gum::HashTable< Key, Val, Alloc > Class Template Reference

The class for generic Hash Tables. More...

#include <agrum/tools/core/hashTable.h>

Collaboration diagram for gum::HashTable< Key, Val, Alloc >:

Public Member Functions
template<typename... Args>
INLINE HashTable< Key, Val, Alloc >::value_type &	emplace (Args &&... args)
template<typename Mount , typename OtherAlloc >
HashTable< Key, Mount, OtherAlloc > INLINE	map (Mount(*f)(Val), Size size, bool resize_pol, bool key_uniqueness_pol) const
template<typename Mount , typename OtherAlloc >
HashTable< Key, Mount, OtherAlloc > INLINE	map (Mount(*f)(Val &), Size size, bool resize_pol, bool key_uniqueness_pol) const
template<typename Mount , typename OtherAlloc >
HashTable< Key, Mount, OtherAlloc > INLINE	map (Mount(*f)(const Val &), Size size, bool resize_pol, bool key_uniqueness_pol) const
template<typename Mount , typename OtherAlloc >
HashTable< Key, Mount, OtherAlloc > INLINE	map (const Mount &val, Size size, bool resize_pol, bool key_uniqueness_pol) const
template<typename OtherAlloc >
INLINE bool	operator!= (const HashTable< Key, Val, OtherAlloc > &from) const
Constructors / Destructors
	HashTable (Size size_param=HashTableConst::default_size, bool resize_pol=HashTableConst::default_resize_policy, bool key_uniqueness_pol=HashTableConst::default_uniqueness_policy)
	Default constructor. More...
	HashTable (std::initializer_list< std::pair< Key, Val > > list)
	Initializer list constructor. More...
	HashTable (const HashTable< Key, Val, Alloc > &from)
	Copy constructor. More...
template<typename OtherAlloc >
	HashTable (const HashTable< Key, Val, OtherAlloc > &from)
	Generalized copy constructor. More...
	HashTable (HashTable< Key, Val, Alloc > &&from)
	Move constructor. More...
	~HashTable ()
	Class destructor. More...
Operators
HashTable< Key, Val, Alloc > &	operator= (const HashTable< Key, Val, Alloc > &from)
	Copy operator. More...
template<typename OtherAlloc >
HashTable< Key, Val, Alloc > &	operator= (const HashTable< Key, Val, OtherAlloc > &from)
	Generalized copy operator. More...
HashTable< Key, Val, Alloc > &	operator= (HashTable< Key, Val, Alloc > &&from)
	Move operator. More...
Val &	operator[] (const Key &key)
	Returns a reference on the value the key of which is passed in argument. More...
const Val &	operator[] (const Key &key) const
	returns a reference on the value the key of which is passed in argument More...
template<typename OtherAlloc >
bool	operator== (const HashTable< Key, Val, OtherAlloc > &from) const
	Checks whether two hashtables contain the same elements. More...
template<typename OtherAlloc >
bool	operator!= (const HashTable< Key, Val, OtherAlloc > &from) const
	Checks whether two hashtables contain different sets of elements. More...
Fine tuning
Size	capacity () const noexcept
	Returns the number of slots in the 'nodes' vector of the hashtable. More...
void	resize (Size new_size)
	Changes the number of slots in the 'nodes' vector of the hash table. More...
void	setResizePolicy (const bool new_policy) noexcept
	Enables the user to change dynamically the resizing policy. More...
bool	resizePolicy () const noexcept
	Returns the current resizing policy. More...
void	setKeyUniquenessPolicy (const bool new_policy) noexcept
	Enables the user to change dynamically the policy for checking whether there can exist several elements in the table with identical keys. More...
bool	keyUniquenessPolicy () const noexcept
	Returns the current checking policy. More...
Accessors / Modifiers
Size	size () const noexcept
	Returns the number of elements stored into the hashtable. More...
bool	exists (const Key &key) const
	Checks whether there exists an element with a given key in the hashtable. More...
value_type &	insert (const Key &key, const Val &val)
	Adds a new element (actually a copy of this element) into the hash table. More...
value_type &	insert (Key &&key, Val &&val)
	Moves a new element in the hash table. More...
value_type &	insert (const std::pair< Key, Val > &elt)
	Adds a new element (actually a copy of this element) into the hash table. More...
value_type &	insert (std::pair< Key, Val > &&elt)
	Moves a new element in the hash table. More...
template<typename... Args>
value_type &	emplace (Args &&... args)
	Emplace a new element into the hashTable. More...
mapped_type &	getWithDefault (const Key &key, const Val &default_value)
	Returns a reference on the element the key of which is passed in argument. More...
mapped_type &	getWithDefault (Key &&key, Val &&default_value)
	Returns a reference on the element the key of which is passed in argument. More...
void	set (const Key &key, const Val &default_value)
	Add a new property or modify it if it already existed. More...
void	reset (const Key &key)
	Removes a property (i.e., remove an element). More...
void	erase (const Key &key)
	Removes a given element from the hash table. More...
void	erase (const iterator_safe &iter)
	Removes a given element from the hash table. More...
void	erase (const const_iterator_safe &iter)
	Removes a given element from the hash table. More...
void	eraseByVal (const Val &val)
	Removes a given element from the hash table. More...
const Key &	keyByVal (const Val &val) const
	Returns a reference on the key given a value. More...
const Key &	key (const Key &key) const
	Returns a reference on a given key. More...
void	eraseAllVal (const Val &val)
	Removes all the elements having a certain value from the hash table. More...
void	clear ()
	Removes all the elements in the hash table. More...
bool	empty () const noexcept
	Indicates whether the hash table is empty. More...
template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>
HashTable< Key, Mount, OtherAlloc >	map (Mount(*f)(Val), Size size=Size(0), bool resize_pol=HashTableConst::default_resize_policy, bool key_uniqueness_pol=HashTableConst::default_uniqueness_policy) const
	Transforms a hashtable of vals into a hashtable of mountains. More...
template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>
HashTable< Key, Mount, OtherAlloc >	map (Mount(*f)(Val &), Size size=Size(0), bool resize_pol=HashTableConst::default_resize_policy, bool key_uniqueness_pol=HashTableConst::default_uniqueness_policy) const
	Transforms a hashtable of vals into a hashtable of mountains. More...
template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>
HashTable< Key, Mount, OtherAlloc >	map (Mount(*f)(const Val &), Size size=Size(0), bool resize_pol=HashTableConst::default_resize_policy, bool key_uniqueness_pol=HashTableConst::default_uniqueness_policy) const
	Transforms a hashtable of vals into a hashtable of mountains. More...
template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>
HashTable< Key, Mount, OtherAlloc >	map (const Mount &val, Size size=Size(0), bool resize_pol=HashTableConst::default_resize_policy, bool key_uniqueness_pol=HashTableConst::default_uniqueness_policy) const
	Creates a hashtable of mounts with a given value from a hashtable of vals. More...

Public Types
using	Bucket = HashTableBucket< Key, Val >
	The buckets where data are stored. More...
using	BucketAllocator = typename Alloc::template rebind< Bucket >::other
	The Bucket allocator. More...
using	key_type = Key
	Types for STL compliance. More...
using	mapped_type = Val
	Types for STL compliance. More...
using	value_type = std::pair< const Key, Val >
	Types for STL compliance. More...
using	reference = value_type &
	Types for STL compliance. More...
using	const_reference = const value_type &
	Types for STL compliance. More...
using	pointer = value_type *
	Types for STL compliance. More...
using	const_pointer = const value_type *
	Types for STL compliance. More...
using	size_type = Size
	Types for STL compliance. More...
using	difference_type = std::ptrdiff_t
	Types for STL compliance. More...
using	allocator_type = Alloc
	Types for STL compliance. More...
using	iterator = HashTableIterator< Key, Val >
	Types for STL compliance. More...
using	const_iterator = HashTableConstIterator< Key, Val >
	Types for STL compliance. More...
using	iterator_safe = HashTableIteratorSafe< Key, Val >
	Types for STL compliance. More...
using	const_iterator_safe = HashTableConstIteratorSafe< Key, Val >
	Types for STL compliance. More...

Friends
template<typename K , typename V , typename A >
class	HashTable
	Friends to optimize the access to data, iterators must be friends. More...
class	HashTableIterator< Key, Val >
	Friends to optimize the access to data, iterators must be friends. More...
class	HashTableConstIterator< Key, Val >
	Friends to optimize the access to data, iterators must be friends. More...
class	HashTableIteratorSafe< Key, Val >
	Friends to optimize the access to data, iterators must be friends. More...
class	HashTableConstIteratorSafe< Key, Val >
	Friends to optimize the access to data, iterators must be friends. More...
template<typename T1 , typename T2 , typename A >
class	Bijection
	For bijections to quickly access data. More...
std::ostream &	operator<< (std::ostream &, const HashTable< Key, Val, Alloc > &)
	Prints the content of a gum::HashTable in the stream. More...
std::ostream &	operator<< (std::ostream &s, const HashTable< Key *, Val, Alloc > &table)
	Prints the content of a gum::HashTable with pointers key in the stream. More...

Iterators
const iterator &	end () noexcept
	Returns the unsafe iterator pointing to the end of the hashtable. More...
const const_iterator &	end () const noexcept
	Returns the unsafe const_iterator pointing to the end of the hashtable. More...
const const_iterator &	cend () const noexcept
	Returns the unsafe const_iterator pointing to the end of the hashtable. More...
iterator	begin ()
	Returns an unsafe iterator pointing to the beginning of the hashtable. More...
const_iterator	begin () const
	Returns an unsafe const_iterator pointing to the beginning of the hashtable. More...
const_iterator	cbegin () const
	Returns an unsafe const_iterator pointing to the beginning of the hashtable. More...
const iterator_safe &	endSafe () noexcept
	Returns the safe iterator pointing to the end of the hashtable. More...
const const_iterator_safe &	endSafe () const noexcept
	Returns the safe const_iterator pointing to the end of the hashtable. More...
const const_iterator_safe &	cendSafe () const noexcept
	Returns the safe const_iterator pointing to the end of the hashtable. More...
iterator_safe	beginSafe ()
	Returns the safe iterator pointing to the beginning of the hashtable. More...
const_iterator_safe	beginSafe () const
	Returns the safe const_iterator pointing to the beginning of the hashtable. More...
const_iterator_safe	cbeginSafe () const
	Returns the safe const_iterator pointing to the beginning of the hashtable. More...
static const iterator &	end4Statics ()
	Returns the end iterator for other classes' statics (read the detailed description of this method). More...
static const const_iterator &	constEnd4Statics ()
	Returns the end iterator for other classes' statics (read the detailed description of this method). More...
static const iterator_safe &	endSafe4Statics ()
	Returns the end iterator for other classes' statics (read the detailed description of this method). More...
static const const_iterator_safe &	constEndSafe4Statics ()
	Returns the end iterator for other classes' statics (read the detailed description of this method). More...

