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: node_hash_set.h
17 // -----------------------------------------------------------------------------
18 //
19 // An `absl::node_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, `node_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 table should be a map of type
28 // `flat_hash_map` or a set of type `flat_hash_set`. However, if you need
29 // pointer stability, a `node_hash_set` should be your preferred choice. As
30 // well, if you are migrating your code from using `std::unordered_set`, a
31 // `node_hash_set` should be an easy migration. Consider migrating to
32 // `node_hash_set` and perhaps converting to a more efficient `flat_hash_set`
33 // upon further review.
34 
35 #ifndef ABSL_CONTAINER_NODE_HASH_SET_H_
36 #define ABSL_CONTAINER_NODE_HASH_SET_H_
37 
38 #include <type_traits>
39 
40 #include "absl/algorithm/container.h"
41 #include "absl/base/macros.h"
42 #include "absl/container/internal/hash_function_defaults.h"  // IWYU pragma: export
43 #include "absl/container/internal/node_slot_policy.h"
44 #include "absl/container/internal/raw_hash_set.h"  // IWYU pragma: export
45 #include "absl/memory/memory.h"
46 
47 namespace absl {
48 ABSL_NAMESPACE_BEGIN
49 namespace container_internal {
50 template <typename T>
51 struct NodeHashSetPolicy;
52 }  // namespace container_internal
53 
54 // -----------------------------------------------------------------------------
55 // absl::node_hash_set
56 // -----------------------------------------------------------------------------
57 //
58 // An `absl::node_hash_set<T>` is an unordered associative container which
59 // has been optimized for both speed and memory footprint in most common use
60 // cases. Its interface is similar to that of `std::unordered_set<T>` with the
61 // following notable differences:
62 //
63 // * Supports heterogeneous lookup, through `find()`, `operator[]()` and
64 //   `insert()`, provided that the set is provided a compatible heterogeneous
65 //   hashing function and equality operator.
66 // * Contains a `capacity()` member function indicating the number of element
67 //   slots (open, deleted, and empty) within the hash set.
68 // * Returns `void` from the `erase(iterator)` overload.
69 //
70 // By default, `node_hash_set` uses the `absl::Hash` hashing framework.
71 // All fundamental and Abseil types that support the `absl::Hash` framework have
72 // a compatible equality operator for comparing insertions into `node_hash_set`.
73 // If your type is not yet supported by the `absl::Hash` framework, see
74 // absl/hash/hash.h for information on extending Abseil hashing to user-defined
75 // types.
76 //
77 // Using `absl::node_hash_set` at interface boundaries in dynamically loaded
78 // libraries (e.g. .dll, .so) is unsupported due to way `absl::Hash` values may
79 // be randomized across dynamically loaded libraries.
80 //
81 // Example:
82 //
83 //   // Create a node hash set of three strings
84 //   absl::node_hash_set<std::string> ducks =
85 //     {"huey", "dewey", "louie"};
86 //
87 //  // Insert a new element into the node hash set
88 //  ducks.insert("donald");
89 //
90 //  // Force a rehash of the node hash set
91 //  ducks.rehash(0);
92 //
93 //  // See if "dewey" is present
94 //  if (ducks.contains("dewey")) {
95 //    std::cout << "We found dewey!" << std::endl;
96 //  }
97 template <class T, class Hash = absl::container_internal::hash_default_hash<T>,
98           class Eq = absl::container_internal::hash_default_eq<T>,
99           class Alloc = std::allocator<T>>
100 class node_hash_set
101     : public absl::container_internal::raw_hash_set<
102           absl::container_internal::NodeHashSetPolicy<T>, Hash, Eq, Alloc> {
103   using Base = typename node_hash_set::raw_hash_set;
104 
105  public:
106   // Constructors and Assignment Operators
107   //
108   // A node_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   //    absl::node_hash_set<std::string> set1;
115   //
116   // * Initializer List constructor
117   //
118   //   absl::node_hash_set<std::string> set2 =
119   //       {{"huey"}, {"dewey"}, {"louie"}};
120   //
121   // * Copy constructor
122   //
123   //   absl::node_hash_set<std::string> set3(set2);
124   //
125   // * Copy assignment operator
126   //
127   //  // Hash functor and Comparator are copied as well
128   //  absl::node_hash_set<std::string> set4;
129   //  set4 = set3;
130   //
131   // * Move constructor
132   //
133   //   // Move is guaranteed efficient
134   //   absl::node_hash_set<std::string> set5(std::move(set4));
135   //
136   // * Move assignment operator
137   //
138   //   // May be efficient if allocators are compatible
139   //   absl::node_hash_set<std::string> set6;
140   //   set6 = std::move(set5);
141   //
142   // * Range constructor
143   //
144   //   std::vector<std::string> v = {"a", "b"};
145   //   absl::node_hash_set<std::string> set7(v.begin(), v.end());
node_hash_set()146   node_hash_set() {}
147   using Base::Base;
148 
149   // node_hash_set::begin()
150   //
151   // Returns an iterator to the beginning of the `node_hash_set`.
