Matrix.h

Go to the documentation of this file.

// # Matrix.h: A 2-D Specialization of the Array Class
// # Copyright (C) 1993,1994,1995,1996,1999,2000,2001,2003
// # Associated Universities, Inc. Washington DC, USA.
// #
// # This library is free software; you can redistribute it and/or modify it
// # under the terms of the GNU Library General Public License as published by
// # the Free Software Foundation; either version 2 of the License, or (at your
// # option) any later version.
// #
// # This library is distributed in the hope that it will be useful, but WITHOUT
// # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
// # FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Library General Public
// # License for more details.
// #
// # You should have received a copy of the GNU Library General Public License
// # along with this library; if not, write to the Free Software Foundation,
// # Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
// #
// # Correspondence concerning AIPS++ should be addressed as follows:
// #        Internet email: casa-feedback@nrao.edu.
// #        Postal address: AIPS++ Project Office
// #                        National Radio Astronomy Observatory
// #                        520 Edgemont Road
// #                        Charlottesville, VA 22903-2475 USA
 
#ifndef CASA_MATRIX_2_H
#define CASA_MATRIX_2_H
 
// # Includes
#include "Array.h"
 
namespace casacore {  // #Begin casa namespace
 
// <summary> A 2-D Specialization of the Array class </summary>
// <reviewed reviewer="UNKNOWN" date="before2004/08/25" tests="" demos="">
// </reviewed>
//
// Matrix objects are two-dimensional specializations (e.g., more convenient
// and efficient indexing) of the general Array class. You might also want
// to look at the Array documentation to see inherited functionality. A
// tutorial on using the array classes in general is available in the
// "AIPS++ Programming Manual".
//
// Generally the member functions of Array are also available in
// Matrix versions which take a pair of integers where the array
// needs an IPosition. Since the Matrix
// is two-dimensional, the IPositions are overkill, although you may
// use those versions if you want to.
// <srcblock>
// Matrix<int> mi(100,100);  // Shape is 100x100
// mi.resize(50,50);         // Shape now 50x50
// </srcblock>
//
// Slices may be taken with the Slice class. To take a slice, one "indexes"
// with one Slice(start, length, inc) for each axis,
// where end and inc are optional.
// Additionally, there are row(), column() and diagonal()
// member functions which return Vector's which refer to the storage back
// in the Matrix:
// <srcblock>
// Matrix<float> mf(100, 100);
// mf.diagonal() = 1;
// </srcblock>
//
// Correct indexing order of a matrix is:
// <srcblock>
// Matrix<int> mi(n1,n2)                  // [nrow, ncolumn]
// for (size_t j=0; j<mi.ncolumn(); j++) {
//    for (size_t i=0; i<mi.nrow(); i++) {
//       mi(i,j) = i*j;
//    }
// }
// </srcblock>
//
//
// Element-by-element arithmetic and logical operations are available (in
// aips/ArrayMath.h and aips/ArrayLogical.h). Other Matrix operations (e.g.
// LU decomposition) are available, and more appear periodically.
//
// As with the Arrays, if the preprocessor symbol AIPS_DEBUG is
// defined at compile time invariants will be checked on entry to most
// member functions. Additionally, if AIPS_ARRAY_INDEX_CHECK is defined
// index operations will be bounds-checked. Neither of these should
// be defined for production code.
 
template <typename T>
class Matrix : public Array<T> {
 public:
  // A Matrix of length zero in each dimension; zero origin.
  Matrix();
 
  // A Matrix with "l1" rows and "l2" columns.
  // Fill it with the initial value.
  Matrix(size_t l1, size_t l2, const T &initialValue = T());
 
  // An uninitialized Matrix with "l1" rows and "l2" columns.
  Matrix(size_t l1, size_t l2, typename Array<T>::uninitializedType);
 
  // A matrix of shape with shape "len".
  // Fill it with the initial value.
  Matrix(const IPosition &len, const T &initialValue = T());
 
  // An uninitialized matrix of shape with shape "len".
  Matrix(const IPosition &len, typename Array<T>::uninitializedType);
 