Detailed Description

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>
class gum::HashTable< Key, Val, Alloc >

The class for generic Hash Tables.

In aGrUM, a hashtable is a vector of chained lists (collision problems are fixed by chaining). Each slot of the vector contains a list of elements sharing the same hashed value. To be computationally efficient, the hash table should not contain too many elements as compared to its number of slots. Therefore, it is sometimes useful to resize the chained lists vector. aGrUM's hash tables are designed to automatically double their size when there is in average more than 3 elements per slot. However, when memory consumption is a concern, this feature can be turned off either by passing false as an optional resize_pol argument to the constructor of the hash table or by using method setResizePolicy when the instance of the class has already been constructed. Similarly, the default number of slots of the hash table may be parameterized as an optional argument of the constructor (size_param). Beware: when inserting elements of a given class into a hash table, unless the element is an r-value, only a copy of this element is stored into the table (this is compulsory if the hashtable is to be generic and can be used to store both complex classes and built-in types like integers). HashTable have different kinds of iterators: HashTableIteratorSafe and HashTableConstIteratorSafe (a.k.a. HashTable<>::iterator_safe and HashTable<>::const_iterator_safe) allow safe parsing of the hash tables. By safe, we mean that whenever the element pointed to by such an iterator is removed from the hashtable, accessing it through the iterator (*iter) does not result in a segmentation fault but rather in an exception being thrown. This safety is ensured at a very low cost (actually, our experiments show that our HashTables and HashTable's safe iterators significantly outperform the standard library unordered_maps). Of course, if there is no possibility for an iterator to point to a deleted element, the user can use "unsafe" iterators HashTableIterator and HashTableConstIterator (a.k.a. HashTable<>::iterator and HashTable<>::const_iterator). These iterators are slightly faster than their safe counterparts. However, as in the standard library, accessing through them a deleted element usually results in a mess (most probably a segfault).

Warning

HashTables guarantee that any element stored within them will have the same location in memory until it is removed from the hashtable (and this holds whatever operation is performed on the hashtable like new insertions, deletions, resizing, etc.).

Usage example:

// creation of an empty hash table
HashTable<int,string> table1;

// insert two elements into the hash table
table1.insert (10,"xxx");
table1.insert (20,"yyy");
table1.emplace (30,"zzz");

// creation of a nonempty hashtable using initializer lists
HashTable<int,bool> table { std::make_pair(3,true), std::make_pair(2,false)
};

// display the content of the hash table
cerr << table1;

// get the number of elements stored into the hash table
cerr << "number of elements in table1 = " << table1.size () << endl;

// create two copies of the hash table
HashTable<int,string> table2, table3 = table1;
table2 = table3;

// get the element whose key is 10
cerr << table1[10] << " = xxx" << endl;

// check whether there exists an element with key 20
if (table1.exists (20)) cerr << "element found" << endl;

// transform the hashtable of string into a hashtable of int assuming f is
// defined as: int f (const string& str) { return str.size (); }
HashTable<int,int> table = table1.map (f);

// remove two elements from table1 and table2
table1.erase (10);         // key = 10
table1.eraseByVal ("yyy"); // val = "yyy"
table2.clear ();

// check whether the hash table is empty
if (!table1.empty ()) cerr << "table not empty" << endl;

// check whether hashtables contain the same elements
if ((table1 == table2) && (table1 != table3))
  cerr << "check for equality/inequality" << endl;

// parse the content of a hashtable using an unsafe iterator
for (HashTable<int,string>::const_iterator iter = table1.cbegin();
      iter != table1.cend(); ++iter)
  cerr << *iter;
HashTable<int,string>::iterator iter = table1.begin();
iter += 2;
cerr << iter.key () << " " << iter.val ();

// use an iterator to point the element we wish to erase
HashTable<int,string>::iterator_safe iterS = table1.beginSafe ();
table1.erase ( table1.beginSafe () + 4 );
table1.erase ( iterS ); // this is safe because the iterator is safe

// check for iterator's safety
for (HashTable<int,string>::iterator_safe iter = table1.beginSafe ();
      iter != table1.endSafe (); ++iter )
  table1.eraseByVal ( *iter );

Template Parameters

Key	The type for keys in a gum::HashTable.
Val	The type for values in a gum::HashTable.
Alloc	The gum::HashTable allocator.

Definition at line 666 of file hashTable.h.

Member Typedef Documentation

allocator_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::allocator_type = Alloc

Types for STL compliance.

Definition at line 679 of file hashTable.h.

Bucket

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::Bucket = HashTableBucket< Key, Val >

The buckets where data are stored.

Definition at line 687 of file hashTable.h.

BucketAllocator

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::BucketAllocator = typename Alloc::template rebind< Bucket >::other

The Bucket allocator.

Definition at line 690 of file hashTable.h.

const_iterator

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::const_iterator = HashTableConstIterator< Key, Val >

Types for STL compliance.

Definition at line 681 of file hashTable.h.

const_iterator_safe

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::const_iterator_safe = HashTableConstIteratorSafe< Key, Val >

Types for STL compliance.

Definition at line 683 of file hashTable.h.

const_pointer

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::const_pointer = const value_type*

Types for STL compliance.

Definition at line 676 of file hashTable.h.

const_reference

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::const_reference = const value_type&

Types for STL compliance.

Definition at line 674 of file hashTable.h.

difference_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::difference_type = std::ptrdiff_t

Types for STL compliance.

Definition at line 678 of file hashTable.h.

iterator

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::iterator = HashTableIterator< Key, Val >

Types for STL compliance.

Definition at line 680 of file hashTable.h.

iterator_safe

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::iterator_safe = HashTableIteratorSafe< Key, Val >

Types for STL compliance.

Definition at line 682 of file hashTable.h.

key_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::key_type = Key

Types for STL compliance.

Definition at line 670 of file hashTable.h.

mapped_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::mapped_type = Val

Types for STL compliance.

Definition at line 671 of file hashTable.h.

pointer

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::pointer = value_type*

Types for STL compliance.

Definition at line 675 of file hashTable.h.

reference

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::reference = value_type&

Types for STL compliance.

Definition at line 673 of file hashTable.h.

size_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::size_type = Size

Types for STL compliance.

Definition at line 677 of file hashTable.h.

value_type

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

using gum::HashTable< Key, Val, Alloc >::value_type = std::pair< const Key, Val >

Types for STL compliance.

Definition at line 672 of file hashTable.h.

Constructor & Destructor Documentation

HashTable() [1/5]

template<typename Key , typename Val , typename Alloc >

gum::HashTable< Key, Val, Alloc >::HashTable ( Size  size_param = HashTableConst::default_size, bool  resize_pol = HashTableConst::default_resize_policy, bool  key_uniqueness_pol = HashTableConst::default_uniqueness_policy  )

explicit

Default constructor.

The true capacity (vector's size) of the hashtable will be the lowest number greater than or equal to size_param that is also a power of 2. The second optional argument is the resizing policy. By default, each time there is an average of 3 elements by node, the size of the hashtable is automatically multiplied by 2. But the user may pass false as argument to resize_pol to disable this feature.

Parameters

size_param	The initial size of the gum::HashTable.
resize_pol	The policy for resizing the hashtable when new elements are added (possible values: true = automatic resize and false = manual resize).
key_uniqueness_pol	Uniqueness policy : should we prevent inserting the same key more than once in the table?

Definition at line 366 of file hashTable_tpl.h.

                                                                   :
      // size must be >= 2 else we lose all the bits of the hash function
      _size_{Size(1) << _hashTableLog2_(std::max(Size(2), size_param))},
      _resize_policy_{resize_pol}, _key_uniqueness_policy_{key_uniqueness_pol} {
    // for debugging purposes
    GUM_CONSTRUCTOR(HashTable);

    // finalize the creation
    _create_(_size_);
  }

HashTable() [2/5]

template<typename Key, typename Val, typename Alloc >

gum::HashTable< Key, Val, Alloc >::HashTable ( std::initializer_list< std::pair< Key, Val > >  list )

explicit

Initializer list constructor.

Parameters

list	The initialized list.

Definition at line 380 of file hashTable_tpl.h.

                                                                                         :
      // size must be >= 2 else we lose all the bits of the hash function
      _size_{Size(1) << _hashTableLog2_(std::max< Size >(Size(2), Size(list.size()) / 2))} {
    // for debugging purposes
    GUM_CONSTRUCTOR(HashTable);

    // setup the  _nodes_ vector (contains only empty lists)
    _create_(_size_);

    // insert all the elements
    for (const auto& elt: list) {
      insert(elt);
    }
  }

HashTable() [3/5]

template<typename Key, typename Val, typename Alloc>

gum::HashTable< Key, Val, Alloc >::HashTable

(

const HashTable< Key, Val, Alloc > &

from

)

Copy constructor.

This creates a new hashtable the content of which is similar to that of the table passed in argument. Beware: similar does not mean that both tables share the same objects, but rather that the objects stored in the newly created table are copies of those of the table passed in argument. In particular, the new hash table inherits the parameters (resize policy, uniqueness policy) of table 'from'.

Parameters

from	The gum::HashTable to copy.

Definition at line 396 of file hashTable_tpl.h.

