Refactor Dictionary and HashSet to use shared base class

Add const iterators
This commit is contained in:
Wojtek Figat
2025-01-23 23:59:31 +01:00
parent 61a731704d
commit 38f74c8cf7
5 changed files with 1024 additions and 1076 deletions

View File

@@ -1,15 +0,0 @@
// Copyright (c) 2012-2024 Wojciech Figat. All rights reserved.
#pragma once
#include "Engine/Core/Types/BaseTypes.h"
/// <summary>
/// Tells if the object is occupied, and if not, if the bucket is a subject of compaction.
/// </summary>
enum class BucketState : byte
{
Empty = 0,
Deleted = 1,
Occupied = 2,
};

File diff suppressed because it is too large Load Diff

View File

@@ -2,12 +2,122 @@
#pragma once
#include "Engine/Core/Memory/Memory.h"
#include "Engine/Core/Memory/Allocation.h"
#include "Engine/Core/Memory/AllocationUtils.h"
#include "Engine/Core/Collections/BucketState.h"
#include "Engine/Core/Collections/HashFunctions.h"
#include "Engine/Core/Collections/Config.h"
#include "HashSetBase.h"
/// <summary>
/// Describes single portion of space for the item in a hash set.
/// </summary>
template<typename T, typename AllocationType>
struct HashSetBucket
{
friend Memory;
friend HashSetBase<AllocationType, HashSetBucket>;
friend HashSet<T, AllocationType>;
/// <summary>The item.</summary>
T Item;
private:
HashSetBucketState _state;
HashSetBucket()
: _state(HashSetBucketState::Empty)
{
}
HashSetBucket(HashSetBucket&& other) noexcept
{
_state = other._state;
if (other._state == HashSetBucketState::Occupied)
{
Memory::MoveItems(&Item, &other.Item, 1);
other._state = HashSetBucketState::Empty;
}
}
HashSetBucket& operator=(HashSetBucket&& other) noexcept
{
if (this != &other)
{
if (_state == HashSetBucketState::Occupied)
{
Memory::DestructItem(&Item);
}
_state = other._state;
if (other._state == HashSetBucketState::Occupied)
{
Memory::MoveItems(&Item, &other.Item, 1);
other._state = HashSetBucketState::Empty;
}
}
return *this;
}
/// <summary>Copying a bucket is useless, because a key must be unique in the dictionary.</summary>
HashSetBucket(const HashSetBucket&) = delete;
/// <summary>Copying a bucket is useless, because a key must be unique in the dictionary.</summary>
HashSetBucket& operator=(const HashSetBucket&) = delete;
~HashSetBucket()
{
if (_state == HashSetBucketState::Occupied)
Memory::DestructItem(&Item);
}
FORCE_INLINE void Free()
{
if (_state == HashSetBucketState::Occupied)
Memory::DestructItem(&Item);
_state = HashSetBucketState::Empty;
}
FORCE_INLINE void Delete()
{
ASSERT(IsOccupied());
_state = HashSetBucketState::Deleted;
Memory::DestructItem(&Item);
}
template<typename ItemType>
FORCE_INLINE void Occupy(const ItemType& item)
{
Memory::ConstructItems(&Item, &item, 1);
_state = HashSetBucketState::Occupied;
}
template<typename ItemType>
FORCE_INLINE void Occupy(ItemType&& item)
{
Memory::MoveItems(&Item, &item, 1);
_state = HashSetBucketState::Occupied;
}
FORCE_INLINE bool IsEmpty() const
{
return _state == HashSetBucketState::Empty;
}
FORCE_INLINE bool IsDeleted() const
{
return _state == HashSetBucketState::Deleted;
}
FORCE_INLINE bool IsOccupied() const
{
return _state == HashSetBucketState::Occupied;
}
FORCE_INLINE bool IsNotOccupied() const
{
return _state != HashSetBucketState::Occupied;
}
FORCE_INLINE const T& GetKey() const
{
return Item;
}
};
/// <summary>
/// Template for unordered set of values (without duplicates with O(1) lookup access).
@@ -15,124 +125,12 @@
/// <typeparam name="T">The type of elements in the set.</typeparam>
/// <typeparam name="AllocationType">The type of memory allocator.</typeparam>
template<typename T, typename AllocationType = HeapAllocation>
API_CLASS(InBuild) class HashSet
API_CLASS(InBuild) class HashSet : public HashSetBase<AllocationType, HashSetBucket<T, AllocationType>>
{
friend HashSet;
public:
/// <summary>
/// Describes single portion of space for the item in a hash map.