  // The copy/move constructor uses reference semantics.
  Matrix(const Matrix<T> &source);
  Matrix(Matrix<T> &&source);
 
  // Construct a Matrix by reference from "source". "source must have
  // ndim() of 2 or less.
  Matrix(const Array<T> &source);
  Matrix(Array<T> &&source);
 
  // Create an Matrix of a given shape from a pointer.
  Matrix(const IPosition &shape, T *storage, StorageInitPolicy policy = COPY);
  // Create an Matrix of a given shape from a pointer. Because the pointer
  // is const, a copy is always made.
  Matrix(const IPosition &shape, const T *storage);
 
  // Create an identity matrix of side length n. (Could not do this as a constructor
  // because of ambiguities with other constructors).
  static Matrix<T> identity(size_t n);
 
  // Resize to the given shape
  // <group>
  using Array<T>::resize;
  void resize(size_t nx, size_t ny, bool copyValues = false);
  // </group>
 
  Matrix<T> &operator=(const Matrix<T> &source) { return assign_conforming(source); }
  Matrix<T> &operator=(Matrix<T> &&source) { return assign_conforming(std::move(source)); }
  Matrix<T> &operator=(const Array<T> &source) { return assign_conforming(source); }
  Matrix<T> &operator=(Array<T> &&source) { return assign_conforming(std::move(source)); }
 
  // Copy the values from other to this Matrix. If this matrix has zero
  // elements then it will resize to be the same shape as other; otherwise
  // other must conform to this.
  // Note that the assign function can be used to assign a
  // non-conforming matrix.
  // <group>
  Matrix<T> &assign_conforming(const Matrix<T> &source) {
    Array<T>::assign_conforming(source);
    return *this;
  }
  Matrix<T> &assign_conforming(Matrix<T> &&source) {
    Array<T>::assign_conforming(std::move(source));
    return *this;
  }
 
  Matrix<T> &assign_conforming(const Array<T> &source) {
    // TODO Should be supported by the Array class,
    // see Cube::operator=(const Array&)
 
    if (source.ndim() == 2) {
      Array<T>::assign_conforming(source);
    } else {
      // This might work if a.ndim == 1 or 2
      (*this) = Matrix<T>(source);
    }
    return *this;
  }
 
  Matrix<T> &assign_conforming(Array<T> &&source) {
    if (source.ndim() == 2) {
      Array<T>::assign_conforming(std::move(source));
    } else {
      (*this) = Matrix<T>(std::move(source));
    }
    return *this;
  }
  // </group>
 
  // Copy val into every element of this Matrix; i.e. behaves as if
  // val were a constant conformant matrix.
  Array<T> &operator=(const T &val) { return Array<T>::operator=(val); }
 
  // Copy to this those values in marray whose corresponding elements
  // in marray's mask are true.
  Matrix<T> &assign_conforming(const MaskedArray<T> &marray) {
    Array<T>::assign_conforming(marray);
    return *this;
  }
 
  // Single-pixel addressing. If AIPS_ARRAY_INDEX_CHECK is defined,
  // bounds checking is performed.
  // <group>
  T &operator()(const IPosition &i) { return Array<T>::operator()(i); }
  const T &operator()(const IPosition &i) const { return Array<T>::operator()(i); }
  T &operator()(size_t i1, size_t i2) {
    return this->contiguous_p ? this->begin_p[i1 + i2 * yinc()]
                              : this->begin_p[i1 * xinc() + i2 * yinc()];
  }
 
  const T &operator()(size_t i1, size_t i2) const {
    return this->contiguous_p ? this->begin_p[i1 + i2 * yinc()]
                              : this->begin_p[i1 * xinc() + i2 * yinc()];
  }
  // </group>
 
  // The array is masked by the input LogicalArray.
  // This mask must conform to the array.
  // <group>
 
  // Return a MaskedArray.
  MaskedArray<T> operator()(const LogicalArray &mask) const { return Array<T>::operator()(mask); }
 
  // Return a MaskedArray.
  MaskedArray<T> operator()(const LogicalArray &mask) { return Array<T>::operator()(mask); }
 
