Oscar Dowson

This is still a problem on Julia v1.8.0-rc4. It's very weird that the only test failing is `test_linear_transform`. Taking a deeper look.

It's a very surprising failure ```julia import Ipopt const MOI = Ipopt.MOI model = Ipopt.Optimizer() x = MOI.add_variables(model, 2) f = 1.0 * x[1] + x[2] MOI.add_constraint(model, f, MOI.GreaterThan(1.0)) MOI.add_constraint(model,...

So this isn't a bug in JuMP, MOI, or Ipopt.jl: ```julia julia> import Ipopt julia> function eval_jac_g( x::Vector{Float64}, rows::Vector{Int32}, cols::Vector{Int32}, values::Union{Nothing,Vector{Float64}}, ) if values === nothing rows .= Cint[1, 1,...

@Fisheryu1234 as a rule of thumb, you just need `libpardiso.so` to be somewhere that `dlopen` can find it, and then set `set_optimizer_attribute(model, "linear_solver", "pardiso")`. The specifics of how to compile...

More recent versions of Ipopt provide the `hsllib` option, https://coin-or.github.io/Ipopt/OPTIONS.html#OPT_hsllib, but I haven't used it. You should be able to set it as the absolute path: ```julia set_optimizer_attribute(model, "hsllib", "/usr/local/lib/libhsl_ma27.dylib")...

> which the opposite should be true Not necessarily. The performance of the linear solvers can vary a lot between problems. See, e.g., https://arxiv.org/pdf/1909.08104.pdf.

Nice reduction in allocations: Current ```julia julia> @time bench_opf_acp(; case = "pglib_opf_case2000_goc.m") # [ ... ] FEASIBLE_POINT 13.729058980941772 15.410575 seconds (21.57 M allocations: 946.675 MiB, 5.64% gc time) (nlp-jump-examples) pkg>...

The recommended way to do this is just to set a new objective. We supported `ScalarCoefficientChange` because many solvers provide an efficient way of updating a linear objective coefficient. In...

So coming back to this, I'm still hesitant to add this to MOI because we don't have a solver that efficiently supports it. One fix to resolve the dualization issue...

cc @guilhermebodin thoughts?