Originally posted by: CPLEX12USER

Hi,

 I am trying to solve a MIP using Java concert with cplex 12.2. I can get an optimal solution to the model that I tried to solve. I wanted to do some analysis and exported the problem just before calling cplex.:solve() with all default CPLEX parameter. If I read the same problem into an interactive optimizer and try to solve the same (using optimize and mipopt), I get the problem as infeasible. Can you please let me know what could be the difference?

If I can  get optimal solution with concert, what am I missing that makes the problem infeasible with interactive optimizer. Should i do anything other than exporting model and reading it within interactive optimizer?

Any help highly appreciated. Thank you!

Originally posted by: T_O

Did you export your model as .sav file? If not, please do so and try again.

Best regards,
Thomas

In case the issue persists even with a SAV file:

1. Can you switch to a more recent version of CPLEX? 12.2 is really old.
2. Do you have any non-default parameters settings in Concert? If so, then export them as well (IloCplex::writeParam()) and apply those in the interactive as well.
3. Do you have any callbacks in Concert? If so, check whether the issue persists if you remove them.
4. Does your model mix large and small numbers? Try reading the model into the interactive and then execute 'disp prob stats'.
5. Can you attach your model and/or at least the solve log and model statistics here?

Originally posted by: CPLEX12USER

Hello,

 Thanks a lot for the directions. Looks like, while naming the constraint in loops, i had inadvertently given same name for a set of constraints. I was able to get an optimal solution though within concert. However, while i exported the model and imported in interactive optimizer, I ended up with issue. Not sure why interactive optimizer caught it and concert could go ahead. After I removed those duplicate ones, optimal solution was obtained through interactive optimizer.

If we have given the same name (not intentionally)  for different constraints , what should be the ideal behavior from CPLEX?

Thank you!

CPLEX explicitly does not check for duplicate constraint or variable names since in the majority of cases this does not pose a problem.

There are two cases in which duplicate names create issues:

1. In an LP format file, duplicate variable names will cause trouble when the file is read: all variables with the same name will be merged into a single variable.
2. In an MPS format file, duplicate constraint names will cause trouble when the file is read: CPLEX will actually refuse to read the file, other solvers may do different things.

There may also be issues with MST, BAS, ORD, ... files in case you attempt to identify variables and/or constraints by name and there are duplicates.

From what you describe above, I am not clear how the duplicate constraint names could cause any trouble.