The OPTMODEL Procedure

When a solver finishes, it reports zero or more solutions for the optimization variables of the current problem. Each solution assigns a value to each of the variables in the problem. The SOLVE statement saves these solutions with the current problem. The first reported solution, if any, is also copied into the optimization variables. The number of solutions is available in the predeclared numeric parameter _NSOL_.

Note: The network solver does not require optimization variables and has its own conventions for returning multiple solutions.

You can access the solutions that are saved with the problem by adding a solution index to the .sol suffix of the variable name. For example, x.sol[2] would reference the second solution saved for variable x in the current problem. Both the variable name and the suffix can be indexed. For example, assign[3,7].sol[1] refers to the first solution for the array variable element assign[3,7]. The solution index must be an integer in the range 1 to _NSOL_.

The following example illustrates the processing of multiple solutions from the CLP solver:

proc optmodel printlevel=0;
   var x{1..2} integer >= 1 <= 3;
   con c: alldiff(x);
   solve with clp / allsolns;
   print _NSOL_;
   print {j in 1..2, i in 1.._NSOL_} x[j].sol[i];
   create data solout from [sol]={i in 1.._NSOL_}
       {j in 1..2} <col("x"||j)=x[j].sol[i]> ;

This program produces the output in Figure 5.64. It also creates a data set, solout, which has each solution in a separate observation.