The CPLEX.jl package provides an unofficial interface for using IBM's CPLEX Optimizer™ from the Julia language. You cannot use CPLEX.jl without having purchased and installed a copy of CPLEX Optimizer™ from IBM. This package is available free of charge and in no way replaces or alters any functionality of IBM's CPLEX Optimizer product.
CPLEX.jl is a Julia interface for the CPLEX optimization software. CPLEX functionality is extensive, so coverage is incomplete, but the basic functionality for solving linear and mixed-integer programs is provided.
CPLEX.jl is intended for use with the MathProgBase solver interface; an internal API, while present, is not documented.
NOTE: CPLEX does not officially support linking to their dynamic C library, which is necessary for use from Julia. However, the steps outlined below have worked for OS-X, Windows, and Linux machines.
-
First, you must obtain a copy of the CPLEX software and a license; trial versions and academic licenses are available here.
-
Once CPLEX is installed on your machine, point the
LD_LIBRARY_PATHvariable to the directory containing the CPLEX library by adding, for example,export LD_LIBRARY_PATH="/path/to/cplex/bin/x86-64_linux":$LD_LIBRARY_PATHto your start-up file (e.g..bash_profile, [adding library path on Ubuntu](http://stackoverflow.com/questions/13428910/how-to-set-the-environmental-variable-ld-library-path-in-linux for a)). On linux, make sure this directory containslibcplexXXX.sowhereXXXis stands for the version number; on OS-X the file should be namedlibcplexXXX.dylib. Alternatively, you can also use theCPLEX_STUDIO_BINARIESenvironment variable as follows:
$ CPLEX_STUDIO_BINARIES=/path/to/cplex/bin/x86-64_linux julia -e 'Pkg.add("CPLEX"); Pkg.build("CPLEX")'
- At the Julia prompt, run
julia> Pkg.add("CPLEX")
(or manually clone this module to your .julia directory).
- Check that your version is included in
deps/build.jlin the aliases for the library dependency; if not, open an issue.
Currently, CPLEX.jl is compatible only with 64-bit CPLEX and 64-bit Julia on Windows. CPLEX.jl attempts to automatically find the CPLEX library based on the CPLEX_STUDIO_BINARIES environmental variable set by the CPLEX installer.
Which version of CPLEX are you trying to install? Currently CPLEX.jl only supports 1260, 1261, 1262, 1263, 1270, 1271, and 1280. If it's not one of those, file an issue with the version number you'd like to support. Some steps need to be taken (like checking for new or renamed parameters) before CPLEX.jl can support new versions.
The most common problem is not setting LD_LIBRARY_PATH correctly. Open a terminal and check the output of
echo $LD_LIBRARY_PATH
is the path to the CPLEX installation. If it's not, did you follow step 2 above?
Hint: on OS X the path should probably be something like
/Users/[username]/Applications/IBM/ILOG/CPLEX_Studio[version number]/cplex/bin/x86-64_osx/
The most common problem is not setting CPLEX_STUDIO_BINARIES correctly. Open a Julia prompt and check that the output of
julia> ENV["CPLEX_STUDIO_BINARIES"]is the path to the CPLEX installation. If you get a key "CPLEX_STUDIO_BINARIES" not found error, make sure the environment variable is set correctly, or just set it from within the Julia prompt
julia> ENV["CPLEX_STUDIO_BINARIES"] = "path/to/cplex/installation"
julia> Pkg.build("CPLEX")Another alternative is to run
CPLEX_STUDIO_BINARIES="path/to/cplex/installation" julia -e 'Pkg.build("CPLEX")'
Solver parameters can be passed through the CplexSolver() object, e.g., CplexSolver(CPX_PARAM_EPINT=1e-8). Parameters match those of the CPLEX documentation. Additionally, the mipstart_effortlevel parameter can be used to tell CPLEX how much effort to put into turning warmstarts into feasible solutions, with possible values CPLEX.CPX_MIPSTART_AUTO, CPLEX.CPX_MIPSTART_CHECKFEAS, CPLEX.CPX_MIPSTART_SOLVEFIXED, CPLEX.CPX_MIPSTART_SOLVEMIP, CPLEX.CPX_MIPSTART_REPAIR, and CPLEX.CPX_MIPSTART_NOCHECK.