boost库----->dynamic_bitset的学习

     c++标准为处理二进制数值提供了两个工具：vector<bool>和bitset。

     vector<bool>是对元素类型为bool的vector特化，它的内部并不真正存储bool值，而是以bit来压缩保存、使用代理技术来操作bit，造成的后果就是它很像容器，大多数情况下和标准容器一致，但它不是容器，不满足容器的定义。

      bitset与vector<bool>类似，同样存储二进制位，但它的大小固定，而且比vector<bool>支持更多的位运算。

      vector<bool>和bitset各有优缺点：vector<bool>可以动态增长，但不能方便地进行位运算；bitset则正好相反，可以方便地容纳的二进制位做位运算，但不能动态增长。

     boost.dynamic_bitset的出现恰好填补了这两者之间的空白，它类似于标准库的bitset，提供丰富的位运算，同时长度又是动态变化的。

    dynamic_bitset位于名字空间boost，为了使用dynamic_bitset组件，需要包含头文件<boost.dynamic_bitset.hpp>，即：

       #include<boost/dynamic_bitset_hpp>

      using namespace boost;

下面我给出部分源码的解析（只是个人的理解）：

//=============================================================================
// dynamic_bitset implementation
   //函数的实现

//-----------------------------------------------------------------------------
// constructors, etc.

template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>::dynamic_bitset(const Allocator& alloc)
  : m_bits(alloc), m_num_bits(0)
{

}

template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>::
dynamic_bitset(size_type num_bits, unsigned long value, const Allocator& alloc)
    : m_bits(alloc),
      m_num_bits(0)
{
    //使用无符号整型进行初始化
    init_from_unsigned_long(num_bits, value);
}

// copy constructor
template <typename Block, typename Allocator>
inline dynamic_bitset<Block, Allocator>::
dynamic_bitset(const dynamic_bitset& b)
  : m_bits(b.m_bits), m_num_bits(b.m_num_bits)
{

}

template <typename Block, typename Allocator>
inline dynamic_bitset<Block, Allocator>::
~dynamic_bitset()
{
    assert(m_check_invariants());
}

template <typename Block, typename Allocator>
inline void dynamic_bitset<Block, Allocator>::
swap(dynamic_bitset<Block, Allocator>& b) // no throw
{
    std::swap(m_bits, b.m_bits);
    std::swap(m_num_bits, b.m_num_bits);
}
//赋值
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>& dynamic_bitset<Block, Allocator>::
operator=(const dynamic_bitset<Block, Allocator>& b)
{
    m_bits = b.m_bits;
    m_num_bits = b.m_num_bits;
    return *this;
}

template <typename Block, typename Allocator>
inline typename dynamic_bitset<Block, Allocator>::allocator_type
dynamic_bitset<Block, Allocator>::get_allocator() const
{
    return m_bits.get_allocator();
}

//-----------------------------------------------------------------------------
// size changing operations
//num_bits:代表扩展多少位二进制
//vlaue为true是代表将每一位的值为都设为1
template <typename Block, typename Allocator>
void dynamic_bitset<Block, Allocator>::
resize(size_type num_bits, bool value) // strong guarantee
{
  //num_blocks()函数返回二进制为占用的Block数量
  //size() / sizeof(Block)*8 + 1
  //size()函数返回二进制的位数
  const size_type old_num_blocks = num_blocks();
  //required_blocks统计需要的Block数量
  const size_type required_blocks = calc_num_blocks(num_bits);

  //根据value的真假，如果value为真时，拿1填充所有的位，
  //否则拿0填充
  const block_type v = value? ~Block(0) : Block(0);

  //当需要的Block数量与原来的Block数量不相等时，进行扩展
  if (required_blocks != old_num_blocks) {
    m_bits.resize(required_blocks, v); // s.g. (copy)
  }


  // At this point:
  //
  //  - if the buffer was shrunk, we have nothing more to do,
  //    except a call to m_zero_unused_bits()
  //
  //  - if it was enlarged, all the (used) bits in the new blocks have
  //    the correct value, but we have not yet touched those bits, if
  //    any, that were 'unused bits' before enlarging: if value == true,
  //    they must be set.

  //当value为1并且所要扩展的位数大于原来的位数
  if (value && (num_bits > m_num_bits)) {
    //
    const block_width_type extra_bits = count_extra_bits();
    if (extra_bits) {
        assert(old_num_blocks >= 1 && old_num_blocks <= m_bits.size());

        // Set them.
        m_bits[old_num_blocks - 1] |= (v << extra_bits);
    }

  }
  //更新比特位的数目
  m_num_bits = num_bits;
  m_zero_unused_bits();

}

//清空
template <typename Block, typename Allocator>
void dynamic_bitset<Block, Allocator>::
clear() // no throw
{
  m_bits.clear();
  m_num_bits = 0;
}


template <typename Block, typename Allocator>
void dynamic_bitset<Block, Allocator>::
push_back(bool bit)
{
  //获取比特位的个数
  const size_type sz = size();
  //扩展一位
  resize(sz + 1);
  //位设置
  set(sz, bit);
}

