C++ STL源码学习（之hash_table篇）

stl_hash_table.h

这不属于C++标准，是SGI STL标准的一部分，用于辅助实现hash_map和hash_set

/// Hashtable class, used to implement the hashed associative containers
/// hash_set, hash_map, hash_multiset, and hash_multimap.

///STL HashTable采用的是所谓的开链哈希法,依靠一个类似vector<list<T>>来实现.
///先通过哈希函数确定所需处理元素应当在vector中的那个位置(可以一步找到),vector
///中的每个元素我们称之为一个桶,则这个过程即是寻找桶的过程,每个桶实际是一个
///list<T>元素,若要处理的元素存在,则必然在这个桶之中,然后顺序遍历这个list即
///可确定该元素是否存在或者所在位置.由于第一步的哈希使得桶中存放的元素一
///般很少,因此遍历查询过程比较高效.

template <class _Val>
struct _Hashtable_node
{
_Hashtable_node* _M_next;
_Val _M_val;
};

///val:存储的值
///key:对应的键
///_HashFcn:所采用的hash函数类型
///_ExtractKey:用于从存储对象值中抽出键对象的函数,当hash_table中
///存储pair类型实现hash_map时最有用.
///_EqualKey:用于确定两个键值是否相等的函数
template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc = alloc>
class hashtable;

template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc>
struct _Hashtable_iterator;

template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc>
struct _Hashtable_const_iterator;

template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc>
struct _Hashtable_iterator {
typedef hashtable<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,_Alloc>
_Hashtable;
typedef _Hashtable_iterator<_Val, _Key, _HashFcn,
_ExtractKey, _EqualKey, _Alloc>
iterator;
typedef _Hashtable_const_iterator<_Val, _Key, _HashFcn,
_ExtractKey, _EqualKey, _Alloc>
const_iterator;

typedef _Hashtable_node<_Val> _Node;

typedef forward_iterator_tag iterator_category;
typedef _Val value_type;
typedef ptrdiff_t difference_type;
typedef size_t size_type;
typedef _Val& reference;
typedef _Val* pointer;

_Node* _M_cur;
_Hashtable* _M_ht;

_Hashtable_iterator(_Node* __n, _Hashtable* __tab)
: _M_cur(__n), _M_ht(__tab) {}

_Hashtable_iterator() {}
reference operator*() const { return _M_cur->_M_val; }

pointer operator->() const { return &(operator*()); }

iterator& operator++();
iterator operator++(int);
bool operator==(const iterator& __it) const
{ return _M_cur == __it._M_cur; }
bool operator!=(const iterator& __it) const
{ return _M_cur != __it._M_cur; }
};

template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc>
struct _Hashtable_const_iterator {
typedef hashtable<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,_Alloc>
_Hashtable;
typedef _Hashtable_iterator<_Val,_Key,_HashFcn,
_ExtractKey,_EqualKey,_Alloc>
iterator;
typedef _Hashtable_const_iterator<_Val, _Key, _HashFcn,
_ExtractKey, _EqualKey, _Alloc>
const_iterator;
typedef _Hashtable_node<_Val> _Node;

typedef forward_iterator_tag iterator_category;
typedef _Val value_type;
typedef ptrdiff_t difference_type;
typedef size_t size_type;
typedef const _Val& reference;
typedef const _Val* pointer;

const _Node* _M_cur;
const _Hashtable* _M_ht;

_Hashtable_const_iterator(const _Node* __n, const _Hashtable* __tab)
: _M_cur(__n), _M_ht(__tab) {}
_Hashtable_const_iterator() {}
_Hashtable_const_iterator(const iterator& __it)
: _M_cur(__it._M_cur), _M_ht(__it._M_ht) {}
reference operator*() const { return _M_cur->_M_val; }
pointer operator->() const { return &(operator*()); }
const_iterator& operator++();
const_iterator operator++(int);
bool operator==(const const_iterator& __it) const
{ return _M_cur == __it._M_cur; }
bool operator!=(const const_iterator& __it) const
{ return _M_cur != __it._M_cur; }
};

