c44657f556
-- dc6d2715f0415082fcc8da8bf74e74bce69b236c by Derek Mauro <dmauro@google.com>: Correctly detect C++ exceptions support on Clang for Windows PiperOrigin-RevId: 294905116 -- b43c44501b4820f4a2f396e426619bd02565707e by Derek Mauro <dmauro@google.com>: Set CMAKE_CXX_STANDARD on the MacOS CMake build PiperOrigin-RevId: 294730418 -- 184a078649167f9738da60b0f12108256bcfd67b by Abseil Team <absl-team@google.com>: No need for custom spec to deal with limited platforms. PiperOrigin-RevId: 294700133 -- b437c7f659b809fc84a45eab284265fec497a3e3 by Abseil Team <absl-team@google.com>: Not calling sigaltstack on WatchOS and TVOS since they don't allow it. PiperOrigin-RevId: 294699951 -- 23ab8dd381ee4104125dece8455bc96b81239789 by Gennadiy Rozental <rogeeff@google.com>: Replace use of atomic+global Mutex+bool with absl::call_once for Flag initialization. This simplifies the initialization logic and helps with upcoming work with value storage rework. PiperOrigin-RevId: 294654938 -- cee576163a2753c6138bc254e81de4800ea3307a by Gennadiy Rozental <rogeeff@google.com>: Separate const bits from mutable bits. Since bit field is not atomic unit for reading/writing, we can't have constant bits which are not protected by data guard to share the space with mutable bits which are protected. This CL just reorder fields in class and does not make any other changes. PiperOrigin-RevId: 294501780 -- b4d0e2ab559d04f655c93f008594562234773c15 by Abseil Team <absl-team@google.com>: Correct the comment. PiperOrigin-RevId: 294499328 -- a788cf71af6247df033298c49939ba0414d71693 by Derek Mauro <dmauro@google.com>: Move the FAQ to the top level directory PiperOrigin-RevId: 294493863 GitOrigin-RevId: dc6d2715f0415082fcc8da8bf74e74bce69b236c Change-Id: I71b0d8cd401b48d41433417858ae0d69398b6602
498 lines
18 KiB
C++
498 lines
18 KiB
C++
// Copyright 2018 The Abseil Authors.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// https://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
//
|
|
// -----------------------------------------------------------------------------
|
|
// File: node_hash_set.h
|
|
// -----------------------------------------------------------------------------
|
|
//
|
|
// An `absl::node_hash_set<T>` is an unordered associative container designed to
|
|
// be a more efficient replacement for `std::unordered_set`. Like
|
|
// `unordered_set`, search, insertion, and deletion of map elements can be done
|
|
// as an `O(1)` operation. However, `node_hash_set` (and other unordered
|
|
// associative containers known as the collection of Abseil "Swiss tables")
|
|
// contain other optimizations that result in both memory and computation
|
|
// advantages.
|
|
//
|
|
// In most cases, your default choice for a hash table should be a map of type
|
|
// `flat_hash_map` or a set of type `flat_hash_set`. However, if you need
|
|
// pointer stability, a `node_hash_set` should be your preferred choice. As
|
|
// well, if you are migrating your code from using `std::unordered_set`, a
|
|
// `node_hash_set` should be an easy migration. Consider migrating to
|
|
// `node_hash_set` and perhaps converting to a more efficient `flat_hash_set`
|
|
// upon further review.
|
|
|
|
#ifndef ABSL_CONTAINER_NODE_HASH_SET_H_
|
|
#define ABSL_CONTAINER_NODE_HASH_SET_H_
|
|
|
|
#include <type_traits>
|
|
|
|
#include "absl/algorithm/container.h"
|
|
#include "absl/container/internal/hash_function_defaults.h" // IWYU pragma: export
|
|
#include "absl/container/internal/node_hash_policy.h"
|
|
#include "absl/container/internal/raw_hash_set.h" // IWYU pragma: export
|
|
#include "absl/memory/memory.h"
|
|
|
|
namespace absl {
|
|
ABSL_NAMESPACE_BEGIN
|
|
namespace container_internal {
|
|
template <typename T>
|
|
struct NodeHashSetPolicy;
|
|
} // namespace container_internal
|
|
|
|
// -----------------------------------------------------------------------------
|
|
// absl::node_hash_set
|
|
// -----------------------------------------------------------------------------
|
|
//
|
|
// An `absl::node_hash_set<T>` is an unordered associative container which
|
|
// has been optimized for both speed and memory footprint in most common use
|
|
// cases. Its interface is similar to that of `std::unordered_set<T>` with the
|
|
// following notable differences:
|
|
//
|
|
// * Supports heterogeneous lookup, through `find()`, `operator[]()` and
|
|
// `insert()`, provided that the map is provided a compatible heterogeneous
|
|
// hashing function and equality operator.
