Originally posted by: Chicoscience

Hi, I am now migrating to CPLEX 12.1 and my code, which works perfectly with CPLEX 10.2, is not working anymore.

When the following code is executed:

cplex.solve();
IloNum status = cplex.getStatus();
try {
IloNum value = cplex.getObjValue();
} catch (IloCplex::Exception e) {
printf("ERROR: %d\n", (int)e.getStatus());
}

On CPLEX 12.1, the satus returned is Optimal (alright) but when I try to retrieve the best objective value the following error code is printed:

ERROR: 1217

which estates that there exists no solution to the problem. This piece of code works fine with CPLEX 10.2. Can anybody help me?

Thank you very much,
Chicoscience
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Chicoscience

Return value of a few functions:

cplex.solve() -> true
cplex.getStatus() -> Optimal
cplex.isPrimalFeasible() -> FALSE
cplex.isDualFeasible() -> FALSE

How can that be possible? There has to be a very simple solution to this problem.

Thank you,
Chicoscience
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

First, I'm not sure I would trust assigning the value of getStatus to an IloNum. IloCplex::getStatus returns an enumeration (type IloAlgorithm::Status), and while I don't know that it actually happens, conceivably the CPLEX developers could change the indexing of that enumeration from one version to the next. You might try

IloAlgorithm::Status status = cplex.getStatus();
if (status == IloAlgorithm::Status::Optimal) { ... }

to verify that you're really getting an optimal solution.

Also, it might be useful to check cplex.getCplexStatus(), which returns an enumeration (IloCplex::CplexStatus) that's a bit more granular.

/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: Chicoscience

Hi Paul, thank you for your answer.

I put the following piece of code:

cplex.out() << "Status is " << cplex.getStatus() << endl;
cplex.out() << "Cplex Status is " << cplex.getCplexStatus() << endl;

And the result was:

Status is Optimal
Cplex Status is OptimalTol

I defined a tolerance value, that's why I got OptimalTol. I still don't get it, why does it happen?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

> Chicoscience wrote:
> Hi Paul, thank you for your answer.
>
> I put the following piece of code:
>
> cplex.out() << "Status is " << cplex.getStatus() << endl;
> cplex.out() << "Cplex Status is " << cplex.getCplexStatus() << endl;
>
> And the result was:
>
> Status is Optimal
> Cplex Status is OptimalTol
>
> I defined a tolerance value, that's why I got OptimalTol. I still >on't get it, why does it happen?

Can you provide a bit more information. Specifically:

1) What type of problems are you solving? LP? MILP?

2) Can you provide the optimizer output from the run with CPLEX 10.2 and 12.1?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Chicoscience

Hi Edklotz, thank you for your answer as well. I am solving a MIP problem, even on CPLEX 12.1 the output is ok, it looks like the problem was solved. The outputs:

Output with CPLEX 10.2

ILOG CPLEX 10.200, licensed to "IBM ILOG Optimization Suite for Academic Initiative", options: e m b q 
MIP start values provide initial solution with objective 369.0000.
Clique table members: 1614.
MIP emphasis: balance optimality and feasibility.
Root relaxation solution time =    0.01 sec.
 
        Nodes                                         Cuts/ 
   Node  Left     Objective  IInf  Best Integer     Best Node    ItCnt     Gap         Variable B Parent  Depth
 