///HashTable的vector长度是有讲究的,为了尽量减少冲突(过多的元素被散列
///到同一个桶中),我们的桶个数一般应为质数个(由于我们是通过hash函数得到
///的值与桶数做mod运算得到需处理元素所在的桶号).这里取53开始的后28个
///质数,他们中的最大质数大于32位内存可存储的值.
/// Note: assumes long is at least 32 bits.
enum { __stl_num_primes = 28 };

static const unsigned long __stl_prime_list[__stl_num_primes] =
{
53ul,         97ul,         193ul,       389ul,       769ul,
1543ul,       3079ul,       6151ul,      12289ul,     24593ul,
49157ul,      98317ul,      196613ul,    393241ul,    786433ul,
1572869ul,    3145739ul,    6291469ul,   12582917ul,  25165843ul,
50331653ul,   100663319ul,  201326611ul, 402653189ul, 805306457ul,
1610612741ul, 3221225473ul, 4294967291ul
};

///给定一个值(实际即HashTable中所要存储的元素个数)得到大于等于
///它的最小质数(实际即HashTable中Array的长度/桶的个数).
inline unsigned long __stl_next_prime(unsigned long __n)
{
const unsigned long* __first = __stl_prime_list;
const unsigned long* __last = __stl_prime_list + (int)__stl_num_primes;
const unsigned long* pos = lower_bound(__first, __last, __n);
return pos == __last ? *(__last - 1) : *pos;
}

/// Forward declaration of operator==.
template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
class hashtable;

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
bool operator==(const hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>& __ht1,
const hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>& __ht2);

/// Hashtables handle allocators a bit differently than other containers
///  do.  If we're using standard-conforming allocators, then a hashtable
///  unconditionally has a member variable to hold its allocator, even if
///  it so happens that all instances of the allocator type are identical.
/// This is because, for hashtables, this extra storage is negligible.
///  Additionally, a base class wouldn't serve any other purposes; it
///  wouldn't, for example, simplify the exception-handling code.

template <class _Val, class _Key, class _HashFcn,
class _ExtractKey, class _EqualKey, class _Alloc>
class hashtable {
public:
typedef _Key key_type;
typedef _Val value_type;
typedef _HashFcn hasher;
typedef _EqualKey key_equal;

typedef size_t            size_type;
typedef ptrdiff_t         difference_type;
typedef value_type*       pointer;
typedef const value_type* const_pointer;
typedef value_type&       reference;
typedef const value_type& const_reference;

hasher hash_funct() const { return _M_hash; }
key_equal key_eq() const { return _M_equals; }

private:
typedef _Hashtable_node<_Val> _Node;

public:
typedef _Alloc allocator_type;
allocator_type get_allocator() const { return allocator_type(); }
private:
typedef simple_alloc<_Node, _Alloc> _M_node_allocator_type;

///分配和回收桶中的一个节点
_Node* _M_get_node() { return _M_node_allocator_type::allocate(1); }
void _M_put_node(_Node* __p) { _M_node_allocator_type::deallocate(__p, 1); }

private:
hasher                _M_hash;
key_equal             _M_equals;
_ExtractKey           _M_get_key;
vector<_Node*,_Alloc> _M_buckets;
size_type             _M_num_elements;

public:
typedef _Hashtable_iterator<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,_Alloc>
iterator;
typedef _Hashtable_const_iterator<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,
_Alloc>
const_iterator;

friend struct
_Hashtable_iterator<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,_Alloc>;
friend struct
_Hashtable_const_iterator<_Val,_Key,_HashFcn,_ExtractKey,_EqualKey,_Alloc>;

public:
hashtable(size_type __n,
const _HashFcn&    __hf,
const _EqualKey&   __eql,
const _ExtractKey& __ext,
const allocator_type& __a = allocator_type())
: _M_hash(__hf),
_M_equals(__eql),
_M_get_key(__ext),
_M_buckets(__a),
_M_num_elements(0)
{
_M_initialize_buckets(__n);
}

hashtable(size_type __n,
const _HashFcn&    __hf,
const _EqualKey&   __eql,
const allocator_type& __a = allocator_type())
:_M_hash(__hf),
_M_equals(__eql),
_M_get_key(_ExtractKey()),
_M_buckets(__a),
_M_num_elements(0)
{
_M_initialize_buckets(__n);
}

hashtable(const hashtable& __ht)
: _M_hash(__ht._M_hash),
_M_equals(__ht._M_equals),
_M_get_key(__ht._M_get_key),
_M_buckets(__ht.get_allocator()),
_M_num_elements(0)
{
_M_copy_from(__ht);
}

hashtable& operator= (const hashtable& __ht)
{
if (&__ht != this) {
clear();
_M_hash = __ht._M_hash;
_M_equals = __ht._M_equals;
_M_get_key = __ht._M_get_key;
_M_copy_from(__ht);
}
return *this;
}

