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 `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 ABSL_CONTAINER_FLAT_HASH_SET_H_
30 #define ABSL_CONTAINER_FLAT_HASH_SET_H_
31
32 #include <type_traits>
33 #include <utility>
34
35 #include "absl/algorithm/container.h"
36 #include "absl/base/macros.h"
37 #include "absl/container/internal/container_memory.h"
38 #include "absl/container/internal/hash_function_defaults.h" // IWYU pragma: export
39 #include "absl/container/internal/raw_hash_set.h" // IWYU pragma: export
40 #include "absl/memory/memory.h"
41
42 namespace absl {
43 ABSL_NAMESPACE_BEGIN
44 namespace container_internal {
45 template <typename T>
46 struct FlatHashSetPolicy;
47 } // namespace container_internal
48
49 // -----------------------------------------------------------------------------
50 // absl::flat_hash_set
51 // -----------------------------------------------------------------------------
52 //
53 // An `absl::flat_hash_set<T>` is an unordered associative container which has
54 // been optimized for both speed and memory footprint in most common use cases.
55 // Its interface is similar to that of `std::unordered_set<T>` with the
56 // following notable differences:
57 //
58 // * Requires keys that are CopyConstructible
59 // * Supports heterogeneous lookup, through `find()` and `insert()`, provided
60 // that the set is provided a compatible heterogeneous hashing function and
61 // equality operator.
62 // * Invalidates any references and pointers to elements within the table after
63 // `rehash()`.
64 // * Contains a `capacity()` member function indicating the number of element
65 // slots (open, deleted, and empty) within the hash set.
66 // * Returns `void` from the `erase(iterator)` overload.
67 //
68 // By default, `flat_hash_set` uses the `absl::Hash` hashing framework. All
69 // fundamental and Abseil types that support the `absl::Hash` framework have a
70 // compatible equality operator for comparing insertions into `flat_hash_set`.
71 // If your type is not yet supported by the `absl::Hash` framework, see
72 // absl/hash/hash.h for information on extending Abseil hashing to user-defined
73 // types.
74 //
75 // Using `absl::flat_hash_set` at interface boundaries in dynamically loaded
76 // libraries (e.g. .dll, .so) is unsupported due to way `absl::Hash` values may
77 // be randomized across dynamically loaded libraries.
78 //
79 // NOTE: A `flat_hash_set` stores its keys directly inside its implementation
80 // array to avoid memory indirection. Because a `flat_hash_set` is designed to
81 // move data when rehashed, set keys will not retain pointer stability. If you
82 // require pointer stability, consider using
83 // `absl::flat_hash_set<std::unique_ptr<T>>`. If your type is not moveable and
84 // you require pointer stability, consider `absl::node_hash_set` instead.
85 //
86 // Example:
87 //
88 // // Create a flat hash set of three strings
89 // absl::flat_hash_set<std::string> ducks =
90 // {"huey", "dewey", "louie"};
91 //
92 // // Insert a new element into the flat hash set
93 // ducks.insert("donald");
94 //
95 // // Force a rehash of the flat hash set
96 // ducks.rehash(0);
97 //
98 // // See if "dewey" is present
99 // if (ducks.contains("dewey")) {
100 // std::cout << "We found dewey!" << std::endl;
101 // }
102 template <class T, class Hash = absl::container_internal::hash_default_hash<T>,
103 class Eq = absl::container_internal::hash_default_eq<T>,
104 class Allocator = std::allocator<T>>
105 class flat_hash_set
106 : public absl::container_internal::raw_hash_set<
107 absl::container_internal::FlatHashSetPolicy<T>, Hash, Eq, Allocator> {
108 using Base = typename flat_hash_set::raw_hash_set;
109
110 public:
111 // Constructors and Assignment Operators
112 //
113 // A flat_hash_set supports the same overload set as `std::unordered_set`
114 // for construction and assignment:
115 //
116 // * Default constructor
117 //
118 // // No allocation for the table's elements is made.