      0     0      341.7500    66      369.0000      341.7500       95    7.39%
                   352.4375    69      369.0000     Cuts:  67      137    4.49%
                   354.0357    83      369.0000     Cuts:  70      160    4.06%
                   354.2237    88      369.0000      User:  1      174    4.00%
                   354.6180    87      369.0000      User:  1      188    3.90%
                   354.8965    85      369.0000      User:  1      206    3.82%
                   355.6667    92      369.0000      User:  1      233    3.61%
                   356.5511    77      369.0000      User:  1      253    3.37%
                   356.6270    89      369.0000      User:  1      263    3.35%
                   357.2500    77      369.0000      User:  1      284    3.18%
                   357.5000    78      369.0000      User:  1      294    3.12%
                   357.5278    87      369.0000      User:  1      309    3.11%
                   357.5972    82      369.0000      User:  1      315    3.09%
                   357.6250    81      369.0000      User:  1      320    3.08%
                   357.7917    77      369.0000      User:  1      327    3.04%
                   357.8750    75      369.0000      User:  1      329    3.02%
                   357.9375    88      369.0000      User:  1      334    3.00%
                   358.0000    86      369.0000      User:  1      336    2.98%
                   358.2339    69      369.0000      User:  1      346    2.92%
                   358.2353    71      369.0000      User:  1      347    2.92%
                   358.2372    71      369.0000      User:  1      350    2.92%
                   358.2730    77      369.0000      User:  1      355    2.91%
                   358.2730    80      369.0000      User:  1      356    2.91%
                   358.2857    79      369.0000      User:  1      359    2.90%
                   358.2976    82      369.0000      User:  1      361    2.90%
                   358.2976    85      369.0000      User:  1      362    2.90%
                   358.3033    76      369.0000      User:  1      364    2.90%
                   358.3073    83      369.0000      User:  1      366    2.90%
                   358.3108    85      369.0000      User:  1      368    2.90%
                   358.3199    82      369.0000      User:  1      373    2.89%
                   358.3333    79      369.0000      User:  1      376    2.89%
                   358.3333    92      369.0000      User:  1      377    2.89%
                   358.3333    95      369.0000      User:  1      378    2.89%
                   358.3514    89      369.0000      User:  1      385    2.89%
                   358.3708    92      369.0000      User:  1      389    2.88%
                   358.3750    76      369.0000      User:  1      391    2.88%
                   358.6207    78      369.0000      User:  1      403    2.81%
                   358.6619    91      369.0000      User:  1      412    2.80%
                   358.7273    86      369.0000      User:  1      415    2.78%
                   358.7500    77      369.0000      User:  1      417    2.78%
                   359.0000    67      369.0000      User:  1      420    2.71%
                   359.0000    76      369.0000      User:  1      424    2.71%
     10     9      362.5000    10      369.0000      360.9167      874    2.19%          id1290 U      9      6
*    13     9                   0      365.0000      360.9167      977    1.12%          id1248 N     12      8
     20     9      363.4722    44      365.0000      361.3750     1315    0.99%          id1319 U     19      8
     30     7      364.1000    40      365.0000      362.9500     1671    0.56%          id1292 U     29      6
*    36     0                   0      364.0000      363.6667     1986    0.09%          id1325 U     35      7
 
Gomory fractional cuts applied:  10
User cuts applied:  207
Return from cplex.solve(): true
Status is Optimal
Cplex Status is 102
cplex.isPrimalFeasible()? true
cplex.isDualFeasible()? true

Output with CPLEX 12.1

IBM ILOG License Manager: "IBM ILOG Optimization Suite for Academic Initiative" is accessing CPLEX 12 with option(s): "e m b q ".
1 of 1 MIP starts provided solutions.
MIP start 'm1' defined initial solution with objective 369.0000.
Warning: Control callbacks may disable some MIP features.
Clique table members: 1614.
MIP emphasis: balance optimality and feasibility.
MIP search method: traditional branch-and-cut.
Parallel mode: none, using 1 thread.
Root relaxation solution time =    0.02 sec.
 
        Nodes                                         Cuts/ 
   Node  Left     Objective  IInf  Best Integer     Best Node    ItCnt     Gap         Variable B NodeID Parent  Depth
 
