flat_hash_set.h 19 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507
  1. // Copyright 2018 The Abseil Authors.
  2. //
  3. // Licensed under the Apache License, Version 2.0 (the "License");
  4. // you may not use this file except in compliance with the License.
  5. // You may obtain a copy of the License at
  6. //
  7. // https://www.apache.org/licenses/LICENSE-2.0
  8. //
  9. // Unless required by applicable law or agreed to in writing, software
  10. // distributed under the License is distributed on an "AS IS" BASIS,
  11. // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12. // See the License for the specific language governing permissions and
  13. // limitations under the License.
  14. //
  15. // -----------------------------------------------------------------------------
  16. // File: flat_hash_set.h
  17. // -----------------------------------------------------------------------------
  18. //
  19. // An `y_absl::flat_hash_set<T>` is an unordered associative container designed to
  20. // be a more efficient replacement for `std::unordered_set`. Like
  21. // `unordered_set`, search, insertion, and deletion of set elements can be done
  22. // as an `O(1)` operation. However, `flat_hash_set` (and other unordered
  23. // associative containers known as the collection of Abseil "Swiss tables")
  24. // contain other optimizations that result in both memory and computation
  25. // advantages.
  26. //
  27. // In most cases, your default choice for a hash set should be a set of type
  28. // `flat_hash_set`.
  29. #ifndef Y_ABSL_CONTAINER_FLAT_HASH_SET_H_
  30. #define Y_ABSL_CONTAINER_FLAT_HASH_SET_H_
  31. #include <type_traits>
  32. #include <utility>
  33. #include "y_absl/algorithm/container.h"
  34. #include "y_absl/base/macros.h"
  35. #include "y_absl/container/internal/container_memory.h"
  36. #include "y_absl/container/internal/hash_function_defaults.h" // IWYU pragma: export
  37. #include "y_absl/container/internal/raw_hash_set.h" // IWYU pragma: export
  38. #include "y_absl/memory/memory.h"
  39. namespace y_absl {
  40. Y_ABSL_NAMESPACE_BEGIN
  41. namespace container_internal {
  42. template <typename T>
  43. struct FlatHashSetPolicy;
  44. } // namespace container_internal
  45. // -----------------------------------------------------------------------------
  46. // y_absl::flat_hash_set
  47. // -----------------------------------------------------------------------------
  48. //
  49. // An `y_absl::flat_hash_set<T>` is an unordered associative container which has
  50. // been optimized for both speed and memory footprint in most common use cases.
  51. // Its interface is similar to that of `std::unordered_set<T>` with the
  52. // following notable differences:
  53. //
  54. // * Requires keys that are CopyConstructible
  55. // * Supports heterogeneous lookup, through `find()` and `insert()`, provided
  56. // that the set is provided a compatible heterogeneous hashing function and
  57. // equality operator.
  58. // * Invalidates any references and pointers to elements within the table after
  59. // `rehash()`.
  60. // * Contains a `capacity()` member function indicating the number of element
  61. // slots (open, deleted, and empty) within the hash set.
  62. // * Returns `void` from the `erase(iterator)` overload.
  63. //
  64. // By default, `flat_hash_set` uses the `y_absl::Hash` hashing framework. All
  65. // fundamental and Abseil types that support the `y_absl::Hash` framework have a
  66. // compatible equality operator for comparing insertions into `flat_hash_set`.
  67. // If your type is not yet supported by the `y_absl::Hash` framework, see
  68. // y_absl/hash/hash.h for information on extending Abseil hashing to user-defined
  69. // types.
  70. //
  71. // Using `y_absl::flat_hash_set` at interface boundaries in dynamically loaded
  72. // libraries (e.g. .dll, .so) is unsupported due to way `y_absl::Hash` values may
  73. // be randomized across dynamically loaded libraries.
  74. //
  75. // NOTE: A `flat_hash_set` stores its keys directly inside its implementation
  76. // array to avoid memory indirection. Because a `flat_hash_set` is designed to
  77. // move data when rehashed, set keys will not retain pointer stability. If you
  78. // require pointer stability, consider using
  79. // `y_absl::flat_hash_set<std::unique_ptr<T>>`. If your type is not moveable and
  80. // you require pointer stability, consider `y_absl::node_hash_set` instead.
