chriscoey / glpkmathproginterface.jl

Interface between the GLPK.jl wrapper and MathProgBase.jl

GLPKMathProgInterface.jl

Interface between the GLPK.jl wrapper and MathProgBase.jl. With this package, you can use the GLPK solver in MathProgBase.jl

The linear programming solver accepts the following optional keywords:

• method: accepts :Simplex (default), :Exact, :InteriorPoint
• presolve: accepts false (default) or true. If set to true, status reporting will always report :Undefined in case of presolver failure (i.e. also for infeasible problems).

Example:

using MathProgBase
using GLPKMathProgInterface
sol = linprog([-1,0], [2 1], '<', 1.5, GLPKSolverLP(method=:Exact, presolve=true))

Additionally, all low-level parameters which can be passed to the GLPK problem objects, such as msg_lev, tol_bnd etc., can also be passed as keyword arguments, for finer control. Refer to the GLPK.jl documentation for more details on the available parameters. Invalid option names will be ignored with a warning (note that the set of valid options is different for the :Simplex and :InteriorPoint methods). This is a low-level feature, therefore options with invalid arguments may cause the solver to crash.

For example, to get a more verbose output in the above case, write:

sol = linprog([-1,0], [2 1], '<', 1.5, GLPKSolverLP(method=:Exact, presolve=true, msg_lev=GLPK.MSG_ON)

The mixed integer programming solver accepts the following optional keywords:

• presolve: accepts false (default) or true. If set to true, status reporting will always report :Undefined in case of presolver failure (i.e. also for infeasible problems). If set to false, presolving will be performed via the simplex method. Example:

using MathProgBase
using GLPKMathProgInterface
sol = mixintprog(-[5,3,3], [1 8 2], '<', 9, :Int, 0, 1, GLPKSolverMIP(presolve=true))

Additionally, as in the case of linear programming solver (see above), all low-level parameters which can be passed to the GLPK problem objects can also be passed as keyword arguments to the GLPKSolverMIP constructor for finer control. The constructor accepts arguments both for the MIP solver and the LP simplex solver (which may be used by the MIP solver internally), so any field of an IntoptParam or of a SimplexParam can be passed; in case of common options (e.g. msg_lev) the parameters are set for both solvers. The only exceptions are the callback pointers cb_func and cb_info, which can not be passed in this way, since the callback interface is only accessible via MathProgBase. As in the LP case, invalid options are ignored with a warning. This is a low level feature, and passing invalid parameter arguments may crash the solver.