//把整数转换为二进制位全部追加到dynamic_bitset末尾（最高位）
//它将把整数转换为一个Block再追加
template <typename Block, typename Allocator>
void dynamic_bitset<Block, Allocator>::
append(Block value) // strong guarantee
{
    const block_width_type r = count_extra_bits();

    if (r == 0) {
        //没有空间时，调用push_back进行尾插
        // the buffer is empty, or all blocks are filled
        m_bits.push_back(value);
    }
    else {
        //存在空间时
        //调用push_back进行追加
        m_bits.push_back(value >> (bits_per_block - r));
        m_bits[m_bits.size() - 2] |= (value << r); // m_bits.size() >= 2
    }

    m_num_bits += bits_per_block;
    assert(m_check_invariants());

}


//-----------------------------------------------------------------------------
// bitset operations
//运算符的实现
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::operator&=(const dynamic_bitset& rhs)
{
    //确保两个对象的二进制位数相同
    assert(size() == rhs.size());
    //num_block()为Block的块数
    for (size_type i = 0; i < num_blocks(); ++i)
        m_bits[i] &= rhs.m_bits[i];
    return *this;
}
//|=的实现
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::operator|=(const dynamic_bitset& rhs)
{
    assert(size() == rhs.size());
    for (size_type i = 0; i < num_blocks(); ++i)
        m_bits[i] |= rhs.m_bits[i];
    //m_zero_unused_bits();
    return *this;
}

//异或等的实现
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::operator^=(const dynamic_bitset& rhs)
{
    assert(size() == rhs.size());
    for (size_type i = 0; i < this->num_blocks(); ++i)
        m_bits[i] ^= rhs.m_bits[i];
    //m_zero_unused_bits();
    return *this;
}

template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::operator-=(const dynamic_bitset& rhs)
{
    assert(size() == rhs.size());
    for (size_type i = 0; i < num_blocks(); ++i)
        m_bits[i] &= ~rhs.m_bits[i];
    //m_zero_unused_bits();
    return *this;
}

//
// NOTE:
//  Note that the 'if (r != 0)' is crucial to avoid undefined
//  behavior when the left hand operand of >> isn't promoted to a
//  wider type (because rs would be too large).
//

//左移等的实现
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::operator<<=(size_type n)
{
    //m_num_bits：二进制的位数
    //当左移的位数大于二进制本身的位数时，相当于将每一位都置为0
    if (n >= m_num_bits)
        return reset();
    //else
    //当n大于0并且小于二进制位数时
    if (n > 0) {
        //num_blocks():统计Block的块数
        size_type    const last = num_blocks() - 1;  // num_blocks() is >= 1
        //bits_per_block:每一块所占的比特数
        //div：移动的位数中占几个Block
        size_type    const div  = n / bits_per_block; // div is <= last
        //根据移动位数计算出二进制位的下标
        //r == 0 说明移动的位数正好是Block的倍数，这时我们可以移动数组的元素
        block_width_type const r = bit_index(n);
        block_type * const b    = &m_bits[0];

        if (r != 0) {
            //r！=0时，r正是n/Block所得的余数
            block_width_type const rs = bits_per_block - r;

            for (size_type i = last-div; i>0; --i) {
                b[i+div] = (b[i] << r) | (b[i-1] >> rs);
            }
            b[div] = b[0] << r;

        }
        else {
            //r == 0时，直接移动数组的元素
            for (size_type i = last-div; i>0; --i) {
                b[i+div] = b[i];
            }
            b[div] = b[0];
        }

        // zero out div blocks at the less significant end
        // 移出的用0填充
        std::fill_n(b, div, static_cast<block_type>(0));

        // zero out any 1 bit that flowed into the unused part
        // 将未使用的比特位置0
        m_zero_unused_bits(); // thanks to Lester Gong

    }

    return *this;


}


//
// NOTE:
//  see the comments to operator <<=
//
//右移和左移的原理是一样的
template <typename B, typename A>
dynamic_bitset<B, A> & dynamic_bitset<B, A>::operator>>=(size_type n) {
    if (n >= m_num_bits) {
        return reset();
    }
    //else
    if (n>0) {

        size_type  const last  = num_blocks() - 1; // num_blocks() is >= 1
        size_type  const div   = n / bits_per_block;   // div is <= last
        block_width_type const r     = bit_index(n);
        block_type * const b   = &m_bits[0];


        if (r != 0) {

            block_width_type const ls = bits_per_block - r;

            for (size_type i = div; i < last; ++i) {
                b[i-div] = (b[i] >> r) | (b[i+1]  << ls);
            }
            // r bits go to zero
            b[last-div] = b[last] >> r;
        }

        else {
            for (size_type i = div; i <= last; ++i) {
                b[i-div] = b[i];
            }
            // note the '<=': the last iteration 'absorbs'
            // b[last-div] = b[last] >> 0;
        }