  81. //
  82. // Example:
  83. //
  84. // // Create a flat hash set of three strings
  85. // y_absl::flat_hash_set<TString> ducks =
  86. // {"huey", "dewey", "louie"};
  87. //
  88. // // Insert a new element into the flat hash set
  89. // ducks.insert("donald");
  90. //
  91. // // Force a rehash of the flat hash set
  92. // ducks.rehash(0);
  93. //
  94. // // See if "dewey" is present
  95. // if (ducks.contains("dewey")) {
  96. // std::cout << "We found dewey!" << std::endl;
  97. // }
  98. template <class T, class Hash = y_absl::container_internal::hash_default_hash<T>,
  99. class Eq = y_absl::container_internal::hash_default_eq<T>,
  100. class Allocator = std::allocator<T>>
  101. class flat_hash_set
  102. : public y_absl::container_internal::raw_hash_set<
  103. y_absl::container_internal::FlatHashSetPolicy<T>, Hash, Eq, Allocator> {
  104. using Base = typename flat_hash_set::raw_hash_set;
  105. public:
  106. // Constructors and Assignment Operators
  107. //
  108. // A flat_hash_set supports the same overload set as `std::unordered_set`
  109. // for construction and assignment:
  110. //
  111. // * Default constructor
  112. //
  113. // // No allocation for the table's elements is made.
  114. // y_absl::flat_hash_set<TString> set1;
  115. //
  116. // * Initializer List constructor
  117. //
  118. // y_absl::flat_hash_set<TString> set2 =
  119. // {{"huey"}, {"dewey"}, {"louie"},};
  120. //
  121. // * Copy constructor
  122. //
  123. // y_absl::flat_hash_set<TString> set3(set2);
  124. //
  125. // * Copy assignment operator
  126. //
  127. // // Hash functor and Comparator are copied as well
  128. // y_absl::flat_hash_set<TString> set4;
  129. // set4 = set3;
  130. //
  131. // * Move constructor
  132. //
  133. // // Move is guaranteed efficient
  134. // y_absl::flat_hash_set<TString> set5(std::move(set4));
  135. //
  136. // * Move assignment operator
  137. //
  138. // // May be efficient if allocators are compatible
  139. // y_absl::flat_hash_set<TString> set6;
  140. // set6 = std::move(set5);
  141. //
  142. // * Range constructor
  143. //
  144. // std::vector<TString> v = {"a", "b"};
  145. // y_absl::flat_hash_set<TString> set7(v.begin(), v.end());
  146. flat_hash_set() {}
  147. using Base::Base;
  148. // flat_hash_set::begin()
  149. //
  150. // Returns an iterator to the beginning of the `flat_hash_set`.
  151. using Base::begin;
  152. // flat_hash_set::cbegin()
  153. //
  154. // Returns a const iterator to the beginning of the `flat_hash_set`.
  155. using Base::cbegin;
  156. // flat_hash_set::cend()
  157. //
  158. // Returns a const iterator to the end of the `flat_hash_set`.
  159. using Base::cend;
  160. // flat_hash_set::end()
  161. //
  162. // Returns an iterator to the end of the `flat_hash_set`.
  163. using Base::end;
  164. // flat_hash_set::capacity()
  165. //
  166. // Returns the number of element slots (assigned, deleted, and empty)
  167. // available within the `flat_hash_set`.
  168. //
  169. // NOTE: this member function is particular to `y_absl::flat_hash_set` and is
  170. // not provided in the `std::unordered_set` API.
  171. using Base::capacity;
  172. // flat_hash_set::empty()
  173. //
  174. // Returns whether or not the `flat_hash_set` is empty.
  175. using Base::empty;
  176. // flat_hash_set::max_size()
  177. //
  178. // Returns the largest theoretical possible number of elements within a
  179. // `flat_hash_set` under current memory constraints. This value can be thought
  180. // of the largest value of `std::distance(begin(), end())` for a
  181. // `flat_hash_set<T>`.
  182. using Base::max_size;
  183. // flat_hash_set::size()
  184. //
  185. // Returns the number of elements currently within the `flat_hash_set`.
