#ifndef HASH_MAP_H #define HASH_MAP_H /*************************************************************************/ /* hash_map.h */ /*************************************************************************/ /* This file is part of: */ /* PANDEMONIUM ENGINE */ /* https://github.com/Relintai/pandemonium_engine */ /*************************************************************************/ /* Copyright (c) 2022-present Péter Magyar. */ /* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ /* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ /* */ /* Permission is hereby granted, free of charge, to any person obtaining */ /* a copy of this software and associated documentation files (the */ /* "Software"), to deal in the Software without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of the Software, and to */ /* permit persons to whom the Software is furnished to do so, subject to */ /* the following conditions: */ /* */ /* The above copyright notice and this permission notice shall be */ /* included in all copies or substantial portions of the Software. */ /* */ /* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ /* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ /* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ /* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ /* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ /* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ /* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /*************************************************************************/ #include "core/containers/hashfuncs.h" #include "core/containers/paged_allocator.h" #include "core/containers/pair.h" #include "core/math/math_funcs.h" #include "core/os/memory.h" /** * A HashMap implementation that uses open addressing with Robin Hood hashing. * Robin Hood hashing swaps out entries that have a smaller probing distance * than the to-be-inserted entry, that evens out the average probing distance * and enables faster lookups. Backward shift deletion is employed to further * improve the performance and to avoid infinite loops in rare cases. * * Keys and values are stored in a double linked list by insertion order. This * has a slight performance overhead on lookup, which can be mostly compensated * using a paged allocator if required. * * The assignment operator copy the pairs from one map to the other. */ template > class HashMap { public: const uint32_t MIN_CAPACITY_INDEX = 2; // Use a prime. const float MAX_OCCUPANCY = 0.75; const uint32_t EMPTY_HASH = 0; public: struct Element { Element *next = nullptr; Element *prev = nullptr; KeyValue data; const TKey &key() const { return data.key; } TValue &value() { return data.value; } const TValue &value() const { return data.value; } TValue &get() { return data.value; }; const TValue &get() const { return data.value; }; Element() {} Element(const TKey &p_key, const TValue &p_value) : data(p_key, p_value) {} }; public: _FORCE_INLINE_ uint32_t get_capacity() const { return hash_table_size_primes[capacity_index]; } _FORCE_INLINE_ uint32_t size() const { return num_elements; } /* Standard Godot Container API */ bool empty() const { return num_elements == 0; } void clear() { if (elements == nullptr || num_elements == 0) { return; } uint32_t capacity = hash_table_size_primes[capacity_index]; for (uint32_t i = 0; i < capacity; i++) { if (hashes[i] == EMPTY_HASH) { continue; } hashes[i] = EMPTY_HASH; memdelete(elements[i]); elements[i] = nullptr; } tail_element = nullptr; head_element = nullptr; num_elements = 0; } TValue &get(const TKey &p_key) { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); CRASH_COND_MSG(!exists, "HashMap key not found."); return elements[pos]->data.value; } const TValue &get(const TKey &p_key) const { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); CRASH_COND_MSG(!exists, "HashMap key not found."); return elements[pos]->data.value; } const