                                                                                   :
      _size_{table._size_}, _resize_policy_{table._resize_policy_},
      _key_uniqueness_policy_{table._key_uniqueness_policy_}, _begin_index_{table._begin_index_} {
    // for debugging purposes
    GUM_CONS_CPY(HashTable);

    // setup the  _nodes_ vector (contains only empty lists)
    _create_(_size_);

    // fill with the content of table
    _copy_(table);
  }

HashTable() [4/5]

template<typename Key, typename Val, typename Alloc >

template<typename OtherAlloc >

gum::HashTable< Key, Val, Alloc >::HashTable

(

const HashTable< Key, Val, OtherAlloc > &

from

)

Generalized copy constructor.

This creates a new hashtable the content of which is similar to that of the table passed in argument. Beware: similar does not mean that both tables share the same objects, but rather that the objects stored in the newly created table are copies of those of the table passed in argument. In particular, the new hash table inherits the parameters (resize policy, uniqueness policy) of table 'table'

Parameters

from	The gum::HashTable to copy.

Definition at line 411 of file hashTable_tpl.h.

                                                                                        :
      _size_{table._size_}, _resize_policy_{table._resize_policy_},
      _key_uniqueness_policy_{table._key_uniqueness_policy_}, _begin_index_{table._begin_index_} {
    // for debugging purposes
    GUM_CONS_CPY(HashTable);

    // setup the  _nodes_ vector (contains only empty lists)
    _create_(_size_);

    // fill with the content of table
    _copy_(table);
  }

HashTable() [5/5]

template<typename Key, typename Val, typename Alloc>

gum::HashTable< Key, Val, Alloc >::HashTable

(

HashTable< Key, Val, Alloc > &&

from

)

Move constructor.

Parameters

from	The gum::HashTable to move.

Definition at line 425 of file hashTable_tpl.h.

                                                                              :
      _nodes_(std::move(table._nodes_)), _size_{table._size_}, _nb_elements_{table._nb_elements_},
      _hash_func_{table._hash_func_}, _resize_policy_{table._resize_policy_},
      _key_uniqueness_policy_{table._key_uniqueness_policy_}, _begin_index_{table._begin_index_},
      _safe_iterators_(std::move(table._safe_iterators_)), _alloc_(std::move(table._alloc_)) {
    // for debugging purposes
    table._size_ = 0;
    GUM_CONS_MOV(HashTable);
  }

~HashTable()

template<typename Key , typename Val , typename Alloc >

INLINE gum::HashTable< Key, Val, Alloc >::~HashTable

(

)

Class destructor.

Definition at line 457 of file hashTable_tpl.h.

                                                  {
    // for debugging purposes
    GUM_DESTRUCTOR(HashTable);

    // update all the registered iterators: they should now point to nullptr
    // and their hashtable should be set to nullptr
    _clearIterators_();
  }

Member Function Documentation

_clearIterators_()

template<typename Key , typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::_clearIterators_ ( )

private

Clear all the safe iterators.

Definition at line 436 of file hashTable_tpl.h.

                                                             {
    const Size len = _safe_iterators_.size();
    for (Size i = Size(0); i < len; ++i)
      _safe_iterators_[i]->clear();
  }

_copy_()

template<typename Key, typename Val, typename Alloc >

template<typename OtherAlloc >

void gum::HashTable< Key, Val, Alloc >::_copy_ ( const HashTable< Key, Val, OtherAlloc > &  table )

private

A function used to perform copies of HashTables.

This code is shared by the copy constructor and the copy operator. The function ensures that when a memory allocation problem occurs:

• no memory leak occurs
• the hashtable returned is empty but in a coherent state
• an exception is thrown

The function assumes that both this and table have arrays ' __nodes' of the same size.

Parameters

table

The gum::HashTable to copy.

Template Parameters

OtherAlloc

The other gum::HashTable allocator.

Definition at line 296 of file hashTable_tpl.h.

                                                                                          {
    // in debug mode, check that this and table have ' __nodes' arrays of the
    // same size
    GUM_ASSERT(table._size_ == _size_);

    // try to fill the array of chained lists
    for (Size i = 0; i < table._size_; ++i) {
      try {
        _nodes_[i] = table._nodes_[i];
      } catch (...) {
        // here we could allocate the  _nodes_[j], j=0..i-1, so we should
        // deallocate them
        for (Size j = 0; j < _size_; ++j)
          _nodes_[j].clear();

        _nb_elements_ = Size(0);

        // propagate the exception
        throw;
      }
    }

    _nb_elements_ = table._nb_elements_;
  }

_create_()

template<typename Key , typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::_create_ ( Size  size )

private

Used by all default constructors (general and specialized).

Parameters

size	The size of the gum::HashTable to create.

Definition at line 349 of file hashTable_tpl.h.

                                                              {
    // setup the  _nodes_ vector (contains only empty lists)
    _nodes_.resize(size);

    for (auto& list: _nodes_) {
      list.setAllocator(_alloc_);
    }

    // set up properly the hash function
    _hash_func_.resize(size);

    // make sure the end() iterator is constructed properly
    end4Statics();
    endSafe4Statics();
  }

_erase_()

template<typename Key, typename Val, typename Alloc >

void gum::HashTable< Key, Val, Alloc >::_erase_ ( HashTableBucket< Key, Val > *  bucket, Size  index  )

private

Erases a given bucket.

Definition at line 954 of file hashTable_tpl.h.

                                                                                            {
    if (bucket == nullptr) return;

    // update the registered iterators pointing to this bucket
    for (auto iter: _safe_iterators_) {
      if (iter->_bucket_ == bucket) {
        iter->operator++();
        iter->_next_bucket_ = iter->_bucket_;
        iter->_bucket_      = nullptr;
      } else if (iter->_next_bucket_ == bucket) {
        iter->_bucket_ = bucket;
        iter->operator++();
        iter->_next_bucket_ = iter->_bucket_;
        iter->_bucket_      = nullptr;
      }
    }

    // remove the element from the  _nodes_ vector
    _nodes_[index].erase(bucket);

    --_nb_elements_;

    if ((index == _begin_index_) && _nodes_[index].empty()) {
      _begin_index_ = std::numeric_limits< Size >::max();
    }
  }

_insert_()

template<typename Key , typename Val , typename Alloc >

void gum::HashTable< Key, Val, Alloc >::_insert_ ( Bucket *  bucket )

private

Adds a new element (actually a copy of this element) in the hash table.

If there already exists an element with the same key in the list and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Parameters

bucket

The bucket inserted in the hash table.

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Definition at line 806 of file hashTable_tpl.h.

                                                                                 {
    Size hash_key = _hash_func_(bucket->key());

    // check that there does not already exist an element with the same key
    if (_key_uniqueness_policy_ && _nodes_[hash_key].exists(bucket->key())) {
      // remove the bucket from memory
      Key k = bucket->key();
      _alloc_.destroy(bucket);
      _alloc_.deallocate(bucket, 1);
      GUM_ERROR(DuplicateElement,
                "the hashtable contains an element with the same key (" << k << ")");
    }

    // check whether there is sufficient space to insert the new pair
    // if not, resize the current hashtable
    if (_resize_policy_ && (_nb_elements_ >= _size_ * HashTableConst::default_mean_val_by_slot)) {
      resize(_size_ << 1);
      hash_key = _hash_func_(bucket->key());
    }

    // add the new pair
    _nodes_[hash_key].insert(bucket);
    ++_nb_elements_;

    // recompute the index of the beginning of the hashtable if possible
    // WARNING: if  _begin_index_ = std::numeric_limits<Size>::max (), we CANNOT
    // recompute the index because we cannot know whether the current index is
    // equal to max because there was no element in the hashtable or whether a
    // previous  _erase_() has set the index to max.
    if (_begin_index_ < hash_key) { _begin_index_ = hash_key; }
  }

begin() [1/2]

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::iterator gum::HashTable< Key, Val, Alloc >::begin

(

)

Returns an unsafe iterator pointing to the beginning of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns an unsafe iterator pointing to the beginning of the hashtable.

Definition at line 606 of file hashTable_tpl.h.

                                                                                           {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return iterator{end()};
    else
      return iterator{*this};
  }

begin() [2/2]

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::const_iterator gum::HashTable< Key, Val, Alloc >::begin

(

)

const

Returns an unsafe const_iterator pointing to the beginning of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns an unsafe const_iterator pointing to the beginning of the hashtable.

Definition at line 616 of file hashTable_tpl.h.

                                               {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return const_iterator{end()};
    else
      return const_iterator{*this};
  }

beginSafe() [1/2]

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::iterator_safe gum::HashTable< Key, Val, Alloc >::beginSafe

(

)

Returns the safe iterator pointing to the beginning of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe iterator pointing to the beginning of the hashtable.

Definition at line 666 of file hashTable_tpl.h.

                                             {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return iterator_safe{endSafe()};
    else
      return iterator_safe{*this};
  }

beginSafe() [2/2]

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::const_iterator_safe gum::HashTable< Key, Val, Alloc >::beginSafe

(

)

const

Returns the safe const_iterator pointing to the beginning of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe const_iterator pointing to the beginning of the hashtable.

Definition at line 676 of file hashTable_tpl.h.

                                                   {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return const_iterator_safe{endSafe()};
    else
      return const_iterator_safe{*this};
  }

capacity()

template<typename Key , typename Val , typename Alloc >

INLINE Size gum::HashTable< Key, Val, Alloc >::capacity ( ) const

noexcept

Returns the number of slots in the 'nodes' vector of the hashtable.

The method runs in constant time.

Returns

Returns the number of slots in the 'nodes' vector of the hashtable.

Definition at line 710 of file hashTable_tpl.h.

                                                                    {
    return _size_;
  }

cbegin()

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::const_iterator gum::HashTable< Key, Val, Alloc >::cbegin

(

)

const

Returns an unsafe const_iterator pointing to the beginning of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns an unsafe const_iterator pointing to the beginning of the hashtable.

Definition at line 626 of file hashTable_tpl.h.

                                                {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return const_iterator{cend()};
    else
      return const_iterator{*this};
  }

cbeginSafe()

template<typename Key , typename Val , typename Alloc >

INLINE HashTable< Key, Val, Alloc >::const_iterator_safe gum::HashTable< Key, Val, Alloc >::cbeginSafe

(

)

const

Returns the safe const_iterator pointing to the beginning of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe const_iterator pointing to the beginning of the hashtable.

Definition at line 686 of file hashTable_tpl.h.

