cds  2.3.1
cds::container::MichaelHashSet< GC, OrderedList, Traits > Class Template Reference

Michael's hash set. More...

#include <cds/container/michael_set.h>

Public Types

typedef GC gc
 Garbage collector.
 
typedef OrderedList ordered_list
 type of ordered list used as a bucket implementation
 
typedef Traits traits
 Set traits.
 
typedef ordered_list::value_type value_type
 type of value to be stored in the list
 
typedef ordered_list::key_comparator key_comparator
 key comparison functor
 
typedef ordered_list::stat stat
 Internal statistics.
 
typedef cds::opt::v::hash_selector< typename traits::hash >::type hash
 Hash functor for value_type and all its derivatives that you use.
 
typedef traits::item_counter item_counter
 Item counter type.
 
typedef traits::allocator allocator
 Bucket table allocator.
 
typedef internal_bucket_type::guarded_ptr guarded_ptr
 Guarded pointer - a result of get() and extract() functions.
 

Public Member Functions

 MichaelHashSet (size_t nMaxItemCount, size_t nLoadFactor)
 Initialize hash set. More...
 
 ~MichaelHashSet ()
 Clears hash set and destroys it.
 
template<typename Q >
bool insert (Q &&val)
 Inserts new node. More...
 
template<typename Q , typename Func >
bool insert (Q &&val, Func f)
 Inserts new node. More...
 
template<typename Q , typename Func >
std::pair< bool, bool > update (Q &&val, Func func, bool bAllowUpdate=true)
 Updates the element. More...
 
template<typename Q >
std::pair< bool, bool > upsert (Q &&val, bool bAllowInsert=true)
 Inserts or updates the node (only for IterableList) More...
 
template<typename... Args>
bool emplace (Args &&... args)
 Inserts data of type value_type constructed from args. More...
 
template<typename Q >
bool erase (Q const &key)
 Deletes key from the set. More...
 
template<typename Q , typename Less >
bool erase_with (Q const &key, Less pred)
 Deletes the item from the set using pred predicate for searching. More...
 
template<typename Q , typename Func >
bool erase (Q const &key, Func f)
 Deletes key from the set. More...
 
template<typename Q , typename Less , typename Func >
bool erase_with (Q const &key, Less pred, Func f)
 Deletes the item from the set using pred predicate for searching. More...
 
bool erase_at (iterator const &iter)
 Deletes the item pointed by iterator iter (only for IterableList based set) More...
 
template<typename Q >
guarded_ptr extract (Q const &key)
 Extracts the item with specified key. More...
 
template<typename Q , typename Less >
guarded_ptr extract_with (Q const &key, Less pred)
 Extracts the item using compare functor pred. More...
 
template<typename Q , typename Func >
bool find (Q &key, Func f)
 Finds the key key. More...
 
template<typename Q >
iterator find (Q &key)
 Finds key and returns iterator pointed to the item found (only for IterableList) More...
 
template<typename Q , typename Less , typename Func >
bool find_with (Q &key, Less pred, Func f)
 Finds the key key using pred predicate for searching. More...
 
template<typename Q , typename Less >
iterator find_with (Q &key, Less pred)
 Finds key using pred predicate and returns iterator pointed to the item found (only for IterableList) More...
 
template<typename Q >
bool contains (Q const &key)
 Checks whether the set contains key. More...
 
template<typename Q , typename Less >
bool contains (Q const &key, Less pred)
 Checks whether the set contains key using pred predicate for searching. More...
 
template<typename Q >
guarded_ptr get (Q const &key)
 Finds the key key and return the item found. More...
 
template<typename Q , typename Less >
guarded_ptr get_with (Q const &key, Less pred)
 Finds the key key and return the item found. More...
 
void clear ()
 Clears the set (non-atomic) More...
 
bool empty () const
 Checks if the set is empty. More...
 
size_t size () const
 Returns item count in the set. More...
 
stat const & statistics () const
 Returns const reference to internal statistics.
 
size_t bucket_count () const
 Returns the size of hash table. More...
 

Static Public Attributes

static constexpr const size_t c_nHazardPtrCount = ordered_list::c_nHazardPtrCount
 Count of hazard pointer required.
 

