integer_sequence Class

Article
08/03/2021

Represents an integer sequence. Can be used to deduce and expand parameter packs in variadic types such as std::tuple<T...> that are passed as arguments to a function.

Syntax

template <class T, T... Vals>
struct integer_sequence

Parameters

T
The type of the values; must be an integral type: bool, char, char16_t, char32_t, wchar_t, or signed or unsigned integer types.

Vals
A non-type parameter pack that represents a sequence of values of integral type T.

Members

Name	Description
`static size_t size() noexcept`	The number of elements in the sequence.
`typedef T value_type`	The type of each element in the sequence. Must be an integral type.

Remarks

A parameter pack that is passed directly to a function can be unpacked without any special library helpers. When a parameter pack is part of a type that is passed to a function, and you need indices to access the elements, then the easiest way to unpack it is to use integer_sequence and its related type aliases make_integer_sequence, index_sequence, make_index_sequence, and index_sequence_for.

Example

The following example is based on the original proposal N3658. It shows how to use an integer_sequence to create a std::tuple from a std::array<T,N>, and how to use an integer_sequence to get at the tuple members.

In the a2t function, an index_sequence is an alias of integer_sequence based on the size_t integral type. make_index_sequence is an alias that at compile time creates a zero-based index_sequence with the same number of elements as the array that is passed in by the caller. a2t passes the index_sequence by value to a2t_ , where the expression a[I]... unpacks I, and then the elements are being fed to make_tuple which consumes them as individual arguments. For example, if the sequence contains three elements, then make_tuple is called as make_tuple(a[0], a[1], a[2]). The array elements themselves can of course be any type.

The apply function accepts a std::tuple, and produces an integer_sequence by using the tuple_size helper class. Note that std::decay_t is necessary because tuple_size does not work with reference types. The apply_ function unpacks the tuple members and forwards them as separate arguments to a function call. In this example the function is a simple lambda expression that prints out the values.

#include <stddef.h>
#include <iostream>
#include <tuple>
#include <utility>
#include <array>
#include <string>

using namespace std;

// Create a tuple from the array and the index_sequence
template<typename Array, size_t... I>
auto a2t_(const Array& a, index_sequence<I...>)
{
    return make_tuple(a[I]...);
}

// Create an index sequence for the array, and pass it to the
// implementation function a2t_
template<typename T, size_t N>
auto a2t(const array<T, N>& a)
{
    return a2t_(a, make_index_sequence<N>());
}

// Call function F with the tuple members as separate arguments.
template<typename F, typename Tuple = tuple<T...>, size_t... I>
decltype(auto) apply_(F&& f, Tuple&& args, index_sequence<I...>)
{
    return forward<F>(f)(get<I>(forward<Tuple>(args))...);
}

// Create an index_sequence for the tuple, and pass it with the
// function object and the tuple to the implementation function apply_
template<typename F, typename Tuple = tuple<T...>>
decltype(auto) apply(F&& f, Tuple&& args)
{
    using Indices = make_index_sequence<tuple_size<decay_t<Tuple>>::value >;
    return apply_(forward<F>(f), forward<Tuple>(args), Indices());
}

int main()
{
    const array<string, 3> arr { "Hello", "from", "C++14" };

    //Create a tuple given a array
    auto tup = a2t(arr);

    // Extract the tuple elements
    apply([](const string& a, const string& b, const string& c) {cout << a << " " << b << " " << c << endl; }, tup);

    char c;
    cin >> c;
}

To make an index_sequence for a parameter pack, use index_sequence_for<T...> which is an alias for make_index_sequence<sizeof...(T)>

Requirements

Header: <type_traits>

Namespace: std