Originally posted by: UDOPS

The distribution example for Lagrangian relaxation has the following pieces,

in the flow control:

norm = 0;
for(i in thisOplModel.cities) {
slack[i] = 0;
for (var j in thisOplModel.cities) {
slack[i]+= m2.Ship[j][i];
}
slack[i] -= thisOplModel.request[i];
norm += Opl.pow(slack[i],2);
}

in the related model:

dexpr float lagrangian_obj = sum(i in cities, j in cities) ship_costi,j * Shipi,j
+ sum(j in cities) mult[j] * (request[j] - sum(i in cities) Shipi,j);

Do you think I could rewrite this as below? In addition to making it easier to read, would it be more efficient?

in the model:

dexpr float request_slackj in cities=sum(j in cities)(request[j] - sum(i in cities) Shipi,j)

dexpr float lagrangian_obj = sum(i in cities, j in cities) ship_costi,j * Shipi,j
+ sum(j in cities) mult[j] * request_slack[j];

changing the flow control:

norm = 0;
for(i in thisOplModel.cities) {
norm += Opl.pow(request_slack[i],2);
}
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

Tried it: it allowed the dexpr change, but flow control did not like the use of dexpr "request_slack[]". Why not?
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Have you tries "thisOplModel.request_slack[]"?

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

Yes, specific to my modification:

/*for(i in thisOplModel.cities) {
slack[i] = 0;
for (var j in thisOplModel.cities) {
slack[i]+= m2.Ship[j][i];
}
slack[i] -= thisOplModel.request[i];
norm += Opl.pow(slack[i],2);
} */

for (i in thisOplModel.cities) {
slack[i]+= m2.request_slack[i];
norm += Opl.pow(slack[i],2);
}
The commented portion is original, and works. The live portion runs, but returns a yellow warning "Not a model element "j", use 'var' to declare local scripting variables", and the values calculated are not the same.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

I'm confused. There is no "j" in the uncommented part ...

Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

Yes, exactly. The "j" referred to in the error message does not exist in any portion of the code that I think is relevant.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

I suppose that
(*) the problem originates from somewhere else
(*) the problem cannot be resolved without having a look at the complete model.

Just my 2 cents.

Have you tried to comment the "active block" in your snippet?
Does the problem still occur?

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer