JuMPConverter.jl

JuMPConverter.jl is a converter for AMPL™ models in .mod or GAMS™ models in .gmx to JuMP models in .jl.

License

JuMPConverter.jl is licensed under the MIT License.

Installation

Install JuMPConverter as follows:

import Pkg
Pkg.add("https://github.com/jump-dev/JuMPConverter.jl")

Use with JuMP

AMPL™ converter

To read an AMPL™ model file.mod with (optional) data in file.dat into a JuMP model, use the following:

using JuMPConverter
model = JuMPConverter.read_from_file("file.mod", "file.dat")

Convert the .mod into .jl

Internally, read_from_file convert the .mod as a .jl file using JuMP to model the problem. To get this .jl file, use the following:

using JuMPConverter
model = JuMPConverter.AMPL.read_model("file.mod")
open("file.jl", "w") do io
    println(io, model)
end

Loading data from a .dat file

AMPL keeps the model structure (.mod) and the data (.dat) in separate files. The emitted Julia file is self-contained: it defines the keyword-argument build_model(; S, W, ...) plus a build_model(path::String) dispatcher that picks between a .dat file and a directory of CSVs based on isdir(path). Once the .jl file is generated, the .mod is no longer needed at runtime:

include("file.jl")

# Load straight from the .dat
jump_model = build_model("file.dat")

Internally, this loads the .dat file into a data dictionary. JuMPConverter.AMPL.read_dat returns a Dict{Symbol,Any} mapping each parameter or set name to a scalar, Vector, Array, JuMP.Containers.DenseAxisArray, or JuMP.Containers.SparseAxisArray. Doing this in two steps can be useful if you want to inspect the dictionary. One caveat is that you still need to keep the model (output of JuMPConverter.AMPL.read_model) around (or you need to use a DatSchema as done in the file.jl):

data = JuMPConverter.AMPL.read_dat("file.dat", model)
jump_model = build_model(; data...)

Loading data from CSV files

You can also export a .dat to a directory of CSVs (one file per parameter or set) and pass that directory back to build_model:

JuMPConverter.AMPL.dat_to_csv("file.dat", model, "data/")
jump_model = build_model("data/")

The CSV directory loader reads each kwarg's CSV when present, so any file you delete falls back to the kwarg's .mod default. CSV shape: scalars and sets are single-column; 1D parameters are index, value; 2D parameters are labeled matrices (column headers from the second axis); 3+D parameters use long form (i1, i2, ..., value).

GAMS™ converter

To read a GAMS™ model file.gms, do:

using JuMPConverter
model = JuMPConverter.GAMS.read_model("file.gms")

To print the JuMP model to the terminal, simply do

println(io, model)

To save the JuMP model in a file file.jl, do

open("file.jl", "w") do io
    println(io, model)
end