/// </summary>
struct Bucket
{
friend HashSet;
friend Memory;
/// <summary>The item.</summary>
T Item;
private:
BucketState _state;
Bucket()
: _state(BucketState::Empty)
{
}
Bucket(Bucket&& other) noexcept
{
_state = other._state;
if (other._state == BucketState::Occupied)
{
Memory::MoveItems(&Item, &other.Item, 1);
other._state = BucketState::Empty;
}
}
Bucket& operator=(Bucket&& other) noexcept
{
if (this != &other)
{
if (_state == BucketState::Occupied)
{
Memory::DestructItem(&Item);
}
_state = other._state;
if (other._state == BucketState::Occupied)
{
Memory::MoveItems(&Item, &other.Item, 1);
other._state = BucketState::Empty;
}
}
return *this;
}
/// <summary>Copying a bucket is useless, because a key must be unique in the dictionary.</summary>
Bucket(const Bucket&) = delete;
/// <summary>Copying a bucket is useless, because a key must be unique in the dictionary.</summary>
Bucket& operator=(const Bucket&) = delete;
~Bucket()
{
if (_state == BucketState::Occupied)
Memory::DestructItem(&Item);
}
FORCE_INLINE void Free()
{
if (_state == BucketState::Occupied)
Memory::DestructItem(&Item);
_state = BucketState::Empty;
}
FORCE_INLINE void Delete()
{
_state = BucketState::Deleted;
Memory::DestructItem(&Item);
}
template<typename ItemType>
FORCE_INLINE void Occupy(const ItemType& item)
{
Memory::ConstructItems(&Item, &item, 1);
_state = BucketState::Occupied;
}
template<typename ItemType>
FORCE_INLINE void Occupy(ItemType&& item)
{
Memory::MoveItems(&Item, &item, 1);
_state = BucketState::Occupied;
}
FORCE_INLINE bool IsEmpty() const
{
return _state == BucketState::Empty;
}
FORCE_INLINE bool IsDeleted() const
{
return _state == BucketState::Deleted;
}
FORCE_INLINE bool IsOccupied() const
{
return _state == BucketState::Occupied;
}
FORCE_INLINE bool IsNotOccupied() const
{
return _state != BucketState::Occupied;
}
};
using AllocationData = typename AllocationType::template Data<Bucket>;
private:
int32 _elementsCount = 0;
int32 _deletedCount = 0;
int32 _size = 0;
AllocationData _allocation;
typedef HashSetBucket<T, AllocationType> Bucket;
typedef HashSetBase<AllocationType, Bucket> Base;
public:
/// <summary>
@@ -148,7 +146,7 @@ public:
/// <param name="capacity">The number of elements that can be added without a need to allocate more memory.</param>
FORCE_INLINE explicit HashSet(const int32 capacity)
{
SetCapacity(capacity);
Base::SetCapacity(capacity);
}
/// <summary>
@@ -157,13 +155,7 @@ public:
/// <param name="other">The other collection to move.</param>
HashSet(HashSet&& other) noexcept
{
_elementsCount = other._elementsCount;
_deletedCount = other._deletedCount;
_size = other._size;
other._elementsCount = 0;
other._deletedCount = 0;
other._size = 0;
AllocationUtils::MoveToEmpty<Bucket, AllocationType>(_allocation, other._allocation, _size, _size);
Base::MoveToEmpty(MoveTemp(other));
}
/// <summary>
@@ -196,15 +188,9 @@ public:
{
if (this != &other)
{
Clear();
_allocation.Free();
_elementsCount = other._elementsCount;
_deletedCount = other._deletedCount;
_size = other._size;
other._elementsCount = 0;
other._deletedCount = 0;
other._size = 0;
AllocationUtils::MoveToEmpty<Bucket, AllocationType>(_allocation, other._allocation, _size, _size);
Base::Clear();
Base::_allocation.Free();
Base::MoveToEmpty(MoveTemp(other));
}
return *this;
}
@@ -214,113 +200,129 @@ public:
/// </summary>
~HashSet()
{
Clear();
}
public:
/// <summary>
/// Gets the amount of the elements in the collection.
/// The read-only hash set collection iterator.
/// </summary>
FORCE_INLINE int32 Count() const
{
return _elementsCount;
}
/// <summary>
/// Gets the amount of the elements that can be contained by the collection.
/// </summary>
FORCE_INLINE int32 Capacity() const
{
return _size;
}
/// <summary>
/// Returns true if collection is empty.
/// </summary>
FORCE_INLINE bool IsEmpty() const
{
return _elementsCount == 0;
}
/// <summary>
/// Returns true if collection has one or more elements.
/// </summary>
FORCE_INLINE bool HasItems() const
{
return _elementsCount != 0;
}
public:
/// <summary>
/// The hash set collection iterator.