|
|
// * Contains a `capacity()` member function indicating the number of element
|
|
// slots (open, deleted, and empty) within the hash set.
|
|
// * Returns `void` from the `erase(iterator)` overload.
|
|
//
|
|
// By default, `node_hash_set` uses the `absl::Hash` hashing framework.
|
|
// All fundamental and Abseil types that support the `absl::Hash` framework have
|
|
// a compatible equality operator for comparing insertions into `node_hash_set`.
|
|
// If your type is not yet supported by the `absl::Hash` framework, see
|
|
// absl/hash/hash.h for information on extending Abseil hashing to user-defined
|
|
// types.
|
|
//
|
|
// Example:
|
|
//
|
|
// // Create a node hash set of three strings
|
|
// absl::node_hash_map<std::string, std::string> ducks =
|
|
// {"huey", "dewey", "louie"};
|
|
//
|
|
// // Insert a new element into the node hash map
|
|
// ducks.insert("donald"};
|
|
//
|
|
// // Force a rehash of the node hash map
|
|
// ducks.rehash(0);
|
|
//
|
|
// // See if "dewey" is present
|
|
// if (ducks.contains("dewey")) {
|
|
// std::cout << "We found dewey!" << std::endl;
|
|
// }
|
|
template <class T, class Hash = absl::container_internal::hash_default_hash<T>,
|
|
class Eq = absl::container_internal::hash_default_eq<T>,
|
|
class Alloc = std::allocator<T>>
|
|
class node_hash_set
|
|
: public absl::container_internal::raw_hash_set<
|
|
absl::container_internal::NodeHashSetPolicy<T>, Hash, Eq, Alloc> {
|
|
using Base = typename node_hash_set::raw_hash_set;
|
|
|
|
public:
|
|
// Constructors and Assignment Operators
|
|
//
|
|
// A node_hash_set supports the same overload set as `std::unordered_map`
|
|
// for construction and assignment:
|
|
//
|
|
// * Default constructor
|
|
//
|
|
// // No allocation for the table's elements is made.
|
|
// absl::node_hash_set<std::string> set1;
|
|
//
|
|
// * Initializer List constructor
|
|
//
|
|
// absl::node_hash_set<std::string> set2 =
|
|
// {{"huey"}, {"dewey"}, {"louie"}};
|
|
//
|
|
// * Copy constructor
|
|
//
|
|
// absl::node_hash_set<std::string> set3(set2);
|
|
//
|
|
// * Copy assignment operator
|
|
//
|
|
// // Hash functor and Comparator are copied as well
|
|
// absl::node_hash_set<std::string> set4;
|
|
// set4 = set3;
|
|
//
|
|
// * Move constructor
|
|
//
|
|
// // Move is guaranteed efficient
|
|
// absl::node_hash_set<std::string> set5(std::move(set4));
|
|
//
|
|
// * Move assignment operator
|
|
//
|
|
// // May be efficient if allocators are compatible
|
|
// absl::node_hash_set<std::string> set6;
|
|
// set6 = std::move(set5);
|
|
//
|
|
// * Range constructor
|
|
//
|
|
// std::vector<std::string> v = {"a", "b"};
|
|
// absl::node_hash_set<std::string> set7(v.begin(), v.end());
|
|
node_hash_set() {}
|
|
using Base::Base;
|
|
|
|
// node_hash_set::begin()
|
|
//
|
|
// Returns an iterator to the beginning of the `node_hash_set`.