~hashtable() { clear(); }

size_type size() const { return _M_num_elements; }
size_type max_size() const { return size_type(-1); }
bool empty() const { return size() == 0; }

///STL容器自身的swap成员函数一般都比较高效,HashTable也不例外
void swap(hashtable& __ht)
{
__STD::swap(_M_hash, __ht._M_hash);
__STD::swap(_M_equals, __ht._M_equals);
__STD::swap(_M_get_key, __ht._M_get_key);
_M_buckets.swap(__ht._M_buckets);
__STD::swap(_M_num_elements, __ht._M_num_elements);
}

iterator begin()
{
///找到第一个非空的桶,得到其第一个元素
for (size_type __n = 0; __n < _M_buckets.size(); ++__n)
if (_M_buckets[__n])
return iterator(_M_buckets[__n], this);
return end();
}

iterator end() { return iterator(0, this); }

const_iterator begin() const
{
for (size_type __n = 0; __n < _M_buckets.size(); ++__n)
if (_M_buckets[__n])
return const_iterator(_M_buckets[__n], this);
return end();
}

const_iterator end() const { return const_iterator(0, this); }

friend bool operator==<> (const hashtable&,const hashtable&);

public:

size_type bucket_count() const { return _M_buckets.size(); }

size_type max_bucket_count() const
{ return __stl_prime_list[(int)__stl_num_primes - 1]; }

///得到第__bucket个桶中的元素个数
size_type elems_in_bucket(size_type __bucket) const
{
size_type __result = 0;
for (_Node* __cur = _M_buckets[__bucket]; __cur; __cur = __cur->_M_next)
__result += 1;
return __result;
}

pair<iterator, bool> insert_unique(const value_type& __obj)
{
resize(_M_num_elements + 1);
return insert_unique_noresize(__obj);
}

iterator insert_equal(const value_type& __obj)
{
resize(_M_num_elements + 1);
return insert_equal_noresize(__obj);
}

pair<iterator, bool> insert_unique_noresize(const value_type& __obj);
iterator insert_equal_noresize(const value_type& __obj);

///下面insert_*中使用到的型别推导技法在STL中很常见
template <class _InputIterator>
void insert_unique(_InputIterator __f, _InputIterator __l)
{
insert_unique(__f, __l, __ITERATOR_CATEGORY(__f));
}

template <class _InputIterator>
void insert_equal(_InputIterator __f, _InputIterator __l)
{
insert_equal(__f, __l, __ITERATOR_CATEGORY(__f));
}

template <class _InputIterator>
void insert_unique(_InputIterator __f, _InputIterator __l,
input_iterator_tag)
{
for ( ; __f != __l; ++__f)
insert_unique(*__f);
}

template <class _InputIterator>
void insert_equal(_InputIterator __f, _InputIterator __l,
input_iterator_tag)
{
for ( ; __f != __l; ++__f)
insert_equal(*__f);
}

template <class _ForwardIterator>
void insert_unique(_ForwardIterator __f, _ForwardIterator __l,
forward_iterator_tag)
{
size_type __n = 0;
distance(__f, __l, __n);
resize(_M_num_elements + __n);
for ( ; __n > 0; --__n, ++__f)
insert_unique_noresize(*__f);
}

template <class _ForwardIterator>
void insert_equal(_ForwardIterator __f, _ForwardIterator __l,
forward_iterator_tag)
{
size_type __n = 0;
distance(__f, __l, __n);
resize(_M_num_elements + __n);
for ( ; __n > 0; --__n, ++__f)
insert_equal_noresize(*__f);
}

reference find_or_insert(const value_type& __obj);

iterator find(const key_type& __key)
{
size_type __n = _M_bkt_num_key(__key);  ///计算键所在的桶序号