/// </summary>
struct Iterator
struct ConstIterator : Base::IteratorBase
{
friend HashSet;
private:
HashSet* _collection;
int32 _index;
public:
Iterator(HashSet* collection, const int32 index)
: _collection(collection)
, _index(index)
ConstIterator(const HashSet* collection, const int32 index)
: Base::IteratorBase(collection, index)
{
}
Iterator(const HashSet* collection, const int32 index)
: _collection(const_cast<HashSet*>(collection))
, _index(index)
ConstIterator()
: Base::IteratorBase(nullptr, -1)
{
}
ConstIterator(const ConstIterator& i)
: Base::IteratorBase(i._collection, i._index)
{
}
ConstIterator(ConstIterator&& i) noexcept
: Base::IteratorBase(i._collection, i._index)
{
}
public:
FORCE_INLINE bool operator!() const
{
return !(bool)*this;
}
FORCE_INLINE bool operator==(const ConstIterator& v) const
{
return this->_index == v._index && this->_collection == v._collection;
}
FORCE_INLINE bool operator!=(const ConstIterator& v) const
{
return this->_index != v._index || this->_collection != v._collection;
}
ConstIterator& operator=(const ConstIterator& v)
{
this->_collection = v._collection;
this->_index = v._index;
return *this;
}
ConstIterator& operator=(ConstIterator&& v) noexcept
{
this->_collection = v._collection;
this->_index = v._index;
return *this;
}
ConstIterator& operator++()
{
this->Next();
return *this;
}
ConstIterator operator++(int) const
{
ConstIterator i = *this;
i.Next();
return i;
}
ConstIterator& operator--()
{
this->Prev();
return *this;
}
ConstIterator operator--(int) const
{
ConstIterator i = *this;
i.Prev();
return i;
}
};
/// <summary>
/// The hash set collection iterator.
/// </summary>
struct Iterator : Base::IteratorBase
{
friend HashSet;
public:
Iterator(HashSet* collection, const int32 index)
: Base::IteratorBase(collection, index)
{
}
Iterator()
: _collection(nullptr)
, _index(-1)
: Base::IteratorBase(nullptr, -1)
{
}
Iterator(const Iterator& i)
: _collection(i._collection)
, _index(i._index)
: Base::IteratorBase(i._collection, i._index)
{
}
Iterator(Iterator&& i) noexcept
: _collection(i._collection)
, _index(i._index)
: Base::IteratorBase(i._collection, i._index)
{
}
public:
FORCE_INLINE int32 Index() const
{
return _index;
}
FORCE_INLINE bool IsEnd() const
{
return _index == _collection->_size;
}
FORCE_INLINE bool IsNotEnd() const
{
return _index != _collection->_size;
}
FORCE_INLINE Bucket& operator*() const
{
return _collection->_allocation.Get()[_index];
return ((HashSet*)this->_collection)->_allocation.Get()[this->_index];
}
FORCE_INLINE Bucket* operator->() const
{
return &_collection->_allocation.Get()[_index];
}
FORCE_INLINE explicit operator bool() const
{
return _index >= 0 && _index < _collection->_size;
return &((HashSet*)this->_collection)->_allocation.Get()[this->_index];
}
FORCE_INLINE bool operator!() const
@@ -330,87 +332,56 @@ public:
FORCE_INLINE bool operator==(const Iterator& v) const
{
return _index == v._index && _collection == v._collection;
return this->_index == v._index && this->_collection == v._collection;
}
FORCE_INLINE bool operator!=(const Iterator& v) const
{
return _index != v._index || _collection != v._collection;
return this->_index != v._index || this->_collection != v._collection;
}
Iterator& operator=(const Iterator& v)
{
_collection = v._collection;
_index = v._index;
this->_collection = v._collection;
this->_index = v._index;
return *this;
}
Iterator& operator=(Iterator&& v) noexcept
{
_collection = v._collection;
_index = v._index;
this->_collection = v._collection;
this->_index = v._index;
return *this;
}
Iterator& operator++()
{
const int32 capacity = _collection->_size;
if (_index != capacity)
{
const Bucket* data = _collection->_allocation.Get();
do
{
++_index;
}
while (_index != capacity && data[_index].IsNotOccupied());
}
this->Next();
return *this;
}
Iterator operator++(int)
Iterator operator++(int) const
{
Iterator i = *this;
++i;
i.Next();
return i;
}
Iterator& operator--()
{
if (_index > 0)
{
const Bucket* data = _collection->_allocation.Get();
do
{
--_index;
}
while (_index > 0 && data[_index].IsNotOccupied());
}
this->Prev();
return *this;
}
Iterator operator--(int)
{
Iterator i = *this;
--i;
i.Prev();
return i;
}
};
public:
/// <summary>
/// Removes all elements from the collection.
/// </summary>
void Clear()
{
if (_elementsCount + _deletedCount != 0)
{
Bucket* data = _allocation.Get();
for (int32 i = 0; i < _size; i++)
data[i].Free();
_elementsCount = _deletedCount = 0;
}
}
/// <summary>
/// Clears the collection and delete value objects.
