rely on generic MPB interface (#81)

src/mpb_interface.jl

@@ -2,119 +2,6 @@
 # Modeled after Miles Lubin's AmplMathProgInterface
 
 import MathProgBase
-import MathProgBase.SolverInterface
 
-export NLPtoMPB
-
-
-type CUTEstNLPEvaluator <: SolverInterface.AbstractNLPEvaluator
-  nlp :: CUTEstModel
-end
-
-MathProgBase.initialize(::CUTEstNLPEvaluator, requested_features) = nothing
-MathProgBase.features_available(::CUTEstNLPEvaluator) = [:Grad, :Jac, :JacVec, :HessVec, :Hess]
-
-MathProgBase.eval_f(d::CUTEstNLPEvaluator, x) = obj(d.nlp, x)
-
-MathProgBase.eval_g(d::CUTEstNLPEvaluator, g, x) = copy!(g, cons(d.nlp, x))
-
-MathProgBase.eval_grad_f(d::CUTEstNLPEvaluator, g, x) = copy!(g, grad(d.nlp, x))
-
-function MathProgBase.jac_structure(d::CUTEstNLPEvaluator)
-  rows, cols, vals = jac_coord(d.nlp, d.nlp.meta.x0)
-  return rows, cols
-end
-
-function MathProgBase.hesslag_structure(d::CUTEstNLPEvaluator)
-  rows, cols, vals = hess_coord(d.nlp, d.nlp.meta.x0, y=ones(d.nlp.meta.ncon))
-  return rows, cols
-end
-
-function MathProgBase.eval_jac_g(d::CUTEstNLPEvaluator, J, x)
-  rows, cols, vals = jac_coord(d.nlp, x)
-  copy!(J, vals)
-end
-
-function MathProgBase.eval_jac_prod(d::CUTEstNLPEvaluator, y, x, w)
-  y = jprod(d.nlp, x, w)
-end
-
-function MathProgBase.eval_jac_prod_t(d::CUTEstNLPEvaluator, y, x, w)
-  y = jtprod(d.nlp, x, w)
-end
-
-function MathProgBase.eval_hesslag_prod(d::CUTEstNLPEvaluator, h, x, v, σ, μ)
-  result = hprod(d.nlp, x, v, μ, obj_weight=σ)
-  copy!(h, result)
-end
-
-function MathProgBase.eval_hesslag(d::CUTEstNLPEvaluator, H, x, σ, μ)
-  rows, cols, vals = hess_coord(d.nlp, x, y=μ, obj_weight=σ)
-  copy!(H, vals)
-end
-
-MathProgBase.isobjlinear(d::CUTEstNLPEvaluator) = d.nlp.meta.nlo == 0
-
-MathProgBase.isobjquadratic(d::CUTEstNLPEvaluator) = false
-
-# MathProgBase.isconstrlinear(d::CUTEstNLPEvaluator, i) = i in d.nlp.meta.lin
-
-function loadcutestproblem!(m::MathProgBase.AbstractMathProgModel, nlp::CUTEstModel)
-  sense = nlp.meta.minimize ? :Min : :Max
-  MathProgBase.loadproblem!(m, nlp.meta.nvar, nlp.meta.ncon, nlp.meta.lvar,
-    nlp.meta.uvar, nlp.meta.lcon, nlp.meta.ucon, sense, CUTEstNLPEvaluator(nlp))
-  MathProgBase.setwarmstart!(m, nlp.meta.x0)
-
-  nnlvar = max(nlp.meta.nlvc, nlp.meta.nlvo)
-  narcvar = nlp.meta.nwv
-  nlinvar = nlp.meta.nvar - (nnlvar + narcvar + nlp.meta.nbv + nlp.meta.niv)
-  nbinvar = nlp.meta.nbv
-  nintvar = nlp.meta.niv
-
-  v = fill(:Cont, nlp.meta.nvar)
-  # First populate Table 4
-  varidx = 1
-  for i = 1:(nlp.meta.nlvb - nlp.meta.nlvbi)
-    varidx += 1
-  end
-  for i = 1:nlp.meta.nlvbi
-    v[varidx] = :Int
-    varidx += 1
-  end
-  for i = 1:(nlp.meta.nlvc - (nlp.meta.nlvb + nlp.meta.nlvci))
-    varidx += 1
-  end
-  for i = 1:nlp.meta.nlvci
-    v[varidx] = :Int
-    varidx += 1
-  end
-  for i = 1:(nlp.meta.nlvo - (nlp.meta.nlvc + nlp.meta.nlvoi))
-    varidx += 1
-  end
-  for i = 1:nlp.meta.nlvoi
-    v[varidx] = :Int
-    varidx += 1
-  end
-  # Now populate Table 3
-  varidx += narcvar + nlinvar
-  for i = 1:nbinvar
-    v[varidx] = :Bin
-    varidx += 1
-  end
-  for i = 1:nintvar
-    v[varidx] = :Int
-    varidx += 1
-  end
-  @assert varidx == nlp.meta.nvar + 1
-  # Set variable types
-  if any(vtype -> vtype == :Int || vtype == :Bin, v)
-      MathProgBase.setvartype!(m, v)
-  end
-end
-
-@doc (@doc NLPModels.NLPtoMPB)
-function NLPtoMPB(nlp :: CUTEstModel, solver :: MathProgBase.AbstractMathProgSolver)
-  model = MathProgBase.NonlinearModel(solver)
-  loadcutestproblem!(model, nlp)
-  return model
-end
+MathProgBase.features_available(::NLPModelEvaluator{CUTEstModel}) =
+  [:Grad, :Jac, :JacVec, :HessVec, :Hess]

test/test_mpb.jl

@@ -1,3 +1,8 @@
+using Ipopt
+using MathProgBase
+using NLPModels
+using CUTEst
+
 # pass a CUTEstModel to IPOPT
 nlp = CUTEstModel("HS6")
 show(nlp.meta)