TValue *getptr(const TKey &p_key) const { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (exists) { return &elements[pos]->data.value; } return nullptr; } TValue *getptr(const TKey &p_key) { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (exists) { return &elements[pos]->data.value; } return nullptr; } const Element *get_element(const TKey &p_key) const { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (exists) { return elements[pos]; } return NULL; } Element *get_element(const TKey &p_key) { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (exists) { return elements[pos]; } return NULL; } _FORCE_INLINE_ const Element *find(const TKey &p_key) const { return get_element(p_key); } _FORCE_INLINE_ Element *find(const TKey &p_key) { return get_element(p_key); } /** * Same as get, except it can return NULL when item was not found. * This version is custom, will take a hash and a custom key (that should support operator==() */ template _FORCE_INLINE_ TValue *custom_getptr(C p_custom_key, uint32_t p_custom_hash) { if (unlikely(!elements)) { return NULL; } const uint32_t capacity = hash_table_size_primes[capacity_index]; const uint64_t capacity_inv = hash_table_size_primes_inv[capacity_index]; uint32_t hash = p_custom_hash; uint32_t pos = fastmod(hash, capacity_inv, capacity); uint32_t distance = 0; while (true) { if (hashes[pos] == EMPTY_HASH) { return NULL; } if (distance > _get_probe_length(pos, hashes[pos], capacity, capacity_inv)) { return NULL; } if (hashes[pos] == hash && Comparator::compare(elements[pos]->data.key, p_custom_key)) { return &elements[pos]->data.value; } pos = fastmod((pos + 1), capacity_inv, capacity); distance++; } return NULL; } template _FORCE_INLINE_ const TValue *custom_getptr(C p_custom_key, uint32_t p_custom_hash) const { if (unlikely(!elements)) { return NULL; } const uint32_t capacity = hash_table_size_primes[capacity_index]; const uint64_t capacity_inv = hash_table_size_primes_inv[capacity_index]; uint32_t hash = p_custom_hash; uint32_t pos = fastmod(hash, capacity_inv, capacity); uint32_t distance = 0; while (true) { if (hashes[pos] == EMPTY_HASH) { return NULL; } if (distance > _get_probe_length(pos, hashes[pos], capacity, capacity_inv)) { return NULL; } if (hashes[pos] == hash && Comparator::compare(elements[pos]->data.key, p_custom_key)) { return &elements[pos]->data.value; } pos = fastmod((pos + 1), capacity_inv, capacity); distance++; } return NULL; } _FORCE_INLINE_ bool has(const TKey &p_key) const { uint32_t _pos = 0; return _lookup_pos(p_key, _pos); } bool erase(const TKey &p_key) { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (!exists) { return false; } const uint32_t capacity = hash_table_size_primes[capacity_index]; const uint64_t capacity_inv = hash_table_size_primes_inv[capacity_index]; uint32_t next_pos = fastmod((pos + 1), capacity_inv, capacity); while (hashes[next_pos] != EMPTY_HASH && _get_probe_length(next_pos, hashes[next_pos], capacity, capacity_inv) != 0) { SWAP(hashes[next_pos], hashes[pos]); SWAP(elements[next_pos], elements[pos]); pos = next_pos; next_pos = fastmod((pos + 1), capacity_inv, capacity); } hashes[pos] = EMPTY_HASH; if (head_element == elements[pos]) { head_element = elements[pos]->next; } if (tail_element == elements[pos]) { tail_element = elements[pos]->prev; } if (elements[pos]->prev) { elements[pos]->prev->next = elements[pos]->next; } if (elements[pos]->next) { elements[pos]->next->prev = elements[pos]->prev; } memdelete(elements[pos]); elements[pos] = nullptr; num_elements--; return true; } // Reserves space for a number of elements, useful to avoid many resizes and rehashes. // If adding a known (possibly large) number of elements at once, must be larger than old capacity. void reserve(uint32_t p_new_capacity) { uint32_t new_index = capacity_index; while (hash_table_size_primes[new_index] < p_new_capacity) { ERR_FAIL_COND_MSG(new_index + 1 == (uint32_t)HASH_TABLE_SIZE_MAX, nullptr); new_index++; } if (new_index == capacity_index) { return; } if (elements == nullptr) { capacity_index = new_index; return; // Unallocated yet. } _resize_and_rehash(new_index); } _FORCE_INLINE_ Element *front() { return head_element; } _FORCE_INLINE_ Element *back() { return tail_element; } _FORCE_INLINE_ const Element *front() const { return head_element; } _FORCE_INLINE_ const Element *back() const { return tail_element; } /** * Get the next key to p_key, and the first key if p_key is null. * Returns a pointer to the next key if found, NULL otherwise. * Adding/Removing elements while iterating will, of course, have unexpected results, don't do it. * * Example: * * const TKey *k=NULL; * * while( (k=table.next(k)) ) { * * print( *k ); * } * * This is for backwards compatibility. Use this syntax instead for new code: * * for (const HashMap::Element *E = map.front(); E; E = E->next) { * ... * } * */ const TKey *next(const TKey *p_key) const { if (unlikely(!elements)) { return nullptr; } if (!p_key) { /* get the first key */ if (unlikely(!front())) { return nullptr; } return &front()->data.key; } else { /* get the next key */ const Element *e = get_element(*p_key); ERR_FAIL_COND_V_MSG(!e, nullptr, "Invalid key supplied."); if (e->next) { /* if there is a "next" in the list, return that */ return &e->next->data.key; } /* nothing found, was at end */ } return nullptr; /* nothing found */ } /* Indexing */ const TValue &operator[](const TKey &p_key) const { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); CRASH_COND(!exists); return elements[pos]->data.value; } TValue &operator[](const TKey &p_key) { uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (!exists) { return _insert(p_key, TValue())->data.value; } else { return elements[pos]->data.value; } } /* Insert */ Element *insert(const TKey &p_key, const TValue &p_value, bool p_front_insert = false) { return _insert(p_key, p_value, p_front_insert); } Element *set(const TKey &p_key, const TValue &p_value, bool p_front_insert = false) { return _insert(p_key, p_value, p_front_insert); } /* Helpers */ void get_key_list(List *p_keys) const { if (unlikely(!elements)) { return; } for (const Element *E = front(); E; E = E->next) { p_keys->push_back(E->data.key); } } /* Constructors */ HashMap(const HashMap &p_other) { reserve(hash_table_size_primes[p_other.capacity_index]); if (p_other.num_elements == 0) { return; } for (const Element *E = p_other.front(); E; E = E->next) { insert(E->data.key, E->data.value); } } void operator=(const HashMap &p_other) { if (this == &p_other) { return; // Ignore self assignment. } if (num_elements != 0) { clear(); } reserve(hash_table_size_primes[p_other.capacity_index]); if (p_other.elements == nullptr) { return; // Nothing to copy. } for (const Element *E = p_other.front(); E; E = E->next) { insert(E->data.key, E->data.value); } } HashMap(uint32_t p_initial_capacity) { // Capacity can't be 0. capacity_index = 0; reserve(p_initial_capacity); } HashMap() { capacity_index = MIN_CAPACITY_INDEX; } uint32_t debug_get_hash(uint32_t p_index) { if (num_elements == 0) { return 0; } ERR_FAIL_INDEX_V(p_index, get_capacity(), 0); return hashes[p_index]; } Element *debug_get_element(uint32_t p_index) { if (num_elements == 0) { return NULL; } ERR_FAIL_INDEX_V(p_index, get_capacity(), NULL); return elements[p_index]; } ~HashMap() { clear(); if (elements != nullptr) { Memory::free_static(elements); Memory::free_static(hashes); } } private: Element **elements = nullptr; uint32_t *hashes = nullptr; Element *head_element = nullptr; Element *tail_element = nullptr; uint32_t capacity_index = 0; uint32_t num_elements = 0; _FORCE_INLINE_ uint32_t _hash(const TKey &p_key) const { uint32_t hash = Hasher::hash(p_key); if (unlikely(hash == EMPTY_HASH)) { hash = EMPTY_HASH + 1; } return hash; } static _FORCE_INLINE_ uint32_t _get_probe_length(const uint32_t