/// Note: collection must contain pointers to the objects that have public destructor and be allocated using New method.
@@ -425,93 +396,7 @@ public:
if (i->Item)
::Delete(i->Item);
}
Clear();
}
/// <summary>
/// Changes capacity of the collection
/// </summary>
/// <param name="capacity">New capacity</param>
/// <param name="preserveContents">Enable/disable preserving collection contents during resizing</param>
void SetCapacity(int32 capacity, const bool preserveContents = true)
{
if (capacity == _size)
return;
ASSERT(capacity >= 0);
AllocationData oldAllocation;
AllocationUtils::MoveToEmpty<Bucket, AllocationType>(oldAllocation, _allocation, _size, _size);
const int32 oldSize = _size;
const int32 oldElementsCount = _elementsCount;
_deletedCount = _elementsCount = 0;
if (capacity != 0 && (capacity & (capacity - 1)) != 0)
capacity = AllocationUtils::AlignToPowerOf2(capacity);
if (capacity)
{
_allocation.Allocate(capacity);
Bucket* data = _allocation.Get();
for (int32 i = 0; i < capacity; i++)
data[i]._state = BucketState::Empty;
}
_size = capacity;
Bucket* oldData = oldAllocation.Get();
if (oldElementsCount != 0 && capacity != 0 && preserveContents)
{
FindPositionResult pos;
for (int32 i = 0; i < oldSize; i++)
{
Bucket& oldBucket = oldData[i];
if (oldBucket.IsOccupied())
{
FindPosition(oldBucket.Item, pos);
ASSERT(pos.FreeSlotIndex != -1);
Bucket& bucket = _allocation.Get()[pos.FreeSlotIndex];
bucket = MoveTemp(oldBucket);
_elementsCount++;
}
}
}
if (oldElementsCount != 0)
{
for (int32 i = 0; i < oldSize; i++)
oldData[i].Free();
}
}
/// <summary>
/// Ensures that collection has given capacity.
/// </summary>
/// <param name="minCapacity">The minimum required capacity.</param>
/// <param name="preserveContents">True if preserve collection data when changing its size, otherwise collection after resize will be empty.</param>
void EnsureCapacity(int32 minCapacity, const bool preserveContents = true)
{
minCapacity *= DICTIONARY_DEFAULT_SLACK_SCALE;
if (_size >= minCapacity)
return;
int32 capacity = _allocation.CalculateCapacityGrow(_size, minCapacity);
if (capacity < DICTIONARY_DEFAULT_CAPACITY)
capacity = DICTIONARY_DEFAULT_CAPACITY;
SetCapacity(capacity, preserveContents);
}
/// <summary>
/// Swaps the contents of collection with the other object without copy operation. Performs fast internal data exchange.
/// </summary>
/// <param name="other">The other collection.</param>
void Swap(HashSet& other)
{
if IF_CONSTEXPR (AllocationType::HasSwap)
{
::Swap(_elementsCount, other._elementsCount);
::Swap(_deletedCount, other._deletedCount);
::Swap(_size, other._size);
_allocation.Swap(other._allocation);
}
else
{
HashSet tmp = MoveTemp(other);
other = *this;
*this = MoveTemp(tmp);
}
Base::Clear();
}
public:
@@ -523,7 +408,7 @@ public:
template<typename ItemType>
bool Add(const ItemType& item)
{
Bucket* bucket = OnAdd(item);
Bucket* bucket = Base::OnAdd(item, false);
if (bucket)
bucket->Occupy(item);
return bucket != nullptr;
@@ -536,7 +421,7 @@ public:
/// <returns>True if element has been added to the collection, otherwise false if the element is already present.</returns>
bool Add(T&& item)
{
Bucket* bucket = OnAdd(item);
Bucket* bucket = Base::OnAdd(item, false);
if (bucket)
bucket->Occupy(MoveTemp(item));
return bucket != nullptr;
@@ -546,7 +431,7 @@ public:
/// Add element at iterator to the collection
/// </summary>
/// <param name="i">Iterator with item to add</param>
void Add(const Iterator& i)
DEPRECATED("Use Add with separate Key and Value from iterator.") void Add(const Iterator& i)
{
ASSERT(i._collection != this && i);
const Bucket& bucket = *i;
@@ -561,15 +446,13 @@ public:
template<typename ItemType>
bool Remove(const ItemType& item)
{
if (IsEmpty())
return false;
FindPositionResult pos;
FindPosition(item, pos);
typename Base::FindPositionResult pos;
Base::FindPosition(item, pos);
if (pos.ObjectIndex != -1)
