Also see #5

If solution was terminated due to a timelimt, ROI terminates with an error

Error in make_OP_solution(solution = solution, objval = optimum, status = status,  : 
  unknown solver status code. Please contact the ROI.plugin.cbc maintainer.

Hi there,

I am using OMPR and CBC via the rcbc binding to solve a MILP problem.
No issues in extracting the primal for a variable cog_power[sm,t], but when I try to extract the dual variables from the OMPR solution object sol I get

get_solution(sol, cog_power[sm,t], type="dual")
Error: The solution from the solver is invalid. It is NULL or contains NAs. Maybe the solver does not export duals?

Is it because duals are not supported by RCBC, or the ROI plugin, or something else?

Thanks a lot

Currently only optimal and solutionlimit are mapped to a successful run and return a result.

CBC arguments are registered with ROI. I think I did it correctly but have not tested it. We need to have tests for that.

Hi Dirk,

I have noticed that any special bounds added with set_bounds in ompr are not taken into account when using cbc solver.

After looking into the code, I have noticed that default bounds are set at 0 and inf or 1 depending on the variable types.
However there is nothing to translate the special bounds from ROI to Cbc. So any bounds stocked in bounds$lower$val or bounds$upper$val in ROI are replaced by the default bounds.

I’ve been trying to understand how to pass a relative or absolute gap to CBC through the ROI plugin but doesn’t seem to be possible. In rcbc I can use cbc_args = list(“ratioGap”=some value) but when I use control = list(“ratioGap”...) in ROI plugin I got the message “ratioGap is not available in solver cbc”. Am I missing something?