Feasible integer problem reported as unbounded

[neil-lindquist]

Feasible integer problem reported as unbounded

[neil-lindquist]

I've fixed the unbounded error; although, it now fails to find an optimal solution before running out of memory.
I'm guessing GLPK uses some advanced tricks to cut down the search space more effectively and know when a solution is effectively optimal.

I don't know what your end use is, but I'm guessing you could likely skip the integer constraints and just round to the nearest integer before you use them. The values of w ORE HKGWZ GPVTF QDVJ KHKGT XJWVT DCFZ NZVS PSHF are 8.289275e7 1.0e12 4.3484155e9 7.5524506e8 8.289275e7 4.1446378e8 3.6472812e9 1.2920908e10 2.7665457e9 1.3357246e10, so you'd be introducing an error less than 00.001% to those variables.

[bd339]

Great! I haven't had the chance to try it yet but it's not a big problem. I already got the solution from GLPK and it was very much a one-off kind of thing. I only opened these bugs to improve the quality of the implementation :)
GLPK found the solution in 0.1 secs. But it does mean that this library probably is not that practical for integer programming for me. Some of my ILPs even choke GLPK and I have to use a commercial solver.

[neil-lindquist]

Good to hear that this wasn't hindering you. And the issues have exposed some embarrassing holes in the test suite, so I really appreciate them.

Yea, this library isn't well tuned for large/hard problems.
Adding a CPLEX or Gurobi backend shouldn't be too hard, I just haven't made it enough of a priority.