C++ 简单的循环缓冲区队列

Question

I have a implemented a queue using an array and treating it as a circular buffer. _tail points to the next element to read and _head points to the next element to write into:

template<int SIZE>
class Q {

bool push(const int item){
    if(false == isFull()){
        _q[_head] = item;
        _head = ++_head % SIZE;
        return true;
    }

    return false;
}

bool pop(int& toReturn){
    if(false == isEmpty()){
        toReturn = _q[_tail];
        _tail = ++_tail % SIZE;  
        return true;
    }

    return false;
}

private:

    std::array<int, SIZE> _q;
    std::atomic<int> _head;
    std::atomic<int> _tail;
};

However, I have a bug in my isFull() and isEmpty() logic, identified by the following questions:

When the queue is initially empty, both _tail and _head will point to _q[0] .

When I have written _q[0] , _q[1] , _q[2] and q[3] again _tail and _head will point to the same, so clearly I cannot use _tail == _head to determine full/empty.

What is the simplest way to implement the isEmpty() and isFull() logic? I wasn't sure whether I should write a "BLANK" enum after a queue item has been read?

Answer 1

Declare a data member (counter) in the class that counts the number of items in the queue. Increment it in (push) and decrement it in (pop). When it is (0), the queue is empty

Answer 2

The logic is relatively straightforward, if you are willing to reduce the effective size of your queue by one element:

• When head and tail are the same, the queue is empty.
• When incrementing tail by one, with wrap-around, produces head , the queue is full.

This makes it impossible to insert N-th element, which serves as a "sentinel". Hence, for an N-element queue you need to allocate an array of N+1 elements, and use (SIZE+1) in all your expressions dealing with wrap-around.

std::array<int,SIZE+1> _q;

Implementation note:

These two lines

_head = ++_head % SIZE;
_tail = ++_tail % SIZE;  

have undefined behavior, because the compiler has flexibility at applying side effects from incrementing _head and _tail either before or after the assignment is over. If the compiler chooses to apply side effects after the assignment, the wrap-around effect will not happen.

Since you do not need compound assignment at all, the fix is easy:

_head = (_head+1) % SIZE;
_tail = (_tail+1) % SIZE;  

Answer 3

How about changing implementation a bit like

Change your head's definition from " _head points to the next element to write into: " to
" _head points to latest element added "

Then isFull() can be

bool isFull()
{
    if((_head + 1) % SIZE == _tail)
        return true;

    return false;
}

and isEmpty could be

bool isEmpty()
{
    if(_head == _tail)
        return true;

    return false;
}

pop() will be same but change push() and don't forget to initialize _head and _tail to -1

bool push(const int item){
    if(false == isFull()){
        if(_head == -1)    //If adding the first element,
            _tail = 0;     // tail will point to it as it was inititiallty 0
        _head = ++_head % SIZE;
         _q[_head] = item;
        return true;
    }

    return false;
}

Answer 4

Since you use modulus to calculate the index of your array, perhaps keeping a running count for each read and write would simplify the logic.

Change the implementation like this ...

bool push(const int item){
    if(false == isFull()){
        _q[_head % SIZE] = item;
        ++_head;
        return true;
    }

    return false;
}

bool pop(int& toReturn){
    if(false == isEmpty()){
        toReturn = _q[_tail % SIZE];
        ++_tail;  
        return true;
    }

    return false;
}

And use these ...

bool isFull()
{
    return _head - _tail == SIZE;
}

bool isEmpty()
{
    if (_head == _tail)
    {
        _head = _tail = 0;
        return true;
    }

    return false;
}

Answer 5

If you need a full-featured ring-buffer, try this:

#pragma once
#include <vector>
#include <stdexcept>
#include <new>
#include <utility>
#include <cstring>
#include <cassert>

#if defined(_MSC_VER)
    #pragma warning(push)
    #pragma warning(disable: 26495) // uninitialized member t at TU::TU()
#endif
#if defined(__llvm__)
    #pragma clang diagnostic push
    #pragma clang diagnostic ignored "-Wdangling-else"
#endif