        // div blocks are zero filled at the most significant end
        std::fill_n(b + (num_blocks()-div), div, static_cast<block_type>(0));
    }

    return *this;
}

//左移和右移
//创建临时对象调用<<= 、>>=，并将对象返回
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
dynamic_bitset<Block, Allocator>::operator<<(size_type n) const
{
    dynamic_bitset r(*this);
    return r <<= n;
}

template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
dynamic_bitset<Block, Allocator>::operator>>(size_type n) const
{
    dynamic_bitset r(*this);
    return r >>= n;
}


//-----------------------------------------------------------------------------
// basic bit operations
//将某一位设置为0或1操作
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::set(size_type pos, bool val)
{
    assert(pos < m_num_bits);

    if (val)
        //val为真是进行置位操作
        //根据pos获取Block的下标
        m_bits[block_index(pos)] |= bit_mask(pos);
    else
        //否则进行清零操作
        reset(pos);

    return *this;
}

//set（）无参数时将二进制的所有位都置为1
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::set()
{
  std::fill(m_bits.begin(), m_bits.end(), ~Block(0));
  //将未使用的比特位清0
  m_zero_unused_bits();
  return *this;
}

//将某一位置清零操作
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::reset(size_type pos)
{
    assert(pos < m_num_bits);
#if defined __MWERKS__ && BOOST_WORKAROUND(__MWERKS__, <= 0x3003) // 8.x
    // CodeWarrior 8 generates incorrect code when the &=~ is compiled,
    // use the |^ variation instead.. <grafik>
    m_bits[block_index(pos)] |= bit_mask(pos);
    m_bits[block_index(pos)] ^= bit_mask(pos);
#else
    m_bits[block_index(pos)] &= ~bit_mask(pos);
#endif
    return *this;
}

//将二进制的所有位都进行清零操作
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::reset()
{
  std::fill(m_bits.begin(), m_bits.end(), Block(0));
  return *this;
}

//将二进制进行翻转某一位操作
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::flip(size_type pos)
{
    assert(pos < m_num_bits);
    m_bits[block_index(pos)] ^= bit_mask(pos);
    return *this;
}
//将二进制的所有位进行翻转操作
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>&
dynamic_bitset<Block, Allocator>::flip()
{
    for (size_type i = 0; i < num_blocks(); ++i)
        m_bits[i] = ~m_bits[i];
    //将未使用的比特位置0
    m_zero_unused_bits();
    return *this;
}

template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::m_unchecked_test(size_type pos) const
{
    //根据pos的值获得Block的下标，将该下标的元素与掩码相与
    return (m_bits[block_index(pos)] & bit_mask(pos)) != 0;
}

//检测某一位为1或0
//为1返回真
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::test(size_type pos) const
{
    assert(pos < m_num_bits);
    return m_unchecked_test(pos);
}

//如果二进制位中存在1时返回真
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::any() const
{
    for (size_type i = 0; i < num_blocks(); ++i)
        if (m_bits[i])
            return true;
    return false;
}

//与any()函数相反
//当二进制中存在0时返回真
template <typename Block, typename Allocator>
inline bool dynamic_bitset<Block, Allocator>::none() const
{
    return !any();
}

//取反
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
dynamic_bitset<Block, Allocator>::operator~() const
{
    //拷贝构造另一个对象
    dynamic_bitset b(*this);
    //调用翻转函数
    b.flip();
    return b;
}

//统计二进制中1的个数
template <typename Block, typename Allocator>
typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::count() const
{
    using detail::dynamic_bitset_impl::table_width;
    using detail::dynamic_bitset_impl::access_by_bytes;
    using detail::dynamic_bitset_impl::access_by_blocks;
    using detail::dynamic_bitset_impl::value_to_type;

#if BOOST_WORKAROUND(__GNUC__, == 4) && (__GNUC_MINOR__ == 3) && (__GNUC_PATCHLEVEL__ == 3)
    // NOTE: Explicit qualification of "bits_per_block"
    //       breaks compilation on gcc 4.3.3
    enum { no_padding = bits_per_block == CHAR_BIT * sizeof(Block) };
#else
    // NOTE: Explicitly qualifying "bits_per_block" to workaround
    //       regressions of gcc 3.4.x
    enum { no_padding =
        dynamic_bitset<Block, Allocator>::bits_per_block
        == CHAR_BIT * sizeof(Block) };
#endif

    enum { enough_table_width = table_width >= CHAR_BIT };

    enum { mode = (no_padding && enough_table_width)
                          ? access_by_bytes
                          : access_by_blocks };

    return do_count(m_bits.begin(), num_blocks(), Block(0),
                    static_cast<value_to_type<(bool)mode> *>(0));
}


//-----------------------------------------------------------------------------
// conversions
//转换函数

template <typename B, typename A, typename stringT>
void to_string_helper(const dynamic_bitset<B, A> & b, stringT & s,
                      bool dump_all)
{
    typedef typename stringT::traits_type Tr;
    typedef typename stringT::value_type  Ch;

