RipQP is failing

[abelsiqueira]

Maybe it's RipQP's fault, but we have it on the docs.

julia> using JSOSuite, RipQP
julia> minimize("RipQP", ones(2), [4.0 1.0; 1.0 2.0])
ERROR: ArgumentError: Invalid buffers for SparseMatrixCSC construction n=2, colptr=3-element Vector{Int64}, rowval=4-element Vector{Int64}, nzval=4-element Vector{Float64}
Stacktrace:
  [1] SparseArrays.SparseMatrixCSC{Float64, Int64}(m::Int64, n::Int64, colptr::Vector{Int64}, rowval::Vector{Int64}, nzval::Vector{Float64})
    @ SparseArrays ~/.julia/juliaup/julia-1.10.4+0.x64.linux.gnu/share/julia/stdlib/v1.10/SparseArrays/src/sparsematrix.jl:29
  [2] create_K2(id::RipQP.QM_IntData, D::Vector{…}, Q::SparseArrays.SparseMatrixCSC{…}, A::SparseArrays.SparseMatrixCSC{…}, diag_Q::SparseArrays.SparseVector{…}, regu::RipQP.Regularization{…}, uplo::Symbol, ::Type{…})
    @ RipQP ~/.julia/packages/RipQP/H6krR/src/iterations/solvers/augmented/K2LDL.jl:388
  [3] get_K2_matrixdata
    @ ~/.julia/packages/RipQP/H6krR/src/iterations/solvers/augmented/K2LDL.jl:461 [inlined]
  [4] PreallocatedData(sp::K2LDLParams{…}, fd::RipQP.QM_FloatData{…}, id::RipQP.QM_IntData, itd::RipQP.IterDataCPU{…}, pt::RipQP.Point{…}, iconf::RipQP.InputConfig{…})
    @ RipQP ~/.julia/packages/RipQP/H6krR/src/iterations/solvers/augmented/K2LDL.jl:95
  [5] RipQP.RipQPMonoSolver(QM::QuadraticModels.PresolvedQuadraticModel{…}; itol::InputTol{…}, scaling::Bool, normalize_rtol::Bool, kc::Int64, perturb::Bool, mode::Symbol, ap::RipQP.RipQPMonoParameters{…}, history::Bool, w::SystemW
rite, display::Bool)
    @ RipQP ~/.julia/packages/RipQP/H6krR/src/mono_solver.jl:40
  [6] RipQPMonoSolver
    @ ~/.julia/packages/RipQP/H6krR/src/mono_solver.jl:9 [inlined]
  [7] #RipQPSolver#168
    @ ~/.julia/packages/RipQP/H6krR/src/mono_solver.jl:6 [inlined]
  [8] ripqp(QM::QuadraticModels.QuadraticModel{…}, ap::RipQP.RipQPMonoParameters{…}; ps::Bool, scaling::Bool, history::Bool, display::Bool, kwargs::@Kwargs{…})
    @ RipQP ~/.julia/packages/RipQP/H6krR/src/RipQP.jl:69
  [9] ripqp
    @ ~/.julia/packages/RipQP/H6krR/src/RipQP.jl:47 [inlined]
 [10] ripqp(QM::QuadraticModels.QuadraticModel{…}; mode::Symbol, sp::K2LDLParams{…}, sp2::Nothing, sp3::Nothing, solve_method::PC, solve_method2::Nothing, solve_method3::Nothing, kwargs::@Kwargs{…})
    @ RipQP ~/.julia/packages/RipQP/H6krR/src/RipQP.jl:156
 [11] minimize(::Val{:RipQP}, nlp::QuadraticModels.QuadraticModel{Float64, Vector{Float64}, Matrix{Float64}, Matrix{Float64}}; max_iter::Int64, max_time::Float64, kwargs::@Kwargs{})
    @ JSOSuite ~/.julia/packages/JSOSuite/itKIP/src/solvers/ripqp_solve.jl:61
 [12] minimize
    @ ~/.julia/packages/JSOSuite/itKIP/src/solvers/ripqp_solve.jl:1 [inlined]
 [13] minimize(solver_name::String, nlp::QuadraticModels.QuadraticModel{Float64, Vector{Float64}, Matrix{Float64}, Matrix{Float64}}; kwargs::@Kwargs{})
    @ JSOSuite ~/.julia/packages/JSOSuite/itKIP/src/solve.jl:44
 [14] minimize
    @ ~/.julia/packages/JSOSuite/itKIP/src/solve.jl:38 [inlined]
 [15] #minimize#16
    @ ~/.julia/packages/JSOSuite/itKIP/src/solve-model.jl:249 [inlined]
 [16] minimize(solver_name::String, c::Vector{Float64}, args::Matrix{Float64})
    @ JSOSuite ~/.julia/packages/JSOSuite/itKIP/src/solve-model.jl:239
 [17] top-level scope
    @ REPL[6]:1

[dpo]

The QuadraticModels documentation states that only the lower triangle of H should be given. This works:

julia> minimize("RipQP", ones(2), [4.0 0.0; 1.0 2.0])

[ Info: Solving in Float64 using K2LDL with LDLFactorizationData
[ Info:   iter       obj      rgap      ‖rb‖      ‖rc‖     α_pri      α_du         μ         ρ         δ
[ Info:      0   0.0e+00   0.0e+00   0.0e+00   7.1e-01   0.0e+00   0.0e+00       NaN   1.5e-03   1.5e-03
[ Info:      1  -2.9e-01   5.2e-04   0.0e+00   9.0e-04   1.0e+00   1.0e+00       NaN   1.5e-03   1.5e-03
[ Info:      2  -2.9e-01   6.4e-08   0.0e+00   1.3e-07   1.0e+00   1.0e+00       NaN   1.5e-04   1.5e-04
[ Info:      3  -2.9e-01   8.7e-13   0.0e+00   2.1e-12   1.0e+00   1.0e+00       NaN   1.5e-05   1.5e-05
"Execution stats: first-order stationary"

However, there is clearly a lot of room for improvements here, in QuadraticModels, RipQP and JSOSuite.