Originally posted by: mathygirl

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

indeed when you have a sum then you cannot change a coef, but what you can do is add an empty decision variable and then change this decision variable with setcoef.

 

Let me give you an example:

range r=1..4;
dvar int x[1..4] in 0..10;
dvar int e;
maximize sum(i in r) x[i];
subject to
{
ct:e==0;

sum(i in r) x[i]+e<=10;
}

execute
{
writeln(x);
}

main
{

thisOplModel.generate();
cplex.solve();
thisOplModel.postProcess();
for(var i in thisOplModel.r ) thisOplModel.ct.setCoef(thisOplModel.x[i],i);
thisOplModel.e.UB=-4;
thisOplModel.e.LB=-4;
cplex.solve();
thisOplModel.postProcess();

}

 

 

 

which gives

 [10 0 0 0]
 [4 0 0 0]

regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: mathygirl

Thank you, this makes sense. What is the equivalent of thisOplModel.x[i] in the OPL Java API?

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi

in CPLEX_Studio1261\opl\examples\opl_interfaces\java\cutstock_change\src\cutstock_change

you have an example : Cutstock_change.java

You will see:

IloObjective masterObj = masterOpl.getObjective();
                masterCplex.setLinearCoef(masterObj, newVar, 1);
                for (int i = 1; i < nWdth + 1; i++) {
                    double coef = subCplex.getValue(subOpl.getElement("Use")
                            .asIntVarMap().get(i));
                    IloForAllRange forAll = (IloForAllRange) masterOpl.getElement(
                            "ctFill").asConstraintMap().get(i);
                    masterCplex.setLinearCoef(forAll, newVar, coef);
                }
                masterVars.add(newVar);

regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: mathygirl

Yes, there is a constraint of the form

forall( i in Items )
    ctFill:
      sum( p in Patterns )
        Fill[p][i] * Cut[p] >= Amount[i];

that is modified with the Java API code above.

In my implementation, I have no problem with the setLinearCoef() method on such constraints.

What I am having problems with is modifying constraints of the form:

myConstraint:

    sum(s in mySet) decisionVar[s] == 47;

where there is no "forall" and I am augmenting the decisionVar array, and therefore the sum, by one more term on each iteration.

 

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

this is why  introduced e yesterday in this thread!

In a sum you cannot change a coef, you get You can not change the coefficient of an aggregate expression

So a workaround is to add a decision variable e that is in an empty constraint and then you CAN change the coef of this cmpty constraint in order to set any decision variable coefficient.

Regards

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: mathygirl

Your workaround works. Most likely you're aware that the workaround is not needed in OPL script. This works:

for(var i in thisOplModel.r ) cplex.setCoef(thisOplModel.ct,thisOplModel.x[i],2.0);

I am trying to confirm that there is no fix in Java that doesn't involve going in and modifying the model.

#DecisionOptimization
#OPLusingCPLEXOptimizer

In 2013 I wrote the REF wish : https://www.ibm.com/developerworks/rfe/execute?use_case=viewRfe&CR_ID=42808

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: qtbgo

Dear Alex,

  I am learning opl java interface. In the expample  located in "opl_interfaces\java\mulprod". I modified the mulprod.mod a little as follows:

dvar float+ obj;

minimize obj;
subject to {
  
  obj ==   sum( p in Products, t in Periods ) 
      (InsideCost[p]*Inside[p][t] + 
       OutsideCost[p]*Outside[p][t] +
       InvCost[p]*Inv[p][t]);

 

Then In Mulprod.java, I want to change the coef of obj to 0 and solve the model. Could you tell me how to do it? The framework is as follows:

        IloOplFactory oplF = new IloOplFactory();
        IloOplErrorHandler errHandler = oplF.createOplErrorHandler();
        IloOplModelSource modelSource = oplF.createOplModelSource(DATADIR
                + "/mulprod.mod");
        IloOplSettings settings = oplF.createOplSettings(errHandler);
        IloOplModelDefinition def = oplF.createOplModelDefinition(modelSource,settings);
        IloCplex cplex = oplF.createCplex();
        cplex.setOut(null);
        IloOplModel opl = oplF.createOplModel(def, cplex);
        IloOplDataSource dataSource = oplF.createOplDataSource(DATADIR
                + "/mulprod.dat");
        opl.addDataSource(dataSource);       
        opl.generate();        
        
     
 //TODO how to set the coef of the obj to 0 ?        
        
        
        if (cplex.solve())

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

why not having a look at cutstock_change where you ll see

IloObjective masterObj = masterOpl.getObjective();
                masterCplex.setLinearCoef(masterObj, newVar, 1);

 

?

 

regards

https://medium.com/@alexfleischer_84755/optimization-simply-do-more-with-less-zoo-buses-and-kids-66940178db6

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: qtbgo

Yes, I had tried  like:

 

        IloObjective masterObj = opl.getObjective();
        IloNumVar newVar = cplex.numVar(0, Double.MAX_VALUE);
        cplex.setLinearCoef(masterObj, newVar, 0);

 

But it doesn't affect the result. I don't understand why introduce the newVar, and how it is related to the obj variable in tht mod file.