///在应在的桶中顺序查找
_Node* __first;
for ( __first = _M_buckets[__n];
__first && !_M_equals(_M_get_key(__first->_M_val), __key);
__first = __first->_M_next)
{}
return iterator(__first, this);
}

const_iterator find(const key_type& __key) const
{
size_type __n = _M_bkt_num_key(__key);
const _Node* __first;
for ( __first = _M_buckets[__n];
__first && !_M_equals(_M_get_key(__first->_M_val), __key);
__first = __first->_M_next)
{}
return const_iterator(__first, this);
}

size_type count(const key_type& __key) const
{
const size_type __n = _M_bkt_num_key(__key);
size_type __result = 0;

for (const _Node* __cur = _M_buckets[__n]; __cur; __cur = __cur->_M_next)
if (_M_equals(_M_get_key(__cur->_M_val), __key))
++__result;
return __result;
}

pair<iterator, iterator>
equal_range(const key_type& __key);

pair<const_iterator, const_iterator>
equal_range(const key_type& __key) const;

size_type erase(const key_type& __key);
void erase(const iterator& __it);
void erase(iterator __first, iterator __last);

void erase(const const_iterator& __it);
void erase(const_iterator __first, const_iterator __last);

void resize(size_type __num_elements_hint);
void clear();

private:
size_type _M_next_size(size_type __n) const
{ return __stl_next_prime(__n); }

void _M_initialize_buckets(size_type __n)
{
const size_type __n_buckets = _M_next_size(__n); ///计算应分配的桶个数

_M_buckets.reserve(__n_buckets);    ///分配桶

///将每个桶清空
_M_buckets.insert(_M_buckets.end(), __n_buckets, (_Node*) 0);
_M_num_elements = 0;
}

///计算键所在的桶序号
size_type _M_bkt_num_key(const key_type& __key) const
{
return _M_bkt_num_key(__key, _M_buckets.size());
}

///计算值所在的桶序号
size_type _M_bkt_num(const value_type& __obj) const
{
return _M_bkt_num_key(_M_get_key(__obj));
}

size_type _M_bkt_num_key(const key_type& __key, size_t __n) const
{
///通过hash_fun得到的值在和桶个数做mod运算得到
return _M_hash(__key) % __n;
}

size_type _M_bkt_num(const value_type& __obj, size_t __n) const
{
return _M_bkt_num_key(_M_get_key(__obj), __n);
}

_Node* _M_new_node(const value_type& __obj)
{
_Node* __n = _M_get_node();
__n->_M_next = 0;
try {
construct(&__n->_M_val, __obj);
return __n;
}catch(...){
_M_put_node(__n);
}

}

void _M_delete_node(_Node* __n)
{
destroy(&__n->_M_val);
_M_put_node(__n);
}

void _M_erase_bucket(const size_type __n, _Node* __first, _Node* __last);
void _M_erase_bucket(const size_type __n, _Node* __last);

void _M_copy_from(const hashtable& __ht);

};

template <class _Val, class _Key, class _HF, class _ExK, class _EqK,
class _All>
_Hashtable_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>&
_Hashtable_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>::operator++()
{
const _Node* __old = _M_cur;
_M_cur = _M_cur->_M_next;
if (!_M_cur) {   ///当前迭代器所指元素为当前桶中的最后一个元素

///得到当前桶序号
size_type __bucket = _M_ht->_M_bkt_num(__old->_M_val);

///从下一个桶开始查找非空桶,查找的的第一个非空桶的第一个元素
///即为所求
while (!_M_cur && ++__bucket < _M_ht->_M_buckets.size())
_M_cur = _M_ht->_M_buckets[__bucket];
}
return *this;
}

template <class _Val, class _Key, class _HF, class _ExK, class _EqK, class _All>
inline _Hashtable_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>
_Hashtable_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>::operator++(int)
{
iterator __tmp = *this;
++*this;
return __tmp;
}

template <class _Val, class _Key, class _HF, class _ExK, class _EqK,
class _All>
_Hashtable_const_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>&
_Hashtable_const_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>::operator++()
{
const _Node* __old = _M_cur;
_M_cur = _M_cur->_M_next;
if (!_M_cur) {
size_type __bucket = _M_ht->_M_bkt_num(__old->_M_val);
while (!_M_cur && ++__bucket < _M_ht->_M_buckets.size())
_M_cur = _M_ht->_M_buckets[__bucket];
}
return *this;
}

template <class _Val, class _Key, class _HF, class _ExK, class _EqK,
class _All>
inline _Hashtable_const_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>
_Hashtable_const_iterator<_Val,_Key,_HF,_ExK,_EqK,_All>::operator++(int)
{
const_iterator __tmp = *this;
++*this;
return __tmp;
}