Forward iterators

typedef michael_set::details::iterator< internal_bucket_type, false > iterator
 Forward iterator. More...
 
typedef michael_set::details::iterator< internal_bucket_type, true > const_iterator
 Const forward iterator.
 
iterator begin ()
 Returns a forward iterator addressing the first element in a set. More...
 
iterator end ()
 Returns an iterator that addresses the location succeeding the last element in a set. More...
 
const_iterator begin () const
 Returns a forward const iterator addressing the first element in a set.
 
const_iterator cbegin () const
 Returns a forward const iterator addressing the first element in a set.
 
const_iterator end () const
 Returns an const iterator that addresses the location succeeding the last element in a set.
 
const_iterator cend () const
 Returns an const iterator that addresses the location succeeding the last element in a set.
 

Detailed Description

template<class GC, class OrderedList, class Traits = michael_set::traits>
class cds::container::MichaelHashSet< GC, OrderedList, Traits >

Michael's hash set.

Source:

  • [2002] Maged Michael "High performance dynamic lock-free hash tables and list-based sets"

Michael's hash table algorithm is based on lock-free ordered list and it is very simple. The main structure is an array T of size M. Each element in T is basically a pointer to a hash bucket, implemented as a singly linked list. The array of buckets cannot be dynamically expanded. However, each bucket may contain unbounded number of items.

Template parameters are:

  • GC - Garbage collector used. You may use any Garbage collector from the libcds library. Note the GC must be the same as the GC used for OrderedList
  • OrderedList - ordered list implementation used as bucket for hash set, possible implementations: MichaelList, LazyList, IterableList. The ordered list implementation specifies the type T to be stored in the hash-set, the comparing functor for the type T and other features specific for the ordered list.
  • Traits - set traits, default is michael_set::traits. Instead of defining Traits struct you may use option-based syntax with michael_set::make_traits metafunction.

There are the specializations:

Hash functor

Some member functions of Michael's hash set accept the key parameter of type Q which differs from node type value_type. It is expected that type Q contains full key of node type value_type, and if keys of type Q and value_type are equal the hash values of these keys must be equal too.

The hash functor Traits::hash should accept parameters of both type:

// Our node type
struct Foo {
std::string key_; // key field
// ... other fields
};
// Hash functor
struct fooHash {
size_t operator()( const std::string& s ) const
{
return std::hash( s );
}
size_t operator()( const Foo& f ) const
{
return (*this)( f.key_ );
}
};

How to use

Suppose, we have the following type Foo that we want to store in our MichaelHashSet:

struct Foo {
int nKey; // key field
int nVal; // value field
};

To use MichaelHashSet for Foo values, you should first choose suitable ordered list class that will be used as a bucket for the set. We will use gc::DHP reclamation schema and MichaelList as a bucket type. Also, for ordered list we should develop a comparator for our Foo struct.

#include <cds/container/michael_list_dhp.h>
#include <cds/container/michael_set.h>
namespace cc = cds::container;
// Foo comparator
struct Foo_cmp {
int operator ()(Foo const& v1, Foo const& v2 ) const
{
if ( std::less( v1.nKey, v2.nKey ))
return -1;
return std::less(v2.nKey, v1.nKey) ? 1 : 0;
}
};
// Our ordered list
typedef cc::MichaelList< cds::gc::DHP, Foo,
typename cc::michael_list::make_traits<
cc::opt::compare< Foo_cmp > // item comparator option
>::type
> bucket_list;
// Hash functor for Foo
struct foo_hash {
size_t operator ()( int i ) const
{
return std::hash( i );
}
size_t operator()( Foo const& i ) const
{
return std::hash( i.nKey );
}
};
// Declare set type.
// Note that \p GC template parameter of ordered list must be equal \p GC for the set.
typedef cc::MichaelHashSet< cds::gc::DHP, bucket_list,
cc::michael_set::make_traits<
cc::opt::hash< foo_hash >
>::type
> foo_set;
// Set variable
foo_set fooSet;

Member Typedef Documentation

◆ iterator

template<class GC , class OrderedList , class Traits = michael_set::traits>
typedef michael_set::details::iterator< internal_bucket_type, false > cds::container::MichaelHashSet< GC, OrderedList, Traits >::iterator

Forward iterator.

The forward iterator for Michael's set has some features:

  • it has no post-increment operator
  • to protect the value, the iterator contains a GC-specific guard + another guard is required locally for increment operator. For some GC (like as gc::HP), a guard is a limited resource per thread, so an exception (or assertion) "no free guard" may be thrown if the limit of guard count per thread is exceeded.
  • The iterator cannot be moved across thread boundary because it contains thread-private GC's guard.