119 // absl::flat_hash_set<std::string> set1;
120 //
121 // * Initializer List constructor
122 //
123 // absl::flat_hash_set<std::string> set2 =
124 // {{"huey"}, {"dewey"}, {"louie"},};
125 //
126 // * Copy constructor
127 //
128 // absl::flat_hash_set<std::string> set3(set2);
129 //
130 // * Copy assignment operator
131 //
132 // // Hash functor and Comparator are copied as well
133 // absl::flat_hash_set<std::string> set4;
134 // set4 = set3;
135 //
136 // * Move constructor
137 //
138 // // Move is guaranteed efficient
139 // absl::flat_hash_set<std::string> set5(std::move(set4));
140 //
141 // * Move assignment operator
142 //
143 // // May be efficient if allocators are compatible
144 // absl::flat_hash_set<std::string> set6;
145 // set6 = std::move(set5);
146 //
147 // * Range constructor
148 //
149 // std::vector<std::string> v = {"a", "b"};
150 // absl::flat_hash_set<std::string> set7(v.begin(), v.end());
flat_hash_set()151 flat_hash_set() {}
152 using Base::Base;
153
154 // flat_hash_set::begin()
155 //
156 // Returns an iterator to the beginning of the `flat_hash_set`.
157 using Base::begin;
158
159 // flat_hash_set::cbegin()
160 //
161 // Returns a const iterator to the beginning of the `flat_hash_set`.
162 using Base::cbegin;
163
164 // flat_hash_set::cend()
165 //
166 // Returns a const iterator to the end of the `flat_hash_set`.
167 using Base::cend;
168
169 // flat_hash_set::end()
170 //
171 // Returns an iterator to the end of the `flat_hash_set`.
172 using Base::end;
173
174 // flat_hash_set::capacity()
175 //
176 // Returns the number of element slots (assigned, deleted, and empty)
177 // available within the `flat_hash_set`.
178 //
179 // NOTE: this member function is particular to `absl::flat_hash_set` and is
180 // not provided in the `std::unordered_set` API.
181 using Base::capacity;
182
183 // flat_hash_set::empty()
184 //
185 // Returns whether or not the `flat_hash_set` is empty.
186 using Base::empty;
187
188 // flat_hash_set::max_size()
189 //
190 // Returns the largest theoretical possible number of elements within a
191 // `flat_hash_set` under current memory constraints. This value can be thought
192 // of the largest value of `std::distance(begin(), end())` for a
193 // `flat_hash_set<T>`.
194 using Base::max_size;
195
196 // flat_hash_set::size()
197 //
198 // Returns the number of elements currently within the `flat_hash_set`.
199 using Base::size;
200
201 // flat_hash_set::clear()
202 //
203 // Removes all elements from the `flat_hash_set`. Invalidates any references,
204 // pointers, or iterators referring to contained elements.
205 //
206 // NOTE: this operation may shrink the underlying buffer. To avoid shrinking
207 // the underlying buffer call `erase(begin(), end())`.
208 using Base::clear;
209
210 // flat_hash_set::erase()
211 //
212 // Erases elements within the `flat_hash_set`. Erasing does not trigger a
213 // rehash. Overloads are listed below.
214 //
215 // void erase(const_iterator pos):
216 //
217 // Erases the element at `position` of the `flat_hash_set`, returning
218 // `void`.
219 //
220 // NOTE: returning `void` in this case is different than that of STL
221 // containers in general and `std::unordered_set` in particular (which
222 // return an iterator to the element following the erased element). If that
223 // iterator is needed, simply post increment the iterator:
224 //
225 // set.erase(it++);
226 //
227 // iterator erase(const_iterator first, const_iterator last):
228 //
229 // Erases the elements in the open interval [`first`, `last`), returning an
230 // iterator pointing to `last`.
231 //
232 // size_type erase(const key_type& key):
233 //
234 // Erases the element with the matching key, if it exists, returning the
235 // number of elements erased (0 or 1).
236 using Base::erase;
237
238 // flat_hash_set::insert()
239 //
240 // Inserts an element of the specified value into the `flat_hash_set`,
241 // returning an iterator pointing to the newly inserted element, provided that
242 // an element with the given key does not already exist. If rehashing occurs
243 // due to the insertion, all iterators are invalidated. Overloads are listed
244 // below.
245 //
246 // std::pair<iterator,bool> insert(const T& value):
247 //
248 // Inserts a value into the `flat_hash_set`. Returns a pair consisting of an
249 // iterator to the inserted element (or to the element that prevented the
250 // insertion) and a bool denoting whether the insertion took place.
251 //
252 // std::pair<iterator,bool> insert(T&& value):
253 //
254 // Inserts a moveable value into the `flat_hash_set`. Returns a pair
255 // consisting of an iterator to the inserted element (or to the element that
256 // prevented the insertion) and a bool denoting whether the insertion took
257 // place.
258 //
259 // iterator insert(const_iterator hint, const T& value):
260 // iterator insert(const_iterator hint, T&& value):
261 //
262 // Inserts a value, using the position of `hint` as a non-binding suggestion
263 // for where to begin the insertion search. Returns an iterator to the
264 // inserted element, or to the existing element that prevented the
265 // insertion.