p_pos, const uint32_t p_hash, const uint32_t p_capacity, const uint64_t p_capacity_inv) { const uint32_t original_pos = fastmod(p_hash, p_capacity_inv, p_capacity); return fastmod(p_pos - original_pos + p_capacity, p_capacity_inv, p_capacity); } bool _lookup_pos(const TKey &p_key, uint32_t &r_pos) const { if (elements == nullptr || num_elements == 0) { return false; // Failed lookups, no elements } const uint32_t capacity = hash_table_size_primes[capacity_index]; const uint64_t capacity_inv = hash_table_size_primes_inv[capacity_index]; uint32_t hash = _hash(p_key); uint32_t pos = fastmod(hash, capacity_inv, capacity); uint32_t distance = 0; while (true) { if (hashes[pos] == EMPTY_HASH) { return false; } if (distance > _get_probe_length(pos, hashes[pos], capacity, capacity_inv)) { return false; } if (hashes[pos] == hash && Comparator::compare(elements[pos]->data.key, p_key)) { r_pos = pos; return true; } pos = fastmod((pos + 1), capacity_inv, capacity); distance++; } } void _insert_with_hash(uint32_t p_hash, Element *p_value) { const uint32_t capacity = hash_table_size_primes[capacity_index]; const uint64_t capacity_inv = hash_table_size_primes_inv[capacity_index]; uint32_t hash = p_hash; Element *value = p_value; uint32_t distance = 0; uint32_t pos = fastmod(hash, capacity_inv, capacity); while (true) { if (hashes[pos] == EMPTY_HASH) { elements[pos] = value; hashes[pos] = hash; num_elements++; return; } // Not an empty slot, let's check the probing length of the existing one. uint32_t existing_probe_len = _get_probe_length(pos, hashes[pos], capacity, capacity_inv); if (existing_probe_len < distance) { SWAP(hash, hashes[pos]); SWAP(value, elements[pos]); distance = existing_probe_len; } pos = fastmod((pos + 1), capacity_inv, capacity); distance++; } } void _resize_and_rehash(uint32_t p_new_capacity_index) { uint32_t old_capacity = hash_table_size_primes[capacity_index]; // Capacity can't be 0. capacity_index = MAX((uint32_t)MIN_CAPACITY_INDEX, p_new_capacity_index); uint32_t capacity = hash_table_size_primes[capacity_index]; Element **old_elements = elements; uint32_t *old_hashes = hashes; num_elements = 0; hashes = reinterpret_cast(Memory::alloc_static(sizeof(uint32_t) * capacity)); elements = reinterpret_cast(Memory::alloc_static(sizeof(Element *) * capacity)); for (uint32_t i = 0; i < capacity; i++) { hashes[i] = 0; elements[i] = nullptr; } if (old_capacity == 0) { // Nothing to do. return; } for (uint32_t i = 0; i < old_capacity; i++) { if (old_hashes[i] == EMPTY_HASH) { continue; } _insert_with_hash(old_hashes[i], old_elements[i]); } Memory::free_static(old_elements); Memory::free_static(old_hashes); } _FORCE_INLINE_ Element *_insert(const TKey &p_key, const TValue &p_value, bool p_front_insert = false) { uint32_t capacity = hash_table_size_primes[capacity_index]; if (unlikely(elements == nullptr)) { // Allocate on demand to save memory. hashes = reinterpret_cast(Memory::alloc_static(sizeof(uint32_t) * capacity)); elements = reinterpret_cast(Memory::alloc_static(sizeof(Element *) * capacity)); for (uint32_t i = 0; i < capacity; i++) { hashes[i] = EMPTY_HASH; elements[i] = nullptr; } } uint32_t pos = 0; bool exists = _lookup_pos(p_key, pos); if (exists) { elements[pos]->data.value = p_value; return elements[pos]; } else { if (num_elements + 1 > MAX_OCCUPANCY * capacity) { ERR_FAIL_COND_V_MSG(capacity_index + 1 == HASH_TABLE_SIZE_MAX, nullptr, "Hash table maximum capacity reached, aborting insertion."); _resize_and_rehash(capacity_index + 1); } Element *elem = memnew(Element(p_key, p_value)); if (tail_element == nullptr) { head_element = elem; tail_element = elem; } else if (p_front_insert) { head_element->prev = elem; elem->next = head_element; head_element = elem; } else { tail_element->next = elem; elem->prev = tail_element; tail_element = elem; } uint32_t hash = _hash(p_key); _insert_with_hash(hash, elem); return elem; } } }; #endif // HASH_MAP_H