从c++ 11开始，引入了一些新的模板来帮助减少处理数组长度时的痛苦。它们都定义在header <type_traits>中。

std::rank<T>::value If T is an array type, provides the member constant value equal to the number of dimensions of the array. For any other type, value is 0. std::extent<T, N>::value If T is an array type, provides the member constant value equal to the number of elements along the Nth dimension of the array, if N is in [0, std::rank<T>::value). For any other type, or if T is array of unknown bound along its first dimension and N is 0, value is 0. std::remove_extent<T>::type If T is an array of some type X, provides the member typedef type equal to X, otherwise type is T. Note that if T is a multidimensional array, only the first dimension is removed. std::remove_all_extents<T>::type If T is a multidimensional array of some type X, provides the member typedef type equal to X, otherwise type is T.

要获得多维数组的任何维度上的长度，可以使用decltype与std::extent结合使用。例如:

#include <iostream>
#include <type_traits> // std::remove_extent std::remove_all_extents std::rank std::extent

template<class T, size_t N>
constexpr size_t length(T(&)[N]) { return N; }

template<class T, size_t N>
constexpr size_t length2(T(&arr)[N]) { return sizeof(arr) / sizeof(*arr); }

int main()
{
    int a[5][4][3]{{{1,2,3}, {4,5,6}}, { }, {{7,8,9}}};

    // New way
    constexpr auto l1 = std::extent<decltype(a)>::value;     // 5
    constexpr auto l2 = std::extent<decltype(a), 1>::value;  // 4
    constexpr auto l3 = std::extent<decltype(a), 2>::value;  // 3
    constexpr auto l4 = std::extent<decltype(a), 3>::value;  // 0

    // Mixed way
    constexpr auto la = length(a);
    //constexpr auto lpa = length(*a);  // compile error
    //auto lpa = length(*a);  // get at runtime
    std::remove_extent<decltype(a)>::type pa;  // get at compile time
    //std::remove_reference<decltype(*a)>::type pa;  // same as above
    constexpr auto lpa = length(pa);
    std::cout << la << ' ' << lpa << '\n';

    // Old way
    constexpr auto la2 = sizeof(a) / sizeof(*a);
    constexpr auto lpa2 = sizeof(*a) / sizeof(**a);
    std::cout << la2 << ' ' << lpa2 << '\n';

    return 0;
}

BTY，获取多维数组中元素的总数:

constexpr auto l = sizeof(a) / sizeof(std::remove_all_extents<decltype(a)>::type);

或者把它放在一个函数模板中:

#include <iostream>
#include <type_traits>
    

template<class T>
constexpr size_t len(T &a)
{
    return sizeof(a) / sizeof(typename std::remove_all_extents<T>::type);
}

int main()
{
    int a[5][4][3]{{{1,2,3}, {4,5,6}}, { }, {{7,8,9}}};
    constexpr auto ttt = len(a);
    int i;
    std::cout << ttt << ' ' << len(i) << '\n';
    
    return 0;
}

更多如何使用它们的例子可以通过以下链接找到。

如果你指的是c风格的数组，那么你可以这样做:

int a[7];
std::cout << "Length of array = " << (sizeof(a)/sizeof(*a)) << std::endl;

这对指针不起作用(即它对以下任何一个都不起作用):

int *p = new int[7];
std::cout << "Length of array = " << (sizeof(p)/sizeof(*p)) << std::endl;

or:

void func(int *p)
{
    std::cout << "Length of array = " << (sizeof(p)/sizeof(*p)) << std::endl;
}

int a[7];
func(a);

在c++中，如果你想要这种行为，那么你应该使用容器类;可能std::向量。

虽然这是一个老问题，但值得将答案更新到c++ 17。在标准库中，现在有一个模板化函数std::size()，它返回std容器或c风格数组中的元素数量。例如:

#include <iterator>

uint32_t data[] = {10, 20, 30, 40};
auto dataSize = std::size(data);
// dataSize == 4

vector有一个size()方法，该方法返回vector中元素的数量。

(是的，这是半开玩笑的回答)

你可以简单地使用这段代码:

#include <iostream>
#include <string>
#include <array>

using namespace std;

int main()
{

  array<int,3> values;
  cout << "No. elements in valuea array: " << values.size() << " elements." << endl;
  cout << "sizeof(myints): " << sizeof(values) << endl;

}

这里是参考资料:http://www.cplusplus.com/reference/array/array/size/

下面是谷歌Protobuf中ArraySize的一个实现。

#define GOOGLE_ARRAYSIZE(a) \
  ((sizeof(a) / sizeof(*(a))) / static_cast<size_t>(!(sizeof(a) % sizeof(*(a)))))

// test codes...
char* ptr[] = { "you", "are", "here" };
int testarr[] = {1, 2, 3, 4};
cout << GOOGLE_ARRAYSIZE(testarr) << endl;
cout << GOOGLE_ARRAYSIZE(ptr) << endl;

ARRAYSIZE(arr) works by inspecting sizeof(arr) (the # of bytes in the array) and sizeof(*(arr)) (the # of bytes in one array element). If the former is divisible by the latter, perhaps arr is indeed an array, in which case the division result is the # of elements in the array. Otherwise, arr cannot possibly be an array, and we generate a compiler error to prevent the code from compiling. Since the size of bool is implementation-defined, we need to cast !(sizeof(a) & sizeof(*(a))) to size_t in order to ensure the final result has type size_t. This macro is not perfect as it wrongfully accepts certain pointers, namely where the pointer size is divisible by the pointee size. Since all our code has to go through a 32-bit compiler, where a pointer is 4 bytes, this means all pointers to a type whose size is 3 or greater than 4 will be (righteously) rejected.