MPSReader

Deprecated in favor of MathOptFormat.

MPSReader is a Julia package that transform a .mps file (MPS format) into a model using JuMP.

The MPS format represents problem of the following form :
optimize cᵀ x + c₀
subject to L ≤ Ax ≤ U
l ≤ x ≤ u

optimize stands for minimize or maximize, MPSReader will use minimize as default but it can be specified in Free MPS Format. c₀ ∈ ℝ is a constant term and c ∈ ℝⁿ is the cost function.

L and U are vectors of constraints bounds. By default the lower bound of a constraint is -Inf and the upper bound is +Inf. l and u are vectors of bounds of x. By default the lower bound of a variable is 0 and the upper bound is +Inf. x can have simultaneously binary elements, integer elements and continuous elements. Semi-continuous variables are not supported and using MPSReader on a file containing it will throw an error.

This package exports 2 functions: readmps() and mpstomodel()

Prerequisites

julia 1.0.0
MathOptInterface v0.8.1
JuMP v0.19.0
LinearAlgebra

If you are using julia 1.0.0 or newer you can install all the needed package as follow:

julia> using Pkg
julia> Pkg.add("MathOptInterface")
julia> Pkg.add("LinearAlgebra")
julia> Pkg.develop("JuMP")

Installing

TODO

Running the tests

julia> include("examples/test.jl")

You should get 7 JuMP models and the expected output is in examples/Test_output.txt

Usage

julia> fixed = true #false
julia> varTypes, bounds, objsense, c, c0, A, b, conTypes = readmps("mympsfile.mps", fixed)
julia> myModel, myVariables = mpstomodel("mympsfile.mps", mySolver, fixed) # mySolver = GLPK.Optimizer

varTypes is a vector of booleans, elements are true if the variable is continuous and at false if it is an integer or a boolean. bounds is a vector of pairs, the first is the lower bound of the variable and the second the upper bound. objsense is a Symbol for the optimizing sense, (:min by default). conTypes is a vector that gives the type of the constraint, 1 for ≤, 2 for = and 3 for ≥. Each constraint having a lower and an upper bound is duplicated into a constraint with a lower bound and another with an upper bound.

myModel is the model in minimization (maximization) with constraints of type ≥ (≤) and =. Therefore it should have this form:
minimize cᵀ x + c₀
subject to Ax ≥ b
Aeq x = beq
l ≤ x ≤ u
or
maximize cᵀ x + c₀
subject to Ax ≤ b
Aeq x = beq
l ≤ x ≤ u

By default fixed is true but if you are using a Free MPS Format you can set fixed variable to false. As lpsolve authors we suggest to try the fixed format and if it fails switch for the free format.

Problem Collections

• MIPLIB: https://miplib.zib.de/

References

• MPS file format: http://lpsolve.sourceforge.net/5.5/mps-format.htm