interesting pointer cast problem

[Gerald]

There are several ways to build 2D array in C++.
Usually I do it using following code:

// ========================================
int rows = 10, columns = 10;
int *array1d = new int[rows*columns];
int (*array2d)[columns] = reinterpret_cast<int (*)[columns]>(array1d);
/* Now we can use array2d like this - array2d[j] */
// ========================================

The key statement is
int (*array2d)[columns] = reinterpret_cast<int (*)
[columns]>(array1d).
I wonder whether it complies with C++ spec and how it is supposed to
work.

Thanks
Gerald

 

[Ian Collins]

There are several ways to build 2D array in C++.
Usually I do it using following code:

// ========================================
int rows = 10, columns = 10;
int *array1d = new int[rows*columns];
int (*array2d)[columns] = reinterpret_cast<int (*)[columns]>(array1d);
/* Now we can use array2d like this - array2d[j] */
// ========================================

The key statement is
int (*array2d)[columns] = reinterpret_cast<int (*)
[columns]>(array1d).
I wonder whether it complies with C++ spec and how it is supposed to
work.

No, an array type requires a constant size (columns isn't).

Why don't you use a vector of vectors?

 

[Ian Collins]

Because it's slower to instantiate and consumes more memory than a one
single array used as a two-dimensional one?

But vectors are also safer and scale better. If the size is small, the
extra memory overhead probably doesn't matter. If the size is large,
the extra memory overhead is insignificant.