quadform objective does not work with OSQP

[a7b]

I was trying to solve the following QP using OSQP:

λs = [0, 0.01, 0.1, 1, 10, 100, 200, 300, 1000]
for i in eachindex(λs)
    n = size(mat, 1) - 1
    x = Variable(n)
    constraints = [sum(x) == 100000, x >= 0]
    p = minimize( 1/2 * quadform(x, Σ) - λs[i] * dot(μr, x), constraints)
    solve!(p, OSQP.Optimizer)
end

but I get the error

UnsupportedConstraint: MathOptInterface.VectorAffineFunction{Float64}-in-MathOptInterface.SecondOrderCone constraints are not supported by the
solver you have chosen, and we could not reformulate your model into a
form that is supported.

Both of the constraints can be written in the form required by OSQP but it seems that Convex.jl is unable to do this. It would be nice if this could be fixed.

[odow]

What is the output of import Pkg; Pkg.status()?

[odow]

A smaller example:

julia> begin
           using Convex, OSQP
           x = Variable(2)
           p = minimize(quadform(x, [1 0; 0 1]))
           solve!(p, OSQP.Optimizer)
       end
ERROR: UnsupportedConstraint: `MathOptInterface.VectorAffineFunction{Float64}`-in-`MathOptInterface.SecondOrderCone` constraints are not supported by the
solver you have chosen, and we could not reformulate your model into a
form that is supported.

To fix this error you must choose a different solver.

I expected this to work. I haven't looked at the details yet.

[ericphanson]

I think this is the thing where Convex unconditionally reformulates things, here quadratic to SOC instead of letting MOI do it so it can be appropriate to the solver

[odow]

Oooh. This is actually a duplicate of #384

I'll close this issue to keep the discussion focused on #384

[a7b]

I see, it is a duplicate. Thanks!