                                                    {
    // if the table is empty, make the begin and end point to the same element
    if (_nb_elements_ == Size(0))
      return const_iterator_safe{cendSafe()};
    else
      return const_iterator_safe{*this};
  }

cend()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator & gum::HashTable< Key, Val, Alloc >::cend ( ) const

noexcept

Returns the unsafe const_iterator pointing to the end of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns the unsafe const_iterator pointing to the end of the hashtable.

Definition at line 597 of file hashTable_tpl.h.

                                                       {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(
       reinterpret_cast< const const_iterator* >(HashTableIteratorStaticEnd::_HashTableIterEnd_));
  }

cendSafe()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator_safe & gum::HashTable< Key, Val, Alloc >::cendSafe ( ) const

noexcept

Returns the safe const_iterator pointing to the end of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe const_iterator pointing to the end of the hashtable.

Definition at line 656 of file hashTable_tpl.h.

                                                           {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(reinterpret_cast< const const_iterator_safe* >(
       HashTableIteratorStaticEnd::_HashTableIterEndSafe_));
  }

clear()

template<typename Key , typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::clear

(

)

Removes all the elements in the hash table.

The function does not resize the nodes vector (even if the size of this one has been increased after the creation of the hash table) and it resets the iterators on the hash table to end. The method runs in linear time w.r.t. the number of iterators pointing to the hash table.

Definition at line 443 of file hashTable_tpl.h.

                                                  {
    // update all the registered iterators: they should now point to nullptr
    // and they are positioned to the end of the hashtable.
    _clearIterators_();

    // remove the buckets
    for (Size i = Size(0); i < _size_; ++i)
      _nodes_[i].clear();

    _nb_elements_ = Size(0);
    _begin_index_ = std::numeric_limits< Size >::max();
  }

constEnd4Statics()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator & gum::HashTable< Key, Val, Alloc >::constEnd4Statics ( )

static

Returns the end iterator for other classes' statics (read the detailed description of this method).

To reduce memory consumption of hash tables (which are heavily used in aGrUM) while allowing fast for(iter=begin(); iter!=end();++iter) loops, end iterators are created just once as a static member of a non-template hashtable. While this scheme is efficient and it works quite effectively when manipulating hashtables, it has a drawback: other classes with static members using the HashTable's end() iterator may fail to work due to the well known "static initialization order fiasco" (see Marshall Cline's C++ FAQ for more details about this C++ feature). OK, so what is the problem? Consider for instance class Set. A Set contains a hashtable that stores all its elements in a convenient way. To reduce memory consumption, Set::end iterator is a static member that is initialized with a HashTable's end iterator. If the compiler decides to initialize Set::end before initializing HashTable::end, then Set::end will be in an incoherent state. Unfortunately, we cannot know for sure in which order static members will be initialized (the order is a compiler's decision). Hence, we shall enforce the fact that HashTable::end is initialized before Set::end. Using method HashTable::end4Statics will ensure this fact: it uses the C++ "construct on first use" idiom (see the C++ FAQ) that ensures that the order fiasco is avoided. More precisely, end4Statics initializes a global variable that is the very end iterator used by all hashtables. Now, this induces a small overhead. So, we also provide a HashTable::end() method that returns the end iterator without this small overhead, but assuming that function end4Statics has already been called once (which is always the case) when a hashtable has been created.

So, to summarize: when initializing static members, use constEnd4Statics() rather than cend(). In all the other cases, use simply the usual method cend().

Returns

Returns the end iterator for other classes' statics (read the detailed description of this method).

Definition at line 329 of file hashTable_tpl.h.

                                                    {
    return *(
       reinterpret_cast< const const_iterator* >(HashTableIteratorStaticEnd::constEnd4Statics()));
  }

constEndSafe4Statics()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator_safe & gum::HashTable< Key, Val, Alloc >::constEndSafe4Statics ( )

static

Returns the end iterator for other classes' statics (read the detailed description of this method).

To reduce memory consumption of hash tables (which are heavily used in aGrUM) while allowing fast for(iter=begin(); iter!=end();++iter) loops, end iterators are created just once as a static member of a non-template hashtable. While this scheme is efficient and it works quite effectively when manipulating hashtables, it has a drawback: other classes with static members using the HashTable's end() iterator may fail to work due to the well known "static initialization order fiasco" (see Marshall Cline's C++ FAQ for more details about this C++ feature). OK, so what is the problem? Consider for instance class Set. A Set contains a hashtable that stores all its elements in a convenient way. To reduce memory consumption, Set::end iterator is a static member that is initialized with a HashTable's end iterator. If the compiler decides to initialize Set::end before initializing HashTable::end, then Set::end will be in an incoherent state. Unfortunately, we cannot know for sure in which order static members will be initialized (the order is a compiler's decision). Hence, we shall enforce the fact that HashTable::end is initialized before Set::end. Using method HashTable::end4Statics will ensure this fact: it uses the C++ "construct on first use" idiom (see the C++ FAQ) that ensures that the order fiasco is avoided. More precisely, end4Statics initializes a global variable that is the very end iterator used by all hashtables. Now, this induces a small overhead. So, we also provide a HashTable::end() method that returns the end iterator without this small overhead, but assuming that function end4Statics has already been called once (which is always the case) when a hashtable has been created.

So, to summarize: when initializing static members, use constEndSafe4Statics() rather than cendSafe(). In all the other cases, use simply the usual method cendSafe().

Returns

Returns the end iterator for other classes' statics (read the detailed description of this method).

Definition at line 343 of file hashTable_tpl.h.

                                                        {
    return *(reinterpret_cast< const const_iterator_safe* >(
       HashTableIteratorStaticEnd::constEndSafe4Statics()));
  }

emplace() [1/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename... Args>

INLINE HashTable< Key, Val, Alloc >::value_type& gum::HashTable< Key, Val, Alloc >::emplace

(

Args &&...

args

)

Definition at line 905 of file hashTable_tpl.h.

                                                         {
    Bucket* bucket = _alloc_.allocate(1);

    try {
      _alloc_.construct(bucket,
                        HashTableBucket< Key, Val >::Emplace::EMPLACE,
                        std::forward< Args >(args)...);
    } catch (...) {
      _alloc_.deallocate(bucket, 1);
      throw;
    }

    _insert_(bucket);
    return bucket->elt();
  }

emplace() [2/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename... Args>

value_type& gum::HashTable< Key, Val, Alloc >::emplace

(

Args &&...

args

)

Emplace a new element into the hashTable.

If there already exists an element with the same key in the list and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Returns

a reference to the pair (key,val) inserted in the hash table.

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Parameters

args	The element to emplace.

empty()

template<typename Key , typename Val , typename Alloc >

INLINE bool gum::HashTable< Key, Val, Alloc >::empty ( ) const

noexcept

Indicates whether the hash table is empty.

Returns

Returns true if the gum::HashTable is empty.

Definition at line 1042 of file hashTable_tpl.h.

                                                                 {
    return (_nb_elements_ == Size(0));
  }

end() [1/2]

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::iterator & gum::HashTable< Key, Val, Alloc >::end ( )

noexcept

Returns the unsafe iterator pointing to the end of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns the unsafe iterator pointing to the end of the hashtable.

Definition at line 578 of file hashTable_tpl.h.

                                                {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(reinterpret_cast< const iterator* >(HashTableIteratorStaticEnd::_HashTableIterEnd_));
  }

end() [2/2]

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator & gum::HashTable< Key, Val, Alloc >::end ( ) const

noexcept

Returns the unsafe const_iterator pointing to the end of the hashtable.

Unsafe iterators are slightly faster than safe iterators. However, BE CAREFUL when using them: they should ONLY be used when you have the guarantee that they will never point to a deleted element. If unsure, prefer using the safe iterators (those are only slightly slower).

Returns

Returns the unsafe const_iterator pointing to the end of the hashtable.

Definition at line 587 of file hashTable_tpl.h.

                                                      {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(
       reinterpret_cast< const const_iterator* >(HashTableIteratorStaticEnd::_HashTableIterEnd_));
  }

end4Statics()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::iterator & gum::HashTable< Key, Val, Alloc >::end4Statics ( )

static

Returns the end iterator for other classes' statics (read the detailed description of this method).

To reduce memory consumption of hash tables (which are heavily used in aGrUM) while allowing fast for(iter=begin(); iter!=end();++iter) loops, end iterators are created just once as a static member of a non-template hashtable. While this scheme is efficient and it works quite effectively when manipulating hashtables, it has a drawback: other classes with static members using the HashTable's end() iterator may fail to work due to the well known "static initialization order fiasco" (see Marshall Cline's C++ FAQ for more details about this C++ feature). OK, so what is the problem? Consider for instance class Set. A Set contains a hashtable that stores all its elements in a convenient way. To reduce memory consumption, Set::end iterator is a static member that is initialized with a HashTable's end iterator. If the compiler decides to initialize Set::end before initializing HashTable::end, then Set::end will be in an incoherent state. Unfortunately, we cannot know for sure in which order static members will be initialized (the order is a compiler's decision). Hence, we shall enforce the fact that HashTable::end is initialized before Set::end. Using method HashTable::end4Statics will ensure this fact: it uses the C++ "construct on first use" idiom (see the C++ FAQ) that ensures that the order fiasco is avoided. More precisely, end4Statics initializes a global variable that is the very end iterator used by all hashtables. Now, this induces a small overhead. So, we also provide a HashTable::end() method that returns the end iterator without this small overhead, but assuming that function end4Statics has already been called once (which is always the case) when a hashtable has been created.

So, to summarize: when initializing static members, use end4Statics() rather than end(). In all the other cases, use simply the usual method end().

Returns

Returns the end iterator for other classes' statics (read the detailed description of this method).

Definition at line 323 of file hashTable_tpl.h.

                                               {
    return *(reinterpret_cast< const iterator* >(HashTableIteratorStaticEnd::end4Statics()));
  }

endSafe() [1/2]

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::iterator_safe & gum::HashTable< Key, Val, Alloc >::endSafe ( )

noexcept

Returns the safe iterator pointing to the end of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe iterator pointing to the end of the hashtable.

Definition at line 636 of file hashTable_tpl.h.