|
|
using Base::begin;
|
|
|
|
// node_hash_set::cbegin()
|
|
//
|
|
// Returns a const iterator to the beginning of the `node_hash_set`.
|
|
using Base::cbegin;
|
|
|
|
// node_hash_set::cend()
|
|
//
|
|
// Returns a const iterator to the end of the `node_hash_set`.
|
|
using Base::cend;
|
|
|
|
// node_hash_set::end()
|
|
//
|
|
// Returns an iterator to the end of the `node_hash_set`.
|
|
using Base::end;
|
|
|
|
// node_hash_set::capacity()
|
|
//
|
|
// Returns the number of element slots (assigned, deleted, and empty)
|
|
// available within the `node_hash_set`.
|
|
//
|
|
// NOTE: this member function is particular to `absl::node_hash_set` and is
|
|
// not provided in the `std::unordered_map` API.
|
|
using Base::capacity;
|
|
|
|
// node_hash_set::empty()
|
|
//
|
|
// Returns whether or not the `node_hash_set` is empty.
|
|
using Base::empty;
|
|
|
|
// node_hash_set::max_size()
|
|
//
|
|
// Returns the largest theoretical possible number of elements within a
|
|
// `node_hash_set` under current memory constraints. This value can be thought
|
|
// of the largest value of `std::distance(begin(), end())` for a
|
|
// `node_hash_set<T>`.
|
|
using Base::max_size;
|
|
|
|
// node_hash_set::size()
|
|
//
|
|
// Returns the number of elements currently within the `node_hash_set`.
|
|
using Base::size;
|
|
|
|
// node_hash_set::clear()
|
|
//
|
|
// Removes all elements from the `node_hash_set`. Invalidates any references,
|
|
// pointers, or iterators referring to contained elements.
|
|
//
|
|
// NOTE: this operation may shrink the underlying buffer. To avoid shrinking
|
|
// the underlying buffer call `erase(begin(), end())`.
|
|
using Base::clear;
|
|
|
|
// node_hash_set::erase()
|
|
//
|
|
// Erases elements within the `node_hash_set`. Erasing does not trigger a
|
|
// rehash. Overloads are listed below.
|
|
//
|
|
// void erase(const_iterator pos):
|
|
//
|
|
// Erases the element at `position` of the `node_hash_set`, returning
|
|
// `void`.
|
|
//
|
|
// NOTE: this return behavior is different than that of STL containers in
|
|
// general and `std::unordered_map` in particular.
|
|
//
|
|
// iterator erase(const_iterator first, const_iterator last):
|
|
//
|
|
// Erases the elements in the open interval [`first`, `last`), returning an
|
|
// iterator pointing to `last`.
|
|
//
|
|
// size_type erase(const key_type& key):
|
|
//
|
|
// Erases the element with the matching key, if it exists.
|
|
using Base::erase;
|
|
|
|
// node_hash_set::insert()
|
|
//
|
|
// Inserts an element of the specified value into the `node_hash_set`,
|
|
// returning an iterator pointing to the newly inserted element, provided that
|
|
// an element with the given key does not already exist. If rehashing occurs
|
|
// due to the insertion, all iterators are invalidated. Overloads are listed
|
|
// below.
|
|
//
|
|
// std::pair<iterator,bool> insert(const T& value):
|
|
//
|
|
// Inserts a value into the `node_hash_set`. Returns a pair consisting of an
|
|
// iterator to the inserted element (or to the element that prevented the
|
|
// insertion) and a bool denoting whether the insertion took place.
|
|
//
|
|
// std::pair<iterator,bool> insert(T&& value):
|
|
//
|
|
// Inserts a moveable value into the `node_hash_set`. Returns a pair
|
|
// consisting of an iterator to the inserted element (or to the element that
|
|
// prevented the insertion) and a bool denoting whether the insertion took
|
|
// place.
|
|
//
|
|
// iterator insert(const_iterator hint, const T& value):
|
|
// iterator insert(const_iterator hint, T&& value):
|
|
//
|
|
// Inserts a value, using the position of `hint` as a non-binding suggestion
|
|
// for where to begin the insertion search. Returns an iterator to the
|
|
// inserted element, or to the existing element that prevented the
|
|
// insertion.
