迭代多维数组

Question

Suppose I want to do something for every number in a multi-dimensional array. I've found that you can get a pointer for the first number and then use use pointer addition. For example, the following code prints the numbers 1 through 12 :

double a[2][3][2] = {{{1, 2}, {3, 4}, {5, 6}}, {{7, 8}, {9, 10}, {11, 12}}};
double *p = &a[0][0][0];
for (int i = 0; i < 12; i++)
    cout << *(p + i) << endl;

Is it unidiomatic to consider a multi-dimensional array as being flat in this way? If so, what is the preferred way of doing it? Also, is there a simpler way to write double *p = &a[0][0][0]; to get a pointer for the first number in a multi-dimensional array (in the same way that you can just write double *p = a; for a one-dimensional array)?

Answer 1

Yes, a multidimensional array is guaranteed to be flat. However, it may be best to kind this sort of thing. If you want to flatwise iterate over a multidimensional array, I think it'd be better to introduce a range view into it:

template <typename T>
struct Flat {
    auto begin() { return first(arr); }
    auto end() {
        return begin() + sizeof(arr)/sizeof(*begin());
    }

    template <typename X> X* first(X& val) { return &val; }
    template <typename X, size_t N> auto first(X(&val)[N]) { return first(*val); }

    T& arr;
};

template <typename T>
Flat<T> flatten(T& arr) {
    return Flat<T>{arr};
}

And simply use that one:

for (double d : flatten(a)) {
    std::cout << d << std::endl;
}

Otherwise, the only other way of declaring p is something like double *p = &***a; I'm not sure if becoming a three star programmer is high on your life achievement list.

Answer 2

While it can be very useful to know that a multidimensional array is actually flat, it would usually be unidiomatic to refer to it using pointers and pointer arithmetic, because it introduces more potential for error and is harder to read than the idiomatic solution, subscript notation. Because of that, I would recommend using this:

double a[2][3][2] = {{{1, 2}, {3, 4}, {5, 6}}, {{7, 8}, {9, 10}, {11, 12}}};

for (int i = 0; i < 2; i++)
{
    for (int j = 0; j < 3; j++)
    {
        for (int k = 0; k < 2; k++)
        {
            cout << a[i][j][k] << endl;
        }
    }
}

Answer 3

Yes, a multi-dimensional array can always be treated as flat. Also, same as a one dimensional array you can say double *p = reinterpret_cast<double*>(a) which is the same as double *p = &a[0][0][0] .

A multi-dimensional array, if dynamically allocated, may not be flat. But then that would be obvious as the allocation will be done by you.