///判断两个HashTable是否相等
template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
bool operator==(const hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>& __ht1,
const hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>& __ht2)
{
typedef typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::_Node _Node;

///(1)桶个数必须相等
if (__ht1._M_buckets.size() != __ht2._M_buckets.size())
return false;

///(2)每个相同桶序号的相同位置的元素必须相等
for (int __n = 0; __n < __ht1._M_buckets.size(); ++__n) {
_Node* __cur1 = __ht1._M_buckets[__n];
_Node* __cur2 = __ht2._M_buckets[__n];
for ( ; __cur1 && __cur2 && __cur1->_M_val == __cur2->_M_val;
__cur1 = __cur1->_M_next, __cur2 = __cur2->_M_next)
{}
if (__cur1 || __cur2)
return false;
}
return true;
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
pair<typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::iterator, bool>
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::insert_unique_noresize(const value_type& __obj)
{
const size_type __n = _M_bkt_num(__obj);   ///计算应在桶序号
_Node* __first = _M_buckets[__n];

///遍历该桶,如果已有键相同元素,返回
for (_Node* __cur = __first; __cur; __cur = __cur->_M_next)
if (_M_equals(_M_get_key(__cur->_M_val), _M_get_key(__obj)))
return pair<iterator, bool>(iterator(__cur, this), false);

///需要插入

_Node* __tmp = _M_new_node(__obj);   ///构建对应值的插入结点

///插入应在桶头部
__tmp->_M_next = __first;
_M_buckets[__n] = __tmp;
++_M_num_elements;
return pair<iterator, bool>(iterator(__tmp, this), true);
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::iterator
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::insert_equal_noresize(const value_type& __obj)
{
const size_type __n = _M_bkt_num(__obj);
_Node* __first = _M_buckets[__n];

for (_Node* __cur = __first; __cur; __cur = __cur->_M_next)
if (_M_equals(_M_get_key(__cur->_M_val), _M_get_key(__obj))) {

///如果应在桶中已有键相同元素,相同元素的最前面
_Node* __tmp = _M_new_node(__obj);
__tmp->_M_next = __cur->_M_next;
__cur->_M_next = __tmp;
++_M_num_elements;
return iterator(__tmp, this);
}

///所在桶中无键相同元素
_Node* __tmp = _M_new_node(__obj);
__tmp->_M_next = __first;
_M_buckets[__n] = __tmp;
++_M_num_elements;
return iterator(__tmp, this);
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::reference
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::find_or_insert(const value_type& __obj)
{
///首先调整hashtable,为了防止桶个数太少以至于冲突太多
resize(_M_num_elements + 1);

size_type __n = _M_bkt_num(__obj);
_Node* __first = _M_buckets[__n];

for (_Node* __cur = __first; __cur; __cur = __cur->_M_next)
if (_M_equals(_M_get_key(__cur->_M_val), _M_get_key(__obj)))
return __cur->_M_val;

_Node* __tmp = _M_new_node(__obj);
__tmp->_M_next = __first;
_M_buckets[__n] = __tmp;
++_M_num_elements;
return __tmp->_M_val;
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
pair<typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::iterator,
typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::iterator>
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::equal_range(const key_type& __key)
{
typedef pair<iterator, iterator> _Pii;
const size_type __n = _M_bkt_num_key(__key);

for (_Node* __first = _M_buckets[__n]; __first; __first = __first->_M_next)
if (_M_equals(_M_get_key(__first->_M_val), __key)) {  ///找到一个键与__key相同的元素

///继续从当前位置遍历该桶,若遇到一个与之不同的键即可得到所求
for (_Node* __cur = __first->_M_next; __cur; __cur = __cur->_M_next)
if (!_M_equals(_M_get_key(__cur->_M_val), __key))
return _Pii(iterator(__first, this), iterator(__cur, this));