template<typename T, typename Alloc = std::allocator<T>>
class ring_buffer
{
    union TU
    {
        T t;
        TU()  {}
        ~TU() {}
    };
    using vec_t = std::vector<TU, typename std::allocator_traits<Alloc>::template rebind_alloc<TU>>;
    using it_t  = typename vec_t::iterator;
public:
    struct iterator;
    struct const_iterator
    {
                        const_iterator();
                        const_iterator( const_iterator const &it );
        const_iterator &operator +=( std::ptrdiff_t offset );
        const_iterator &operator -=( std::ptrdiff_t offset );
        const_iterator &operator  =( const_iterator const &it );
        bool            operator ==( const_iterator const &it ) const;
        bool            operator !=( const_iterator const &it ) const;
        bool            operator  <( const_iterator const &it ) const;
        bool            operator <=( const_iterator const &it ) const;
        bool            operator  >( const_iterator const &it ) const;
        bool            operator >=( const_iterator const &it ) const;
        const_iterator  operator  +( std::ptrdiff_t offset ) const;
        const_iterator  operator  -( std::ptrdiff_t offset ) const;
        T const        &operator  *() const;
        const_iterator &operator ++();
        const_iterator &operator --();
        const_iterator  operator ++( int );
        const_iterator  operator --( int );
        T const        &operator []( std::ptrdiff_t offset ) const;
        T const        &at( std::ptrdiff_t offset ) const;
    private:
        friend class ring_buffer;
                        const_iterator( ring_buffer &ring, it_t it );
        std::ptrdiff_t  idx() const;
        template<bool THROW = false>
        it_t         offset_it( std::ptrdiff_t offset ) const;
        ring_buffer *m_ring;
        it_t         m_it;
    };
    struct iterator : public const_iterator
    {
                 iterator();
                 iterator( iterator const &it );
        iterator &operator +=( std::ptrdiff_t offset );
        iterator &operator -=( std::ptrdiff_t offset );
        iterator &operator  =( iterator const &it );
        iterator  operator  +( std::ptrdiff_t offset ) const;
        iterator  operator  -( std::ptrdiff_t offset ) const;
        T        &operator  *() const;
        iterator &operator ++();
        iterator &operator --();
        iterator  operator ++( int );
        iterator  operator --( int );
        T        &operator []( std::ptrdiff_t offset ) const;
        T        &at(          std::ptrdiff_t offset ) const;
    private:
                  iterator( ring_buffer &ring, it_t it );
        explicit  iterator( const_iterator const &it );
        friend class ring_buffer;
    };
                    ring_buffer( std::size_t cap );
                    ring_buffer( ring_buffer const &other );
                    ~ring_buffer();
    T              &operator []( std::size_t offset );
    T              &at( std::size_t offset );
    T              &front() const;
    iterator        begin() const;
    iterator        end() const;
    const_iterator  cbegin() const;
    const_iterator  cend() const;
    void            pop_front();
    template<typename ... Args>
    T              &emplace_back( Args &&...args );
    T              &push_back( T const &other );
    T              &push_back( T &&other );
    bool            full() const;
    std::size_t     size() const;
    void            capacity( std::size_t newCap );
private:
    vec_t  m_ring;
    it_t   m_read, m_write;
    size_t m_size, m_capacity;
};

template<typename T, typename Alloc>
inline
std::ptrdiff_t ring_buffer<T, Alloc>::const_iterator::idx() const
{
    if( m_it >= m_ring->m_read )
    {
        assert(   m_ring->m_read  <= m_ring->m_write && m_it <= m_ring->m_write
               || m_ring->m_write <  m_ring->m_read   & m_it >= m_ring->m_ring.begin() && m_it <= m_ring->m_write);
        return m_it - m_ring->m_read;
    }
    else
    {
        assert(m_it >= m_ring->m_ring.begin() && m_it <= m_ring->m_write);
        return (m_it - m_ring->m_ring.begin()) + (m_ring->m_ring.end() - m_ring->m_read);
    }
}