    BOOST_DYNAMIC_BITSET_CTYPE_FACET(Ch, fac, std::locale());
    const Ch zero = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '0');
    const Ch one  = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '1');

    // Note that this function may access (when
    // dump_all == true) bits beyond position size() - 1

    typedef typename dynamic_bitset<B, A>::size_type size_type;

    const size_type len = dump_all?
         dynamic_bitset<B, A>::bits_per_block * b.num_blocks():
         b.size();
    s.assign (len, zero);

    for (size_type i = 0; i < len; ++i) {
        if (b.m_unchecked_test(i))
            Tr::assign(s[len - 1 - i], one);

    }

}


// A comment similar to the one about the constructor from
// basic_string can be done here. Thanks to James Kanze for
// making me (Gennaro) realize this important separation of
// concerns issue, as well as many things about i18n.
//
//将二进制转换位子符串
template <typename Block, typename Allocator, typename stringT>
inline void
to_string(const dynamic_bitset<Block, Allocator>& b, stringT& s)
{
    to_string_helper(b, s, false);
}


// Differently from to_string this function dumps out
// every bit of the internal representation (may be
// useful for debugging purposes)
//
template <typename B, typename A, typename stringT>
inline void
dump_to_string(const dynamic_bitset<B, A>& b, stringT& s)
{
    to_string_helper(b, s, true /* =dump_all*/);
}

template <typename Block, typename Allocator, typename BlockOutputIterator>
inline void
to_block_range(const dynamic_bitset<Block, Allocator>& b,
               BlockOutputIterator result)
{
    // note how this copies *all* bits, including the
    // unused ones in the last block (which are zero)
    std::copy(b.m_bits.begin(), b.m_bits.end(), result);
}

template <typename Block, typename Allocator>
unsigned long dynamic_bitset<Block, Allocator>::
to_ulong() const
{

  if (m_num_bits == 0)
      return 0; // convention

  // Check for overflows. This may be a performance burden on very
  // large bitsets but is required by the specification, sorry
  if (find_next(ulong_width - 1) != npos)
    throw std::overflow_error("boost::dynamic_bitset::to_ulong overflow");


  // Ok, from now on we can be sure there's no "on" bit
  // beyond the "allowed" positions
  typedef unsigned long result_type;

  const size_type max_size =
            (std::min)(m_num_bits, static_cast<size_type>(ulong_width));

  const size_type last_block = block_index( max_size - 1 );

  assert((last_block * bits_per_block) < static_cast<size_type>(ulong_width));

  result_type result = 0;
  for (size_type i = 0; i <= last_block; ++i) {
    const size_type offset = i * bits_per_block;
    result |= (static_cast<result_type>(m_bits[i]) << offset);
  }

  return result;
}

//返回二进制的位数
template <typename Block, typename Allocator>
inline typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::size() const
{
    return m_num_bits;
}

//返回二进制所占的块数
template <typename Block, typename Allocator>
inline typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::num_blocks() const
{
    return m_bits.size();
}

template <typename Block, typename Allocator>
inline typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::max_size() const
{
    // Semantics of vector<>::max_size() aren't very clear
    // (see lib issue 197) and many library implementations
    // simply return dummy values, _unrelated_ to the underlying
    // allocator.
    //
    // Given these problems, I was tempted to not provide this
    // function at all but the user could need it if he provides
    // his own allocator.
    //

    const size_type m = detail::dynamic_bitset_impl::
                        vector_max_size_workaround(m_bits);

    return m <= (size_type(-1)/bits_per_block) ?
        m * bits_per_block :
        size_type(-1);
}

//判断二进制是否为空
template <typename Block, typename Allocator>
inline bool dynamic_bitset<Block, Allocator>::empty() const
{
  return size() == 0;
}


//检测一个对象是否为另一个对象的子集
//子集就是两个二进制对象一模一样
//例如 10010 和 10010
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::
is_subset_of(const dynamic_bitset<Block, Allocator>& a) const
{
    //确保两个对象的二进制位数相同
    assert(size() == a.size());
    for (size_type i = 0; i < num_blocks(); ++i)
        //不为0说明两个对象不相同
        if (m_bits[i] & ~a.m_bits[i])
            return false;
    return true;
}

//检测一个对象是否为另一个对象的真子集
//例如 10101和 00101
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::
is_proper_subset_of(const dynamic_bitset<Block, Allocator>& a) const
{
    //确保二进制的位数相同
    assert(size() == a.size());
   //确保两个对象的块数相同
    assert(num_blocks() == a.num_blocks());

    bool proper = false;
    for (size_type i = 0; i < num_blocks(); ++i) {
        const Block & bt =   m_bits[i];
        const Block & ba = a.m_bits[i];

        if (bt & ~ba)
            return false; // not a subset at all
        if (ba & ~bt)
            proper = true;
    }
    return proper;
}

