Originally posted by: Maichel_Aguayo

Hello all,

I am trying to inject a new incumbent solution using HEURISTICCALLBACK (I am solving a MIP model in CPLEX/C++).  I only have a two-dimensional variable in the model.  How can I use setSolution for this case ?

I have reproduced below some part of my code for convenience. I got the following error: no matching function for call to "Non_fixedI::setSolution(Arcs&, FloatMatrix&, IloNum&)"

Thanks,

typedef IloArray<IloIntVarArray> Arcs;

...

ILOHEURISTICCALLBACK1(Non_fixed,Arcs, x) {

 FloatMatrix sol;   // contains the value of the variables

IloNum objval=new_incumbentObjValue;      // contains the objective value of the incumbent solution

...

 setSolution(x, sol, objval);  //error: no matching function for call to 'Non_fixedI::setSolution(Arcs&, FloatMatrix&, IloNum&)'

}

int main(int argc, char **argv)

{

 Arcs x(masterEnv, numNodes);

IloCplex masterCplex(masterMod);

  masterCplex.use(Non_fixed(masterEnv, x));

....

}

You have to copy the variables and values into flat arrays. Something like this:

IloNumVarArray vars(getEnv());
IloNumArray vals(getEnv());
for (IloInt i = 0; i < x.getSize(); ++x)
   for (IloInt j = 0; j < x[i].getSize(); ++j) {
      vars.add(x[i][j]);
      vals.add(sol[i][j]);
   }

Then you can call setSolution() with the flat arrays (don't forget to end() the arrays after that).

Originally posted by: Maichel_Aguayo

Thanks Daniel for your help. It did work as suggested.

Just in case someone else faces the same issue, you can use setBound and solve in the heuristic callback to check if the solution injected is feasible.  I have reproduced below some part of my code for convenience.  

ILOHEURISTICCALLBACK1(Non_fixed, Arcs, x) {
  ...
// New_sol[i][j] contain the solution values

     for (IloInt i = 0; i < x.getSize(); ++i) 
        for (IloInt j = 0; j < x[i].getSize(); ++j) { 
            vars.add(x[i][j]); 
            vals.add(New_sol[i][j]); 
             setBounds(x[i][j], New_sol[i][j] ,New_sol[i][j] );
   }

     if ( solve() ) {
        setSolution( vars, vals, objval);
     }
     
        vars.end();
        vals.end();
      
   }

Calling setBounds() and solve() is actually the recommended way to go! Thanks for mentioning that and posting your code!