  186. using Base::size;
  187. // flat_hash_set::clear()
  188. //
  189. // Removes all elements from the `flat_hash_set`. Invalidates any references,
  190. // pointers, or iterators referring to contained elements.
  191. //
  192. // NOTE: this operation may shrink the underlying buffer. To avoid shrinking
  193. // the underlying buffer call `erase(begin(), end())`.
  194. using Base::clear;
  195. // flat_hash_set::erase()
  196. //
  197. // Erases elements within the `flat_hash_set`. Erasing does not trigger a
  198. // rehash. Overloads are listed below.
  199. //
  200. // void erase(const_iterator pos):
  201. //
  202. // Erases the element at `position` of the `flat_hash_set`, returning
  203. // `void`.
  204. //
  205. // NOTE: returning `void` in this case is different than that of STL
  206. // containers in general and `std::unordered_set` in particular (which
  207. // return an iterator to the element following the erased element). If that
  208. // iterator is needed, simply post increment the iterator:
  209. //
  210. // set.erase(it++);
  211. //
  212. // iterator erase(const_iterator first, const_iterator last):
  213. //
  214. // Erases the elements in the open interval [`first`, `last`), returning an
  215. // iterator pointing to `last`. The special case of calling
  216. // `erase(begin(), end())` resets the reserved growth such that if
  217. // `reserve(N)` has previously been called and there has been no intervening
  218. // call to `clear()`, then after calling `erase(begin(), end())`, it is safe
  219. // to assume that inserting N elements will not cause a rehash.
  220. //
  221. // size_type erase(const key_type& key):
  222. //
  223. // Erases the element with the matching key, if it exists, returning the
  224. // number of elements erased (0 or 1).
  225. using Base::erase;
  226. // flat_hash_set::insert()
  227. //
  228. // Inserts an element of the specified value into the `flat_hash_set`,
  229. // returning an iterator pointing to the newly inserted element, provided that
  230. // an element with the given key does not already exist. If rehashing occurs
  231. // due to the insertion, all iterators are invalidated. Overloads are listed
  232. // below.
  233. //
  234. // std::pair<iterator,bool> insert(const T& value):
  235. //
  236. // Inserts a value into the `flat_hash_set`. Returns a pair consisting of an
  237. // iterator to the inserted element (or to the element that prevented the
  238. // insertion) and a bool denoting whether the insertion took place.
  239. //
  240. // std::pair<iterator,bool> insert(T&& value):
  241. //
  242. // Inserts a moveable value into the `flat_hash_set`. Returns a pair
  243. // consisting of an iterator to the inserted element (or to the element that
  244. // prevented the insertion) and a bool denoting whether the insertion took
  245. // place.
  246. //
  247. // iterator insert(const_iterator hint, const T& value):
  248. // iterator insert(const_iterator hint, T&& value):
  249. //
  250. // Inserts a value, using the position of `hint` as a non-binding suggestion
  251. // for where to begin the insertion search. Returns an iterator to the
  252. // inserted element, or to the existing element that prevented the
  253. // insertion.
  254. //
  255. // void insert(InputIterator first, InputIterator last):
  256. //
  257. // Inserts a range of values [`first`, `last`).
  258. //
  259. // NOTE: Although the STL does not specify which element may be inserted if
  260. // multiple keys compare equivalently, for `flat_hash_set` we guarantee the
  261. // first match is inserted.
  262. //
  263. // void insert(std::initializer_list<T> ilist):
  264. //
  265. // Inserts the elements within the initializer list `ilist`.
  266. //
  267. // NOTE: Although the STL does not specify which element may be inserted if
  268. // multiple keys compare equivalently within the initializer list, for
  269. // `flat_hash_set` we guarantee the first match is inserted.
  270. using Base::insert;
  271. // flat_hash_set::emplace()
  272. //
  273. // Inserts an element of the specified value by constructing it in-place
  274. // within the `flat_hash_set`, provided that no element with the given key
  275. // already exists.
  276. //
  277. // The element may be constructed even if there already is an element with the
  278. // key in the container, in which case the newly constructed element will be
  279. // destroyed immediately.