152   using Base::begin;
153 
154   // node_hash_set::cbegin()
155   //
156   // Returns a const iterator to the beginning of the `node_hash_set`.
157   using Base::cbegin;
158 
159   // node_hash_set::cend()
160   //
161   // Returns a const iterator to the end of the `node_hash_set`.
162   using Base::cend;
163 
164   // node_hash_set::end()
165   //
166   // Returns an iterator to the end of the `node_hash_set`.
167   using Base::end;
168 
169   // node_hash_set::capacity()
170   //
171   // Returns the number of element slots (assigned, deleted, and empty)
172   // available within the `node_hash_set`.
173   //
174   // NOTE: this member function is particular to `absl::node_hash_set` and is
175   // not provided in the `std::unordered_set` API.
176   using Base::capacity;
177 
178   // node_hash_set::empty()
179   //
180   // Returns whether or not the `node_hash_set` is empty.
181   using Base::empty;
182 
183   // node_hash_set::max_size()
184   //
185   // Returns the largest theoretical possible number of elements within a
186   // `node_hash_set` under current memory constraints. This value can be thought
187   // of the largest value of `std::distance(begin(), end())` for a
188   // `node_hash_set<T>`.
189   using Base::max_size;
190 
191   // node_hash_set::size()
192   //
193   // Returns the number of elements currently within the `node_hash_set`.
194   using Base::size;
195 
196   // node_hash_set::clear()
197   //
198   // Removes all elements from the `node_hash_set`. Invalidates any references,
199   // pointers, or iterators referring to contained elements.
200   //
201   // NOTE: this operation may shrink the underlying buffer. To avoid shrinking
202   // the underlying buffer call `erase(begin(), end())`.
203   using Base::clear;
204 
205   // node_hash_set::erase()
206   //
207   // Erases elements within the `node_hash_set`. Erasing does not trigger a
208   // rehash. Overloads are listed below.
209   //
210   // void erase(const_iterator pos):
211   //
212   //   Erases the element at `position` of the `node_hash_set`, returning
213   //   `void`.
214   //
215   //   NOTE: this return behavior is different than that of STL containers in
216   //   general and `std::unordered_set` in particular.
217   //
218   // iterator erase(const_iterator first, const_iterator last):
219   //
220   //   Erases the elements in the open interval [`first`, `last`), returning an
221   //   iterator pointing to `last`.
222   //
223   // size_type erase(const key_type& key):
224   //
225   //   Erases the element with the matching key, if it exists, returning the
226   //   number of elements erased (0 or 1).
227   using Base::erase;
228 
229   // node_hash_set::insert()
230   //
231   // Inserts an element of the specified value into the `node_hash_set`,
232   // returning an iterator pointing to the newly inserted element, provided that
233   // an element with the given key does not already exist. If rehashing occurs
234   // due to the insertion, all iterators are invalidated. Overloads are listed
235   // below.
236   //
237   // std::pair<iterator,bool> insert(const T& value):
238   //
239   //   Inserts a value into the `node_hash_set`. Returns a pair consisting of an
240   //   iterator to the inserted element (or to the element that prevented the
241   //   insertion) and a bool denoting whether the insertion took place.
242   //
243   // std::pair<iterator,bool> insert(T&& value):
244   //
245   //   Inserts a moveable value into the `node_hash_set`. Returns a pair
246   //   consisting of an iterator to the inserted element (or to the element that
247   //   prevented the insertion) and a bool denoting whether the insertion took
248   //   place.
249   //
250   // iterator insert(const_iterator hint, const T& value):
251   // iterator insert(const_iterator hint, T&& value):
252   //
253   //   Inserts a value, using the position of `hint` as a non-binding suggestion
254   //   for where to begin the insertion search. Returns an iterator to the
255   //   inserted element, or to the existing element that prevented the
256   //   insertion.
257   //
258   // void insert(InputIterator first, InputIterator last):
259   //
260   //   Inserts a range of values [`first`, `last`).
261   //
262   //   NOTE: Although the STL does not specify which element may be inserted if
263   //   multiple keys compare equivalently, for `node_hash_set` we guarantee the
264   //   first match is inserted.
265   //
266   // void insert(std::initializer_list<T> ilist):
267   //
268   //   Inserts the elements within the initializer list `ilist`.
