std::adjacent_difference(3)

🎀 std::adjacent_difference 🎀

okay so bestie, this is the std::adjacent_difference(3) page from the C++ Standard Libary manual, and we are literally about to make adjacent differences make sense. ✨

🌸 NAME 🌸

std::adjacent_difference — std::adjacent_difference. yes, the name is the name, iykyk 💅

✨ Synopsis ✨

Defined in header <numeric>.

• (1) constexpr since C++20:

template< class InputIt, class OutputIt >
OutputIt adjacent_difference( InputIt first, InputIt last,
                              OutputIt d_first );

• (2) since C++17:

template< class ExecutionPolicy, class ForwardIt1, class ForwardIt2 >
ForwardIt2 adjacent_difference( ExecutionPolicy&& policy,
                                ForwardIt1 first, ForwardIt1 last,
                                ForwardIt2 d_first );

• (3) constexpr since C++20:

template< class InputIt, class OutputIt, class BinaryOp >
OutputIt adjacent_difference( InputIt first, InputIt last,
                              OutputIt d_first, BinaryOp op );

• (4) since C++17:

template< class ExecutionPolicy, class ForwardIt1, class ForwardIt2, class BinaryOp >
ForwardIt2 adjacent_difference( ExecutionPolicy&& policy,
                                ForwardIt1 first, ForwardIt1 last,
                                ForwardIt2 d_first, BinaryOp op );

🦋 What the overloads actually do 🦋

(1) If [first, last) is empty, does nothing. Otherwise, performs the following operations in order:

1. Creates an accumulator acc of type T, and initializes it with *first.
2. Assigns acc to *d_first.
3. For each iterator iter in [++first, last) in order, performs the following operations in order:
   • a) Creates an object val of type T, and initializes it with *iter.
   • b) Computes val - acc (until C++11) or val - std::move(acc) (since C++11).
   • c) Assigns the result to *++d_first.
   • d) Copy assigns from val to acc (until C++11) or move assigns from val to acc (since C++11).

(2) If [first, last) is empty, does nothing. Otherwise, performs the following operations in order:

1. Assigns *first to *d_first.
2. For each integer i in [1, std::distance(first, last)), performs the following operations in order:
   • a) Computes curr - prev, where curr is the next ith iterator of first, and prev is the next i - 1th iterator of first.
   • b) Assigns the result to *dest, where dest is the next ith iterator of d_first.

(3) Same as (1), but computes op(val, acc) (until C++11) or op(val, std::move(acc)) (since C++11) instead.

(4) Same as (2), but computes op(curr, prev) instead.

Given binary_op as the actual binary operation:

• If any of the following conditions is satisfied, the program is ill-formed:
  • For overloads (1,3):
    • T is not constructible from *first.
    • acc is not writable to d_first.
    • The result of binary_op(val, acc) (until C++11) or binary_op(val, std::move(acc)) (since C++11) is not writable to d_first.
  • For overloads (2,4):
    • *first is not writable to d_first.
    • The result of binary_op(*first, *first) is not writable to d_first.
• Given d_last as the iterator to be returned, if any of the following conditions is satisfied, the behavior is undefined:
  • For overloads (1,3), T is not MoveAssignable. (since C++11)
  • For overloads (2,4), [first, last) and [d_first, d_last) overlaps.
  • binary_op modifies any element of [first, last) or [d_first, d_last).
  • binary_op invalidates any iterator or subrange in [first, last] or [d_first, d_last].

🦋 Parameters 🦋

• first, last — the range of elements.
• d_first — the beginning of the destination range.
• policy — the execution policy to use. See execution policy for details.
• op — binary operation function object that will be applied.

The signature of the function should be equivalent to the following:

Ret fun(const Type1 &a, const Type2 &b);

The signature does not need to have const &. The types Type1 and Type2 must be such that an object of type iterator_traits<InputIt>::value_type can be implicitly converted to both of them. The type Ret must be such that an object of type OutputIt can be dereferenced and assigned a value of type Ret.

🌈 Type requirements 🌈

• InputIt must meet the requirements of LegacyInputIterator.
• OutputIt must meet the requirements of LegacyOutputIterator.
• ForwardIt1, ForwardIt2 must meet the requirements of LegacyForwardIterator.

💖 Return value 💖

Iterator to the element past the last element written, or d_first if [first, last) is empty.

🔥 Complexity 🔥

Given N as std::distance(first, last):

• 1,2: Exactly N-1 applications of operator-.
• 3,4: Exactly N-1 applications of the binary function op.

😭 Exceptions 😭

The overloads with a template parameter named ExecutionPolicy report errors as follows:

• If execution of a function invoked as part of the algorithm throws an exception and ExecutionPolicy is one of the standard policies, std::terminate is called. For any other ExecutionPolicy, the behavior is implementation-defined.
• If the algorithm fails to allocate memory, std::bad_alloc is thrown.