template<typename T, typename Alloc>
template<bool THROW>
inline
typename ring_buffer<T, Alloc>::it_t ring_buffer<T, Alloc>::const_iterator::offset_it( std::ptrdiff_t offset ) const
{
    using namespace std;
    ptrdiff_t rel;
    static
    char const outOfRangeStr[] = "ring-buffer access out of range";
    if( m_ring->m_read <= m_ring->m_write )
    {
        rel = m_it - m_ring->m_read + offset;
        if constexpr( !THROW )
            assert(rel >= 0 && rel <= m_ring->m_write - m_ring->m_read);
        else
            if( rel < 0 || rel > m_ring->m_write - m_ring->m_read )
                throw out_of_range( outOfRangeStr );
        return m_it + offset;
    }
    if( offset >= 0 )
    {
        if( m_it >= m_ring->m_read )
        {
            rel = m_it - m_ring->m_read + offset;
            if( rel < m_ring->m_ring.end() - m_ring->m_read )
                return m_it + rel;
            rel -= m_ring->m_ring.end() - m_ring->m_read;
            if constexpr( !THROW )
                assert(rel <= m_ring->m_write - m_ring->m_ring.begin());
            else
                if( rel > m_ring->m_write - m_ring->m_ring.begin() )
                    throw out_of_range( outOfRangeStr );
            return m_ring->m_ring.begin() + rel;
        }
        if constexpr( !THROW )
            assert(m_it - m_ring->m_ring.begin() + offset <= m_ring->m_write - m_ring->m_ring.begin());
        else
            if( m_it - m_ring->m_ring.begin() + offset > m_ring->m_write - m_ring->m_ring.begin() )
                throw out_of_range( outOfRangeStr );
        return m_it + offset;
    }
    if( m_it >= m_ring->m_read )
    {
        rel = m_it - m_ring->m_read + offset;
        if constexpr( !THROW )
            assert(rel >= 0);
        else
            if( rel < 0 )
                throw out_of_range( outOfRangeStr );
        return m_it + offset;
    }
    rel = m_it - m_ring->m_ring.begin() + offset;
    if( rel >= 0 )
        return m_it + rel;
    rel += m_it - m_ring->m_ring.begin();
    if constexpr( !THROW )
        assert(m_ring->m_ring.end() - m_ring->m_read >= -rel);
    else
        if( m_ring->m_ring.end() - m_ring->m_read < -rel )
            throw out_of_range( outOfRangeStr );
    return m_ring->m_ring.end() + rel;
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::const_iterator::const_iterator()
#if !defined(NDEBUG)
    : m_ring( nullptr )
#endif
{
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::const_iterator::const_iterator( const_iterator const &it ) :
    m_ring( it.m_ring ),
    m_it( it.m_it )
{
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::const_iterator::const_iterator( ring_buffer &ring, it_t it ) :
    m_ring( &ring ),
    m_it( it )
{
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator &ring_buffer<T, Alloc>::const_iterator::operator +=( std::ptrdiff_t offset )
{
    m_it = offset_it( offset );
    return *this;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator &ring_buffer<T, Alloc>::const_iterator::operator -=( std::ptrdiff_t offset )
{
    m_it = offset_it( -offset );
    return *this;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator &ring_buffer<T, Alloc>::const_iterator::operator =( const_iterator const &it )
{
    m_ring = it.m_ring;
    m_it   = it.m_it;
    return *this;
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator ==( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return m_it == it.m_it;
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator !=( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return m_it != it.m_it;
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator  <( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return idx() < it.idx();
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator <=( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return idx() <= it.idx();
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator >( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return idx() > it.idx();
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::const_iterator::operator >=( const_iterator const &it ) const
{
    assert(m_ring == it.m_ring);
    return idx() >= it.idx();
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::const_iterator::operator +( std::ptrdiff_t offset ) const
{
    return const_iterator( *m_ring, offset_it( offset ) );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::const_iterator::operator -( std::ptrdiff_t offset ) const
{
    return const_iterator( *m_ring, offset_it( -offset ) );
}

template<typename T, typename Alloc>
inline
T const &ring_buffer<T, Alloc>::const_iterator::operator *() const
{
    assert(idx() < (std::ptrdiff_t)m_ring->m_size);
    return m_it->t;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator &ring_buffer<T, Alloc>::const_iterator::operator ++()
{
    m_it = offset_it( 1 );
    return *this;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator &ring_buffer<T, Alloc>::const_iterator::operator --()
{
    m_it = offset_it( -1 );
    return *this;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::const_iterator::operator ++( int )
{
    const_iterator ret( *this );
    m_it = offset_it( 1 );
    return ret;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::const_iterator::operator --( int )
{
    const_iterator ret( *this );
    m_it = offset_it( -1 );
    return ret;
}

template<typename T, typename Alloc>
inline
T const &ring_buffer<T, Alloc>::const_iterator::operator []( std::ptrdiff_t offset ) const
{
    assert(offset_it( offset ) != m_ring->m_write);
    return offset_it( offset )->t;
}

template<typename T, typename Alloc>
inline
T const &ring_buffer<T, Alloc>::const_iterator::at( std::ptrdiff_t offset ) const
{
    assert(offset_it( offset ) != m_ring->m_write);
    return offset_it<true>( offset )->t;
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::iterator::iterator() :
    const_iterator()
{
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::iterator::iterator( iterator const &it ) :
    const_iterator( it )
{
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::iterator::iterator( ring_buffer &ring, it_t it ) :
    const_iterator( ring, it )
{
}