//将个对象中以二进制位数较少的进行两个对象相与
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::intersects(const dynamic_bitset & b) const
{
    //common_block：取两个对象中块数较小的数
    size_type common_blocks = num_blocks() < b.num_blocks()
                              ? num_blocks() : b.num_blocks();

    for(size_type i = 0; i < common_blocks; ++i) {
        if(m_bits[i] & b.m_bits[i])
            return true;
    }
    return false;
}

// --------------------------------
// lookup


// look for the first bit "on", starting
// from the block with index first_block
//
//从某一块开始寻找1的位置
template <typename Block, typename Allocator>
typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::m_do_find_from(size_type first_block) const
{
    size_type i = first_block;

    // skip null blocks
    // 跳过为0的块
    while (i < num_blocks() && m_bits[i] == 0)
        ++i;

    //当i大于块数时，说明没找到
    if (i >= num_blocks())
        return npos; // not found

    //找到后返回位置
    return i * bits_per_block + boost::lowest_bit(m_bits[i]);

}

//从最低位找二进制位1
template <typename Block, typename Allocator>
typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::find_first() const
{
    return m_do_find_from(0);
}

//从pos位置开始找下一个二进制位1
template <typename Block, typename Allocator>
typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::find_next(size_type pos) const
{
    //获取二进制位的个数
    const size_type sz = size();
    //当pos大于二进制位的个数或者二进制位的个数为0时
    //返回 没找到
    if (pos >= (sz-1) || sz == 0)
        return npos;

    ++pos;
    //根据pos的值获取块的下标
    const size_type blk = block_index(pos);
    //根据pos的值获取二进制位的下标
    const block_width_type ind = bit_index(pos);

    // mask out bits before pos
    //将pos所在的块和pos在块中的下标，将该块pos之前的二进制位都清零
    //如果fore为1，说明存在1
    //不为1时，说明该块不存在二进制位为1，从下一块开始找
    const Block fore = m_bits[blk] & ( ~Block(0) << ind );

    return fore?
        blk * bits_per_block + lowest_bit(fore)
        :
        m_do_find_from(blk + 1);

}



//-----------------------------------------------------------------------------
// comparison
//重载比较运算符（全局函数）

//== 运算符
template <typename Block, typename Allocator>
bool operator==(const dynamic_bitset<Block, Allocator>& a,
                const dynamic_bitset<Block, Allocator>& b)
{
    return (a.m_num_bits == b.m_num_bits)
           && (a.m_bits == b.m_bits);
}

//！= 运算符
template <typename Block, typename Allocator>
inline bool operator!=(const dynamic_bitset<Block, Allocator>& a,
                       const dynamic_bitset<Block, Allocator>& b)
{
    return !(a == b);
}

//< 运算符
template <typename Block, typename Allocator>
bool operator<(const dynamic_bitset<Block, Allocator>& a,
               const dynamic_bitset<Block, Allocator>& b)
{
    //确保两个对象二进制的位数相同
    assert(a.size() == b.size());
    typedef typename dynamic_bitset<Block, Allocator>::size_type size_type;

    //if (a.size() == 0)
    //  return false;

    // Since we are storing the most significant bit
    // at pos == size() - 1, we need to do the comparisons in reverse.
    //
    //从高位开始判断，首次比较小的对象就返回真
    for (size_type ii = a.num_blocks(); ii > 0; --ii) {
      size_type i = ii-1;
      if (a.m_bits[i] < b.m_bits[i])
        return true;
      else if (a.m_bits[i] > b.m_bits[i])
        return false;
    }
    //相等时返回false
    return false;
}

//<= 运算符
template <typename Block, typename Allocator>
inline bool operator<=(const dynamic_bitset<Block, Allocator>& a,
                       const dynamic_bitset<Block, Allocator>& b)
{
    return !(a > b);
}

//> 运算符
template <typename Block, typename Allocator>
inline bool operator>(const dynamic_bitset<Block, Allocator>& a,
                      const dynamic_bitset<Block, Allocator>& b)
{
    return b < a;
}

//>= 运算符
template <typename Block, typename Allocator>
inline bool operator>=(const dynamic_bitset<Block, Allocator>& a,
                       const dynamic_bitset<Block, Allocator>& b)
{
    return !(a < b);
}

//-----------------------------------------------------------------------------
// stream operations

//重载输出<< 运算符
#ifdef BOOST_OLD_IOSTREAMS
template < typename Block, typename Alloc>
std::ostream&
operator<<(std::ostream& os, const dynamic_bitset<Block, Alloc>& b)
{
    // NOTE: since this is aimed at "classic" iostreams, exception
    // masks on the stream are not supported. The library that
    // ships with gcc 2.95 has an exceptions() member function but
    // nothing is actually implemented; not even the class ios::failure.