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

so let me share a complete tiny example.

Consider the very good example in opl\examples\opl_interfaces\java\mulprod and change the model into

{string} Products = ...;
{string} Resources = ...;
int NbPeriods = ...;
range Periods = 1..NbPeriods;

float Consumption[Resources][Products] = ...;
float Capacity[Resources] = ...;
float Demand[Products][Periods] = ...;
float InsideCost[Products] = ...;
float OutsideCost[Products]  = ...;
float Inventory[Products]  = ...;
float InvCost[Products]  = ...;
range Periods0 = 0..NbPeriods;

dvar float+ Inside[Products][Periods];
dvar float+ Outside[Products][Periods];
dvar float+ Inv[Products][Periods0];

dvar float costInside;
dvar float costOutside;
dvar float costInv;

/*minimize
  sum( p in Products, t in Periods )
      (InsideCost[p]*Inside[p][t] +
       OutsideCost[p]*Outside[p][t] +
       InvCost[p]*Inv[p][t]);*/

minimize costInside+costOutside+costInv;

subject to {

costInside==
sum( p in Products, t in Periods )
      InsideCost[p]*Inside[p][t];
costOutside==
sum( p in Products, t in Periods )
      OutsideCost[p]*Outside[p][t] ;

costInv==
sum( p in Products, t in Periods )
             InvCost[p]*Inv[p][t];

  forall( r in Resources, t in Periods )
    ctCapacity:
      sum( p in Products )
        Consumption[r][p] * Inside[p][t] <= Capacity[r];
  forall( p in Products , t in Periods )
    ctDemand:
      Inv[p][t-1] + Inside[p][t] + Outside[p][t] == Demand[p][t] + Inv[p][t];
  forall( p in Products )
    ctInventory:
      Inv[p][0] == Inventory[p];
};
tuple plan {
   float inside;
   float outside;
   float inv;
}
plan Plan[p in Products][t in Periods] = <Inside[p,t],Outside[p,t],Inv[p,t]>;
execute DISPLAY {
  writeln("plan=",Plan);
}

tuple InsideSolutionT{
    string Products;
    int Periods;
    float value;
};
{InsideSolutionT} InsideSolution = {<i0,i1,Inside[i0][i1]> | i0 in Products,i1 in Periods};
tuple OutsideSolutionT{
    string Products;
    int Periods;
    float value;
};
{OutsideSolutionT} OutsideSolution = {<i0,i1,Outside[i0][i1]> | i0 in Products,i1 in Periods};
tuple InvSolutionT{
    string Products;
    int Periods0;
    float value;
};
{InvSolutionT} InvSolution = {<i0,i1,Inv[i0][i1]> | i0 in Products,i1 in Periods0};

 

in order to have 3 KPIs in the objective

 

Then when you run you get

 

 [java] costInside = 22.333;
     [java] costOutside = 434.67;
     [java] costInv = 0;
     [java] Inside = [[10 0 0]
     [java]              [0 0 0]
     [java]              [50 66.667 66.667]];
     [java] Outside = [[0 100 50]
     [java]              [20 200 100]
     [java]              [0 33.333 33.333]];
     [java] Inv = [[0 0 0 0]
     [java]              [0 0 0 0]
     [java]              [0 0 0 0]];

but then if in mulprod.java

 

you change

opl.generate();

      

        if (cplex.solve())
        {
            System.out.println("OBJECTIVE: " + opl.getCplex().getObjValue());
            opl.postProcess();
            opl.printSolution(System.out);
        }

into

opl.generate();

        IloObjective masterObj = opl.getObjective();
        IloNumVar insideCost = opl.getElement("costInside").asNumVar();
        cplex.setLinearCoef(masterObj, insideCost, 10);

        if (cplex.solve())
        {
            System.out.println("OBJECTIVE: " + opl.getCplex().getObjValue());
            opl.postProcess();
            opl.printSolution(System.out);
        }

then you ll get

 

 [java] costInside = 0;
     [java] costOutside = 516;
     [java] costInv = 0;
     [java] Inside = [[0 0 0]
     [java]              [0 0 0]
     [java]              [0 0 0]];
     [java] Outside = [[10 100 50]
     [java]              [20 200 100]
     [java]              [50 100 100]];
     [java] Inv = [[0 0 0 0]
     [java]              [0 0 0 0]
     [java]              [0 0 0 0]];

we simply changed the coefficient of the inside cost in the orbjective

 

regards

 

https://www.linkedin.com/pulse/making-decision-optimization-simple-alex-fleischer/

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: qtbgo

Thank you Alex so much. This is exactly what I want.

#DecisionOptimization
#OPLusingCPLEXOptimizer