template<typename T, typename Alloc>
inline
ring_buffer<T, Alloc>::iterator::iterator( const_iterator const &it ) :
    const_iterator( it )
{
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator &ring_buffer<T, Alloc>::iterator::operator +=( std::ptrdiff_t offset )
{
    return (iterator &)const_iterator::operator +=( offset );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator &ring_buffer<T, Alloc>::iterator::operator -=( std::ptrdiff_t offset )
{
    return (iterator &)const_iterator::operator -=( -offset );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator &ring_buffer<T, Alloc>::iterator::operator =( iterator const &it )
{
    return (iterator &)const_iterator::operator =( it );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::iterator::operator +( std::ptrdiff_t offset ) const
{
    return (iterator)const_iterator::operator +( offset );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::iterator::operator -( std::ptrdiff_t offset ) const
{
    return (iterator)const_iterator::operator -( offset );
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::iterator::operator *() const
{
    return (T &)const_iterator::operator *();
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator &ring_buffer<T, Alloc>::iterator::operator ++()
{
    return (iterator &)const_iterator::operator ++();
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator &ring_buffer<T, Alloc>::iterator::operator --()
{
    return (iterator &)const_iterator::operator --();
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::iterator::operator ++( int )
{
    return (iterator)const_iterator::operator ++( 1 );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::iterator::operator --( int )
{
    return (iterator)const_iterator::operator --( 1 );
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::iterator::operator []( std::ptrdiff_t offset ) const
{
    return (T &)const_iterator::operator []( offset );
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::iterator::at( std::ptrdiff_t offset ) const
{
    return (T &)const_iterator::at( offset );
}

template<typename T, typename Alloc>
ring_buffer<T, Alloc>::ring_buffer( std::size_t cap ) :
    m_ring( cap + 1 ),
    m_read( m_ring.begin() ),
    m_write( m_ring.begin() ),
    m_size( 0 ),
    m_capacity( cap )
{
}

template<typename T, typename Alloc>
ring_buffer<T, Alloc>::ring_buffer( ring_buffer const &other ) :
    m_ring( other.m_capacity + 1 ),
    m_size( other.m_size ),
    m_capacity( other.m_capacity ),
    m_read( m_ring.begin() ),
    m_write( m_ring.begin() + m_size )
{
    it_t write = m_read;
    try
    {
        for( TU const &tu : other.m_ring )
            new( (void *)&*write++ )T( tu.t );
    }
    // if an exception occurs, destroy the objects which we've copied so far
    catch( ... )
    {
        for( it_t begin = m_read; write > begin; (--write)->t.~T() );
        throw;
    }
}

template<typename T, typename Alloc>
ring_buffer<T, Alloc>::~ring_buffer()
{
    auto destrRange = []( it_t destr, it_t end )
    {
        for( ; destr != end; destr++->t.~T() );
    };
    if( m_read <= m_write )
    {
        destrRange( m_read, m_write );
        return;
    }
    destrRange( m_read,  m_ring.end() );
    destrRange( m_ring.begin(), m_write );
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::operator []( std::size_t offset )
{
    using namespace std;
    assert(offset < m_size);
    if( m_write >= m_read )
    {
        assert((size_t)(m_ring.end() - m_ring.begin()) > offset);
        return m_read[offset].t;
    }
    if( offset < (size_t)(m_ring.end() - m_read) )
        return m_read[offset].t;
    offset -= m_ring.end() - m_read;
    assert((size_t)(m_write - m_ring.begin()) > offset);
    return m_ring[offset].t;
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::at( std::size_t offset )
{
    using namespace std;
    if( m_size == 0 )
        throw logic_error( "ring-buffer empty" );
    if( m_write >= m_read )
        if( offset < (size_t)(m_write - m_read) )
            return m_read[offset].t;
        else
            throw out_of_range( "ring-buffer index too large" );
    if( offset < (size_t)(m_ring.end() - m_read) )
        return m_read[offset].t;
    if( (offset -= m_ring.end() - m_read) >= (size_t)(m_write - m_ring.begin()) )
        throw out_of_range( "ring-buffer index too large" );
    return m_ring[offset].t;
}

template<typename T, typename Alloc>
inline
T &ring_buffer<T, Alloc>::front() const
{
    assert(m_size);
    return m_read->t;
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::begin() const
{
    return iterator( *(ring_buffer *)this, m_read );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::iterator ring_buffer<T, Alloc>::end() const
{
    return iterator( *(ring_buffer *)this, m_write );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::cbegin() const
{
    return const_iterator( *(ring_buffer *)this, m_read );
}

template<typename T, typename Alloc>
inline
typename ring_buffer<T, Alloc>::const_iterator ring_buffer<T, Alloc>::cend() const
{
    return const_iterator( *(ring_buffer *)this, m_write );
}