*     0+    0                          369.0000                     95     --- 
      0     0      341.7500    66      369.0000      341.7500       95    7.38%
      0     0      353.8333    55      369.0000      Cuts: 70      148    4.11%
      0     0      355.5000    74      369.0000       Cuts: 5      161    3.66%
      0     0      356.2500    85      369.0000       Cuts: 3      175    3.46%
      0     0      357.0135    60      369.0000       User: 1      191    3.25%
      0     0      358.2500    46      369.0000       Cuts: 4      202    2.91%
      0     0      358.5000    22      369.0000       Cuts: 3      204    2.85%
      0     0      358.6111    80      369.0000       Cuts: 3      211    2.82%
      0     0      359.0000    40      369.0000       User: 1      220    2.71%
      0     0      359.5000    69      369.0000       Cuts: 3      236    2.57%
      0     0      359.5000    77      369.0000       User: 1      239    2.57%
      0     0      359.5000    44      369.0000       User: 1      241    2.57%
      0     0      359.5000    82      369.0000       User: 1      242    2.57%
      0     0      359.5000    37      369.0000       User: 1      246    2.57%
      0     0      359.5000    42      369.0000       User: 1      247    2.57%
      0     0      359.5000    30      369.0000       User: 1      251    2.57%
      0     0      359.5000    34      369.0000       User: 1      252    2.57%
      0     0      359.5333    81      369.0000       User: 1      259    2.57%
      0     0      359.5714    73      369.0000       User: 1      262    2.56%
      0     0      359.5789    82      369.0000       User: 1      263    2.55%
      0     0      359.6000    52      369.0000       User: 1      267    2.55%
      0     0      359.6250    58      369.0000       User: 1      269    2.54%
      0     0      359.6667    41      369.0000       User: 1      270    2.53%
      0     0      359.6785    87      369.0000       User: 1      280    2.53%
      0     0      359.6849    86      369.0000       User: 1      283    2.52%
      0     0      359.7273    85      369.0000       User: 1      284    2.51%
      0     0      359.7778    55      369.0000       User: 1      288    2.50%
      0     0      359.7778    56      369.0000       User: 1      291    2.50%
      0     0      360.0000    30      369.0000       User: 1      298    2.44%
      0     0      360.1364    61      369.0000       User: 1      304    2.40%
      0     0      360.1667    60      369.0000       User: 1      305    2.39%
      0     0      360.2209    84      369.0000       User: 1      309    2.38%
      0     0      360.2500    61      369.0000       User: 1      311    2.37%
      0     0      360.2500    55      369.0000       User: 1      313    2.37%
      0     0      360.2500    45      369.0000       User: 1      314    2.37%
      0     0      360.2500    63      369.0000       User: 1      315    2.37%
      0     0      360.2500    53      369.0000       User: 1      316    2.37%
      0     0      360.2500    53      369.0000       User: 1      317    2.37%
      0     0      360.2632    56      369.0000       User: 1      318    2.37%
      0     0      360.2652    89      369.0000       User: 1      319    2.37%
      0     0      360.2667    59      369.0000       User: 1      321    2.37%
      0     2      360.2667    59      369.0000      360.3333      321    2.35%                        0             0
     10    12      363.2500    18      369.0000      362.8333      696    1.67%          id1307 D     10      9      7
*    12    12      integral     0      365.0000      362.8333      738    0.59%            id10 U     12     11      9
     20    10      364.7031    58      365.0000      363.2647     1090    0.48%          id1284 U     20     19      3
*    23     5      integral     0      364.0000      363.3333     1177    0.18%
 
Zero-half cuts applied:  17
Gomory fractional cuts applied:  1
User cuts applied:  151
 
Return from cplex.solve(): true
Status is Optimal
Cplex Status is OptimalTol
cplex.isPrimalFeasible()? false
cplex.isDualFeasible()? false
Error 1217 when calling cplex.getObjValue()

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Chicoscience

As you can see by the output, it is a Branch-and-Cut algorithm and the Cplex Status is different, it is 102 when using Cplex 10.2 and OptimalTol when using Cplex 12.1... what 102 means?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Status 102 is CPXMIP_OPTIMAL_TOL, so the same as in CPLEX 12.1. Maybe, the << operator in the Concert version shipped with CPLEX 10 did not understand status 102?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Chicoscience

Probably yes... very strange, all seems fine with CPLEX 12.1 as the output shows, but I just can't read the solution and cplex thinks the problem is not primal feasible!! Very very strange!
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Does this happen for all problem instances that you try, or is this specific to this one? Can you post the problem instance such that we can take a look?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Chicoscience

Hi Tobias, I found it! As usual it was the programmer's fault :-), but it is still interesting to report it.

After cplex.solve(), I was calling cplex.clearCuts() - this is because this code (and the object it represents) is also used as a inner solver of a Local Branching algorithm.

When I removed the execution of cplex.clearCuts(), it worked properly with CPLEX 12.1. I still don't get why did it work with CPLEX 10.2 but not with 12.1, what changed in clearCuts() function?

Thank you very much,
Chicoscience
#CPLEXOptimizers
#DecisionOptimization