Programmatically process GLPK solution file

[willu47]

To programmatically process the results from OSeMOSYS using GLPK (as an alternative to manually writing out tabular files my modifying the model files), I would prefer to use the solution file generated by GLPK as we are doing with the CBC and CPLEX solvers. Apparently, using the --wglp option to write a model file in GLPK format and the solution with the --write (instead of --output) generates a machine readable text file which can be more easily processed, and the variable names can be extracted from the glpk model file as the identifiers match.

From https://en.wikibooks.org/wiki/GLPK/Interoperability :

"To access the solution found by GLPSOL from a computer program, one could write the model with the --wglp option and the solution with the --write option. The ordering of rows and columns is the same in both files. It is enough to obtain all necessary information. (The only improvements I would like to make here is to use a DIMACS-like format for solution files.)" Source: help-glpk, 19 Jan 2011, grammar corrected, emphasis added.

The following command can be used to generate the solution and model file which can be parsed programmatically:

glpsol --wglp model.lp -m src/osemosys.txt -d tests/utopia.txt --write results.sol

[willu47]

Names in the model file

In the model file, the names of the problem, objective function, variables and constraints are listed in the lines beginning with n where:

n p NAME # p = problem instance
n z NAME # z = objective function
n i ROW NAME # i = constraint name, ROW is the row ordinal number
n j COL NAME # j = variable name, COL is the column ordinal number

Rows in the solution file

In the solution file, the rows take on the following format:

i ROW ST PRIM DUAL

where:

• ROW is the ordinal number of the row
• ST is one of:
  • b = inactive constraint;
  • l = inequality constraint active on its lower bound;
  • u = inequality constraint active on its upper bound;
  • f = active free (unounded) row;
  • s = active equality constraint.
• PRIM specifies the row primal value (float)
• DUAL specifies the row dual value (float)

Columns in the solution file

Columns take on the following format:

j COL ST PRIM DUAL

where:

• COL specifies the column ordinal number
• ST contains one of the following lower-case letters that specifies the column status in the basic solution:
  • b = basic variable
  • l = non-basic variable having its lower bound active
  • u = non-basic variable having its upper bound active
  • f = non-basic free (unbounded) variable
  • s = non-basic fixed variable.
• PRIM field contains column primal value (float)
• DUAL field contains the column dual value (float)