///该桶中键与__key相同的元素为最后一波元素
for (size_type __m = __n + 1; __m < _M_buckets.size(); ++__m)
if (_M_buckets[__m])
return _Pii(iterator(__first, this),iterator(_M_buckets[__m], this));

return _Pii(iterator(__first, this), end());
}
return _Pii(end(), end());
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
pair<typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::const_iterator,
typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::const_iterator>
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::equal_range(const key_type& __key) const
{
typedef pair<const_iterator, const_iterator> _Pii;
const size_type __n = _M_bkt_num_key(__key);

for (const _Node* __first = _M_buckets[__n] ;
__first;
__first = __first->_M_next) {
if (_M_equals(_M_get_key(__first->_M_val), __key)) {
for (const _Node* __cur = __first->_M_next;
__cur;
__cur = __cur->_M_next)
if (!_M_equals(_M_get_key(__cur->_M_val), __key))
return _Pii(const_iterator(__first, this),
const_iterator(__cur, this));
for (size_type __m = __n + 1; __m < _M_buckets.size(); ++__m)
if (_M_buckets[__m])
return _Pii(const_iterator(__first, this),
const_iterator(_M_buckets[__m], this));
return _Pii(const_iterator(__first, this), end());
}
}
return _Pii(end(), end());
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
typename hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::size_type
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::erase(const key_type& __key)
{
///找到对应桶以后的删除操作和单链表中删除元素相同,
///最后记着对头结点需另行处理
const size_type __n = _M_bkt_num_key(__key);
_Node* __first = _M_buckets[__n];
size_type __erased = 0;

if (__first) {
_Node* __cur = __first;
_Node* __next = __cur->_M_next;
while (__next) {
if (_M_equals(_M_get_key(__next->_M_val), __key)) {
__cur->_M_next = __next->_M_next;
_M_delete_node(__next);
__next = __cur->_M_next;
++__erased;
--_M_num_elements;
}
else {
__cur = __next;
__next = __cur->_M_next;
}
}

if (_M_equals(_M_get_key(__first->_M_val), __key)) {
_M_buckets[__n] = __first->_M_next;
_M_delete_node(__first);
++__erased;
--_M_num_elements;
}
}
return __erased;
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::erase(const iterator& __it)
{
///同样找到对应桶后和单链表删除操作相同
_Node* __p = __it._M_cur;
if (__p) {
const size_type __n = _M_bkt_num(__p->_M_val);
_Node* __cur = _M_buckets[__n];

if (__cur == __p) {
_M_buckets[__n] = __cur->_M_next;
_M_delete_node(__cur);
--_M_num_elements;
}
else {
_Node* __next = __cur->_M_next;
while (__next) {
if (__next == __p) {
__cur->_M_next = __next->_M_next;
_M_delete_node(__next);
--_M_num_elements;
break;
}
else {
__cur = __next;
__next = __cur->_M_next;
}
}
}
}
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::erase(iterator __first, iterator __last)
{
size_type __f_bucket = __first._M_cur ?
_M_bkt_num(__first._M_cur->_M_val) : _M_buckets.size();

size_type __l_bucket = __last._M_cur ?
_M_bkt_num(__last._M_cur->_M_val) : _M_buckets.size();

if (__first._M_cur == __last._M_cur)
return;
else if (__f_bucket == __l_bucket)   ///删除区间位于同一个桶内
_M_erase_bucket(__f_bucket, __first._M_cur, __last._M_cur);
else {
///区间内的每个桶分别作合适的删除
_M_erase_bucket(__f_bucket, __first._M_cur, 0);
for (size_type __n = __f_bucket + 1; __n < __l_bucket; ++__n)
_M_erase_bucket(__n, 0);
if (__l_bucket != _M_buckets.size())
_M_erase_bucket(__l_bucket, __last._M_cur);
}
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
inline void
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::erase(const_iterator __first,
const_iterator __last)
{
erase(iterator(const_cast<_Node*>(__first._M_cur),
const_cast<hashtable*>(__first._M_ht)),
iterator(const_cast<_Node*>(__last._M_cur),
const_cast<hashtable*>(__last._M_ht)));
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
inline void
hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::erase(const const_iterator& __it)
{
erase(iterator(const_cast<_Node*>(__it._M_cur),
const_cast<hashtable*>(__it._M_ht)));
}

///对hashtable的重新调整,为了避免桶个数太少以至于冲突太多.这是整个
///hashtable中最关键也最复杂的一个成员函数.
template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::resize(size_type __num_elements_hint)
{
const size_type __old_n = _M_buckets.size();
if (__num_elements_hint > __old_n) {
const size_type __n = _M_next_size(__num_elements_hint);///得到下一个质数
if (__n > __old_n) {
vector<_Node*, _All> __tmp(__n, (_Node*)(0),
_M_buckets.get_allocator());   ///重新分配合适大小的vector<list>
try {
///按序号遍历每个桶
for (size_type __bucket = 0; __bucket < __old_n; ++__bucket) {