269   //
270   //   NOTE: Although the STL does not specify which element may be inserted if
271   //   multiple keys compare equivalently within the initializer list, for
272   //   `node_hash_set` we guarantee the first match is inserted.
273   using Base::insert;
274 
275   // node_hash_set::emplace()
276   //
277   // Inserts an element of the specified value by constructing it in-place
278   // within the `node_hash_set`, provided that no element with the given key
279   // already exists.
280   //
281   // The element may be constructed even if there already is an element with the
282   // key in the container, in which case the newly constructed element will be
283   // destroyed immediately.
284   //
285   // If rehashing occurs due to the insertion, all iterators are invalidated.
286   using Base::emplace;
287 
288   // node_hash_set::emplace_hint()
289   //
290   // Inserts an element of the specified value by constructing it in-place
291   // within the `node_hash_set`, using the position of `hint` as a non-binding
292   // suggestion for where to begin the insertion search, and only inserts
293   // provided that no element with the given key already exists.
294   //
295   // The element may be constructed even if there already is an element with the
296   // key in the container, in which case the newly constructed element will be
297   // destroyed immediately.
298   //
299   // If rehashing occurs due to the insertion, all iterators are invalidated.
300   using Base::emplace_hint;
301 
302   // node_hash_set::extract()
303   //
304   // Extracts the indicated element, erasing it in the process, and returns it
305   // as a C++17-compatible node handle. Overloads are listed below.
306   //
307   // node_type extract(const_iterator position):
308   //
309   //   Extracts the element at the indicated position and returns a node handle
310   //   owning that extracted data.
311   //
312   // node_type extract(const key_type& x):
313   //
314   //   Extracts the element with the key matching the passed key value and
315   //   returns a node handle owning that extracted data. If the `node_hash_set`
316   //   does not contain an element with a matching key, this function returns an
317   // empty node handle.
318   using Base::extract;
319 
320   // node_hash_set::merge()
321   //
322   // Extracts elements from a given `source` node hash set into this
323   // `node_hash_set`. If the destination `node_hash_set` already contains an
324   // element with an equivalent key, that element is not extracted.
325   using Base::merge;
326 
327   // node_hash_set::swap(node_hash_set& other)
328   //
329   // Exchanges the contents of this `node_hash_set` with those of the `other`
330   // node hash set, avoiding invocation of any move, copy, or swap operations on
331   // individual elements.
332   //
333   // All iterators and references on the `node_hash_set` remain valid, excepting
334   // for the past-the-end iterator, which is invalidated.
335   //
336   // `swap()` requires that the node hash set's hashing and key equivalence
337   // functions be Swappable, and are exchaged using unqualified calls to
338   // non-member `swap()`. If the set's allocator has
339   // `std::allocator_traits<allocator_type>::propagate_on_container_swap::value`
340   // set to `true`, the allocators are also exchanged using an unqualified call
341   // to non-member `swap()`; otherwise, the allocators are not swapped.
342   using Base::swap;
343 
344   // node_hash_set::rehash(count)
345   //
346   // Rehashes the `node_hash_set`, setting the number of slots to be at least
347   // the passed value. If the new number of slots increases the load factor more
348   // than the current maximum load factor
349   // (`count` < `size()` / `max_load_factor()`), then the new number of slots
350   // will be at least `size()` / `max_load_factor()`.
351   //
352   // To force a rehash, pass rehash(0).
353   //
354   // NOTE: unlike behavior in `std::unordered_set`, references are also
355   // invalidated upon a `rehash()`.
356   using Base::rehash;
357 
358   // node_hash_set::reserve(count)
359   //
360   // Sets the number of slots in the `node_hash_set` to the number needed to
361   // accommodate at least `count` total elements without exceeding the current
362   // maximum load factor, and may rehash the container if needed.
363   using Base::reserve;
364 
365   // node_hash_set::contains()
366   //
367   // Determines whether an element comparing equal to the given `key` exists
368   // within the `node_hash_set`, returning `true` if so or `false` otherwise.
369   using Base::contains;
370 
371   // node_hash_set::count(const Key& key) const
372   //
373   // Returns the number of elements comparing equal to the given `key` within
374   // the `node_hash_set`. note that this function will return either `1` or `0`
375   // since duplicate elements are not allowed within a `node_hash_set`.
376   using Base::count;
377 
378   // node_hash_set::equal_range()
379   //
380   // Returns a closed range [first, last], defined by a `std::pair` of two
381   // iterators, containing all elements with the passed key in the
382   // `node_hash_set`.
383   using Base::equal_range;
384 
385   // node_hash_set::find()
386   //
387   // Finds an element with the passed `key` within the `node_hash_set`.