  // </group>
 
  // The array is masked by the input MaskedLogicalArray.
  // The mask is effectively the AND of the internal LogicalArray
  // and the internal mask of the MaskedLogicalArray.
  // The MaskedLogicalArray must conform to the array.
  // <group>
 
  // Return a MaskedArray.
  MaskedArray<T> operator()(const MaskedLogicalArray &mask) const {
    return Array<T>::operator()(mask);
  }
 
  // Return a MaskedArray.
  MaskedArray<T> operator()(const MaskedLogicalArray &mask) { return Array<T>::operator()(mask); }
 
  // </group>
 
  // Returns a reference to the i'th row.
  // <group>
  Vector<T> row(size_t i);
  const Vector<T> row(size_t i) const;
  // </group>
 
  // Returns a reference to the j'th column
  // <group>
  Vector<T> column(size_t j);
  const Vector<T> column(size_t j) const;
  // </group>
 
  // Returns a diagonal from the Matrix. The Matrix must be square.
  // n==0 is the main diagonal. n>0 is above the main diagonal, n<0
  // is below it.
  // <group>
  Vector<T> diagonal(long long n = 0);
  const Vector<T> diagonal(long long n = 0) const;
  // </group>
 
  // Take a slice of this matrix. Slices are always indexed starting
  // at zero. This uses reference semantics, i.e. changing a value
  // in the slice changes the original.
  // <srcblock>
  // Matrix<double> vd(100,100);
  // //...
  // vd(Slice(0,10),Slice(10,10)) = -1.0; // 10x10 sub-matrix set to -1.0
  // </srcblock>
  // <group>
  Matrix<T> operator()(const Slice &sliceX, const Slice &sliceY);
  const Matrix<T> operator()(const Slice &sliceX, const Slice &sliceY) const;
  // </group>
 
  // Slice using IPositions. Required to be defined, otherwise the base
  // class versions are hidden.
  // <group>
  Array<T> operator()(const IPosition &blc, const IPosition &trc, const IPosition &incr) {
    return Array<T>::operator()(blc, trc, incr);
  }
  const Array<T> operator()(const IPosition &blc, const IPosition &trc,
                            const IPosition &incr) const {
    return Array<T>::operator()(blc, trc, incr);
  }
  Array<T> operator()(const IPosition &blc, const IPosition &trc) {
    return Array<T>::operator()(blc, trc);
  }
  const Array<T> operator()(const IPosition &blc, const IPosition &trc) const {
    return Array<T>::operator()(blc, trc);
  }
  Array<T> operator()(const Slicer &slicer) { return Array<T>::operator()(slicer); }
  const Array<T> operator()(const Slicer &slicer) const { return Array<T>::operator()(slicer); }
  // </group>
 
  // The length of each axis of the Matrix.
  const IPosition &shape() const { return this->length_p; }
  void shape(int &s1, int &s2) const {
    s1 = this->length_p(0);
    s2 = this->length_p(1);
  }
 
  // The number of rows in the Matrix, i.e. the length of the first axis.
  size_t nrow() const { return this->length_p(0); }
 
  // The number of columns in the Matrix, i.e. the length of the 2nd axis.
  size_t ncolumn() const { return this->length_p(1); }
 
  // Checks that the Matrix is consistent (invariants check out).
  virtual bool ok() const override;
 
 protected:
  virtual void preTakeStorage(const IPosition &shape) override;
  // Remove the degenerate axes from other and store result in this matrix.
  // An exception is thrown if removing degenerate axes does not result
  // in a matrix.
  virtual void doNonDegenerate(const Array<T> &other, const IPosition &ignoreAxes) override;
 
  virtual size_t fixedDimensionality() const override { return 2; }
 
 private:
  // Cached constants to improve indexing.
  // size_t xinc_p, yinc_p;
 
  size_t xinc() const { return this->inc_p(0); }
  size_t yinc() const { return this->inc_p(1) * this->originalLength_p(0); }
};
 
// # Declare extern templates for often used types.
extern template class Matrix<bool>;
extern template class Matrix<float>;
extern template class Matrix<double>;
 
}  // namespace casacore
 
#include "Matrix.tcc"
 
#endif