💅 Possible implementation 💅

adjacent_difference (1)

template<class InputIt, class OutputIt>
constexpr // since C++20
OutputIt adjacent_difference(InputIt first, InputIt last, OutputIt d_first)
{
    if (first == last)
        return d_first;

    typedef typename std::iterator_traits<InputIt>::value_type value_t;
    value_t acc = *first;
    *d_first = acc;

    while (++first != last)
    {
        value_t val = *first;
        *++d_first = val - std::move(acc); // std::move since C++11
        acc = std::move(val);
    }

    return ++d_first;
}

adjacent_difference (3)

template<class InputIt, class OutputIt, class BinaryOp>
constexpr // since C++20
OutputIt adjacent_difference(InputIt first, InputIt last,
                             OutputIt d_first, BinaryOp op)
{
    if (first == last)
        return d_first;

    typedef typename std::iterator_traits<InputIt>::value_type value_t;
    value_t acc = *first;
    *d_first = acc;

    while (++first != last)
    {
        value_t val = *first;
        *++d_first = op(val, std::move(acc)); // std::move since C++11
        acc = std::move(val);
    }

    return ++d_first;
}

🧚 Notes 🧚

acc was introduced because of the resolution of LWG issue 539. The reason for using acc rather than directly calculating the differences is because the semantic of the latter is confusing if the following types mismatch:

• the value type of InputIt
• the writable type(s) of OutputIt
• the types of the parameters of operator- or op
• the return type of operator- or op

acc serves as the intermediate object to cache values of the iterated elements:

• its type is the value type of InputIt
• the value written to d_first (which is the return value of operator- or op) is assigned to it
• its value is passed to operator- or op

char i_array[4] = {100, 100, 100, 100};
int  o_array[4];

// OK: performs conversions when needed
// 1. creates “acc” of type char (the value type)
// 2. “acc” is assigned to the first element of “o_array”
// 3. the char arguments are used for long multiplication (char -> long)
// 4. the long product is assigned to the output range (long -> int)
// 5. the next value of “i_array” is assigned to “acc”
// 6. go back to step 3 to process the remaining elements in the input range
std::adjacent_difference(i_array, i_array + 4, o_array, std::multiplies<long>{});

🌟 Example 🌟

#include <array>
#include <functional>
#include <iostream>
#include <iterator>
#include <numeric>
#include <vector>

void println(auto comment, const auto& sequence)
{
    std::cout << comment;
    for (const auto& n : sequence)
        std::cout << n << ' ';
    std::cout << '\n';
};

int main()
{
    // Default implementation - the difference between two adjacent items
    std::vector v{4, 6, 9, 13, 18, 19, 19, 15, 10};
    println("Initially, v = ", v);
    std::adjacent_difference(v.begin(), v.end(), v.begin());
    println("Modified v = ", v);

    // Fibonacci
    std::array<int, 10> a {1};
    std::adjacent_difference(std::begin(a), std::prev(std::end(a)),
                             std::next(std::begin(a)), std::plus<>{});
    println("Fibonacci, a = ", a);
}

🌸 Output 🌸

Initially, v = 4 6 9 13 18 19 19 15 10
Modified v = 4 2 3 4 5 1 0 -4 -5
Fibonacci, a = 1 1 2 3 5 8 13 21 34 55

🧚 Defect reports 🧚

The following behavior-changing defect reports were applied retroactively to previously published C++ standards.

• LWG 242 — Applied to: C++98. Behavior as published: op could not have side effects. Correct behavior: it cannot modify the ranges involved.
• LWG 539 — Applied to: C++98. Behavior as published: the type requirements needed for the result evaluations and assignments to be valid were missing. Correct behavior: added.
• LWG 2055 (P0616R0) — Applied to: C++11. Behavior as published: acc was not moved while being accumulated. Correct behavior: it is moved.
• LWG 3058 — Applied to: C++17. Behavior as published: for overloads (2,4), the result of each invocation of operator- or op was assigned to a temporary object, and that object is assigned to the output range. Correct behavior: assign the results to the output range directly.

✨ See also ✨

okay bestie, if adjacent_difference was your gateway drug, here's the rest of the friend group 👑

• partial_sum — computes the partial sum of a range of elements (function template). it's giving "running total main character energy" 🧮✨
• accumulate — sums up or folds a range of elements (function template). like, it just takes everything in and folds it into one gorgeous value 🫶

ngl, these two are the ones you'll keep reaching for when adjacent_difference has done its little transformation moment. same header, same family, same slay 💅

---

and that's a wrap on this page, bestie!! 🎀 you made it to the end and honestly? obsessed with your commitment. go forth and difference some adjacents 💖

source: http://cppreference.com · version/date: 2024.06.10 · the page footer, but make it cute 🦋