{
_allocation.Get()[pos.ObjectIndex].Delete();
--_elementsCount;
++_deletedCount;
Base::_allocation.Get()[pos.ObjectIndex].Delete();
--Base::_elementsCount;
++Base::_deletedCount;
return true;
}
return false;
@@ -585,10 +468,9 @@ public:
ASSERT(i._collection == this);
if (i)
{
ASSERT(_allocation.Get()[i._index].IsOccupied());
_allocation.Get()[i._index].Delete();
--_elementsCount;
++_deletedCount;
Base::_allocation.Get()[i._index].Delete();
--Base::_elementsCount;
++Base::_deletedCount;
return true;
}
return false;
@@ -601,15 +483,26 @@ public:
/// <param name="item">Item to find</param>
/// <returns>Iterator for the found element or End if cannot find it</returns>
template<typename ItemType>
Iterator Find(const ItemType& item) const
Iterator Find(const ItemType& item)
{
if (IsEmpty())
return End();
FindPositionResult pos;
FindPosition(item, pos);
typename Base::FindPositionResult pos;
Base::FindPosition(item, pos);
return pos.ObjectIndex != -1 ? Iterator(this, pos.ObjectIndex) : End();
}
/// <summary>
/// Find element with given item in the collection
/// </summary>
/// <param name="item">Item to find</param>
/// <returns>Iterator for the found element or End if cannot find it</returns>
template<typename ItemType>
ConstIterator Find(const ItemType& item) const
{
typename Base::FindPositionResult pos;
Base::FindPosition(item, pos);
return pos.ObjectIndex != -1 ? ConstIterator(this, pos.ObjectIndex) : End();
}
/// <summary>
/// Determines whether a collection contains the specified element.
/// </summary>
@@ -618,39 +511,60 @@ public:
template<typename ItemType>
bool Contains(const ItemType& item) const
{
if (IsEmpty())
return false;
FindPositionResult pos;
FindPosition(item, pos);
typename Base::FindPositionResult pos;
Base::FindPosition(item, pos);
return pos.ObjectIndex != -1;
}
public:
/// <summary>
/// Clones other collection into this
/// Clones other collection into this.
/// </summary>
/// <param name="other">Other collection to clone</param>
/// <param name="other">The other collection to clone.</param>
void Clone(const HashSet& other)
{
Clear();
SetCapacity(other.Capacity(), false);
for (Iterator i = other.Begin(); i != other.End(); ++i)
Add(i);
ASSERT(Count() == other.Count());
ASSERT(Capacity() == other.Capacity());
Base::Clear();
Base::SetCapacity(other.Capacity(), false);
for (ConstIterator i = other.Begin(); i != other.End(); ++i)
Add(i->Item);
ASSERT(Base::Count() == other.Count());
ASSERT(Base::Capacity() == other.Capacity());
}
/// <summary>
/// Gets the items collection to the output array (will contain unique items).
/// </summary>
/// <param name="result">The result.</param>
template<typename ArrayAllocation>
void GetItems(Array<T, ArrayAllocation>& result) const
{
for (ConstIterator i = Begin(); i.IsNotEnd(); ++i)
result.Add(i->Item);
}
public:
Iterator Begin() const
Iterator Begin()
{
Iterator i(this, -1);
++i;
return i;
}
Iterator End() const
Iterator End()
{
return Iterator(this, _size);
return Iterator(this, Base::_size);
}
ConstIterator Begin() const
{
ConstIterator i(this, -1);
++i;
return i;
}
ConstIterator End() const
{
return ConstIterator(this, Base::_size);
}
Iterator begin()
@@ -662,141 +576,18 @@ public:
FORCE_INLINE Iterator end()
{
return Iterator(this, _size);
return Iterator(this, Base::_size);
}
Iterator begin() const
ConstIterator begin() const
{
Iterator i(this, -1);
ConstIterator i(this, -1);
++i;
return i;
}
FORCE_INLINE Iterator end() const
FORCE_INLINE ConstIterator end() const
{
return Iterator(this, _size);
}
private:
/// <summary>
/// The result container of the set item lookup searching.
/// </summary>
struct FindPositionResult
{
int32 ObjectIndex;
int32 FreeSlotIndex;
};
/// <summary>
/// Returns a pair of positions: 1st where the object is, 2nd where
/// it would go if you wanted to insert it. 1st is -1
/// if object is not found; 2nd is -1 if it is.