266 //
267 // void insert(InputIterator first, InputIterator last):
268 //
269 // Inserts a range of values [`first`, `last`).
270 //
271 // NOTE: Although the STL does not specify which element may be inserted if
272 // multiple keys compare equivalently, for `flat_hash_set` we guarantee the
273 // first match is inserted.
274 //
275 // void insert(std::initializer_list<T> ilist):
276 //
277 // Inserts the elements within the initializer list `ilist`.
278 //
279 // NOTE: Although the STL does not specify which element may be inserted if
280 // multiple keys compare equivalently within the initializer list, for
281 // `flat_hash_set` we guarantee the first match is inserted.
282 using Base::insert;
283
284 // flat_hash_set::emplace()
285 //
286 // Inserts an element of the specified value by constructing it in-place
287 // within the `flat_hash_set`, provided that no element with the given key
288 // 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;
296
297 // flat_hash_set::emplace_hint()
298 //
299 // Inserts an element of the specified value by constructing it in-place
300 // within the `flat_hash_set`, using the position of `hint` as a non-binding
301 // suggestion for where to begin the insertion search, and only inserts
302 // provided that no element with the given key already exists.
303 //
304 // The element may be constructed even if there already is an element with the
305 // key in the container, in which case the newly constructed element will be
306 // destroyed immediately.
307 //
308 // If rehashing occurs due to the insertion, all iterators are invalidated.
309 using Base::emplace_hint;
310
311 // flat_hash_set::extract()
312 //
313 // Extracts the indicated element, erasing it in the process, and returns it
314 // as a C++17-compatible node handle. Overloads are listed below.
315 //
316 // node_type extract(const_iterator position):
317 //
318 // Extracts the element at the indicated position and returns a node handle
319 // owning that extracted data.
320 //
321 // node_type extract(const key_type& x):
322 //
323 // Extracts the element with the key matching the passed key value and
324 // returns a node handle owning that extracted data. If the `flat_hash_set`
325 // does not contain an element with a matching key, this function returns an
326 // empty node handle.
327 using Base::extract;
328
329 // flat_hash_set::merge()
330 //
331 // Extracts elements from a given `source` flat hash set into this
332 // `flat_hash_set`. If the destination `flat_hash_set` already contains an
333 // element with an equivalent key, that element is not extracted.
334 using Base::merge;
335
336 // flat_hash_set::swap(flat_hash_set& other)
337 //
338 // Exchanges the contents of this `flat_hash_set` with those of the `other`
339 // flat hash set, avoiding invocation of any move, copy, or swap operations on
340 // individual elements.
341 //
342 // All iterators and references on the `flat_hash_set` remain valid, excepting
343 // for the past-the-end iterator, which is invalidated.
344 //
345 // `swap()` requires that the flat hash set's hashing and key equivalence
346 // functions be Swappable, and are exchaged using unqualified calls to
347 // non-member `swap()`. If the set's allocator has
348 // `std::allocator_traits<allocator_type>::propagate_on_container_swap::value`
349 // set to `true`, the allocators are also exchanged using an unqualified call
350 // to non-member `swap()`; otherwise, the allocators are not swapped.
351 using Base::swap;
352
353 // flat_hash_set::rehash(count)
354 //
355 // Rehashes the `flat_hash_set`, setting the number of slots to be at least
356 // the passed value. If the new number of slots increases the load factor more
357 // than the current maximum load factor
358 // (`count` < `size()` / `max_load_factor()`), then the new number of slots
359 // will be at least `size()` / `max_load_factor()`.
360 //
361 // To force a rehash, pass rehash(0).
362 //
363 // NOTE: unlike behavior in `std::unordered_set`, references are also
364 // invalidated upon a `rehash()`.
365 using Base::rehash;
366
367 // flat_hash_set::reserve(count)
368 //
369 // Sets the number of slots in the `flat_hash_set` to the number needed to
370 // accommodate at least `count` total elements without exceeding the current
371 // maximum load factor, and may rehash the container if needed.
372 using Base::reserve;
373
374 // flat_hash_set::contains()
375 //
376 // Determines whether an element comparing equal to the given `key` exists
377 // within the `flat_hash_set`, returning `true` if so or `false` otherwise.
378 using Base::contains;
379
380 // flat_hash_set::count(const Key& key) const
381 //
382 // Returns the number of elements comparing equal to the given `key` within
383 // the `flat_hash_set`. note that this function will return either `1` or `0`
384 // since duplicate elements are not allowed within a `flat_hash_set`.