    using namespace std;

    const ios::iostate ok = ios::goodbit;
    ios::iostate err = ok;

    if (os.opfx()) {

        //try
        typedef typename dynamic_bitset<Block, Alloc>::size_type bitsetsize_type;

        const bitsetsize_type sz = b.size();
        std::streambuf * buf = os.rdbuf();
        size_t npad = os.width() <= 0  // careful: os.width() is signed (and can be < 0)
            || (bitsetsize_type) os.width() <= sz? 0 : os.width() - sz;

        const char fill_char = os.fill();
        const ios::fmtflags adjustfield = os.flags() & ios::adjustfield;

        // if needed fill at left; pad is decresed along the way
        if (adjustfield != ios::left) {
            for (; 0 < npad; --npad)
                if (fill_char != buf->sputc(fill_char)) {
                    err |= ios::failbit;
                    break;
                }
        }

        if (err == ok) {
            // output the bitset
            for (bitsetsize_type i = b.size(); 0 < i; --i) {
                const char dig = b.test(i-1)? '1' : '0';
                if (EOF == buf->sputc(dig)) {
                    err |= ios::failbit;
                    break;
                }
            }
        }

        if (err == ok) {
            // if needed fill at right
            for (; 0 < npad; --npad) {
                if (fill_char != buf->sputc(fill_char)) {
                    err |= ios::failbit;
                    break;
                }
            }
        }

        os.osfx();
        os.width(0);

    } // if opfx

    if(err != ok)
        os.setstate(err); // assume this does NOT throw
    return os;

}
#else

template <typename Ch, typename Tr, typename Block, typename Alloc>
std::basic_ostream<Ch, Tr>&
operator<<(std::basic_ostream<Ch, Tr>& os,
           const dynamic_bitset<Block, Alloc>& b)
{

    using namespace std;

    const ios_base::iostate ok = ios_base::goodbit;
    ios_base::iostate err = ok;

    typename basic_ostream<Ch, Tr>::sentry cerberos(os);
    if (cerberos) {

        BOOST_DYNAMIC_BITSET_CTYPE_FACET(Ch, fac, os.getloc());
        const Ch zero = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '0');
        const Ch one  = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '1');

        try {

            typedef typename dynamic_bitset<Block, Alloc>::size_type bitsetsize_type;
            typedef basic_streambuf<Ch, Tr> buffer_type;

            buffer_type * buf = os.rdbuf();
            size_t npad = os.width() <= 0  // careful: os.width() is signed (and can be < 0)
                || (bitsetsize_type) os.width() <= b.size()? 0 : os.width() - b.size();

            const Ch fill_char = os.fill();
            const ios_base::fmtflags adjustfield = os.flags() & ios_base::adjustfield;

            // if needed fill at left; pad is decresed along the way
            if (adjustfield != ios_base::left) {
                for (; 0 < npad; --npad)
                    if (Tr::eq_int_type(Tr::eof(), buf->sputc(fill_char))) {
                          err |= ios_base::failbit;
                          break;
                    }
            }

            if (err == ok) {
                // output the bitset
                for (bitsetsize_type i = b.size(); 0 < i; --i) {
                    typename buffer_type::int_type
                        ret = buf->sputc(b.test(i-1)? one : zero);
                    if (Tr::eq_int_type(Tr::eof(), ret)) {
                        err |= ios_base::failbit;
                        break;
                    }
                }
            }

            if (err == ok) {
                // if needed fill at right
                for (; 0 < npad; --npad) {
                    if (Tr::eq_int_type(Tr::eof(), buf->sputc(fill_char))) {
                        err |= ios_base::failbit;
                        break;
                    }
                }
            }


            os.width(0);

        } catch (...) { // see std 27.6.1.1/4
            bool rethrow = false;
            try { os.setstate(ios_base::failbit); } catch (...) { rethrow = true; }

            if (rethrow)
                throw;
        }
    }

    if(err != ok)
        os.setstate(err); // may throw exception
    return os;

}
#endif


#ifdef BOOST_OLD_IOSTREAMS

    // A sentry-like class that calls isfx in its destructor.
    // "Necessary" because bit_appender::do_append may throw.
    class pseudo_sentry {
        std::istream & m_r;
        const bool m_ok;
    public:
        explicit pseudo_sentry(std::istream & r) : m_r(r), m_ok(r.ipfx(0)) { }
        ~pseudo_sentry() { m_r.isfx(); }
        operator bool() const { return m_ok; }
    };

//重载输入>>运算符
template <typename Block, typename Alloc>
std::istream&
operator>>(std::istream& is, dynamic_bitset<Block, Alloc>& b)
{

// Extractor for classic IO streams (libstdc++ < 3.0)
// ----------------------------------------------------//
//  It's assumed that the stream buffer functions, and
//  the stream's setstate() _cannot_ throw.


