[Coin-SMI] OSLSE vs COIN (problem with readSMPS)

[Alois Geyer]

I have noticed some discrepancies between solutions obtained from OSLSE and COIN (e.g. using the 
KandW3 example and others). to investigate this issue further I have looked more closely to the results of 
readSMPS (among other things). what seems strange is the following:  

if the row sense in the core-file is 'G' or 'L', the stoch-file uses the REPLACE option and the RHS is 
stochastic "osiStoch->getRowSense()" returns 'R' rather than 'G' or 'L' (see code excerpt from unitTest.cpp 
below). I understand that 'R' could be used on purpose to indicate that the RHS is stochastic/random. but 
'R' does not appear if the row sense is 'E' or the RHS is non-stochastic.  

in addition, the RHS of 'L' constraints is not replaced by the stoch-file entry. the RHS of 'G' constraints is 
replaced!  

another observation: using ADD in the the stoch-file does not change 'G' or 'L' to 'R' (as above) but the 
stochastic RHS is not added to the RHS from the core-file. it is only added for equality constraints ('E'). 

has anyone else made similar observations?  how can this problem be resolved? 

~~Alois 

{
            // test SMPS files KandW3R
            printf(" *** testing KandW3.\n");
            SmiScnModel smi;  
            smi.readSmps("../../Mps/Stochastic/KandW3R");        
            OsiClpSolverInterface *clp = new OsiClpSolverInterface();
            smi.setOsiSolverHandle(*clp);
            OsiSolverInterface *osiStoch = smi.loadOsiSolverData();
     
            //+AG*
            int nrows = osiStoch->getNumRows();
            int ncols = osiStoch->getNumCols();
            const char * rs  = osiStoch->getRowSense();
            const double * rhs = osiStoch->getRightHandSide();

            int i;    
            for (i=0;i<nrows;i++){
                  printf("i %d rs %c rhs %f\n",i,rs[i],rhs[i]);
            }
            //-AG*

           osiStoch->initialSolve();    
           assert(fabs(osiStoch->getObjValue()2613.0) < 0.0001);
      }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/smi/attachments/20060215/8beb8655/attachment.html