/**********************************************************************************\ * Name: Trimloss.c Fair Isaac 14/03/2014 * * Purpose: Modify a problem by adding an extra variable within an additional * * constraint. * * Description: We take a trimloss problem in which each integer variable x(p) * * represents the number of rolls cut to pattern p. * * We define a new integer variable y=SUM(p)x(p) and add the associated* * constraint * * x(1)+x(2)+...+x(N)-y = 0 * * We do this by first adding a row containing the (unitary) * * coefficients of the x(p), and then a column corresponding to y. * * We output the revised matrix to a file and then solve the revised * * MIP, giving y the highest branching priority. Finally, we output * * the solution, both to the screen and to an ASCII solution file. * * Input: trimloss.mps original matrix file * * Output: trimloss.log log file * * Trimloss.sol binary solution file - produced by default * * Revised.mps revised matrix file * * Revised.asc ASCII solution file * * Revised.hdr header for the ASCII solution file * \**********************************************************************************/ import java.io.*; import com.dashoptimization.*; public class Trimloss implements AutoCloseable { /** * Execute example from command-line **/ public static void main(String[] args) { String model = args.length == 0 ? "../data/trimloss" : args[0]; try (Trimloss trimloss = new Trimloss()) { trimloss.run(model); } } /** * XPRSprob for the trimloss problem **/ private XPRSprob probg; /** * The index of the newly-added now **/ private int addedRowIndex; /** * The index of the newly-added column **/ private int addedColIndex; /** * Initialize Trimloss problem **/ public Trimloss() { probg = new XPRSprob(); } /** * Destroy the Trimloss problem **/ public void close() { probg.close(); } /** * Execute Trimloss example **/ public void run(String model) { /* Setup the logfile */ File logfile = new File("trimloss.log"); if (logfile.exists()) logfile.delete(); probg.setLogFile(logfile.toString()); /* Load initial problem */ loadInitialProblem(model); /* Add new row */ addNewRow(); /* Add new column */ addNewColumn(); /* Output the revised matrix */ probg.writeProb("revised"); /* Solve the MIP */ solveProblem(); /* Display solution */ printSolution(); /* Write solution to an ASCII file */ probg.writeSol("revised"); } /** * Loads the initial trimloss matrix into the Xpress-Optimizer problem **/ private void loadInitialProblem(String model) { probg.readProb(model); } /** * Adds a new row to the matrix **/ private void addNewRow() { int nCol = probg.attributes().getCols(); /* nCol = number of columns in the matrix */ int nNewRows = 1; /* Adding 1 new row */ double[] dRHS = {0.0}; int[] nRowStart = {0}; byte[] sRowType = {'E'}; /* Store the coefficients of x(p); */ int nRowElem = nCol; /* New row has a coefficient in every column */ double[] pRowVal = new double[nRowElem]; int[] pColInd = new int[nRowElem]; for (int i=0;i<nCol;i++) {
			pRowVal[i] = 1;
			pColInd[i] = i;
		}
		/* Add the new row */
		probg.addRows( nNewRows, nRowElem, sRowType, dRHS, null, nRowStart, pColInd, pRowVal );
	
		/* New row will have been added at end of matrix */
		this.addedRowIndex = probg.attributes().getRows() - 1;
		
		/* Set new row name */
		String[] newNames = {"NewRow"};
		probg.addNames( 1 /* 1 = row names */ , newNames, addedRowIndex, addedRowIndex );
	}
	
	/** 
	 * Adds a new column to the matrix
	 **/
	private void addNewColumn() {
		int nRow = probg.attributes().getRows(); /* nRow = number of rows in the matrix */
		int nNewCols = 1; /* Adding 1 new column */
		int nColElem = 1; /* New column has a single nonzero */
		double[] dObjCoef = {0.0};
		int[] nColStart = {0};
		double[] pColVal = {-1}; /* The coefficient of -y */
		int[] pRowInd = {nRow-1};
		double[] dLowerBnd = {0.0};
		double[] dUpperBnd = {90}; /* The sum of the bounds on the x(p) */
		
		/* Add the new column */
		probg.addCols( nNewCols, nColElem, dObjCoef, nColStart, pRowInd, pColVal, dLowerBnd, dUpperBnd );
		
		/* New column will have been added at the end of the matrix */
		this.addedColIndex = probg.attributes().getCols() - 1;
		
		/* Make the new column an integer variable */
		int nTypeChg = 1;
		int[] nNewColInd = {addedColIndex};
		byte[] sColType = {'I'};
		probg.chgColType( nTypeChg, nNewColInd, sColType );
		
		/* Set the new column name */
		String[] newNames = {"y"};
		probg.addNames( 2 /* 2 = column names */, newNames, addedColIndex, addedColIndex );
	}
	
	/**
	 * Solves the mixed-integer problem
	 **/
	private void solveProblem() {
		/* Turn off presolve and permit no cuts */
		probg.controls().setPresolve(0);
		probg.controls().setCutStrategy(0);

		/* Assign the y variable a high branching priority (1)
		   - by default the original x(p) variables have low priority (500) */
		int[] nNewColIndex = {addedRowIndex};
		int[] nColPrior = {1};
		/* Load this directive */
		probg.loadDirs( 1, nNewColIndex, nColPrior, null, null, null );
		
		probg.mipOptimize();
		
	}
	
	/** 
	 * Displays the solution
	 **/
	private void printSolution() {
		if (probg.attributes().getMIPStatus()==XPRSenumerations.MIPStatus.OPTIMAL) {
			System.out.println("  Problem solved to optimality");

			int nCol = probg.attributes().getCols();
			/* Get the primal solution values */
			double[] x = probg.getMipSolX();

			/* Display the objective and solution values */
			System.out.println();
			System.out.println("  Minimum loss is " + probg.attributes().getMIPObjVal());
			System.out.println();
			System.out.println("  The solution values are");
			for (int i=0;i<nCol-1;i++)
				System.out.println("    x("+(i+1)+") = "+x[i]);
			System.out.println("    y    = "+x[nCol-1]);
		}
		else {
			System.out.println("  Problem was "+probg.attributes().getMIPStatus());
		}
	}
		
}