385 using Base::count;
386
387 // flat_hash_set::equal_range()
388 //
389 // Returns a closed range [first, last], defined by a `std::pair` of two
390 // iterators, containing all elements with the passed key in the
391 // `flat_hash_set`.
392 using Base::equal_range;
393
394 // flat_hash_set::find()
395 //
396 // Finds an element with the passed `key` within the `flat_hash_set`.
397 using Base::find;
398
399 // flat_hash_set::bucket_count()
400 //
401 // Returns the number of "buckets" within the `flat_hash_set`. Note that
402 // because a flat hash set contains all elements within its internal storage,
403 // this value simply equals the current capacity of the `flat_hash_set`.
404 using Base::bucket_count;
405
406 // flat_hash_set::load_factor()
407 //
408 // Returns the current load factor of the `flat_hash_set` (the average number
409 // of slots occupied with a value within the hash set).
410 using Base::load_factor;
411
412 // flat_hash_set::max_load_factor()
413 //
414 // Manages the maximum load factor of the `flat_hash_set`. Overloads are
415 // listed below.
416 //
417 // float flat_hash_set::max_load_factor()
418 //
419 // Returns the current maximum load factor of the `flat_hash_set`.
420 //
421 // void flat_hash_set::max_load_factor(float ml)
422 //
423 // Sets the maximum load factor of the `flat_hash_set` to the passed value.
424 //
425 // NOTE: This overload is provided only for API compatibility with the STL;
426 // `flat_hash_set` will ignore any set load factor and manage its rehashing
427 // internally as an implementation detail.
428 using Base::max_load_factor;
429
430 // flat_hash_set::get_allocator()
431 //
432 // Returns the allocator function associated with this `flat_hash_set`.
433 using Base::get_allocator;
434
435 // flat_hash_set::hash_function()
436 //
437 // Returns the hashing function used to hash the keys within this
438 // `flat_hash_set`.
439 using Base::hash_function;
440
441 // flat_hash_set::key_eq()
442 //
443 // Returns the function used for comparing keys equality.
444 using Base::key_eq;
445 };
446
447 // erase_if(flat_hash_set<>, Pred)
448 //
449 // Erases all elements that satisfy the predicate `pred` from the container `c`.
450 // Returns the number of erased elements.
451 template <typename T, typename H, typename E, typename A, typename Predicate>
erase_if(flat_hash_set<T,H,E,A> & c,Predicate pred)452 typename flat_hash_set<T, H, E, A>::size_type erase_if(
453 flat_hash_set<T, H, E, A>& c, Predicate pred) {
454 return container_internal::EraseIf(pred, &c);
455 }
456
457 namespace container_internal {
458
459 template <class T>
460 struct FlatHashSetPolicy {
461 using slot_type = T;
462 using key_type = T;
463 using init_type = T;
464 using constant_iterators = std::true_type;
465
466 template <class Allocator, class... Args>
constructFlatHashSetPolicy467 static void construct(Allocator* alloc, slot_type* slot, Args&&... args) {
468 absl::allocator_traits<Allocator>::construct(*alloc, slot,
469 std::forward<Args>(args)...);
470 }
471
472 template <class Allocator>
destroyFlatHashSetPolicy473 static void destroy(Allocator* alloc, slot_type* slot) {
474 absl::allocator_traits<Allocator>::destroy(*alloc, slot);
475 }
476
elementFlatHashSetPolicy477 static T& element(slot_type* slot) { return *slot; }
478
479 template <class F, class... Args>
decltypeFlatHashSetPolicy480 static decltype(absl::container_internal::DecomposeValue(
481 std::declval<F>(), std::declval<Args>()...))
482 apply(F&& f, Args&&... args) {
483 return absl::container_internal::DecomposeValue(
484 std::forward<F>(f), std::forward<Args>(args)...);
485 }
486
space_usedFlatHashSetPolicy487 static size_t space_used(const T*) { return 0; }
488 };
489 } // namespace container_internal
490
491 namespace container_algorithm_internal {
492
493 // Specialization of trait in absl/algorithm/container.h
494 template <class Key, class Hash, class KeyEqual, class Allocator>
495 struct IsUnorderedContainer<absl::flat_hash_set<Key, Hash, KeyEqual, Allocator>>
496 : std::true_type {};
497
498 } // namespace container_algorithm_internal
499
500 ABSL_NAMESPACE_END
501 } // namespace absl
502
503 #endif // ABSL_CONTAINER_FLAT_HASH_SET_H_
504