Re: [Help-glpk] OPB output file anomaly

[Andrew Makhorin]

> Just to follow-up on a this question regarding MPS format that may
> have been lost in the mix while discussing OPB format:

>> > It turns out the the difference in solution between the OPB solver
>> > and glpk is actually this constant.  I found this by noticing the same
>> > discrepancy when I generate an MPS and provide the resulting MPS file
>> > to various solvers.  The objective constant is not translated to the
>> > resulting OPB or MPS files.  Is this a feature or bug?  If it's a
>> > feature, could you explain the logic... it isn't clear to me why this
>> > happens.

> Missing obj. constant in MPS:  Bug or no?

This is neither a feature nor bug. By default the glpk routine
generating mps file writes the objective row only if there are no other
free rows in the instance; otherwise the first free row is considered
as the objective row. To check that the constant term is presented you
can write the instance in cplex format, and if the term is missing in
mps file (it should be in rhs section), there is at least one free row
in the instance due to which the actual objective row was not written.