JuliaOpt
passing parameters
How does one pass in time limit, absolute optimality gap, and relative optimality gap stopping condition parameters to GPLK when using const solver = GLPKSolverMIP() ?
You should be able to pass options as keyword arguments to the constructor. The available ones are documented in the GLPK manual (available at http://www.gnu.org/software/glpk), section 2.10.5 "glp intopt — solve MIP problem with the branch-and-cut method", paragraph "Control parameters". For example, the time limit is tm_lim (passed as an integer number of milliseconds).
Great thanks a lot! Does GLPK have an absolute optimality gap? The user manual states that mip_gap is the relative optimality gap.
Are there plans to implement methods setparameters! from MathProgBase as well?
This would mean handling the two cases of TimeLimit and Silent.
No plans I'm aware of; MPB is deprecated. A MOI wrapper for GLPK is not far off (https://github.com/JuliaOpt/GLPK.jl/pull/53). It will then be a bit more work to expose these as MOI optimizer attributes.
OK. I just saw that setparameters!(s::GLPKSolverMIP; mpboptions...) and setparameters!(m::GLPKMathProgModelMIP are implemented, but the corresponding methods for LP are not. I will try to add them in my own code.
Unfortunately, I will still need to run code based off MPB during the next couple months, because of callbacks and because I need it now :smirk:
I will still need to run code based off MPB during the next couple months, because of callbacks and because I need it now
That's understandable :). PRs accepted if you'd like to add setparameters! for the LP solver.