/// Note: because of deletions where-to-insert is not trivial: it's the
/// first deleted bucket we see, as long as we don't find the item later
/// </summary>
/// <param name="item">The item to find</param>
/// <param name="result">Pair of values: where the object is and where it would go if you wanted to insert it</param>
template<typename ItemType>
void FindPosition(const ItemType& item, FindPositionResult& result) const
{
ASSERT(_size);
const int32 tableSizeMinusOne = _size - 1;
int32 bucketIndex = GetHash(item) & tableSizeMinusOne;
int32 insertPos = -1;
int32 numChecks = 0;
const Bucket* data = _allocation.Get();
result.FreeSlotIndex = -1;
while (numChecks < _size)
{
// Empty bucket
const Bucket& bucket = data[bucketIndex];
if (bucket.IsEmpty())
{
// Found place to insert
result.ObjectIndex = -1;
result.FreeSlotIndex = insertPos == -1 ? bucketIndex : insertPos;
return;
}
// Deleted bucket
if (bucket.IsDeleted())
{
// Keep searching but mark to insert
if (insertPos == -1)
insertPos = bucketIndex;
}
// Occupied bucket by target item
else if (bucket.Item == item)
{
// Found item
result.ObjectIndex = bucketIndex;
return;
}
numChecks++;
bucketIndex = (bucketIndex + DICTIONARY_PROB_FUNC(_size, numChecks)) & tableSizeMinusOne;
}
result.ObjectIndex = -1;
result.FreeSlotIndex = insertPos;
}
template<typename ItemType>
Bucket* OnAdd(const ItemType& key)
{
// Check if need to rehash elements (prevent many deleted elements that use too much of capacity)
if (_deletedCount > _size / DICTIONARY_DEFAULT_SLACK_SCALE)
Compact();
// Ensure to have enough memory for the next item (in case of new element insertion)
EnsureCapacity(((_elementsCount + 1) * DICTIONARY_DEFAULT_SLACK_SCALE + _deletedCount) / DICTIONARY_DEFAULT_SLACK_SCALE);
// Find location of the item or place to insert it
FindPositionResult pos;
FindPosition(key, pos);
// Check if object has been already added
if (pos.ObjectIndex != -1)
return nullptr;
// Insert
ASSERT(pos.FreeSlotIndex != -1);
++_elementsCount;
return &_allocation.Get()[pos.FreeSlotIndex];
}
void Compact()
{
if (_elementsCount == 0)
{
// Fast path if it's empty
Bucket* data = _allocation.Get();
for (int32 i = 0; i < _size; ++i)
data[i]._state = BucketState::Empty;
}
else
{
// Rebuild entire table completely
AllocationData oldAllocation;
AllocationUtils::MoveToEmpty<Bucket, AllocationType>(oldAllocation, _allocation, _size, _size);
_allocation.Allocate(_size);
Bucket* data = _allocation.Get();
for (int32 i = 0; i < _size; ++i)
data[i]._state = BucketState::Empty;
Bucket* oldData = oldAllocation.Get();
FindPositionResult pos;
for (int32 i = 0; i < _size; ++i)
{
Bucket& oldBucket = oldData[i];
if (oldBucket.IsOccupied())
{
FindPosition(oldBucket.Item, pos);
ASSERT(pos.FreeSlotIndex != -1);
Bucket& bucket = _allocation.Get()[pos.FreeSlotIndex];
bucket = MoveTemp(oldBucket);
}
}
for (int32 i = 0; i < _size; ++i)
oldData[i].Free();
}
_deletedCount = 0;
return ConstIterator(this, Base::_size);
}
};

View File

@@ -0,0 +1,400 @@
// Copyright (c) 2012-2024 Wojciech Figat. All rights reserved.
#pragma once
#include "Engine/Core/Memory/Memory.h"
#include "Engine/Core/Memory/Allocation.h"
#include "Engine/Core/Memory/AllocationUtils.h"
#include "Engine/Core/Collections/HashFunctions.h"
#include "Engine/Core/Collections/Config.h"
/// <summary>
/// Tells if the object is occupied, and if not, if the bucket is a subject of compaction.
/// </summary>
enum class HashSetBucketState : byte
{
Empty = 0,
Deleted = 1,
Occupied = 2,
};
/// <summary>
/// Base class for unordered set of values (without duplicates with O(1) lookup access).
/// </summary>
/// <typeparam name="BucketType">The type of bucket structure that stores element data and state.</typeparam>
/// <typeparam name="AllocationType">The type of memory allocator.</typeparam>
template<typename AllocationType, typename BucketType>
class HashSetBase
{
friend HashSetBase;
public:
// Type of allocation data used to store hash set buckets.
using AllocationData = typename AllocationType::template Data<BucketType>;
protected:
int32 _elementsCount = 0;
int32 _deletedCount = 0;
int32 _size = 0;
AllocationData _allocation;
HashSetBase()
{
}
void MoveToEmpty(HashSetBase&& other)
{
_elementsCount = other._elementsCount;
_deletedCount = other._deletedCount;
_size = other._size;
other._elementsCount = 0;
other._deletedCount = 0;
other._size = 0;
AllocationUtils::MoveToEmpty<BucketType, AllocationType>(_allocation, other._allocation, _size, _size);
}
~HashSetBase()
{
Clear();
}
public:
/// <summary>
/// Gets the amount of the elements in the collection.
/// </summary>
FORCE_INLINE int32 Count() const
{
return _elementsCount;
}
/// <summary>
/// Gets the amount of the elements that can be contained by the collection.