                                                    {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(reinterpret_cast< const iterator_safe* >(
       HashTableIteratorStaticEnd::_HashTableIterEndSafe_));
  }

endSafe() [2/2]

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::const_iterator_safe & gum::HashTable< Key, Val, Alloc >::endSafe ( ) const

noexcept

Returns the safe const_iterator pointing to the end of the hashtable.

Safe iterators are slightly slower than unsafe ones but they guarantee that you will never get a segfault if they try to access to a deleted element or if they try a ++ operation from a deleted element.

Returns

Returns the safe const_iterator pointing to the end of the hashtable.

Definition at line 646 of file hashTable_tpl.h.

                                                          {
    // note that, here, we know for sure that HashTableIterEnd has been properly
    // initialized as it is initialized by end4Statics, which is called by
    // all hashtables' constructors
    return *(reinterpret_cast< const const_iterator_safe* >(
       HashTableIteratorStaticEnd::_HashTableIterEndSafe_));
  }

endSafe4Statics()

template<typename Key , typename Val , typename Alloc >

INLINE const HashTable< Key, Val, Alloc >::iterator_safe & gum::HashTable< Key, Val, Alloc >::endSafe4Statics ( )

static

Returns the end iterator for other classes' statics (read the detailed description of this method).

To reduce memory consumption of hash tables (which are heavily used in aGrUM) while allowing fast for(iter=begin(); iter!=end();++iter) loops, end iterators are created just once as a static member of a non-template hashtable. While this scheme is efficient and it works quite effectively when manipulating hashtables, it has a drawback: other classes with static members using the HashTable's end() iterator may fail to work due to the well known "static initialization order fiasco" (see Marshall Cline's C++ FAQ for more details about this C++ feature). OK, so what is the problem? Consider for instance class Set. A Set contains a hashtable that stores all its elements in a convenient way. To reduce memory consumption, Set::end iterator is a static member that is initialized with a HashTable's end iterator. If the compiler decides to initialize Set::end before initializing HashTable::end, then Set::end will be in an incoherent state. Unfortunately, we cannot know for sure in which order static members will be initialized (the order is a compiler's decision). Hence, we shall enforce the fact that HashTable::end is initialized before Set::end. Using method HashTable::end4Statics will ensure this fact: it uses the C++ "construct on first use" idiom (see the C++ FAQ) that ensures that the order fiasco is avoided. More precisely, end4Statics initializes a global variable that is the very end iterator used by all hashtables. Now, this induces a small overhead. So, we also provide a HashTable::end() method that returns the end iterator without this small overhead, but assuming that function end4Statics has already been called once (which is always the case) when a hashtable has been created.

So, to summarize: when initializing static members, use endSafe4Statics() rather than endSafe(). In all the other cases, use simply the usual method endSafe().

Returns

Returns the end iterator for other classes' statics (read the detailed description of this method).

Definition at line 336 of file hashTable_tpl.h.

                                                   {
    return *(
       reinterpret_cast< const iterator_safe* >(HashTableIteratorStaticEnd::endSafe4Statics()));
  }

erase() [1/3]

template<typename Key, typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::erase

(

const Key &

key

)

Removes a given element from the hash table.

The element is the first one encountered in the list (from begin() to end()) having the specified key. If no such element can be found, nothing is done (in particular, it does not throw any exception). The function never resizes the nodes vector (even if the resizing policy would enable to decrease this size). The method runs in average in time linear to the number of iterators pointing to the table if the automatic resizing policy is set (else it is in linear time in the number of elements of the hash table plus the number of iterators).

Parameters

key	The key of the element to remove.

Definition at line 982 of file hashTable_tpl.h.

                                                                {
    // get the hashed key
    Size hash = _hash_func_(key);

    // get the bucket containing the element to erase
    HashTableBucket< Key, Val >* bucket = _nodes_[hash].bucket(key);

    _erase_(bucket, hash);
  }

erase() [2/3]

template<typename Key, typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::erase

(

const iterator_safe &

iter

)

Removes a given element from the hash table.

This method updates all the safe iterators pointing to the deleted element, i.e., when trying to dereference those iterators, an exception will be raised because they will know that the element they point to no longer exists.

Parameters

iter	An iterator over the element to remove.

Definition at line 993 of file hashTable_tpl.h.

                                                                           {
    _erase_(iter._getBucket_(), iter._getIndex_());
  }

erase() [3/3]

template<typename Key, typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::erase

(

const const_iterator_safe &

iter

)

Removes a given element from the hash table.

This method updates all the safe iterators pointing to the deleted element, i.e., when trying to dereference those iterators, an exception will be raised because they will know that the element they point to no longer exists.

Parameters

iter	An iterator over the element to remove.

Definition at line 998 of file hashTable_tpl.h.

                                                                                 {
    _erase_(iter._getBucket_(), iter._getIndex_());
  }

eraseAllVal()

template<typename Key , typename Val, typename Alloc >

void gum::HashTable< Key, Val, Alloc >::eraseAllVal

(

const Val &

val

)

Removes all the elements having a certain value from the hash table.

If no such element can be found, nothing is done (in particular, it does not throw any exception). The function never resizes the nodes vector (even if the resizing policy would enable to decrease this size). Comparisons between Val instances are performed through == operators.

Parameters

val	The value to remove.

Definition at line 1035 of file hashTable_tpl.h.

                                                               {
    for (auto iterAll = cbeginSafe(); iterAll != cendSafe(); ++iterAll) {
      if (iterAll._bucket_->val() == val) { _erase_(iterAll._bucket_, iterAll._index_); }
    }
  }

eraseByVal()

template<typename Key , typename Val, typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::eraseByVal

(

const Val &

val

)

Removes a given element from the hash table.

The element is the first one encountered in the list (from begin() to end()) having the specified value. If no such element can be found, nothing is done (in particular, it does not throw any exception). The function never resizes the nodes vector (even if the resizing policy would enable to decrease this size). Comparisons between Val instances are performed through == operators. Logically, this method should have been named "erase", however, this would have prevented creating hash tables where both keys and vals have the same type. Hence we chose to add "ByVal" after erase to make a difference between erasing by key and erasing by val.

Parameters

val	The value to remove.

Definition at line 1003 of file hashTable_tpl.h.

                                                                     {
    for (auto iter = cbegin(); iter != cend(); ++iter)
      if (iter._bucket_->val() == val) {
        _erase_(iter._getBucket_(), iter._getIndex_());
        return;
      }
  }

exists()

template<typename Key, typename Val , typename Alloc >

INLINE bool gum::HashTable< Key, Val, Alloc >::exists

(

const Key &

key

)

const

Checks whether there exists an element with a given key in the hashtable.

The method runs in average in constant time if the resizing policy is set.

Parameters

key	The key to test for existence.

Returns

True if key is in this gum::HashTable.

Definition at line 715 of file hashTable_tpl.h.

                                                                       {
    return _nodes_[_hash_func_(key)].exists(key);
  }

getWithDefault() [1/2]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::mapped_type & gum::HashTable< Key, Val, Alloc >::getWithDefault	(	const Key &	key,
		const Val &	default_value
	)

Returns a reference on the element the key of which is passed in argument.

In case of multiple identical keys in the hash table, the first value encountered is returned. The method runs in constant time. In case of not found key, (key,default_value) is inserted in *this.

Parameters

key	The key for wich we want the value.
default_value	The default value to return if key does not match any value.

Returns

Returns a reference on the element the key of which is passed in argument.

Definition at line 923 of file hashTable_tpl.h.

                                                                                          {
    Bucket* bucket = _nodes_[_hash_func_(key)].bucket(key);

    if (bucket == nullptr)
      return insert(key, default_value).second;
    else
      return bucket->val();
  }

getWithDefault() [2/2]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::mapped_type & gum::HashTable< Key, Val, Alloc >::getWithDefault	(	Key &&	key,
		Val &&	default_value
	)

Returns a reference on the element the key of which is passed in argument.

In case of multiple identical keys in the hash table, the first value encountered is returned. The method runs in constant time. In case of not found key, (key,default_value) is inserted in *this.

Parameters

key	The key for wich we want the value.
default_value	The default value to return if key does not match any value.

Returns

Returns a reference on the element the key of which is passed in argument.

Definition at line 934 of file hashTable_tpl.h.

                                                                                {
    Bucket* bucket = _nodes_[_hash_func_(key)].bucket(key);

    if (bucket == nullptr)
      return insert(std::move(key), std::move(default_value)).second;
    else
      return bucket->val();
  }

insert() [1/4]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::value_type & gum::HashTable< Key, Val, Alloc >::insert	(	const Key &	key,
		const Val &	val
	)

Adds a new element (actually a copy of this element) into the hash table.

If there already exists an element with the same key in the table and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Returns

As only a copy of val is inserted into the hashtable, the method returns a reference on a copy of the pair (key,val).

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Parameters

key	The key to add.
val	The value to add.

Returns

The value added by copy to this gum::HashTable.

Definition at line 840 of file hashTable_tpl.h.

                                                                              {
    Bucket* bucket = _alloc_.allocate(1);

    try {
      _alloc_.construct(bucket, thekey, theval);
    } catch (...) {
      _alloc_.deallocate(bucket, 1);
      throw;
    }

    _insert_(bucket);
    return bucket->elt();
  }

insert() [2/4]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::value_type & gum::HashTable< Key, Val, Alloc >::insert	(	Key &&	key,
		Val &&	val
	)

Moves a new element in the hash table.

If there already exists an element with the same key in the table and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Returns

a reference to the pair (key,val) in the hashtable.

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Parameters

key	The key to move.
val	The value to move.

Returns

The value moved to this gum::HashTable.

Definition at line 856 of file hashTable_tpl.h.

                                                                    {
    Bucket* bucket = _alloc_.allocate(1);

    try {
      _alloc_.construct(bucket, std::move(thekey), std::move(theval));
    } catch (...) {
      _alloc_.deallocate(bucket, 1);
      throw;
    }

    _insert_(bucket);
    return bucket->elt();
  }

insert() [3/4]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::value_type & gum::HashTable< Key, Val, Alloc >::insert

(

const std::pair< Key, Val > &

elt

)

Adds a new element (actually a copy of this element) into the hash table.

