Laboratoire ICAR » Plateforme TXM

/*

 * To change this template, choose Tools | Templates

 * and open the template in the editor.

 */

package JamaPlus;

import JamaPlus.SparseSVD.SparseSVDException;

import java.util.*;

import java.util.logging.Level;

import java.util.logging.Logger;

/**

 *

 * @author Bernard Victorri

 */

public class SparseMatrix {

  private int m, n, nnz;

  private int[] colpts, rowinds;

  private double[] vals;

  private double[] tempRow;  // stockage temporaire pour certains calculs (opb)

  public SparseMatrix(int m, int n) { // null matrix

    this.m = m;

    this.n = n;

    nnz = 0;

    colpts = new int[n+1];

    rowinds = new int[0];

    vals = new double[0];

  }

   private static class Triple implements Comparable<Triple> {

    int row;

    int col;

    double val;

    private Triple(int row, int col, double val) {

      this.row = row;

      this.col = col;

      this.val = val;

    }

    private Triple(int row, int col) {

      this.row = row;

      this.col = col;

    }

    public int compareTo(Triple t) {

      int res = col - t.col;

      if(res==0) res = row - t.row;

      return res;

    }

  }

 public SparseMatrix(int m, int n, ArrayList<Integer> rows, ArrayList<Integer> cols,

          ArrayList<Double> values) {

    this.m = m;

    this.n = n;

    nnz = rows.size();

    if (nnz!=cols.size()||nnz!=rows.size())

      throw new ArrayIndexOutOfBoundsException(

              "Size of lines, cols and values don't agree");

    colpts = new int[n+1];

    rowinds = new int[nnz];

    vals = new double[nnz];

    Triple[] triples = new Triple[nnz];

    for(int i=0; i<nnz; i++) 

     triples[i] = new Triple(rows.get(i), cols.get(i), values.get(i));    

    Arrays.sort(triples);

    colpts[0] = 0;

    int c = 0;

    for(int i=0; i<nnz; i++) {

      if(i>0 && triples[i].row == triples[i-1].row && triples[i].col == triples[i-1].col) 

        throw new ArrayIndexOutOfBoundsException(

              "Two values with same row and column numbers");

      while(c<triples[i].col) colpts[++c] = i;

      rowinds[i] = triples[i].row;

      vals[i] = triples[i].val;

    }

    while(c<n) colpts[++c] = nnz;

  }

public SparseMatrix(int m, int n, ArrayList<Integer> rows, ArrayList<Integer> cols,

          double value) {

    this.m = m;

    this.n = n;

    nnz = rows.size();

    if (nnz!=cols.size()||nnz!=rows.size())

      throw new ArrayIndexOutOfBoundsException(

              "Size of lines, cols and values don't agree");

    colpts = new int[n+1];

    rowinds = new int[nnz];

        vals = new double[nnz];

    Triple[] triples = new Triple[nnz];

    for(int i=0; i<nnz; i++) 

     triples[i] = new Triple(rows.get(i), cols.get(i));    

    Arrays.sort(triples);

    colpts[0] = 0;

    int c = 0;

    for(int i=0; i<nnz; i++) {

      if(i>0 && triples[i].row == triples[i-1].row && triples[i].col == triples[i-1].col) 

        throw new ArrayIndexOutOfBoundsException(

              "Two values with same row and column numbers");

      while(c<triples[i].col) colpts[++c] = i;

      rowinds[i] = triples[i].row;

      vals[i] = value;

    }

    while(c<n) colpts[++c] = nnz;

  }

  private SparseMatrix(int m, int n, int nnz, int[] colpts, int[] rowinds, double[] vals) {

    this.m = m;

    this.n = n;

    this.nnz = nnz;

    this.colpts = colpts;

    this.rowinds = rowinds;

    this.vals = vals;

  }

  public SparseMatrix(Matrix mat) {  // convert a dense matrix

    m = mat.getRowDimension();

    n = mat.getColumnDimension();

    long nbval = mat.getNumberNonZeroValues();

    if (nbval>Integer.MAX_VALUE) throw new ArrayIndexOutOfBoundsException(

              "Number of non-zero values exceeds 32 bits integer");

    nnz = (int) nbval;

    colpts = new int[n+1];

    rowinds = new int[nnz];

    vals = new double[nnz];

    int nval = 0;

    for (int j = 0; j<n; j++) {

      colpts[j] = nval;

      for (int i = 0; i<m; i++) if (mat.get(i, j)!=0) {

          rowinds[nval] = i;

          vals[nval++] = mat.get(i, j);