  280. //
  281. // If rehashing occurs due to the insertion, all iterators are invalidated.
  282. using Base::emplace;
  283. // flat_hash_set::emplace_hint()
  284. //
  285. // Inserts an element of the specified value by constructing it in-place
  286. // within the `flat_hash_set`, using the position of `hint` as a non-binding
  287. // suggestion for where to begin the insertion search, and only inserts
  288. // provided that no element with the given key already exists.
  289. //
  290. // The element may be constructed even if there already is an element with the
  291. // key in the container, in which case the newly constructed element will be
  292. // destroyed immediately.
  293. //
  294. // If rehashing occurs due to the insertion, all iterators are invalidated.
  295. using Base::emplace_hint;
  296. // flat_hash_set::extract()
  297. //
  298. // Extracts the indicated element, erasing it in the process, and returns it
  299. // as a C++17-compatible node handle. Overloads are listed below.
  300. //
  301. // node_type extract(const_iterator position):
  302. //
  303. // Extracts the element at the indicated position and returns a node handle
  304. // owning that extracted data.
  305. //
  306. // node_type extract(const key_type& x):
  307. //
  308. // Extracts the element with the key matching the passed key value and
  309. // returns a node handle owning that extracted data. If the `flat_hash_set`
  310. // does not contain an element with a matching key, this function returns an
  311. // empty node handle.
  312. using Base::extract;
  313. // flat_hash_set::merge()
  314. //
  315. // Extracts elements from a given `source` flat hash set into this
  316. // `flat_hash_set`. If the destination `flat_hash_set` already contains an
  317. // element with an equivalent key, that element is not extracted.
  318. using Base::merge;
  319. // flat_hash_set::swap(flat_hash_set& other)
  320. //
  321. // Exchanges the contents of this `flat_hash_set` with those of the `other`
  322. // flat hash set, avoiding invocation of any move, copy, or swap operations on
  323. // individual elements.
  324. //
  325. // All iterators and references on the `flat_hash_set` remain valid, excepting
  326. // for the past-the-end iterator, which is invalidated.
  327. //
  328. // `swap()` requires that the flat hash set's hashing and key equivalence
  329. // functions be Swappable, and are exchanged using unqualified calls to
  330. // non-member `swap()`. If the set's allocator has
  331. // `std::allocator_traits<allocator_type>::propagate_on_container_swap::value`
  332. // set to `true`, the allocators are also exchanged using an unqualified call
  333. // to non-member `swap()`; otherwise, the allocators are not swapped.
  334. using Base::swap;
  335. // flat_hash_set::rehash(count)
  336. //
  337. // Rehashes the `flat_hash_set`, setting the number of slots to be at least
  338. // the passed value. If the new number of slots increases the load factor more
  339. // than the current maximum load factor
  340. // (`count` < `size()` / `max_load_factor()`), then the new number of slots
  341. // will be at least `size()` / `max_load_factor()`.
  342. //
  343. // To force a rehash, pass rehash(0).
  344. //
  345. // NOTE: unlike behavior in `std::unordered_set`, references are also
  346. // invalidated upon a `rehash()`.
  347. using Base::rehash;
  348. // flat_hash_set::reserve(count)
  349. //
  350. // Sets the number of slots in the `flat_hash_set` to the number needed to
  351. // accommodate at least `count` total elements without exceeding the current
  352. // maximum load factor, and may rehash the container if needed.
  353. using Base::reserve;
  354. // flat_hash_set::contains()
  355. //
  356. // Determines whether an element comparing equal to the given `key` exists
  357. // within the `flat_hash_set`, returning `true` if so or `false` otherwise.
  358. using Base::contains;
  359. // flat_hash_set::count(const Key& key) const
  360. //
  361. // Returns the number of elements comparing equal to the given `key` within
  362. // the `flat_hash_set`. note that this function will return either `1` or `0`
  363. // since duplicate elements are not allowed within a `flat_hash_set`.
  364. using Base::count;
  365. // flat_hash_set::equal_range()
  366. //
  367. // Returns a closed range [first, last], defined by a `std::pair` of two
  368. // iterators, containing all elements with the passed key in the
  369. // `flat_hash_set`.