If there already exists an element with the same key in the table and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Returns

As only a copy of val is inserted into the hashtable, the method returns a reference on a copy of the pair (key,val).

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Parameters

elt	The pair of key value to add.

Returns

The value added by copy to this gum::HashTable.

Definition at line 872 of file hashTable_tpl.h.

                                                                        {
    Bucket* bucket = _alloc_.allocate(1);

    try {
      _alloc_.construct(bucket, reinterpret_cast< const value_type& >(elt));
    } catch (...) {
      _alloc_.deallocate(bucket, 1);
      throw;
    }

    _insert_(bucket);
    return bucket->elt();
  }

insert() [4/4]

template<typename Key, typename Val, typename Alloc >

INLINE HashTable< Key, Val, Alloc >::value_type & gum::HashTable< Key, Val, Alloc >::insert

(

std::pair< Key, Val > &&

elt

)

Moves a new element in the hash table.

If there already exists an element with the same key in the table and the uniqueness policy prevents multiple identical keys to belong to the same hashtable, an exception DuplicateElement is thrown. If the uniqueness policy is not set, the method runs in the worst case in constant time, else if the automatic resizing policy is set, it runs in constant time in average linear in the number of elements by slot.

Returns

a reference to the pair (key,val) in the hashtable.

Exceptions

DuplicateElement

is thrown when attempting to insert a pair (key,val) in a hash table containing already a pair with the same key and when the hash table's uniqueness policy is set.

Parameters

elt	The pair of key value to move in this gum::HashTable.

Returns

The value moved to this gum::HashTable.

Definition at line 888 of file hashTable_tpl.h.

                                                                   {
    Bucket* bucket = _alloc_.allocate(1);

    try {
      _alloc_.construct(bucket, std::move(reinterpret_cast< value_type& >(elt)));
    } catch (...) {
      _alloc_.deallocate(bucket, 1);
      throw;
    }

    _insert_(bucket);
    return bucket->elt();
  }

key()

template<typename Key, typename Val , typename Alloc >

INLINE const Key & gum::HashTable< Key, Val, Alloc >::key

(

const Key &

key

)

const

Returns a reference on a given key.

Some complex structures use pointers on keys of hashtables. These structures thus require that we do not only get a copy of a given key, but the key stored in the hashtable itself. This is the very purpose of this function.

Parameters

key	The key to return.

Returns

Returns a reference on a given key.

Exceptions

NotFound

Raised if the element cannot be found.

Definition at line 1025 of file hashTable_tpl.h.

                                                                          {
    // get the bucket corresponding to the key
    Bucket* bucket = _nodes_[_hash_func_(key)].bucket(key);

    if (bucket == nullptr) { GUM_ERROR(NotFound, "key does not belong to the hashtable") }

    return bucket->key();
  }

keyByVal()

template<typename Key , typename Val, typename Alloc >

INLINE const Key & gum::HashTable< Key, Val, Alloc >::keyByVal

(

const Val &

val

)

const

Returns a reference on the key given a value.

In case of multiple identical values in the hash table, the first key encountered is returned. The method runs in linear time.

Parameters

val	The value for which the key is returned.

Returns

Returns a reference on the key given a value.

Exceptions

NotFound

Raised if the element cannot be found.

Definition at line 1017 of file hashTable_tpl.h.

                                                                               {
    for (auto iter = begin(); iter != end(); ++iter)
      if (iter._bucket_->val() == val) return iter.key();

    GUM_ERROR(NotFound, "not enough elements in the chained list")
  }

keyUniquenessPolicy()

template<typename Key , typename Val , typename Alloc >

INLINE bool gum::HashTable< Key, Val, Alloc >::keyUniquenessPolicy ( ) const

noexcept

Returns the current checking policy.

Returns

Returns the current checking policy.

Definition at line 735 of file hashTable_tpl.h.

                                                                               {
    return _key_uniqueness_policy_;
  }

map() [1/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc >

HashTable< Key, Mount, OtherAlloc > INLINE gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(Val)	f,
		Size	size,
		bool	resize_pol,
		bool	key_uniqueness_pol
	)		const

Definition at line 1049 of file hashTable_tpl.h.

                                                                             {
    // determine the proper size of the hashtable
    // by default, the size of the table is set so that the table does not take
    // too much space while allowing to add a few elements without needing to
    // resize in autmatic resizing mode
    if (size == 0) size = std::max(Size(2), _nb_elements_ / 2);

    // create a new table
    HashTable< Key, Mount, OtherAlloc > table(size, resize_pol, key_uniqueness_pol);

    // fill the new hash table
    for (auto iter = begin(); iter != end(); ++iter) {
      table.insert(iter.key(), f(iter.val()));
    }

    return table;
  }

map() [2/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc >

HashTable< Key, Mount, OtherAlloc > INLINE gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(Val &)	f,
		Size	size,
		bool	resize_pol,
		bool	key_uniqueness_pol
	)		const

Definition at line 1073 of file hashTable_tpl.h.

                                                                             {
    // determine the proper size of the hashtable
    // by default, the size of the table is set so that the table does not take
    // too much space while allowing to add a few elements without needing to
    // resize in autmatic resizing mode
    if (size == Size(0)) size = std::max(Size(2), _nb_elements_ / 2);

    // create a new table
    HashTable< Key, Mount, OtherAlloc > table(size, resize_pol, key_uniqueness_pol);

    // fill the new hash table
    for (auto iter = begin(); iter != end(); ++iter) {
      table.insert(iter.key(), f(const_cast< Val& >(iter.val())));
    }

    return table;
  }

map() [3/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc >

HashTable< Key, Mount, OtherAlloc > INLINE gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(const Val &)	f,
		Size	size,
		bool	resize_pol,
		bool	key_uniqueness_pol
	)		const

Definition at line 1097 of file hashTable_tpl.h.

                                                                             {
    // determine the proper size of the hashtable
    // by default, the size of the table is set so that the table does not take
    // too much space while allowing to add a few elements without needing to
    // resize in autmatic resizing mode
    if (size == Size(0)) size = std::max(Size(2), _nb_elements_ / 2);

    // create a new table
    HashTable< Key, Mount, OtherAlloc > table(size, resize_pol, key_uniqueness_pol);

    // fill the new hash table
    for (auto iter = begin(); iter != end(); ++iter) {
      table.insert(iter.key(), f(iter.val()));
    }

    return table;
  }

map() [4/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc >

HashTable< Key, Mount, OtherAlloc > INLINE gum::HashTable< Key, Val, Alloc >::map	(	const Mount &	val,
		Size	size,
		bool	resize_pol,
		bool	key_uniqueness_pol
	)		const

Definition at line 1121 of file hashTable_tpl.h.

                                                                                     {
    // determine the proper size of the hashtable
    // by default, the size of the table is set so that the table does not take
    // too much space while allowing to add a few elements without needing to
    // resize in autmatic resizing mode
    if (size == Size(0)) size = std::max(Size(2), _nb_elements_ / 2);

    // create a new table
    HashTable< Key, Mount, OtherAlloc > table(size, resize_pol, key_uniqueness_pol);

    // fill the new hash table
    for (auto iter = begin(); iter != end(); ++iter) {
      table.insert(iter.key(), val);
    }

    return table;
  }

map() [5/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>

HashTable< Key, Mount, OtherAlloc > gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(Val)	f,
		Size	size = Size(0),
		bool	resize_pol = HashTableConst::default_resize_policy,
		bool	key_uniqueness_pol = HashTableConst::default_uniqueness_policy
	)		const

Transforms a hashtable of vals into a hashtable of mountains.

Warning

Although the resulting hashtable has the same number of elements as the original hashtable, by default, the size of the former may not be equal to that of the latter. Hence iterators on the original hashtable may not parse it in the same order as iterators on the resulting hashtable. To guarrantee that both hashtables have the same size (and thus have the elements in the same order), set the size argument to the size of the original hashtable.

Parameters

f	A function that maps any Val element into a Mount.
size	The size of the resulting hashtable. When equal to 0, a default size is computed that is a good trade-off between space consumption and efficiency of new elements insertions
resize_pol	the resizing policy (automatic or manual resizing)
key_uniqueness_pol	uniqueness policy

Returns

Returns the gum::HashTable of mountains.

map() [6/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>

HashTable< Key, Mount, OtherAlloc > gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(Val &)	f,
		Size	size = Size(0),
		bool	resize_pol = HashTableConst::default_resize_policy,
		bool	key_uniqueness_pol = HashTableConst::default_uniqueness_policy
	)		const

Transforms a hashtable of vals into a hashtable of mountains.

Warning

Although the resulting hashtable has the same number of elements as the original hashtable, by default, the size of the former may not be equal to that of the latter. Hence iterators on the original hashtable may not parse it in the same order as iterators on the resulting hashtable. To guarrantee that both hashtables have the same size (and thus have the elements in the same order), set the size argument to the size of the original hashtable.

Parameters

f	A function that maps any Val element into a Mount.
size	The size of the resulting hashtable. When equal to 0, a default size is computed that is a good trade-off between space consumption and efficiency of new elements insertions
resize_pol	the resizing policy (automatic or manual resizing)
key_uniqueness_pol	uniqueness policy

Returns

Returns the gum::HashTable of mountains.

map() [7/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>

HashTable< Key, Mount, OtherAlloc > gum::HashTable< Key, Val, Alloc >::map	(	Mount(*)(const Val &)	f,
		Size	size = Size(0),
		bool	resize_pol = HashTableConst::default_resize_policy,
		bool	key_uniqueness_pol = HashTableConst::default_uniqueness_policy
	)		const

Transforms a hashtable of vals into a hashtable of mountains.

Warning

Although the resulting hashtable has the same number of elements as the original hashtable, by default, the size of the former may not be equal to that of the latter. Hence iterators on the original hashtable may not parse it in the same order as iterators on the resulting hashtable. To guarrantee that both hashtables have the same size (and thus have the elements in the same order), set the size argument to the size of the original hashtable.

