Bug with right-hand side

[tmigot]

I am trying to update OptimizationProblems.jl with the new JuMP interface, and run into the following issue for some problems:

using ADNLPModels, JuMP, NLPModels, NLPModelsJuMP

default_nvar = 100
T = Float64

function ju_hs100(args...; kwargs...)
  nlp = Model()
  x0 = [1, 2, 0, 4, 0, 1, 1]
  @variable(nlp, x[i = 1:7], start = x0[i])

  @constraint(nlp, 282 - 7 * x[1] - 3 * x[2] - 10 * x[3]^2 - x[4] + x[5] ≥ 0)
  @constraint(nlp, 196 - 23 * x[1] - x[2]^2 - 6 * x[6]^2 + 8 * x[7] ≥ 0)
  @constraint(nlp, 127 - 2 * x[1]^2 - 3 * x[2]^4 - x[3] - 4 * x[4]^2 - 5 * x[5] ≥ 0)
  @constraint(nlp, -4 * x[1]^2 - x[2]^2 + 3 * x[1] * x[2] - 2 * x[3]^2 - 5 * x[6] + 11 * x[7] ≥ 0)

  @objective(
    nlp,
    Min,
    (x[1] - 10)^2 +
    5 * (x[2] - 12)^2 +
    x[3]^4 +
    3 * (x[4] - 11)^2 +
    10 * x[5]^6 +
    7 * x[6]^2 +
    x[7]^4 - 4 * x[6] * x[7] - 10 * x[6] - 8 * x[7]
  )

  return nlp
end

function ad_hs100(; n::Int = default_nvar, type::Type{T} = Float64, kwargs...) where {T}
  function f(x)
    n = length(x)
    return (x[1] - 10)^2 +
           5 * (x[2] - 12)^2 +
           x[3]^4 +
           3 * (x[4] - 11)^2 +
           10 * x[5]^6 +
           7 * x[6]^2 +
           x[7]^4 - 4 * x[6] * x[7] - 10 * x[6] - 8 * x[7]
  end
  x0 = T[1, 2, 0, 4, 0, 1, 1]
  function c!(cx, x)
    cx[1] = 282 - 7 * x[1] - 3 * x[2] - 10 * x[3]^2 - x[4] + x[5]
    cx[2] = 196 - 23 * x[1] - x[2]^2 - 6 * x[6]^2 + 8 * x[7]
    cx[3] = 127 - 2 * x[1]^2 - 3 * x[2]^4 - x[3] - 4 * x[4]^2 - 5 * x[5]
    cx[4] = -4 * x[1]^2 - x[2]^2 + 3 * x[1] * x[2] - 2 * x[3]^2 - 5 * x[6] + 11 * x[7]
    return cx
  end
  lcon = T[-282.0, -196.0, 0.0, 0.0]
  ucon = T(Inf) * ones(T, 4)
  return ADNLPModels.ADNLPModel!(f, x0, c!, lcon, ucon, name = "hs100"; kwargs...)
end

nlp_jump = MathOptNLPModel(ju_hs100())
nlp_ad = ad_hs100()

using Test, LinearAlgebra
@test nlp_jump.meta.lcon == nlp_ad.meta.lcon
@test nlp_jump.meta.ucon == nlp_ad.meta.ucon
x1 = rand(7)
@test norm(cons(nlp_jump, x1) - cons(nlp_ad, x1), Inf) # doesn't work

#=
julia> cons(nlp_jump, x1)
4-element Vector{Float64}:
 -11.9220041710068
  -0.969071724901412
   3.424652495769024
 119.9921154237291

julia> cons(nlp_ad, x1)
4-element Vector{Float64}:
 270.0779958289932
 195.03092827509857
 119.9921154237291
   3.4246524957690236
=#

Any idea @amontoison ?

[tmigot]

Sorry, not a good question. JuMP is doing some permutations with the order of constraints that I am not 100% getting...