Namespaces
Variants
Views
Actions

std::ranges::sample

From cppreference.com
< cpp‎ | algorithm‎ | ranges
Revision as of 12:14, 24 September 2023 by Andreas Krug (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
 
 
Algorithm library
Constrained algorithms and algorithms on ranges (C++20)
Constrained algorithms, e.g. ranges::copy, ranges::sort, ...
Execution policies (C++17)
Non-modifying sequence operations
Batch operations
(C++17)
Search operations
(C++11)                (C++11)(C++11)

Modifying sequence operations
Copy operations
(C++11)
(C++11)
Swap operations
Transformation operations
Generation operations
Removing operations
Order-changing operations
(until C++17)(C++11)
(C++20)(C++20)
Sampling operations
(C++17)

Sorting and related operations
Partitioning operations
Sorting operations
Binary search operations
(on partitioned ranges)
Set operations (on sorted ranges)
Merge operations (on sorted ranges)
Heap operations
Minimum/maximum operations
(C++11)
(C++17)
Lexicographical comparison operations
Permutation operations
C library
Numeric operations
Operations on uninitialized memory
 
Constrained algorithms
All names in this menu belong to namespace std::ranges
Non-modifying sequence operations
Modifying sequence operations
Partitioning operations
Sorting operations
Binary search operations (on sorted ranges)
       
       
Set operations (on sorted ranges)
Heap operations
Minimum/maximum operations
       
       
Permutation operations
Fold operations
Numeric operations
(C++23)            
Operations on uninitialized storage
Return types
 
Defined in header <algorithm>
Call signature
template< std::input_iterator I, std::sentinel_for<I> S,

          std::weakly_incrementable O, class Gen >
requires (std::forward_iterator<I> || std::random_access_iterator<O>) &&
          std::indirectly_copyable<I, O> &&
          std::uniform_random_bit_generator<std::remove_reference_t<Gen>>

O sample( I first, S last, O out, std::iter_difference_t<I> n, Gen&& gen );
(1) (since C++20)
template< ranges::input_range R, std::weakly_incrementable O, class Gen >

requires (ranges::forward_range<R> || std::random_access_iterator<O>) &&
          std::indirectly_copyable<ranges::iterator_t<R>, O> &&
          std::uniform_random_bit_generator<std::remove_reference_t<Gen>>

O sample( R&& r, O out, ranges::range_difference_t<R> n, Gen&& gen );
(2) (since C++20)
1) Selects M = min(n, last - first) elements from the sequence [firstlast) (without replacement) such that each possible sample has equal probability of appearance, and writes those selected elements into the range beginning at out.
The algorithm is stable (preserves the relative order of the selected elements) only if I models std::forward_iterator.
The behavior is undefined if out is in [firstlast).
2) Same as (1), but uses r as the source range, as if using ranges::begin(r) as first, and ranges::end(r) as last.

The function-like entities described on this page are niebloids, that is:

In practice, they may be implemented as function objects, or with special compiler extensions.

Contents

[edit] Parameters

first1, last1 - the range from which to make the sampling (the population)
r - the range from which to make the sampling (the population)
out - the output iterator where the samples are written
n - number of samples to take
gen - the random number generator used as the source of randomness

[edit] Return value

An iterator equal to out + M, that is the end of the resulting sample range.

[edit] Complexity

Linear: 𝓞(last - first).

[edit] Notes

This function may implement selection sampling or reservoir sampling.

[edit] Possible implementation

struct sample_fn
{
    template<std::input_iterator I, std::sentinel_for<I> S,
             std::weakly_incrementable O, class Gen>
    requires (std::forward_iterator<I> or
              std::random_access_iterator<O>) &&
              std::indirectly_copyable<I, O> &&
              std::uniform_random_bit_generator<std::remove_reference_t<Gen>>
    O operator()(I first, S last, O out, std::iter_difference_t<I> n, Gen&& gen) const
    {
        using diff_t = std::iter_difference_t<I>;
        using distrib_t = std::uniform_int_distribution<diff_t>;
        using param_t = typename distrib_t::param_type;
        distrib_t D{};
 
        if constexpr (std::forward_iterator<I>)
        {
            // this branch preserves "stability" of the sample elements
            auto rest{ranges::distance(first, last)};
            for (n = ranges::min(n, rest); n != 0; ++first)
                if (D(gen, param_t(0, --rest)) < n)
                {
                    *out++ = *first;
                    --n;
                }
            return out;
        }
        else
        {
            // O is a random_access_iterator
            diff_t sample_size{};
            // copy [first, first + M) elements to "random access" output
            for (; first != last && sample_size != n; ++first)
                out[sample_size++] = *first;
            // overwrite some of the copied elements with randomly selected ones
            for (auto pop_size{sample_size}; first != last; ++first, ++pop_size)
            {
                const auto i{D(gen, param_t{0, pop_size})};
                if (i < n)
                    out[i] = *first;
            }
            return out + sample_size;
        }
    }
 
    template<ranges::input_range R, std::weakly_incrementable O, class Gen>
    requires (ranges::forward_range<R> or std::random_access_iterator<O>) &&
              std::indirectly_copyable<ranges::iterator_t<R>, O> &&
              std::uniform_random_bit_generator<std::remove_reference_t<Gen>>
    O operator()(R&& r, O out, ranges::range_difference_t<R> n, Gen&& gen) const
    {
        return (*this)(ranges::begin(r), ranges::end(r), std::move(out), n,
                       std::forward<Gen>(gen));
    }
};
 
inline constexpr sample_fn sample {};

[edit] Example

#include <algorithm>
#include <iomanip>
#include <iostream>
#include <iterator>
#include <random>
#include <vector>
 
void print(auto const& rem, auto const& v)
{
    std::cout << rem << " = [" << std::size(v) << "] { ";
    for (auto const& e : v)
        std::cout << e << ' ';
    std::cout << "}\n";
}
 
int main()
{
    const auto in = {1, 2, 3, 4, 5, 6};
    print("in", in);
 
    std::vector<int> out;
    const int max = in.size() + 2;
    auto gen = std::mt19937{std::random_device{}()};
 
    for (int n{}; n != max; ++n)
    {
        out.clear();
        std::ranges::sample(in, std::back_inserter(out), n, gen);
        std::cout << "n = " << n;
        print(", out", out);
    }
}

Possible output:

in = [6] { 1 2 3 4 5 6 }
n = 0, out = [0] { }
n = 1, out = [1] { 5 }
n = 2, out = [2] { 4 5 }
n = 3, out = [3] { 2 3 5 }
n = 4, out = [4] { 2 4 5 6 }
n = 5, out = [5] { 1 2 3 5 6 }
n = 6, out = [6] { 1 2 3 4 5 6 }
n = 7, out = [6] { 1 2 3 4 5 6 }

[edit] See also

randomly re-orders elements in a range
(niebloid)[edit]
(C++17)
selects N random elements from a sequence
(function template) [edit]