Iterator thread safety depends on type of OrderedList:

  • for MichaelList and LazyList: iterator guarantees safety even if you delete the item that iterator points to because that item is guarded by hazard pointer. However, in case of concurrent deleting operations it is no guarantee that you iterate all item in the set. Moreover, a crash is possible when you try to iterate the next element that has been deleted by concurrent thread. Use this iterator on the concurrent container for debugging purpose only.
  • for IterableList: iterator is thread-safe. You may use it freely in concurrent environment.

The iterator interface:

class iterator {
public:
// Default constructor
// Copy construtor
iterator( iterator const& src );
// Dereference operator
value_type * operator ->() const;
// Dereference operator
value_type& operator *() const;
// Preincrement operator
iterator& operator ++();
// Assignment operator
iterator& operator = (iterator const& src);
// Equality operators
bool operator ==(iterator const& i ) const;
bool operator !=(iterator const& i ) const;
};

Forward iterator

Constructor & Destructor Documentation

◆ MichaelHashSet()

template<class GC , class OrderedList , class Traits = michael_set::traits>
cds::container::MichaelHashSet< GC, OrderedList, Traits >::MichaelHashSet ( size_t  nMaxItemCount,
size_t  nLoadFactor 
)
inline

Initialize hash set.

The Michael's hash set is non-expandable container. You should point the average count of items nMaxItemCount when you create an object. nLoadFactor parameter defines average count of items per bucket and it should be small number between 1 and 10. Remember, since the bucket implementation is an ordered list, searching in the bucket is linear [O(nLoadFactor)].

The ctor defines hash table size as rounding nMaxItemCount / nLoadFactor up to nearest power of two.

Parameters
nMaxItemCountestimation of max item count in the hash set
nLoadFactorload factor: estimation of max number of items in the bucket

Member Function Documentation

◆ begin()

template<class GC , class OrderedList , class Traits = michael_set::traits>
iterator cds::container::MichaelHashSet< GC, OrderedList, Traits >::begin ( )
inline

Returns a forward iterator addressing the first element in a set.

For empty set

begin() == end()

◆ bucket_count()

template<class GC , class OrderedList , class Traits = michael_set::traits>
size_t cds::container::MichaelHashSet< GC, OrderedList, Traits >::bucket_count ( ) const
inline

Returns the size of hash table.

Since MichaelHashSet cannot dynamically extend the hash table size, the value returned is an constant depending on object initialization parameters; see MichaelHashSet::MichaelHashSet for explanation.

◆ clear()

template<class GC , class OrderedList , class Traits = michael_set::traits>
void cds::container::MichaelHashSet< GC, OrderedList, Traits >::clear ( )
inline

Clears the set (non-atomic)

The function erases all items from the set.

The function is not atomic. It cleans up each bucket and then resets the item counter to zero. If there are a thread that performs insertion while clear is working the result is undefined in general case: empty() may return true but the set may contain item(s). Therefore, clear may be used only for debugging purposes.

◆ contains() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::contains ( Q const &  key)
inline

Checks whether the set contains key.

The function searches the item with key equal to key and returns true if the key is found, and false otherwise.

Note the hash functor specified for class Traits template parameter should accept a parameter of type Q that can be not the same as value_type.

◆ contains() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::contains ( Q const &  key,
Less  pred 
)
inline

Checks whether the set contains key using pred predicate for searching.

The function is an analog of contains( key ) but pred is used for key comparing. Less functor has the interface like std::less. Less must imply the same element order as the comparator used for building the set.

◆ emplace()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename... Args>
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::emplace ( Args &&...  args)
inline

Inserts data of type value_type constructed from args.

Returns true if inserting successful, false otherwise.

◆ empty()

template<class GC , class OrderedList , class Traits = michael_set::traits>
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::empty ( ) const
inline

Checks if the set is empty.

Warning
If you use atomicity::empty_item_counter in traits::item_counter, the function always returns true.

◆ end()

template<class GC , class OrderedList , class Traits = michael_set::traits>
iterator cds::container::MichaelHashSet< GC, OrderedList, Traits >::end ( )
inline

Returns an iterator that addresses the location succeeding the last element in a set.

Do not use the value returned by end function to access any item. The returned value can be used only to control reaching the end of the set. For empty set

begin() == end()

◆ erase() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::erase ( Q const &  key)
inline

Deletes key from the set.

Since the key of MichaelHashSet's item type value_type is not explicitly specified, template parameter Q defines the key type searching in the list. The set item comparator should be able to compare the type value_type and the type Q.

Return true if key is found and deleted, false otherwise.

◆ erase() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Func >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::erase ( Q const &  key,
Func  f 
)
inline

Deletes key from the set.

