Originally posted by: jawadomari

Hello everyone, 

I have a small question about creating arrays of variables. I have a 2D array of continuous variables representing the arrival times of each train at each station. I know that some trains do not arrive at some stations.

If I write something like this:

IloArray<IloNumVarArray> arrivalTimes(env, noOfTrains);

for(t = 0; t < noOfTrains; t++)

{

    arrivalTimes[t] = IloNumVarArray(env, noOfStations);

    if(t arrives at s)

    {

       arrivalTimes[t][s] = IloNumVar(env);

       model.add(arrivalTimes[t][s]);

    }

}

Would I have created noOfStations x noOfTrains variables, or just the ones that were added to the model? If it is the latter, would memory have been allocated to the parts of the array where a train does not arrive at a station?

Thank you. 

You would only create as many variables as the number of times the condition "t arrives at s" was satisfied. As for you second question, I believe that memory would be allocated forthe unused variables, but only enough memory in each case for a pointer, not the amount used by a variable. The unused pointers would probably default to zero values. (At least they would in Java; it's been a while since I programmed in C++).

One caveat: if you take this approach, do not pass arrivalTimes[t] as an argument to a CPLEX method (for instance, getValues). CPLEX will attempt to find a value for each variable, including the unallocated ones, and a null pointer adventure will occur.

Originally posted by: jawadomari

Thanks Paul, this is helpful! Have a goo day.