template<typename T, typename Alloc>
inline
void ring_buffer<T, Alloc>::pop_front()
{
    assert(m_size);
    m_read->t.~T();
    if( ++m_read == m_ring.end() )
        m_read = m_ring.begin();
    --m_size;
}

template<typename T, typename Alloc>
template<typename ... Args>
T &ring_buffer<T, Alloc>::emplace_back( Args &&...args )
{
    using namespace std;
    assert(m_capacity);
    if( m_size == m_capacity )
        pop_front();
    T &t = m_write->t;
    new( (void *)&t )T( forward<Args>( args ) ... );
    if( ++m_write == m_ring.end() )
        m_write = m_ring.begin();
    ++m_size;
    return t;
}

template<typename T, typename Alloc>
T &ring_buffer<T, Alloc>::push_back( T const &other )
{
    using namespace std;
    assert(m_capacity);
    if( m_size == m_capacity )
        pop_front();
    T &t = m_write->t;
    new( (void *)&t )T( other );
    if( ++m_write == m_ring.end() )
        m_write = m_ring.begin();
    ++m_size;
    return t;
}

template<typename T, typename Alloc>
T &ring_buffer<T, Alloc>::push_back( T &&other )
{
    using namespace std;
    assert(m_capacity);
    if( m_size == m_capacity )
        pop_front();
    T &t = m_write->t;
    new( (void *)&t )T( move( other ) );
    if( ++m_write == m_ring.end() )
        m_write = m_ring.begin();
    ++m_size;
    return t;
}

template<typename T, typename Alloc>
inline
std::size_t ring_buffer<T, Alloc>::size() const
{
    return m_size;
}

template<typename T, typename Alloc>
inline
bool ring_buffer<T, Alloc>::full() const
{
    return m_size == m_capacity;
}

template<typename T, typename Alloc>
void ring_buffer<T, Alloc>::capacity( std::size_t newCap )
{
    using namespace std;
    if( newCap < m_size )
        throw invalid_argument( "new ring-buffer size too small" );
    vec_t newRing( newCap + 1 );
    it_t  to = newRing.begin();
    auto moveRange = [&to]( it_t from, it_t fromEnd )
    {
        for( ; from != fromEnd; ++from, ++to )
            new( (void *)&*to )T( move( from->t ) );
    };
    auto destrRange = []( it_t destr, it_t end )
    {
        for( ; destr != end; (--destr)->t.~T() );
    };
    try
    {
        // destroy after copying / moving completely because whe might copy only if there isn't
        // move-constructor and have an exception; so the whole thing becomes transactional
        if( m_read <= m_write )
            moveRange(  m_read,         m_write        ),
            destrRange( m_write,        m_read         );
        else
            moveRange(  m_read,         m_ring.end()   ),
            moveRange(  m_ring.begin(), m_write        ),
            destrRange( m_write,        m_ring.begin() ),
            destrRange( m_ring.end(),   m_read         );
    }
    // if there's an exception, the objects haven't been moved but copied;
    // so we can destroy the objects in the destination-vector created so far
    catch( ... )
    {
        destrRange( to, newRing.begin() );
        throw;
    }
    swap( m_ring, newRing );
    m_read     = m_ring.begin();
    m_write    = to;
    m_capacity = newCap;
}

template<typename T, typename Alloc>
void swap( ring_buffer<T, Alloc> &a, ring_buffer<T, Alloc> &b )
{
    swap( a.m_ring,     b.m_ring     );
    swap( a.m_read,     b.m_read     );
    swap( a.m_write,    b.m_write    );
    swap( a.m_capacity, b.m_capacity );
    swap( a.m_size,     b.m_size     );
}

#if defined(_MSC_VER)
    #pragma warning(pop)
#endif
#if defined(__llvm__)
    #pragma clang diagnostic pop
#endif

Answer 6

There's a lot disinformation circling about the circular buffer. First of all, it's a deque. Whenever you see 2 pointers or indices, you might be dealing with a deque and you see head and tail and read and write pointers frequently in relation to the circular buffer. But it's all wrong, the circular buffer is a deque and as such is related to the doubly linked list, which can also serve as a deque. So, what you should think about are the first and last nodes, not head and tail, just like with a doubly linked list. Then you need to decide, whether first == last means, that the buffer is empty. If so, you can implement the buffer using 2 pointers (indices), but if not, then 3. The most important thing is to check out the literature about deques as different people tend to implement their circular buffers differently (head and tail frequently switch roles, for example) and deques are usually dealt with more academically, unambiguously.