Parameters

f	A function that maps any Val element into a Mount.
size	The size of the resulting hashtable. When equal to 0, a default size is computed that is a good trade-off between space consumption and efficiency of new elements insertions
resize_pol	the resizing policy (automatic or manual resizing)
key_uniqueness_pol	uniqueness policy

Returns

Returns the gum::HashTable of mountains.

map() [8/8]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename Mount , typename OtherAlloc = typename Alloc::template rebind< std::pair< Key, Mount > >::other>

HashTable< Key, Mount, OtherAlloc > gum::HashTable< Key, Val, Alloc >::map	(	const Mount &	val,
		Size	size = Size(0),
		bool	resize_pol = HashTableConst::default_resize_policy,
		bool	key_uniqueness_pol = HashTableConst::default_uniqueness_policy
	)		const

Creates a hashtable of mounts with a given value from a hashtable of vals.

Warning

Although the resulting hashtable has the same number of elements as the original hashtable, by default, the size of the former may not be equal to that of the latter. Hence iterators on the original hashtable may not parse it in the same order as iterators on the resulting hashtable. To guarrantee that both hashtables have the same size (and thus have the elements in the same order), set the size argument to the size of the original hashtable.

Parameters

val	The value taken by all the elements of the resulting hashtable.
size	The size of the resulting hashtable. When equal to 0, a default size is computed that is a good trade-off between space consumption and efficiency of new elements insertions
resize_pol	the resizing policy (automatic or manual resizing)
key_uniqueness_pol	uniqueness policy

Returns

Returns the gum::HashTable of mountains.

operator!=() [1/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename OtherAlloc >

INLINE bool gum::HashTable< Key, Val, Alloc >::operator!=

(

const HashTable< Key, Val, OtherAlloc > &

from

)

const

Definition at line 1162 of file hashTable_tpl.h.

                                                                                                 {
    return !operator==(from);
  }

operator!=() [2/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename OtherAlloc >

bool gum::HashTable< Key, Val, Alloc >::operator!=

(

const HashTable< Key, Val, OtherAlloc > &

from

)

const

Checks whether two hashtables contain different sets of elements.

Two hashtables are considered different if they contain different pairs (key,val). Two pairs are different if their keys have different hashed values, or if they are different in the sense of !=, or if their val's are different in the sense of !=.

Parameters

from	The gum::HashTable to test for inequality.

Returns

True if this and from are not equal.

operator=() [1/3]

template<typename Key, typename Val, typename Alloc>

HashTable< Key, Val, Alloc > & gum::HashTable< Key, Val, Alloc >::operator=

(

const HashTable< Key, Val, Alloc > &

from

)

Copy operator.

The copy operators ensures that whenever a memory allocation problem occurs, no memory leak occurs as well and it also guarantees that in this case the hashtable returned is in a coherent state (it is an empty hashtable). Note that the copy not only involves copying pairs (key,value) but also the copy of the resize and key uniqueness policies.

Parameters

from	The gum::HashTable to copy.

Returns

Returns this gum::HashTable.

Definition at line 468 of file hashTable_tpl.h.

                                                                                     {
    // avoid self assignment
    if (this != &from) {
      // for debugging purposes
      GUM_OP_CPY(HashTable);

      // first remove the current content of the hashtable and make
      // the iterators point to end
      clear();

      // if sizes of from's and this'  _nodes_ vectors are not the same,
      // we need to remove the current  _nodes_' array and to create a
      // new array with the correct size
      if (_size_ != from._size_) {
        _nodes_.resize(from._size_);

        for (Size i = Size(0); i < from._size_; ++i) {
          _nodes_[i].setAllocator(_alloc_);
        }

        _size_ = from._size_;

        // update the hash function : this is important as the computation of
        // the hash values heavily depends on the size of the hash table
        _hash_func_.resize(_size_);
      }

      _resize_policy_         = from._resize_policy_;
      _key_uniqueness_policy_ = from._key_uniqueness_policy_;
      _begin_index_           = from._begin_index_;

      // perform the copy
      _copy_(from);
    }

    return *this;
  }

operator=() [2/3]

template<typename Key, typename Val, typename Alloc >

template<typename OtherAlloc >

HashTable< Key, Val, Alloc > & gum::HashTable< Key, Val, Alloc >::operator=

(

const HashTable< Key, Val, OtherAlloc > &

from

)

Generalized copy operator.

The copy operators ensures that whenever a memory allocation problem occurs, no memory leak occurs as well and it also guarantees that in this case the hashtable returned is in a coherent state (it is an empty hashtable). Note that the copy not only involves copying pairs (key,value) but also the copy of the resize and key uniqueness policies.

Parameters

from	The gum::HashTable to copy.

Returns

Returns this gum::HashTable.

Definition at line 509 of file hashTable_tpl.h.

                                                                                          {
    // avoid self assignment
    if (this != reinterpret_cast< const HashTable< Key, Val, Alloc >* >(&from)) {
      // for debugging purposes
      GUM_OP_CPY(HashTable);

      // first remove the current content of the hashtable and make
      // the iterators point to end
      clear();

      // if sizes of from's and this'  _nodes_ vectors are not the same,
      // we need to remove the current  _nodes_' array and to create a
      // new array with the correct size
      if (_size_ != from._size_) {
        _nodes_.resize(from._size_);

        for (Size i = 0; i < from._size_; ++i) {
          _nodes_[i].setAllocator(_alloc_);
        }

        _size_ = from._size_;

        // update the hash function : this is important as the computation of
        // the hash values heavily depends on the size of the hash table
        _hash_func_.resize(_size_);
      }

      _resize_policy_         = from._resize_policy_;
      _key_uniqueness_policy_ = from._key_uniqueness_policy_;
      _begin_index_           = from._begin_index_;

      // perform the copy
      _copy_(from);
    }

    return *this;
  }

operator=() [3/3]

template<typename Key, typename Val, typename Alloc>

HashTable< Key, Val, Alloc > & gum::HashTable< Key, Val, Alloc >::operator=

(

HashTable< Key, Val, Alloc > &&

from

)

Move operator.

Parameters

from	The gum::HashTable to move.

Returns

Returns this gum::HashTable.

Definition at line 549 of file hashTable_tpl.h.

                                                                                 {
    // avoid self assignment
    if (this != &table) {
      // for debugging purposes
      GUM_OP_MOV(HashTable);

      // first remove the current content of the hashtable and make
      // the iterators point to end
      clear();

      _nodes_                 = std::move(table._nodes_);
      _safe_iterators_        = std::move(table._safe_iterators_);
      _alloc_                 = std::move(table._alloc_);
      _size_                  = table._size_;
      _nb_elements_           = table._nb_elements_;
      _hash_func_             = table._hash_func_;
      _resize_policy_         = table._resize_policy_;
      _key_uniqueness_policy_ = table._key_uniqueness_policy_;
      _begin_index_           = table._begin_index_;

      table._size_ = 0;   // necessary if we wish to perform moves iteratively,
                          // i.e. x = std::move ( y ); y = std::move ( z ); ...
    }

    return *this;
  }

operator==()

template<typename Key, typename Val, typename Alloc >

template<typename OtherAlloc >

bool gum::HashTable< Key, Val, Alloc >::operator==

(

const HashTable< Key, Val, OtherAlloc > &

from

)

const

Checks whether two hashtables contain the same elements.

Two hashtables are considered equal if they contain the identical pairs (key,val). Two pairs are identical if their keys have the same hashed value, these two keys are equal in the sense of ==, and their val's are also equal in the sense of ==.

Parameters

from	The gum::HashTable to test for equality.

Returns

True if this and from are equal.

Definition at line 1145 of file hashTable_tpl.h.

                                                                                                 {
    // checks whether the two hashtables contain the same number of elements
    if (from._nb_elements_ != _nb_elements_) return false;

    // parse this and check that each element also belongs to from
    for (auto iter = begin(); iter != end(); ++iter) {
      try {
        if (iter.val() != from[iter.key()]) return false;
      } catch (NotFound&) { return false; }
    }

    return true;
  }

operator[]() [1/2]

template<typename Key, typename Val , typename Alloc >

INLINE Val & gum::HashTable< Key, Val, Alloc >::operator[]

(

const Key &

key

)

Returns a reference on the value the key of which is passed in argument.

In case of multiple identical keys in the hash table, the first value encountered is returned. The method runs in constant time.

Parameters

key	The key of the value to return.

Returns

Returns the value matching the given key.

Exceptions

NotFound

exception is thrown if the element cannot be found.

Definition at line 695 of file hashTable_tpl.h.

                                                                     {
    return _nodes_[_hash_func_(key)][key];
  }

operator[]() [2/2]

template<typename Key, typename Val , typename Alloc >

INLINE const Val & gum::HashTable< Key, Val, Alloc >::operator[]

(

const Key &

key

)

const

returns a reference on the value the key of which is passed in argument

In case of multiple identical keys in the hash table, the first value encountered is returned. The method runs in constant time.

Exceptions

NotFound

exception is thrown if the element cannot be found.

Definition at line 700 of file hashTable_tpl.h.

                                                                                 {
    return _nodes_[_hash_func_(key)][key];
  }

reset()

template<typename Key, typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::reset

(

const Key &

key

)

Removes a property (i.e., remove an element).

Reset removes a property (i.e., a pair (key,val)) if it exists. This is an alias for erase but it is quite convenient when dealing with "dynamic property lists".

Parameters

key	The property to remove.

Definition at line 1012 of file hashTable_tpl.h.

                                                                {
    erase(key);
  }

resize()

template<typename Key , typename Val , typename Alloc >

void gum::HashTable< Key, Val, Alloc >::resize

(

Size

new_size

)

Changes the number of slots in the 'nodes' vector of the hash table.

Usually, method resize enables the user to resize manually the hashtable. When in automatic resize mode, the function will actually resize the table only if resizing policy is compatible with the new size, i.e., the new size is not so small that there would be too many elements per slot in the table (this would lead to a significant loss in performance). However, the resizing policy may be changed by using method setResizePolicy. The method runs in linear time in the size of the hashtable. Upon memory allocation problem, the fuction guarantees that no data is lost and that the hash table and its iterators are in a coherent state. In such a case, a bad_alloc exception is thrown.

Parameters

new_size

The new number of slots in the gum::HashTable.

Definition at line 740 of file hashTable_tpl.h.