The function searches an item with key key, calls f functor and deletes the item. If key is not found, the functor is not called.

The functor Func interface:

struct extractor {
void operator()(value_type& item);
};

where item - the item found.

Since the key of MichaelHashSet's value_type is not explicitly specified, template parameter Q defines the key type searching in the list. The list item comparator should be able to compare the type T of list item and the type Q.

Return true if key is found and deleted, false otherwise

◆ erase_at()

template<class GC , class OrderedList , class Traits = michael_set::traits>
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::erase_at ( iterator const &  iter)
inline

Deletes the item pointed by iterator iter (only for IterableList based set)

Returns true if the operation is successful, false otherwise. The function can return false if the node the iterator points to has already been deleted by other thread.

The function does not invalidate the iterator, it remains valid and can be used for further traversing.

Note
erase_at() is supported only for MichaelHashSet based on IterableList.

◆ erase_with() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::erase_with ( Q const &  key,
Less  pred 
)
inline

Deletes the item from the set using pred predicate for searching.

The function is an analog of erase(Q const&) but pred is used for key comparing. Less functor has the interface like std::less. Less must imply the same element order as the comparator used for building the set.

◆ erase_with() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less , typename Func >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::erase_with ( Q const &  key,
Less  pred,
Func  f 
)
inline

Deletes the item from the set using pred predicate for searching.

The function is an analog of erase(Q const&, Func) but pred is used for key comparing. Less functor has the interface like std::less. Less must imply the same element order as the comparator used for building the set.

◆ extract()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
guarded_ptr cds::container::MichaelHashSet< GC, OrderedList, Traits >::extract ( Q const &  key)
inline

Extracts the item with specified key.

The function searches an item with key equal to key, unlinks it from the set, and returns it as guarded_ptr. If key is not found the function returns an empty guadd pointer.

Note the compare functor should accept a parameter of type Q that may be not the same as value_type.

The extracted item is freed automatically when returned guarded_ptr object will be destroyed or released.

Note
Each guarded_ptr object uses the GC's guard that can be limited resource.

Usage:

michael_set theSet;
// ...
{
typename michael_set::guarded_ptr gp( theSet.extract( 5 ));
if ( gp ) {
// Deal with gp
// ...
}
// Destructor of gp releases internal HP guard
}

◆ extract_with()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less >
guarded_ptr cds::container::MichaelHashSet< GC, OrderedList, Traits >::extract_with ( Q const &  key,
Less  pred 
)
inline

Extracts the item using compare functor pred.

The function is an analog of extract(Q const&) but pred predicate is used for key comparing.

Less functor has the semantics like std::less but should take arguments of type value_type and Q in any order. pred must imply the same element order as the comparator used for building the set.

◆ find() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Func >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::find ( Q &  key,
Func  f 
)
inline

Finds the key key.

The function searches the item with key equal to key and calls the functor f for item found. The interface of Func functor is:

struct functor {
void operator()( value_type& item, Q& key );
};

where item is the item found, key is the find function argument.

The functor may change non-key fields of item. Note that the functor is only guarantee that item cannot be disposed during functor is executing. The functor does not serialize simultaneous access to the set's item. If such access is possible you must provide your own synchronization schema on item level to exclude unsafe item modifications.

The key argument is non-const since it can be used as f functor destination i.e., the functor can modify both arguments.

Note the hash functor specified for class Traits template parameter should accept a parameter of type Q that may be not the same as value_type.

The function returns true if key is found, false otherwise.

◆ find() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
iterator cds::container::MichaelHashSet< GC, OrderedList, Traits >::find ( Q &  key)
inline

Finds key and returns iterator pointed to the item found (only for IterableList)

If key is not found the function returns end().

Note
This function is supported only for the set based on IterableList

◆ find_with() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less , typename Func >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::find_with ( Q &  key,
Less  pred,
Func  f 
)
inline

Finds the key key using pred predicate for searching.

The function is an analog of find(Q&, Func) but pred is used for key comparing. Less functor has the interface like std::less. Less must imply the same element order as the comparator used for building the set.

◆ find_with() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less >
iterator cds::container::MichaelHashSet< GC, OrderedList, Traits >::find_with ( Q &  key,
Less  pred 
)
inline

Finds key using pred predicate and returns iterator pointed to the item found (only for IterableList)

The function is an analog of find(Q&) but pred is used for key comparing. Less functor has the interface like std::less. pred must imply the same element order as the comparator used for building the set.

If key is not found the function returns end().

