_level`...

Perfect. Well, to start things off, lets start getting the gradients of the objectve in there were possible! MOI is a good one to start.

If you don't liek the word "sense" then `objective_goal` is another one. e.g. https://www.artelys.com/docs/knitro/3_referenceManual/callableLibraryAPI.html#basic-problem-construction Actually, looks like ipopt MOI wrapper uses `FEASIBILITY_SENSE ` directly. For example, set the objective to...

Just a reminder on this one for API design If you haven't already, and I wasn't sure: The further you get, the more inconvenient it becomes to add.