    typedef dynamic_bitset<Block, Alloc> bitset_type;
    typedef typename bitset_type::size_type size_type;

    std::ios::iostate err = std::ios::goodbit;
    pseudo_sentry cerberos(is); // skips whitespaces
    if(cerberos) {

        b.clear();

        const std::streamsize w = is.width();
        const size_type limit = w > 0 && static_cast<size_type>(w) < b.max_size()
                                                         ? w : b.max_size();
        typename bitset_type::bit_appender appender(b);
        std::streambuf * buf = is.rdbuf();
        for(int c = buf->sgetc(); appender.get_count() < limit; c = buf->snextc() ) {

            if (c == EOF) {
                err |= std::ios::eofbit;
                break;
            }
            else if (char(c) != '0' && char(c) != '1')
                break; // non digit character

            else {
                try {
                    appender.do_append(char(c) == '1');
                }
                catch(...) {
                    is.setstate(std::ios::failbit); // assume this can't throw
                    throw;
                }
            }

        } // for
    }

    is.width(0);
    if (b.size() == 0)
        err |= std::ios::failbit;
    if (err != std::ios::goodbit)
        is.setstate (err); // may throw

    return is;
}

#else // BOOST_OLD_IOSTREAMS

template <typename Ch, typename Tr, typename Block, typename Alloc>
std::basic_istream<Ch, Tr>&
operator>>(std::basic_istream<Ch, Tr>& is, dynamic_bitset<Block, Alloc>& b)
{

    using namespace std;

    typedef dynamic_bitset<Block, Alloc> bitset_type;
    typedef typename bitset_type::size_type size_type;

    const streamsize w = is.width();
    const size_type limit = 0 < w && static_cast<size_type>(w) < b.max_size()?
                                         w : b.max_size();

    ios_base::iostate err = ios_base::goodbit;
    typename basic_istream<Ch, Tr>::sentry cerberos(is); // skips whitespaces
    if(cerberos) {

        // in accordance with prop. resol. of lib DR 303 [last checked 4 Feb 2004]
        BOOST_DYNAMIC_BITSET_CTYPE_FACET(Ch, fac, is.getloc());
        const Ch zero = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '0');
        const Ch one  = BOOST_DYNAMIC_BITSET_WIDEN_CHAR(fac, '1');

        b.clear();
        try {
            typename bitset_type::bit_appender appender(b);
            basic_streambuf <Ch, Tr> * buf = is.rdbuf();
            typename Tr::int_type c = buf->sgetc();
            for( ; appender.get_count() < limit; c = buf->snextc() ) {

                if (Tr::eq_int_type(Tr::eof(), c)) {
                    err |= ios_base::eofbit;
                    break;
                }
                else {
                    const Ch to_c = Tr::to_char_type(c);
                    const bool is_one = Tr::eq(to_c, one);

                    if (!is_one && !Tr::eq(to_c, zero))
                        break; // non digit character

                    appender.do_append(is_one);

                }

            } // for
        }
        catch (...) {
            // catches from stream buf, or from vector:
            //
            // bits_stored bits have been extracted and stored, and
            // either no further character is extractable or we can't
            // append to the underlying vector (out of memory)

            bool rethrow = false;   // see std 27.6.1.1/4
            try { is.setstate(ios_base::badbit); }
            catch(...) { rethrow = true; }

            if (rethrow)
                throw;

        }
    }

    is.width(0);
    if (b.size() == 0 /*|| !cerberos*/)
        err |= ios_base::failbit;
    if (err != ios_base::goodbit)
        is.setstate (err); // may throw

    return is;

}


#endif


//-----------------------------------------------------------------------------
// bitset operations（全局函数）
//位比较函数

//按位与（求交集）
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
operator&(const dynamic_bitset<Block, Allocator>& x,
          const dynamic_bitset<Block, Allocator>& y)
{
    //创建对象（拷贝构造）
    dynamic_bitset<Block, Allocator> b(x);
    //调用与等函数
    return b &= y;
}

//按位或（求并集）
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
operator|(const dynamic_bitset<Block, Allocator>& x,
          const dynamic_bitset<Block, Allocator>& y)
{
    dynamic_bitset<Block, Allocator> b(x);
    return b |= y;
}
//按位异或
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
operator^(const dynamic_bitset<Block, Allocator>& x,
          const dynamic_bitset<Block, Allocator>& y)
{
    dynamic_bitset<Block, Allocator> b(x);
    return b ^= y;
}
//求差集
template <typename Block, typename Allocator>
dynamic_bitset<Block, Allocator>
operator-(const dynamic_bitset<Block, Allocator>& x,
          const dynamic_bitset<Block, Allocator>& y)
{
    dynamic_bitset<Block, Allocator> b(x);
    return b -= y;
}

//-----------------------------------------------------------------------------
// namespace scope swap

//交换函数
template<typename Block, typename Allocator>
inline void
swap(dynamic_bitset<Block, Allocator>& left,
     dynamic_bitset<Block, Allocator>& right) // no throw
{
    left.swap(right);
}


//-----------------------------------------------------------------------------
// private (on conforming compilers) member functions

