Originally posted by: qtbgo

Hello, I have a large mxn matrix A, elements in it is 0 or 1 or -1, which is very sparse.  The file storing A is 300M large (if I zip this file, the zipped file becomes 3M large). I use java concert to model AX<=M.  X is a var array including 30000 binary variable.

1  I wonder if cplex can compress matrix A automatically in memory? if  IloLPMatrix can compress the matrix?

2  can I preprocess the file to compress it so that it becomes smaller and can be read into model eaiser?

1. Just don't submit the terms with coefficient 0 to CPLEX. That is the best you/CPLEX can do.

2. Remove all 0s from the file and store the remaining data in a sparse way, for example as triples (row, column, coefficient).

Originally posted by: qtbgo

about question 1:

I read the manual here. it talks about sparse. I am not sure if I understand it well. I guess if I use  IloLPMatrix to store sparse matrix, then it will remove all 0s automaticly, am I right?

about question 2:

Is there an example demonstrating your idea?

1. No, IloLPMatrix will not automatically remove 0's from your data. This is illustrated by the following example

import ilog.cplex.*;
import ilog.concert.*;

public final class Sparse {
   public static void main(String[] args) throws IloException {
      final IloCplex cplex = new IloCplex();
      try {
         IloNumVar x = cplex.boolVar("x");
         IloNumVar y = cplex.boolVar("y");
         IloNumVar z = cplex.boolVar("z");

         IloLinearNumExpr expr = cplex.linearNumExpr();
         expr.addTerm(1.0, x);
         expr.addTerm(0.0, y);
         expr.addTerm(1.0, z);

         System.out.println(expr);

         IloRange rng = cplex.range(1.0, expr, 2.0);
         System.out.println(rng);

         IloLPMatrix matrix = cplex.LPMatrix();
         matrix.addRow(rng);
         matrix.addRow(2.0, 3.0, new int[]{ 0, 1, 2 },
                       new double[]{ 0.0, 1.0, 2.0 });
         System.out.println(matrix);
      }
      finally {
         cplex.end();
      }
   }
}

which prints

(1.0*x + 0.0*y + 1.0*z)
IloRange  : 1.0 <= (1.0*x + 0.0*y + 1.0*z) <= 2.0
IloLPMatrix  {
  IloRange  : 1.0 <= (1.0*x + 0.0*y + 1.0*z) <= 2.0
  IloRange  : 2.0 <= (0.0*x + 1.0*y + 2.0*z) <= 3.0
}

2. I don't think there is an example shipped with CPLEX. But it should be rather obvious what to do: instead of storing a matrix in the file, only store the triples of non-zeros.