/// </summary>
FORCE_INLINE int32 Capacity() const
{
return _size;
}
/// <summary>
/// Returns true if collection is empty.
/// </summary>
FORCE_INLINE bool IsEmpty() const
{
return _elementsCount == 0;
}
/// <summary>
/// Returns true if collection has one or more elements.
/// </summary>
FORCE_INLINE bool HasItems() const
{
return _elementsCount != 0;
}
public:
/// <summary>
/// Removes all elements from the collection.
/// </summary>
void Clear()
{
if (_elementsCount + _deletedCount != 0)
{
BucketType* data = _allocation.Get();
for (int32 i = 0; i < _size; i++)
data[i].Free();
_elementsCount = _deletedCount = 0;
}
}
/// <summary>
/// Changes the capacity of the collection.
/// </summary>
/// <param name="capacity">The new capacity.</param>
/// <param name="preserveContents">True if preserve collection data when changing its size, otherwise collection after resize will be empty.</param>
void SetCapacity(int32 capacity, const bool preserveContents = true)
{
if (capacity == _size)
return;
ASSERT(capacity >= 0);
AllocationData oldAllocation;
AllocationUtils::MoveToEmpty<BucketType, AllocationType>(oldAllocation, _allocation, _size, _size);
const int32 oldSize = _size;
const int32 oldElementsCount = _elementsCount;
_deletedCount = _elementsCount = 0;
if (capacity != 0 && (capacity & (capacity - 1)) != 0)
capacity = AllocationUtils::AlignToPowerOf2(capacity);
if (capacity)
{
_allocation.Allocate(capacity);
BucketType* data = _allocation.Get();
for (int32 i = 0; i < capacity; i++)
data[i]._state = HashSetBucketState::Empty;
}
_size = capacity;
BucketType* oldData = oldAllocation.Get();
if (oldElementsCount != 0 && capacity != 0 && preserveContents)
{
FindPositionResult pos;
for (int32 i = 0; i < oldSize; i++)
{
BucketType& oldBucket = oldData[i];
if (oldBucket.IsOccupied())
{
FindPosition(oldBucket.GetKey(), pos);
ASSERT(pos.FreeSlotIndex != -1);
BucketType& bucket = _allocation.Get()[pos.FreeSlotIndex];
bucket = MoveTemp(oldBucket);
_elementsCount++;
}
}
}
if (oldElementsCount != 0)
{
for (int32 i = 0; i < oldSize; i++)
oldData[i].Free();
}
}
/// <summary>
/// Ensures that collection has given capacity.
/// </summary>
/// <param name="minCapacity">The minimum required capacity.</param>
/// <param name="preserveContents">True if preserve collection data when changing its size, otherwise collection after resize will be empty.</param>
void EnsureCapacity(int32 minCapacity, const bool preserveContents = true)
{
minCapacity *= DICTIONARY_DEFAULT_SLACK_SCALE;
if (_size >= minCapacity)
return;
int32 capacity = _allocation.CalculateCapacityGrow(_size, minCapacity);
if (capacity < DICTIONARY_DEFAULT_CAPACITY)
capacity = DICTIONARY_DEFAULT_CAPACITY;
SetCapacity(capacity, preserveContents);
}
/// <summary>
/// Swaps the contents of collection with the other object without copy operation. Performs fast internal data exchange.
/// </summary>
/// <param name="other">The other collection.</param>
void Swap(HashSetBase& other)
{
if IF_CONSTEXPR (AllocationType::HasSwap)
{
::Swap(_elementsCount, other._elementsCount);
::Swap(_deletedCount, other._deletedCount);
::Swap(_size, other._size);
_allocation.Swap(other._allocation);
}
else
{
::Swap(other, *this);
}
}
public:
/// <summary>
/// The collection iterator base implementation.
/// </summary>
struct IteratorBase
{
protected:
const HashSetBase* _collection;
int32 _index;
IteratorBase(const HashSetBase* collection, const int32 index)
: _collection(collection)
, _index(index)
{
}
void Next()
{
const int32 capacity = _collection->_size;
if (_index != capacity)
{
const BucketType* data = _collection->_allocation.Get();
do
{
++_index;
}
while (_index != capacity && data[_index].IsNotOccupied());
}
}
void Prev()
{
if (_index > 0)
{
const BucketType* data = _collection->_allocation.Get();
do
{
--_index;
}
while (_index > 0 && data[_index].IsNotOccupied());
}
}
public:
FORCE_INLINE int32 Index() const
{
return _index;
}
FORCE_INLINE bool IsEnd() const
{
return _index == _collection->_size;
}
FORCE_INLINE bool IsNotEnd() const
{
return _index != _collection->_size;
}
FORCE_INLINE const BucketType& operator*() const
{
return _collection->_allocation.Get()[_index];
}
FORCE_INLINE const BucketType* operator->() const
{
return &_collection->_allocation.Get()[_index];
}
FORCE_INLINE explicit operator bool() const
{
return _index >= 0 && _index < _collection->_size;
}
};
protected:
/// <summary>
/// The result container of the set item lookup searching.