  370. using Base::equal_range;
  371. // flat_hash_set::find()
  372. //
  373. // Finds an element with the passed `key` within the `flat_hash_set`.
  374. using Base::find;
  375. // flat_hash_set::bucket_count()
  376. //
  377. // Returns the number of "buckets" within the `flat_hash_set`. Note that
  378. // because a flat hash set contains all elements within its internal storage,
  379. // this value simply equals the current capacity of the `flat_hash_set`.
  380. using Base::bucket_count;
  381. // flat_hash_set::load_factor()
  382. //
  383. // Returns the current load factor of the `flat_hash_set` (the average number
  384. // of slots occupied with a value within the hash set).
  385. using Base::load_factor;
  386. // flat_hash_set::max_load_factor()
  387. //
  388. // Manages the maximum load factor of the `flat_hash_set`. Overloads are
  389. // listed below.
  390. //
  391. // float flat_hash_set::max_load_factor()
  392. //
  393. // Returns the current maximum load factor of the `flat_hash_set`.
  394. //
  395. // void flat_hash_set::max_load_factor(float ml)
  396. //
  397. // Sets the maximum load factor of the `flat_hash_set` to the passed value.
  398. //
  399. // NOTE: This overload is provided only for API compatibility with the STL;
  400. // `flat_hash_set` will ignore any set load factor and manage its rehashing
  401. // internally as an implementation detail.
  402. using Base::max_load_factor;
  403. // flat_hash_set::get_allocator()
  404. //
  405. // Returns the allocator function associated with this `flat_hash_set`.
  406. using Base::get_allocator;
  407. // flat_hash_set::hash_function()
  408. //
  409. // Returns the hashing function used to hash the keys within this
  410. // `flat_hash_set`.
  411. using Base::hash_function;
  412. // flat_hash_set::key_eq()
  413. //
  414. // Returns the function used for comparing keys equality.
  415. using Base::key_eq;
  416. };
  417. // erase_if(flat_hash_set<>, Pred)
  418. //
  419. // Erases all elements that satisfy the predicate `pred` from the container `c`.
  420. // Returns the number of erased elements.
  421. template <typename T, typename H, typename E, typename A, typename Predicate>
  422. typename flat_hash_set<T, H, E, A>::size_type erase_if(
  423. flat_hash_set<T, H, E, A>& c, Predicate pred) {
  424. return container_internal::EraseIf(pred, &c);
  425. }
  426. namespace container_internal {
  427. template <class T>
  428. struct FlatHashSetPolicy {
  429. using slot_type = T;
  430. using key_type = T;
  431. using init_type = T;
  432. using constant_iterators = std::true_type;
  433. template <class Allocator, class... Args>
  434. static void construct(Allocator* alloc, slot_type* slot, Args&&... args) {
  435. y_absl::allocator_traits<Allocator>::construct(*alloc, slot,
  436. std::forward<Args>(args)...);
  437. }
  438. template <class Allocator>
  439. static void destroy(Allocator* alloc, slot_type* slot) {
  440. y_absl::allocator_traits<Allocator>::destroy(*alloc, slot);
  441. }
  442. static T& element(slot_type* slot) { return *slot; }
  443. template <class F, class... Args>
  444. static decltype(y_absl::container_internal::DecomposeValue(
  445. std::declval<F>(), std::declval<Args>()...))
  446. apply(F&& f, Args&&... args) {
  447. return y_absl::container_internal::DecomposeValue(
  448. std::forward<F>(f), std::forward<Args>(args)...);
  449. }
  450. static size_t space_used(const T*) { return 0; }
  451. };
  452. } // namespace container_internal
  453. namespace container_algorithm_internal {
  454. // Specialization of trait in y_absl/algorithm/container.h
  455. template <class Key, class Hash, class KeyEqual, class Allocator>
  456. struct IsUnorderedContainer<y_absl::flat_hash_set<Key, Hash, KeyEqual, Allocator>>
  457. : std::true_type {};
  458. } // namespace container_algorithm_internal
  459. Y_ABSL_NAMESPACE_END
  460. } // namespace y_absl
  461. #endif // Y_ABSL_CONTAINER_FLAT_HASH_SET_H_