///得到桶中第一个元素
_Node* __first = _M_buckets[__bucket];

///依次将该桶中的元素插入到新hashtable中对应的桶中,直到该桶为空
while (__first) {
///或得该元素在新的hashtable内应在的桶序号
size_type __new_bucket = _M_bkt_num(__first->_M_val, __n);

///将该元素从旧的位置摘下,插入新的hashtable应在的桶内
_M_buckets[__bucket] = __first->_M_next;
__first->_M_next = __tmp[__new_bucket];
__tmp[__new_bucket] = __first;

///将first指向旧桶中的第一个元素
__first = _M_buckets[__bucket];
}
}

///将得到的新hashtable和原有hashtable替换
_M_buckets.swap(__tmp);
}
catch(...) {
///如果操作失败,需要依次删除所有新hashtable内的元素,
///以防内存泄露
for (size_type __bucket = 0; __bucket < __tmp.size(); ++__bucket) {
while (__tmp[__bucket]) {
_Node* __next = __tmp[__bucket]->_M_next;
_M_delete_node(__tmp[__bucket]);
__tmp[__bucket] = __next;
}
}
throw;
}
}
}
}

///删除序号为__n的桶内[first,last)区间内的元素
template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::_M_erase_bucket(const size_type __n, _Node* __first, _Node* __last)
{
_Node* __cur = _M_buckets[__n];

///从头结点开始删除,需要特殊处理
if (__cur == __first)
_M_erase_bucket(__n, __last);
else {
_Node* __next;

///找到需要删除的起点
for (__next = __cur->_M_next;
__next != __first;
__cur = __next, __next = __cur->_M_next)
;

///类似单链表删除
while (__next != __last) {
__cur->_M_next = __next->_M_next;
_M_delete_node(__next);
__next = __cur->_M_next;
--_M_num_elements;
}
}
}

///删除序号为__n的桶内自头结点至__last的元素,不包括__last
template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::_M_erase_bucket(const size_type __n, _Node* __last)
{
_Node* __cur = _M_buckets[__n];
while (__cur != __last) {
_Node* __next = __cur->_M_next;
_M_delete_node(__cur);
__cur = __next;
_M_buckets[__n] = __cur;   ///记着调整_M_buckets[__n]的指向
--_M_num_elements;
}
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>::clear()
{
///挨个清空单链表,并将桶清空
for (size_type __i = 0; __i < _M_buckets.size(); ++__i) {
_Node* __cur = _M_buckets[__i];
while (__cur != 0) {
_Node* __next = __cur->_M_next;
_M_delete_node(__cur);
__cur = __next;
}
_M_buckets[__i] = 0;
}
_M_num_elements = 0;
}

template <class _Val, class _Key, class _HF, class _Ex, class _Eq, class _All>
void hashtable<_Val,_Key,_HF,_Ex,_Eq,_All>
::_M_copy_from(const hashtable& __ht)
{
///挨个复制单链表
_M_buckets.clear();
_M_buckets.reserve(__ht._M_buckets.size());
_M_buckets.insert(_M_buckets.end(), __ht._M_buckets.size(), (_Node*) 0);
try {
for (size_type __i = 0; __i < __ht._M_buckets.size(); ++__i) {
const _Node* __cur = __ht._M_buckets[__i];
if (__cur) {
_Node* __copy = _M_new_node(__cur->_M_val);
_M_buckets[__i] = __copy;

for (_Node* __next = __cur->_M_next;
__next;
__cur = __next, __next = __cur->_M_next) {
__copy->_M_next = _M_new_node(__next->_M_val);
__copy = __copy->_M_next;
}
}
}
_M_num_elements = __ht._M_num_elements;
}catch(...){
clear();
throw;
}
}

hash_table是STL中hash_map 和 hash_set 的内部数据结构，hash_table的插入/删除/查找的时间复杂度都为O(1),是查找速度最快的一种数据结构，但是hash_table中的数据是无序的，一般也只有在数据不需要排序，只需要满足快速查找/插入/删除的时候使用hash_table。hash_table的扩展是将原hash_table中的数据摘下来插入到一个临时的hash_table中，因为每个桶都使用list来实现的，因此插入删除都不存在内存copy，所以也是很高效的，最后再将临时hash_table和原来的hash_table（此时已经为空）交换。