388   using Base::find;
389 
390   // node_hash_set::bucket_count()
391   //
392   // Returns the number of "buckets" within the `node_hash_set`. Note that
393   // because a node hash set contains all elements within its internal storage,
394   // this value simply equals the current capacity of the `node_hash_set`.
395   using Base::bucket_count;
396 
397   // node_hash_set::load_factor()
398   //
399   // Returns the current load factor of the `node_hash_set` (the average number
400   // of slots occupied with a value within the hash set).
401   using Base::load_factor;
402 
403   // node_hash_set::max_load_factor()
404   //
405   // Manages the maximum load factor of the `node_hash_set`. Overloads are
406   // listed below.
407   //
408   // float node_hash_set::max_load_factor()
409   //
410   //   Returns the current maximum load factor of the `node_hash_set`.
411   //
412   // void node_hash_set::max_load_factor(float ml)
413   //
414   //   Sets the maximum load factor of the `node_hash_set` to the passed value.
415   //
416   //   NOTE: This overload is provided only for API compatibility with the STL;
417   //   `node_hash_set` will ignore any set load factor and manage its rehashing
418   //   internally as an implementation detail.
419   using Base::max_load_factor;
420 
421   // node_hash_set::get_allocator()
422   //
423   // Returns the allocator function associated with this `node_hash_set`.
424   using Base::get_allocator;
425 
426   // node_hash_set::hash_function()
427   //
428   // Returns the hashing function used to hash the keys within this
429   // `node_hash_set`.
430   using Base::hash_function;
431 
432   // node_hash_set::key_eq()
433   //
434   // Returns the function used for comparing keys equality.
435   using Base::key_eq;
436 };
437 
438 // erase_if(node_hash_set<>, Pred)
439 //
440 // Erases all elements that satisfy the predicate `pred` from the container `c`.
441 // Returns the number of erased elements.
442 template <typename T, typename H, typename E, typename A, typename Predicate>
erase_if(node_hash_set<T,H,E,A> & c,Predicate pred)443 typename node_hash_set<T, H, E, A>::size_type erase_if(
444     node_hash_set<T, H, E, A>& c, Predicate pred) {
445   return container_internal::EraseIf(pred, &c);
446 }
447 
448 namespace container_internal {
449 
450 template <class T>
451 struct NodeHashSetPolicy
452     : absl::container_internal::node_slot_policy<T&, NodeHashSetPolicy<T>> {
453   using key_type = T;
454   using init_type = T;
455   using constant_iterators = std::true_type;
456 
457   template <class Allocator, class... Args>
new_elementNodeHashSetPolicy458   static T* new_element(Allocator* alloc, Args&&... args) {
459     using ValueAlloc =
460         typename absl::allocator_traits<Allocator>::template rebind_alloc<T>;
461     ValueAlloc value_alloc(*alloc);
462     T* res = absl::allocator_traits<ValueAlloc>::allocate(value_alloc, 1);
463     absl::allocator_traits<ValueAlloc>::construct(value_alloc, res,
464                                                   std::forward<Args>(args)...);
465     return res;
466   }
467 
468   template <class Allocator>
delete_elementNodeHashSetPolicy469   static void delete_element(Allocator* alloc, T* elem) {
470     using ValueAlloc =
471         typename absl::allocator_traits<Allocator>::template rebind_alloc<T>;
472     ValueAlloc value_alloc(*alloc);
473     absl::allocator_traits<ValueAlloc>::destroy(value_alloc, elem);
474     absl::allocator_traits<ValueAlloc>::deallocate(value_alloc, elem, 1);
475   }
476 
477   template <class F, class... Args>
decltypeNodeHashSetPolicy478   static decltype(absl::container_internal::DecomposeValue(
479       std::declval<F>(), std::declval<Args>()...))
480   apply(F&& f, Args&&... args) {
481     return absl::container_internal::DecomposeValue(
482         std::forward<F>(f), std::forward<Args>(args)...);
483   }
484 
element_space_usedNodeHashSetPolicy485   static size_t element_space_used(const T*) { return sizeof(T); }
486 };
487 }  // namespace container_internal
488 
489 namespace container_algorithm_internal {
490 
491 // Specialization of trait in absl/algorithm/container.h
492 template <class Key, class Hash, class KeyEqual, class Allocator>
493 struct IsUnorderedContainer<absl::node_hash_set<Key, Hash, KeyEqual, Allocator>>
494     : std::true_type {};
495 
496 }  // namespace container_algorithm_internal
497 ABSL_NAMESPACE_END
498 }  // namespace absl
499 
500 #endif  // ABSL_CONTAINER_NODE_HASH_SET_H_
501