        }

    }

    colpts[n] = nnz;

  }

  public SparseMatrix copy() {

    int[] colpts1 = new int[n+1];

    int[] rowinds1 = new int[nnz];

    double[] vals1 = new double[nnz];

    System.arraycopy(colpts, 0, colpts1, 0, n+1);

    System.arraycopy(rowinds, 0, rowinds1, 0, nnz);

    System.arraycopy(vals, 0, vals1, 0, nnz);

    return new SparseMatrix(m, n, nnz, colpts1, rowinds1, vals1);

  }

  public int getRowDimension() {

    return m;

  }

  public int getColumnDimension() {

    return n;

  }

  public double get(int i, int j) {

    try {

      for (int nval = colpts[j]; nval<colpts[j+1]; nval++)

        if (rowinds[nval]==i) return vals[nval];

      return 0;

    } catch (ArrayIndexOutOfBoundsException ex) {

      throw new ArrayIndexOutOfBoundsException("Sparse matrix : "

              +ex.getLocalizedMessage());

    }

  }

  public SparseMatrix transpose() {

    int[] colpts1 = new int[m+1];

    int[] rowinds1 = new int[nnz];

    double[] vals1 = new double[nnz];

    for (int v = 0; v<nnz; v++) colpts1[rowinds[v]]++;  // nnz in each row

    colpts1[m] = nnz-colpts1[m-1];

    for (int r = m-1; r>0; r--)

      colpts1[r] = colpts1[r+1]-colpts1[r-1]; // starting point of the previous row

    colpts1[0] = 0;

    int i = 0;

    for (int c = 0; c<n; c++) while (i<colpts[c+1]) {

        int j = colpts1[rowinds[i]+1]++;  // j : index of the value in the new vectors

        rowinds1[j] = c;

        vals1[j] = vals[i++];

      }

    return new SparseMatrix(n, m, nnz, colpts1, rowinds1, vals1);

  }

  public void nulColumnSuppression() {

    ArrayList<Integer> colpts1 = new ArrayList<Integer>();

    colpts1.add(0);

    for (int i = 0; i<n; i++) if (colpts[i+1]>colpts[i])

        colpts1.add(colpts[i+1]);

    n = colpts1.size()-1;

    colpts = new int[n+1];

    for (int i = 0; i<=n; i++) colpts[i] = colpts1.get(i);

  }

  public double elementSum() {

    double s = 0;

    for (int i = 0; i<nnz; i++) s += vals[i];

    return s;

  }

  public Matrix lineSum() {

    double[] S = new double[m];

    for (int i = 0; i<nnz; i++) S[rowinds[i]] += vals[i];

    return new Matrix(S, m);

  }

  public Matrix columnSum() {

    double[] S = new double[n];

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        S[i] += vals[j];

    return new Matrix(S, 1);

  }

  public SparseMatrix timesEquals(double s) {

    for (int i = 0; i<nnz; i++) {

      vals[i] *= s;

    }

    return this;

  }

  public SparseMatrix firstColumnTimesEqual(Matrix B) {

    double[][] A = B.getArray();

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        vals[j] *= A[rowinds[j]][0];

    return this;

  }

  public SparseMatrix firstLineTimesEqual(Matrix B) {

    double[][] A = B.getArray();

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        vals[j] *= A[0][i];

    return this;

  }

  public void opa(double x[], double y[]) {

    // Computes y = Ax (x et y : length n)

    Arrays.fill(y, 0);

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        y[rowinds[j]] += vals[j]*x[i];

  }

  public void opb(double x[], double y[]) {

    // Computes y = A'Ax (x et y : length n)

    if (tempRow==null) tempRow = new double[m];

    else Arrays.fill(tempRow, 0);

    Arrays.fill(y, 0);

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        tempRow[rowinds[j]] += vals[j]*x[i];

    for (int i = 0; i<n; i++) for (int j = colpts[i]; j<colpts[i+1]; j++)

        y[i] += vals[j]*tempRow[rowinds[j]];

  }

  public SparseSVD svd(int dim) throws SparseSVD.SparseSVDException {

    return new SparseSVD(this, dim);

  }

  /** Analyse factorielle de correspondances

  @return     Analyse factorielle de correspondances

  @see Analyse factorielle de correspondances

   */

  public SparseAFC afc() {

    try {

      return new SparseAFC(this);

    } catch (SparseSVDException ex) {

      return null;

    }

  }

}