                                                         {
    // new_size must be >= 2 else all the bits of the hash function are lost
    new_size = std::max(Size(2), new_size);

    // find the real size for allocation (the smallest power of 2 greater
    // than or equal to new_size) and get its base-2 logarithm
    int log_size = _hashTableLog2_(new_size);
    new_size     = Size(1) << log_size;

    // check if the new size is different from the actual size
    // if not, nothing else need be done

    if (new_size != _size_) {
      // under automatic resize policy, check if the new size leaves
      // enough space for storing all the current elements
      if (!_resize_policy_
          || (_nb_elements_ <= new_size * HashTableConst::default_mean_val_by_slot)) {
        // create a new array of  _nodes_ to store the elements
        std::vector< HashTableList< Key, Val, Alloc > > new_nodes(new_size);

        for (auto& list: new_nodes) {
          list.setAllocator(_alloc_);
        }

        // set the new hash function
        _hash_func_.resize(new_size);

        // put all the elements of the current  _nodes_ array into the new one
        Bucket* bucket;
        Size    new_hashed_key;

        for (Size i = Size(0); i < _size_; ++i) {
          while ((bucket = _nodes_[i]._deb_list_) != nullptr) {
            // compute the new hashed key
            new_hashed_key = _hash_func_(bucket->key());

            // remove the bucket from the list of buckets of the current
            // node vector
            _nodes_[i]._deb_list_ = bucket->next;

            // put the bucket into the new  _nodes_ vector
            new_nodes[new_hashed_key].insert(bucket);
          }
        }

        // update the size of the hash table
        _size_        = new_size;
        _begin_index_ = std::numeric_limits< Size >::max();

        // substitute the current  _nodes_ array by the new one
        std::swap(_nodes_, new_nodes);

        // update the iterators
        for (auto iter: _safe_iterators_) {
          if (iter->_bucket_)
            iter->_index_ = _hash_func_(iter->_bucket_->key());
          else {
            iter->_next_bucket_ = nullptr;
            iter->_index_       = 0;
          }
        }
      }
    }
  }

resizePolicy()

template<typename Key , typename Val , typename Alloc >

INLINE bool gum::HashTable< Key, Val, Alloc >::resizePolicy ( ) const

noexcept

Returns the current resizing policy.

Returns

Returns the current resizing policy.

Definition at line 725 of file hashTable_tpl.h.

                                                                        {
    return _resize_policy_;
  }

set()

template<typename Key, typename Val, typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::set	(	const Key &	key,
		const Val &	default_value
	)

Add a new property or modify it if it already existed.

When used as a "dynamic property list", it may be convenient to use this function. Function set inserts a new pair (key,val) if the key does not already exists, or it changes the value associated with key if a pair (key,val) already exists in the hash table.

Parameters

key	The key of the value to add or set.
default_value	The value to set or add.

Definition at line 944 of file hashTable_tpl.h.

                                                                                {
    Bucket* bucket = _nodes_[_hash_func_(key)].bucket(key);

    if (bucket == nullptr)
      insert(key, value);
    else
      bucket->val() = value;
  }

setKeyUniquenessPolicy()

template<typename Key , typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::setKeyUniquenessPolicy ( const bool  new_policy )

noexcept

Enables the user to change dynamically the policy for checking whether there can exist several elements in the table with identical keys.

By default, we should always check that there does not exist duplicate keys. However, this test slows the insertion of elements in the table. So, when we know for sure that no duplicate key will be entered into the table, we may avoid uniqueness checks.

Warning

When setting the key policy to "uniqueness", the function does not check whether there are already different elements with identical keys in the table. It thus only ensures that elements inserted from now on will have unique keys.

Definition at line 730 of file hashTable_tpl.h.

                                                                                                 {
    _key_uniqueness_policy_ = new_policy;
  }

setResizePolicy()

template<typename Key , typename Val , typename Alloc >

INLINE void gum::HashTable< Key, Val, Alloc >::setResizePolicy ( const bool  new_policy )

noexcept

Enables the user to change dynamically the resizing policy.

In most cases, this should be useless. However, when available memory becomes rare, avoiding automatic resizing may speed-up new insertions in the table.

Warning

This function never resizes the hashtable by itself: even if you set the new policy to be an automatic resizing and the number of elements in the table is sufficiently high that we should resize the table, function setResizePolicy won't perform this resizing. The resizing will happen only if you insert a new element or if use method resize.

Parameters

new_policy

The new resizing policy, true implies automatic resizing.

Definition at line 720 of file hashTable_tpl.h.

                                                                                          {
    _resize_policy_ = new_policy;
  }

size()

template<typename Key , typename Val , typename Alloc >

INLINE Size gum::HashTable< Key, Val, Alloc >::size ( ) const

noexcept

Returns the number of elements stored into the hashtable.

The method runs in constant time.

Returns

Returns the number of elements stored into the hashtable.

Definition at line 705 of file hashTable_tpl.h.

                                                                {
    return _nb_elements_;
  }

Friends And Related Function Documentation

HashTable

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename K , typename V , typename A >

friend class HashTable

friend

Friends to optimize the access to data, iterators must be friends.

Definition at line 1682 of file hashTable.h.

Bijection

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

template<typename T1 , typename T2 , typename A >

friend class Bijection

friend

For bijections to quickly access data.

Definition at line 1693 of file hashTable.h.

HashTableConstIterator< Key, Val >

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

friend class HashTableConstIterator< Key, Val >

friend

Friends to optimize the access to data, iterators must be friends.

Definition at line 1684 of file hashTable.h.

HashTableConstIteratorSafe< Key, Val >

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

friend class HashTableConstIteratorSafe< Key, Val >

friend

Friends to optimize the access to data, iterators must be friends.

Definition at line 1686 of file hashTable.h.

HashTableIterator< Key, Val >

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

friend class HashTableIterator< Key, Val >

friend

Friends to optimize the access to data, iterators must be friends.

Definition at line 1683 of file hashTable.h.

HashTableIteratorSafe< Key, Val >

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

friend class HashTableIteratorSafe< Key, Val >

friend

Friends to optimize the access to data, iterators must be friends.

Definition at line 1685 of file hashTable.h.

operator<< [1/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

std::ostream& operator<< ( std::ostream &  s, const HashTable< Key, Val, Alloc > &  table  )

friend

Prints the content of a gum::HashTable in the stream.

Definition at line 1201 of file hashTable_tpl.h.

                                                                                        {
    bool deja = false;
    stream << "{";

    for (Size i = Size(0); i < table._size_; ++i)
      for (auto ptr = table._nodes_[i]._deb_list_; ptr; ptr = ptr->next) {
        if (deja) stream << " , ";

        stream << ptr->key() << "=>" << ptr->val();

        deja = true;
      }

    stream << "}";

    return stream;
  }

operator<< [2/2]

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

std::ostream& operator<< ( std::ostream &  s, const HashTable< Key *, Val, Alloc > &  table  )

friend

Prints the content of a gum::HashTable with pointers key in the stream.

Definition at line 1220 of file hashTable_tpl.h.

                                                                                         {
    bool deja = false;
    stream << "{";

    for (Size i = Size(0); i < table._size_; ++i)
      for (auto ptr = table._nodes_[i]._deb_list_; ptr; ptr = ptr->next) {
        if (deja) stream << " , ";

        stream << ptr->key() << "=>" << ptr->val();

        deja = true;
      }

    stream << "}";

    return stream;
  }

Member Data Documentation

_alloc_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

BucketAllocator gum::HashTable< Key, Val, Alloc >::_alloc_

private

The allocator for the buckets.

Warning

the allocator field should compulsorily be the last of field of the class. As such, for K and V fixed, all hashTable<K,V,A> are the same (up to the allocator) for all allocators A. This feature proves useful to avoid passing the allocator as a template parameter to iterators.

Definition at line 1744 of file hashTable.h.

_begin_index_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

Size gum::HashTable< Key, Val, Alloc >::_begin_index_ {std::numeric_limits< Size >::max()}

mutableprivate

Returns where the begin index should be.

Beware: the beginning of a HashTable is the end of its nodes vector, i.e., the Bucket at the highest index in nodes. This enables a slightly faster parsing than if it were the lowest index.

Warning

std::numeric_limits<Size>::max() means that we do not know where the beginning of the table really is (this can mean either that there is not yet any element in the hash table or that an erase operation has been performed and that we lost track of the element that should correspond to the begin().

Returns

Returns where the begin index should be.

Definition at line 1731 of file hashTable.h.

_hash_func_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

HashFunc< Key > gum::HashTable< Key, Val, Alloc >::_hash_func_

private

The function used to hash keys (may change when the table is resized).

Definition at line 1709 of file hashTable.h.

_key_uniqueness_policy_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

bool gum::HashTable< Key, Val, Alloc >::_key_uniqueness_policy_ {true}

private

Shall we check for key uniqueness in the table?

Definition at line 1715 of file hashTable.h.

_nb_elements_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

Size gum::HashTable< Key, Val, Alloc >::_nb_elements_ {Size(0)}

private

Number of elements of type Val stored in the hash table.

Definition at line 1706 of file hashTable.h.

_nodes_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

std::vector< HashTableList< Key, Val, Alloc > > gum::HashTable< Key, Val, Alloc >::_nodes_

private

The hash table is represented as a vector of chained lists.

' __nodes' is this very vector.

Definition at line 1700 of file hashTable.h.

_resize_policy_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

bool gum::HashTable< Key, Val, Alloc >::_resize_policy_ {true}

private

Is resizing performed automatically?

Definition at line 1712 of file hashTable.h.

_safe_iterators_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

std::vector< HashTableConstIteratorSafe< Key, Val >* > gum::HashTable< Key, Val, Alloc >::_safe_iterators_

mutableprivate

The list of safe iterators pointing to the hash table.

Definition at line 1734 of file hashTable.h.

_size_

template<typename Key, typename Val, typename Alloc = std::allocator< std::pair< Key, Val > >>

Size gum::HashTable< Key, Val, Alloc >::_size_

private

The number of nodes in vector ' __nodes'.

Definition at line 1703 of file hashTable.h.

---

The documentation for this class was generated from the following files:

• agrum/tools/core/hashTable.h
• agrum/tools/core/hashTable_tpl.h