Main Page   Namespace List   Compound List   File List   Namespace Members   Compound Members   File Members  

TNT::Array2D Class Template Reference

#include <tnt_array2d.h>

List of all members.

Public Types

typedef T value_type

Public Methods

 Array2D ()
 Array2D (int m, int n)
 Array2D (int m, int n, T *a)
 Array2D (int m, int n, const T &a)
 Array2D (const Array2D &A)
 operator T ** ()
 operator const T ** ()
Array2D& operator= (const T &a)
Array2D& operator= (const Array2D &A)
Array2D& ref (const Array2D &A)
Array2D copy () const
Array2D& inject (const Array2D &A)
T* operator[] (int i)
const T* operator[] (int i) const
int dim1 () const
int dim2 () const
 ~Array2D ()


Detailed Description

template<class T> class TNT::Array2D

Templated two-dimensional, numerical array which looks like a conventional C multiarray. Storage corresponds to C (row-major) ordering. Elements are accessed via A[i][j] notation.

Array assignment is by reference (i.e. shallow assignment). That is, B=A implies that the A and B point to the same array, so modifications to the elements of A will be reflected in B. If an independent copy is required, then B = A.copy() can be used. Note that this facilitates returning arrays from functions without relying on compiler optimizations to eliminate extensive data copying.

The indexing and layout of this array object makes it compatible with C and C++ algorithms that utilize the familiar C[i][j] notation. This includes numerous textbooks, such as Numercial Recipes, and various public domain codes.


Member Typedef Documentation

template<class T>
typedef T TNT::Array2D<T>::value_type
 

Used to determined the data type of array entries. This is most commonly used when requiring scalar temporaries in templated algorithms that have TNT arrays as input. For example,

	template <class ArrayTwoD>
	void foo(ArrayTwoD &A)
	{
		A::value_type first_entry = A[0][0];
		...
	}
	


Constructor & Destructor Documentation

template<class T>
TNT::Array2D<T>::Array2D<T> ( )
 

Create a null array. This is not the same as Array2D(0,0), which consumes some memory overhead.

template<class T>
TNT::Array2D<T>::Array2D<T> ( int m,
int n )
 

Copy constructor. Array data is NOT copied, but shared. Thus, in Array2D B(A), subsequent changes to A will be reflected in B. For an indepent copy of A, use Array2D B(A.copy()), or B = A.copy(), instead.

template<class T>
TNT::Array2D<T>::Array2D<T> ( int m,
int n,
T * a )
 

Create a new (m x n) array, WIHOUT initializing array elements. To create an initialized array of constants, see Array2D(m,n,value).

This version avoids the O(m*n) initialization overhead and is used just before manual assignment.

Parameters:
m   the first (row) dimension of the new matrix.
n   the second (column) dimension of the new matrix.

template<class T>
TNT::Array2D<T>::Array2D<T> ( int m,
int n,
const T & a )
 

Create a new (m x n) array, initializing array elements to constant specified by argument. Most often used to create an array of zeros, as in A(m, n, 0.0).

Parameters:
m   the first (row) dimension of the new matrix.
n   the second (column) dimension of the new matrix.
val   the constant value to set all elements of the new array to.

template<class T>
TNT::Array2D<T>::Array2D<T> ( const Array2D<T> & A ) [inline]
 

Create a new (m x n) array, as a view of an existing one-dimensional array stored in C order, i.e. right-most dimension varying fastest. (Often referred to as "row-major" ordering.) Note that the storage for this pre-existing array will never be garbage collected by the Array2D class.

Parameters:
m   the first (row) dimension of the new matrix.
n   the second (column) dimension of the new matrix.
a   the one dimensional C array to use as data storage for the array.

template<class T>
TNT::Array2D<T>::~Array2D<T> ( )
 

Destructor. Reclaim resouces associated with this view of the array data. Note that elements are still intact if other array views of this data exist.


Member Function Documentation

template<class T>
Array2D<T> TNT::Array2D<T>::copy ( ) const
 

Create a new copy of existing matrix. Used in B = A.copy() or in the construction of B, e.g. Array2D B(A.copy()), to create a new array that does not share data.

template<class T>
int TNT::Array2D<T>::dim1 ( ) const [inline]
 

Returns:
the size of the first dimension of the array, i.e. the number of rows.

template<class T>
int TNT::Array2D<T>::dim2 ( ) const [inline]
 

Returns:
the size of the second dimension of the array, i.e. the number of columns.

template<class T>
Array2D<T> & TNT::Array2D<T>::inject ( const Array2D<T> & A )
 

Copy the elements to from one array to another, in place. That is B.inject(A), both A and B must conform (i.e. have identical row and column dimensions).

This differs from B = A.copy() in that references to B before this assignment are also affected. That is, if we have

	Array2D A(m,n);
	Array2D C(m,n);
	Array2D B(C);        // elements of B and C are shared. 

then B.inject(A) affects both and C, while B=A.copy() creates a new array B which shares no data with C or A.
Parameters:
A   the array from elements will be copied
Returns:
an instance of the modifed array. That is, in B.inject(A), it returns B. If A and B are not conformat, no modifications to B are made.

template<class T>
TNT::Array2D<T>::operator T ** ( ) [inline]
 

Automatic conversion to a two-dimensional C array. Useful for integrating with numerical codes that utilize the A[i][j] interface. Thus, if a function declared as

	void row_max(double **D);
is called with an Array2D<double>, as in
	Array2D<double> A(M,N);
	f(A);
then A is automatically converted. Likewise,
	double **d = A;
also converts A into a regular C pointer.

template<class T>
TNT::Array2D<T>::operator const T ** ( ) [inline]
 

Automatic conversion to a const two-dimensional C array. Useful for integrating with numerical codes that utilize the A[i][j] interface. Thus, if a function declared as

	void row_max(const double **D);
is called with an Array2D<double>, as in
	Array2D<double> A(M,N);
	f(A);
then A is automatically converted. Likewise,
	const double **d = A;
also converts A into a regular C pointer.

template<class T>
Array2D<T> & TNT::Array2D<T>::operator= ( const Array2D<T> & A ) [inline]
 

B = A is shorthand notation for B.ref(A).

template<class T>
Array2D<T> & TNT::Array2D<T>::operator= ( const T & a ) [inline]
 

Assign all elemnts of A to a constant scalar.

template<class T>
const T * TNT::Array2D<T>::operator[] ( int i ) const [inline]
 

template<class T>
T * TNT::Array2D<T>::operator[] ( int i ) [inline]
 

Used for A[i][j] indexing. The first [] operator returns a conventional pointer which can be dereferenced using the same [] notation.

If TNT_BOUNDS_CHECK macro is define, the left-most index (row index) is checked that it falls within the array bounds (via the assert() macro.) Note that bounds checking can occur in the row dimension, but the not column, since this is just a C pointer.

template<class T>
Array2D<T> & TNT::Array2D<T>::ref ( const Array2D<T> & A ) [inline]
 

Create a reference (shallow assignment) to another existing array. In B.ref(A), B and A shared the same data and subsequent changes to the array elements of one will be reflected in the other.

This is what operator= calls, and B=A and B.ref(A) are equivalent operations.

Returns:
The new referenced array: in B.ref(A), it returns B.


The documentation for this class was generated from the following file:
Generated at Thu Jun 26 17:26:22 2003 for Template Numerical Toolkit (TNT) by doxygen1.2.5 written by Dimitri van Heesch, © 1997-2001