Default to MUMPS if MA57 is not available

[odow]

This is going to become annoying:

julia> model = Model() do
           return AmplNLWriter.Optimizer(UnoSolverCommand())
       end
A JuMP Model
├ solver: AmplNLWriter
├ objective_sense: FEASIBILITY_SENSE
├ num_variables: 0
├ num_constraints: 0
└ Names registered in the model: none

julia> set_attribute(model, "preset", "ipopt")

julia> set_attribute(model, "print_solution", "yes")

julia> @variable(model, x)
x

julia> @objective(model, Min, (x - 1)^2)
x² - 2 x + 1

julia> optimize!(model)
libc++abi: terminating due to uncaught exception of type std::invalid_argument: Linear solver name is unknown

The solution is

julia> set_attribute(model, "linear_solver", "MUMPS")

julia> optimize!(model)
Problem /var/folders/bg/dzq_hhvx1dxgy6gb5510pxj80000gn/T/jl_VfYlG8/model.nl_scaled_equalityconstrained_boundrelaxed
1 variables, 0 constraints

Options:
... etc

But better options would be to:

• default to MUMPS if MA57 is not available and MUMPS is
• improve the error message to print the name of the available solvers and to tell the user to set the linear_solver option

[cvanaret]

I will change the default solvers based on what's available.

[odow]

Maybe the same with the presets? If we haven't compiled BQPD then make -preset ipopt the default?

[cvanaret]

Quick question: you used the command set_attribute(model, "preset", "ipopt") in your code. This should be an attribute of the solver/optimizer, rather than of the model.
Would it be possible to have something like:

solver = AmplNLWriter.Optimizer(UnoSolverCommand())
set_attribute(solver, "preset", "ipopt")
...

?

[odow]

There are multiple ways that you can pass options to a solver: https://jump.dev/JuMP.jl/stable/manual/models/#solver_options

None of this needs to concern Uno though. We'll pack them into the appropriate place in the CLI.

[cvanaret]

Setting the default subproblem solvers + improving the error messages was addressed in commit d8571b6.
I will address the default preset in a separate issue.

[odow]

Great. I confirm that this worked in the latest build of Uno_jll