Note
This function is supported only for the set based on IterableList

◆ get()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
guarded_ptr cds::container::MichaelHashSet< GC, OrderedList, Traits >::get ( Q const &  key)
inline

Finds the key key and return the item found.

The function searches the item with key equal to key and returns the guarded pointer to the item found. If key is not found the functin returns an empty guarded pointer.

Note
Each guarded_ptr object uses one GC's guard which can be limited resource.

Usage:

typedef cds::container::MichaeHashSet< your_template_params > michael_set;
michael_set theSet;
// ...
{
typename michael_set::guarded_ptr gp( theSet.get( 5 ));
if ( gp ) {
// Deal with gp
//...
}
// Destructor of guarded_ptr releases internal HP guard
}

Note the compare functor specified for OrderedList template parameter should accept a parameter of type Q that can be not the same as value_type.

◆ get_with()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Less >
guarded_ptr cds::container::MichaelHashSet< GC, OrderedList, Traits >::get_with ( Q const &  key,
Less  pred 
)
inline

Finds the key key and return the item found.

The function is an analog of get( Q const&) but pred is used for comparing the keys.

Less functor has the semantics like std::less but should take arguments of type value_type and Q in any order. pred must imply the same element order as the comparator used for building the set.

◆ insert() [1/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::insert ( Q &&  val)
inline

Inserts new node.

The function creates a node with copy of val value and then inserts the node created into the set.

The type Q should contain as minimum the complete key for the node. The object of value_type should be constructible from a value of type Q. In trivial case, Q is equal to value_type.

Returns true if val is inserted into the set, false otherwise.

◆ insert() [2/2]

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Func >
bool cds::container::MichaelHashSet< GC, OrderedList, Traits >::insert ( Q &&  val,
Func  f 
)
inline

Inserts new node.

The function allows to split creating of new item into two part:

  • create item with key only
  • insert new item into the set
  • if inserting is success, calls f functor to initialize value-fields of val.

The functor signature is:

void func( value_type& val );

where val is the item inserted. The user-defined functor is called only if the inserting is success.

Warning
For MichaelList and IterableList as the bucket see insert item troubleshooting. LazyList provides exclusive access to inserted item and does not require any node-level synchronization.

◆ size()

template<class GC , class OrderedList , class Traits = michael_set::traits>
size_t cds::container::MichaelHashSet< GC, OrderedList, Traits >::size ( ) const
inline

Returns item count in the set.

Warning
If you use atomicity::empty_item_counter in traits::item_counter, the function always returns 0.

◆ update()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q , typename Func >
std::pair<bool, bool> cds::container::MichaelHashSet< GC, OrderedList, Traits >::update ( Q &&  val,
Func  func,
bool  bAllowUpdate = true 
)
inline

Updates the element.

The operation performs inserting or changing data with lock-free manner.

If the item val not found in the set, then val is inserted iff bAllowInsert is true. Otherwise, the functor func is called with item found.

The functor func signature depends of OrderedList:

for MichaelList, LazyList

struct functor {
void operator()( bool bNew, value_type& item, Q const& val );
};

with arguments:

  • bNew - true if the item has been inserted, false otherwise
  • item - item of the set
  • val - argument val passed into the update() function

The functor may change non-key fields of the item.

for IterableList

void func( value_type& val, value_type * old );

where

  • val - a new data constructed from key
  • old - old value that will be retired. If new item has been inserted then old is nullptr.
Returns
std::pair<bool, bool> where first is true if operation is successful, second is true if new item has been added or false if the item with key already is in the set.
Warning
For MichaelList and IterableList as the bucket see insert item troubleshooting. LazyList provides exclusive access to inserted item and does not require any node-level synchronization.

◆ upsert()

template<class GC , class OrderedList , class Traits = michael_set::traits>
template<typename Q >
std::pair<bool, bool> cds::container::MichaelHashSet< GC, OrderedList, Traits >::upsert ( Q &&  val,
bool  bAllowInsert = true 
)
inline

Inserts or updates the node (only for IterableList)

The operation performs inserting or changing data with lock-free manner.

If the item val is not found in the set, then val is inserted iff bAllowInsert is true. Otherwise, the current element is changed to val, the old element will be retired later.

Returns std::pair<bool, bool> where first is true if operation is successful, second is true if val has been added or false if the item with that key already in the set.


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

cds 2.3.1 Developed by Maxim Khizhinsky aka khizmax 2007 - 2017
Autogenerated Fri Sep 1 2017 08:47:03 by Doxygen 1.8.13