/// </summary>
struct FindPositionResult
{
int32 ObjectIndex;
int32 FreeSlotIndex;
};
/// <summary>
/// Returns a pair of positions: 1st where the object is, 2nd where it would go if you wanted to insert it.
/// 1st is -1 if object is not found; 2nd is -1 if it is.
/// Because of deletions where-to-insert is not trivial: it's the first deleted bucket we see, as long as we don't find the item later.
/// </summary>
/// <param name="key">The key to find</param>
/// <param name="result">A pair of values: where the object is and where it would go if you wanted to insert it</param>
template<typename KeyComparableType>
void FindPosition(const KeyComparableType& key, FindPositionResult& result) const
{
result.FreeSlotIndex = -1;
if (_size == 0)
{
result.ObjectIndex = -1;
return;
}
const int32 tableSizeMinusOne = _size - 1;
int32 bucketIndex = GetHash(key) & tableSizeMinusOne;
int32 insertPos = -1;
int32 checksCount = 0;
const BucketType* data = _allocation.Get();
while (checksCount < _size)
{
// Empty bucket
const BucketType& bucket = data[bucketIndex];
if (bucket.IsEmpty())
{
// Found place to insert
result.ObjectIndex = -1;
result.FreeSlotIndex = insertPos == -1 ? bucketIndex : insertPos;
return;
}
// Deleted bucket
if (bucket.IsDeleted())
{
// Keep searching but mark to insert
if (insertPos == -1)
insertPos = bucketIndex;
}
// Occupied bucket by target item
else if (bucket.GetKey() == key)
{
// Found item
result.ObjectIndex = bucketIndex;
return;
}
// Move to the next bucket
checksCount++;
bucketIndex = (bucketIndex + DICTIONARY_PROB_FUNC(_size, checksCount)) & tableSizeMinusOne;
}
result.ObjectIndex = -1;
result.FreeSlotIndex = insertPos;
}
template<typename KeyComparableType>
BucketType* OnAdd(const KeyComparableType& key, bool checkUnique = true)
{
// Check if need to rehash elements (prevent many deleted elements that use too much of capacity)
if (_deletedCount > _size / DICTIONARY_DEFAULT_SLACK_SCALE)
Compact();
// Ensure to have enough memory for the next item (in case of new element insertion)
EnsureCapacity(((_elementsCount + 1) * DICTIONARY_DEFAULT_SLACK_SCALE + _deletedCount) / DICTIONARY_DEFAULT_SLACK_SCALE);
// Find location of the item or place to insert it
FindPositionResult pos;
FindPosition(key, pos);
// Check if object has been already added
if (pos.ObjectIndex != -1)
{
if (checkUnique)
{
Platform::CheckFailed("That key has been already added to the collection.", __FILE__, __LINE__);
return nullptr;
}
return &_allocation.Get()[pos.ObjectIndex];
}
// Insert
ASSERT(pos.FreeSlotIndex != -1);
++_elementsCount;
return &_allocation.Get()[pos.FreeSlotIndex];
}
void Compact()
{
if (_elementsCount == 0)
{
// Fast path if it's empty
BucketType* data = _allocation.Get();
for (int32 i = 0; i < _size; ++i)
data[i]._state = HashSetBucketState::Empty;
}
else
{
// Rebuild entire table completely
AllocationData oldAllocation;
AllocationUtils::MoveToEmpty<BucketType, AllocationType>(oldAllocation, _allocation, _size, _size);
_allocation.Allocate(_size);
BucketType* data = _allocation.Get();
for (int32 i = 0; i < _size; ++i)
data[i]._state = HashSetBucketState::Empty;
BucketType* oldData = oldAllocation.Get();
FindPositionResult pos;
for (int32 i = 0; i < _size; ++i)
{
BucketType& oldBucket = oldData[i];
if (oldBucket.IsOccupied())
{
FindPosition(oldBucket.GetKey(), pos);
ASSERT(pos.FreeSlotIndex != -1);
BucketType& bucket = _allocation.Get()[pos.FreeSlotIndex];
bucket = MoveTemp(oldBucket);
}
}
for (int32 i = 0; i < _size; ++i)
oldData[i].Free();
}
_deletedCount = 0;
}
};

View File

@@ -83,6 +83,8 @@ template<typename T, typename U>
class Pair;
template<typename KeyType, typename ValueType, typename AllocationType>
class Dictionary;
template<typename T, typename AllocationType>
class HashSet;
template<typename>
class Function;
template<typename... Params>