//返回需要的Block数量
template <typename Block, typename Allocator>
inline typename dynamic_bitset<Block, Allocator>::size_type
dynamic_bitset<Block, Allocator>::calc_num_blocks(size_type num_bits)
{
    //num_bit代表要扩展的位数
    //bits_per_block代表每一个Block有几个比特
    //例如int 有32个比特
    return num_bits / bits_per_block
           + static_cast<int>( num_bits % bits_per_block != 0 );
}

// gives a reference to the highest block
//引用返回
template <typename Block, typename Allocator>
inline Block& dynamic_bitset<Block, Allocator>::m_highest_block()
{
    return const_cast<Block &>
           (static_cast<const dynamic_bitset *>(this)->m_highest_block());
}

// gives a const-reference to the highest block
//返回vector的最后一个元素（常引用返回）
template <typename Block, typename Allocator>
inline const Block& dynamic_bitset<Block, Allocator>::m_highest_block() const
{
    assert(size() > 0 && num_blocks() > 0);
    return m_bits.back();
}


// If size() is not a multiple of bits_per_block
// then not all the bits in the last block are used.
// This function resets the unused bits (convenient
// for the implementation of many member functions)
//
template <typename Block, typename Allocator>
inline void dynamic_bitset<Block, Allocator>::m_zero_unused_bits()
{
    //calc_num_block():返回Block的数量
    assert (num_blocks() == calc_num_blocks(m_num_bits));

    // if != 0 this is the number of bits used in the last block
    //count_extra_bits:判断二进制位是否为Block的倍数 size() / 每个块所占的比特数
    const block_width_type extra_bits = count_extra_bits();

    //当extra_bits不为0时说明二进制位数不是Block的倍数，存在未使用的比特位
    //将未使用的比特置0
    if (extra_bits != 0)
        m_highest_block() &= ~(~static_cast<Block>(0) << extra_bits);

}

// check class invariants
// 检查是否越界
template <typename Block, typename Allocator>
bool dynamic_bitset<Block, Allocator>::m_check_invariants() const
{
    //count_extra_bits:判断二进制位是否为Block的倍数 size() / 每个块所占的比特数
    const block_width_type extra_bits = count_extra_bits();
    //当extra_bits不为0时说明二进制位数不是Block的倍数，存在未使用的比特位
    if (extra_bits > 0) {
        //获取掩码
        block_type const mask = (~static_cast<Block>(0) << extra_bits);
        //将高位与掩码相与，当不为0时返回false
        if ((m_highest_block() & mask) != 0)
            return false;
    }
    if (m_bits.size() > m_bits.capacity() || num_blocks() != calc_num_blocks(size()))
        return false;

    return true;

}


} // namespace boost


#undef BOOST_BITSET_CHAR

#endif // include guard

下面对于整个类的实现做一个简单的说明：

    整个类的私有数据成员包括两个：    （1）m_ bits （2）m_num_bits

     (1)m_bits是一个vector<Block>的数组，其类型参数Block默认为unsigned long，代表m_bits其实就是一个数组，其每个元素的类型就是unsigned long

  （2）m_num_bits:二进制的位数，

     数组的每一个比特就是用来保存二进制位，所有的操作都是围绕这这个数组进行的。

  下面对于该类的一些操作函数简单的说明：

   （a）构造函数：

        该类提供了多种构造方式：

            构建空的对象。

            使用二进制进行构造。

            使用“01”字符串进行构造。

            使用unsigned long 数进行构造

            使用另一个对象进行拷贝构造。

     注：dynamic_bitset内部按照由高到低的顺序存储二进制位，也就是说，第0个元素存储最低位。

     （b）容器的操作：
          resize（）：调整容器的大小，扩展或收缩都是允许的，如果是扩展，原有的二进制位保持不变，新增加的二进制位用指定值置位，如果是收缩，收缩后容器的二进制位保持不变，多余的位被抛弃。

          push_back（）：向容器末尾追加一个值；

          append（）：把整数转换位二进制位全部追加到末尾。

      （c）比较操作与位运算

           ==、&= 、|=、-=；

           &、|、^等

       （d）访问元素：

            test（）：检测第n位是否为1

            any（）：二进制位中存在1返回true

            none（）：二进制位中存在0返回true

            count（）：统计二进制位为1的数量

            set（）：将某一位或所有位置1或置0

            reset（）：将某一位或所有位置0

            flip（）：反转某一位或所有位

            find_first（）：从第0位置开始查找，返回第一个为1的位置

            find_next（）：从第pos位置开始查找，返回第一个为1的位置，找不到返回npos

       （e）类型转换：

            to_ulong（）：将二进制转换为unsigned long

            to_string（）：将二进制转换为一个标准字符串

         （f）集合操作：

             is_subset_of（）：检测一个对象是否为另一个对象的子集

                  is_proper_subset_of（）：检测一个对象是否为另一个对象的真子集

              并集、交集、差集

    上述就是类的一些基本的操作，每个函数的实现也都进行了注释，希望对大家有帮助。。。。