Originally posted by: LeandroCC

Is it possible to have a logical constraint (IloIfThen in my case) as a lazy constraint?

When creating my model I have:

IloConstraintArray lazy(env);
//add bunch of constraints to lazy as
lazy.add(constraint);
 
IloCplex cplex(model);
for (int i=0 ; i < lazy.getSize() ; i++)
   cplex.addLazyConstraint(lazy[i]);
 
for (int i=0 ; i < lazy.getSize() ; i++)
   lazy[i].end();
lazy.end();

Some of these constraints are logical ones, using IloIfThen.

When running, the program crashes in the line cplex.addLazyConstraint(lazy[i]); with the following error:

Warning: Lazy constraints freed during column deletion.
terminate called after throwing an instance of 'IloCplex::InvalidCutException'
Aborted (core dumped)
What is it trying to delete?? Is the invalid cut a lazy constraint containing a logical one?

Is there a way to do it?

Thanks
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

If you print lazy[i] immediately before you add it, does it look correct?

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: LeandroCC

Yes, it looks correct:

IloIfThenI (57157)
  left (my if statement)
  right (my then statement)
 
Warning:  Lazy constraints freed during column deletion.
terminate called after throwing an instance of 'IloCplex::InvalidCutException'

Other information:

I added a regular constraint to the same array, which is added normally to the model.
If I add the IloIfThen constraint via model.add instead of addLazyConstraint then it works fine.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Does the exception occur with i = 0? If yes, does switching lazy[0] with lazy[1] before adding cause the exception to remain at i = 0 or switch to i = 1? (I'm trying to ascertain whether it's a specific constraint causing problems or just the first encounter with a lazy if-then. p

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

It may not be possible to do this after all.

Both the C++ and Java API docs (for CPLEX 12.4) say that ifThen generates an instance of IloConstraint and that addLazyConstraint adds an instance of IloConstraint. I just tried doing essentially the same thing you did, but in Java. It threw an exception, but not the one you saw. The Java exception was

java.lang.ClassCastException: ilog.cplex.CpxIfThen cannot be cast to ilog.cplex.CpxRange

That suggests that, documentation notwithstanding, addLazyConstraint is limited to instances of IloRange, and if-then constraints do not convert to ranges. (That last part is totally plausible to me.)

If I'm right, then (a) the docs are a bit less than totally forthcoming and (b) the C++ exception is a bit more cryptic than one might hope.

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: LeandroCC

Answering your previous post, i=1 continues to throw the exception. It happens in the first IloIfThen, not a particular one.

Following your second post, I guess we will have to wait for someone from IBM to answer this one. I really don't see why this shouldn't be possible.

Thanks for the interest, Paul!
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Lazy constraints must be linear constraints. Logical constraints are not allowed here. That the function signature says 'IloConstraint' is actually a bug in the API that we fixed recently. The functions accept only 'IloRange' arguments. Everything else will result in runtime errors.
#CPLEXOptimizers
#DecisionOptimization