Hi everyone,

I have found the situation when function alwaysConstant is not working with OPL. The problem occurs when constraints using alwaysConstant are enumerated in a loop. However, such problem doesn't occur when modeling with Python.

1 case (OPL with loop)

using CP;

int period_size = 50;
range steps = 0..3;

stateFunction my_state_function;

subject to 
{
  forall( h in steps ) 
  {
    alwaysConstant(my_state_function, h * period_size, (h + 1) * period_size);
  }
}

execute
{
  writeln(my_state_function);
}

gives

Computation error: an overflow occurred.

2 case (OPL without loop)

using CP;

int period_size = 50;
range steps = 0..3;

stateFunction my_state_function;

subject to 
{
  alwaysConstant (my_state_function, 0, 50);
  alwaysConstant (my_state_function, 50, 100);
  alwaysConstant (my_state_function, 100, 150);
  alwaysConstant (my_state_function, 150, 200);
}

execute
{
  writeln(my_state_function);
}

gives

stepwise{ -1 -> 0; 0 -> 200; -1 }

3 case (Python with loop)

import docplex.cp.model as cp_model


if __name__ == "__main__":
    model = cp_model.CpoModel("alwaysConstant with cycle")

    period_size = 50
    step_max = 3

    my_state_function = cp_model.state_function(name="my state function")
    for h in range(step_max + 1):
        model.add(cp_model.always_constant(my_state_function, (h * period_size, (h + 1) * period_size)))

    solution = model.solve(TimeLimit=10, trace_log=False)

    print(solution.get_var_solution(my_state_function))

gives

my state function: ((-4503599627370494, 0, -1), (0, 50, 0), (50, 100, 0), (100, 150, 0), (150, 200, 0), (200, 4503599627370494, -1))

4 case (Python without loop)

import docplex.cp.model as cp_model


if __name__ == "__main__":
    model = cp_model.CpoModel("alwaysConstant without cycle")

    period_size = 50
    step_max = 3

    my_state_function = cp_model.state_function(name="my state function")

    model.add(cp_model.always_constant(my_state_function, (0, 50)))
    model.add(cp_model.always_constant(my_state_function, (50, 100)))
    model.add(cp_model.always_constant(my_state_function, (100, 150)))
    model.add(cp_model.always_constant(my_state_function, (150, 200)))

    solution = model.solve(TimeLimit=10, trace_log=False)

    print(solution.get_var_solution(my_state_function))

gives

my state function: ((-4503599627370494, 0, -1), (0, 50, 0), (50, 100, 0), (100, 150, 0), (150, 200, 0), (200, 4503599627370494, -1))

------------------------------
Rustam Salikhov
------------------------------

Hello Rustam,

Unfortunately, you've found a bug in the OPL.
We are working to fix it.
We are very sorry for the inconvenience.
In the meantime you can use the following workaround.

range steps1 = 0..4;
int k[i in steps1]= i*period_size;
...
   alwaysConstant(my_state_function, k[h], k[h+1]);

I hope this helps.

Dear Christiane,

Thank you for your support. I hope it will be easy to fix it.

range steps1 = 0..4;int k[i in steps1]= i*period_size;...   alwaysConstant(my_state_function, k[h], k[h+1]);