[Clp] C++ Ok, C - dont know Re. CLP from a C prog.

[Alexey Lvov]

>>I am a beginner regarding CLP, and I would like to ask you whether CLP
can be called from a C program. If so, is there some documentation that
explains how to do that?

  Hi,

Do you mean exactly C, not C++? For C I do not know the answer (the chances
are it may not be possible because COIN it written in C++). For C++ please
see below. CLP documentation starts with a simple example which shows how
to solve problem from a file. I would also put a reference to file
Clp-1.10.0/Clp/examples/addRows.cpp (or similar) to the front page of CLP
doc. It shows how to create and solve CLP problem in memory. Please find
two examples below. #1 follows CLP doc. #2 follows what John Forrest sent
me over lately. The second one worked much better for me. ( Also to be able
to "make" your program, first follow the instructions in Clp-1.10.0/INSTALL
).

----------------------------------------------------
          Dr. Alexey Lvov
IBM T.J. Watson Research Institute
1101 Kitchawan Rd. (Rt. 134) Office 34-159
Yorktown Heights  NY  10598
Tel: (914)-945-3732.





1) (a piece from   Clp-1.10.0/Clp/examples/addRows.cpp    )

//     make DRIVER=NNN      makes file NNN.cpp    //

#include "ClpSimplex.hpp"
#include "ClpPrimalColumnSteepest.hpp"
#include "ClpDualRowSteepest.hpp"



int main ()
{
  // Minimize  x0  +  2 * x1  +  4 * x2  +  5 * x3,
  //
  //  0 <= x0 <= 15,
  //  0 <= x1      ,
  //       x2 is not bound by any constraints,
  // -1 <= x3      ,
  //
  // x0  -  x2  <=  3,
  //
  // x0  +  x1  +  x2  +  x3  =  23.



  // --- BEGIN add the objective function and bounds on individual var.
--- //

  const int num_variables = 4;

  ClpSimplex model;
  model.resize(0,num_variables); // The first argument must be 0.

  model.setObjectiveCoefficient(0,  1.0);
  // Lower bounds for variables are 0.0 by default.
  model.setColumnUpper(         0, 15.0);

  model.setObjectiveCoefficient(1,  2.0);
  // Lower bounds for variables are 0.0 by default
  model.setColumnUpper(         1,  COIN_DBL_MAX);

  model.setObjectiveCoefficient(2,  4.0);
  model.setColumnLower(         2,  -COIN_DBL_MAX);
  model.setColumnUpper(         2,  COIN_DBL_MAX);

  model.setObjectiveCoefficient(3,  5.0);
  model.setColumnLower(         3, -1.0);
  model.setColumnUpper(         3,  COIN_DBL_MAX);

  // ---  END  add the objective function and bounds on individual var.
--- //


  // -- BEGIN add equality and ineq. constr. on lin. combinations of var.
-- //

  int *    rowIndex;
  double * rowValue;
  int      num_nonzero_coef;

  num_nonzero_coef = 2;
  rowIndex = new int   [num_nonzero_coef];
  rowValue = new double[num_nonzero_coef];
  rowIndex[0] = 0; rowValue[0] =  1.0;
  rowIndex[1] = 2; rowValue[1] = -1.0;
  // The 4-th argument is the lower bound,
  // The 5-th argument is the upper bound.
  model.addRow(num_nonzero_coef, rowIndex, rowValue, -COIN_DBL_MAX, 3.0);
  delete[] rowValue;
  delete[] rowIndex;

  num_nonzero_coef = 4;
  rowIndex = new int   [num_nonzero_coef];
  rowValue = new double[num_nonzero_coef];
  rowIndex[0] = 0; rowValue[0] = 1.0;
  rowIndex[1] = 1; rowValue[1] = 1.0;
  rowIndex[2] = 2; rowValue[2] = 1.0;
  rowIndex[3] = 3; rowValue[3] = 1.0;
  // The standard way of adding an equality constraint is to set both
  // the lower and upper bound to the same value.
  model.addRow(num_nonzero_coef, rowIndex, rowValue, 23.0, 23.0);
  delete[] rowValue;
  delete[] rowIndex;

  // --  END  add equality and ineq. constr. on lin. combinations of var.
-- //



  // Increase level of detail
  model.setLogLevel(4);

  // The model has been set up. Now solve!
  model.primal();



  char const * const status_message[5] = {
    "optimal",
    "primal infeasible",
    "primal unbounded, dual infeasible",
    "stopped on iterations etc",
    "stopped due to errors"};
  std::cout << "Model status is \""
            << status_message[model.status()] << "\" after "
            << model.numberIterations() << " iterations - objective is "
            << model.objectiveValue() << std::endl;

  std::cout << "The model has " << model.numberRows()
            << " rows and "     << model.numberColumns()
            << " variables." << std::endl;

  if (num_variables != model.numberColumns()) {
    std::cout << "ERROR : num_variables != model.numberColumns() .\n";
    return -1;
  }

  double * columnPrimal = model.primalColumnSolution();
  int iColumn;
  printf("Solution:\n");
  for (iColumn = 0; iColumn < num_variables; iColumn++)
    printf("x%d = %8.5f\n", iColumn, columnPrimal[iColumn]);



  return 0;
}





2)

#include "ClpSimplex.hpp"
#include "ClpSolve.hpp"
#include "CoinModel.hpp"

.......
..........
  ClpSimplex * model = new ClpSimplex;
  model->resize(0, lp_num_variables); // The first argument must be 0.

  for (i = 0; i < lp_num_variables; i++) {
    model->setObjectiveCoefficient(i, lp_objective_function[i]);
    model->setColumnLower(i, (i < lp_num_s_variables) ? 0.0 :
-COIN_DBL_MAX);
    model->setColumnUpper(i, COIN_DBL_MAX);
  }

  {
    // use {} so model2 will vanish
    CoinModel model2;
    for (i = 0; i < lp_num_constr_on_s; i++) { // equality constraints
      model2.addRow(lp_constraints[i].num_coefficients,
                  lp_constraints[i].indexes,
                  lp_constraints[i].coefficients,
                  lp_constraints[i].right_part,
                  lp_constraints[i].right_part);
    }
    for (     ; i < lp_num_constraints;    i++) { // inequality constraints
      model2.addRow(lp_constraints[i].num_coefficients,
                  lp_constraints[i].indexes,
                  lp_constraints[i].coefficients,
                  lp_constraints[i].right_part,
                  COIN_DBL_MAX);
    }
    model->addRows(model2,false);
  }

  model->setLogLevel(1);
  ClpSolve solveOptions;
  model->setPerturbation(50);
  model->initialSolve(solveOptions);
...........
..........

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20090515/9ac17acf/attachment.html>