|
|
//
|
|
// void insert(InputIterator first, InputIterator last):
|
|
//
|
|
// Inserts a range of values [`first`, `last`).
|
|
//
|
|
// NOTE: Although the STL does not specify which element may be inserted if
|
|
// multiple keys compare equivalently, for `node_hash_set` we guarantee the
|
|
// first match is inserted.
|
|
//
|
|
// void insert(std::initializer_list<T> ilist):
|
|
//
|
|
// Inserts the elements within the initializer list `ilist`.
|
|
//
|
|
// NOTE: Although the STL does not specify which element may be inserted if
|
|
// multiple keys compare equivalently within the initializer list, for
|
|
// `node_hash_set` we guarantee the first match is inserted.
|
|
using Base::insert;
|
|
|
|
// node_hash_set::emplace()
|
|
//
|
|
// Inserts an element of the specified value by constructing it in-place
|
|
// within the `node_hash_set`, provided that no element with the given key
|
|
// already exists.
|
|
//
|
|
// The element may be constructed even if there already is an element with the
|
|
// key in the container, in which case the newly constructed element will be
|
|
// destroyed immediately.
|
|
//
|
|
// If rehashing occurs due to the insertion, all iterators are invalidated.
|
|
using Base::emplace;
|
|
|
|
// node_hash_set::emplace_hint()
|
|
//
|
|
// Inserts an element of the specified value by constructing it in-place
|
|
// within the `node_hash_set`, using the position of `hint` as a non-binding
|
|
// suggestion for where to begin the insertion search, and only inserts
|
|
// provided that no element with the given key already exists.
|
|
//
|
|
// The element may be constructed even if there already is an element with the
|
|
// key in the container, in which case the newly constructed element will be
|
|
// destroyed immediately.
|
|
//
|
|
// If rehashing occurs due to the insertion, all iterators are invalidated.
|
|
using Base::emplace_hint;
|
|
|
|
// node_hash_set::extract()
|
|
//
|
|
// Extracts the indicated element, erasing it in the process, and returns it
|
|
// as a C++17-compatible node handle. Overloads are listed below.
|
|
//
|
|
// node_type extract(const_iterator position):
|
|
//
|
|
// Extracts the element at the indicated position and returns a node handle
|
|
// owning that extracted data.
|
|
//
|
|
// node_type extract(const key_type& x):
|
|
//
|
|
// Extracts the element with the key matching the passed key value and
|
|
// returns a node handle owning that extracted data. If the `node_hash_set`
|
|
// does not contain an element with a matching key, this function returns an
|
|
// empty node handle.
|
|
using Base::extract;
|
|
|
|
// node_hash_set::merge()
|
|
//
|
|
// Extracts elements from a given `source` flat hash map into this
|
|
// `node_hash_set`. If the destination `node_hash_set` already contains an
|
|
// element with an equivalent key, that element is not extracted.
|
|
using Base::merge;
|
|
|
|
// node_hash_set::swap(node_hash_set& other)
|
|
//
|
|
// Exchanges the contents of this `node_hash_set` with those of the `other`
|
|
// flat hash map, avoiding invocation of any move, copy, or swap operations on
|
|
// individual elements.
|
|
//
|
|
// All iterators and references on the `node_hash_set` remain valid, excepting
|
|
// for the past-the-end iterator, which is invalidated.
|
|
//
|
|
// `swap()` requires that the flat hash set's hashing and key equivalence
|
|
// functions be Swappable, and are exchaged using unqualified calls to
|
|
// non-member `swap()`. If the map's allocator has
|
|
// `std::allocator_traits<allocator_type>::propagate_on_container_swap::value`
|
|
// set to `true`, the allocators are also exchanged using an unqualified call
|
|
// to non-member `swap()`; otherwise, the allocators are not swapped.
|
|
using Base::swap;
|
|
|
|
// node_hash_set::rehash(count)
|
|
//
|
|
// Rehashes the `node_hash_set`, setting the number of slots to be at least
|
|
// the passed value. If the new number of slots increases the load factor more
|
|
// than the current maximum load factor
|
|
// (`count` < `size()` / `max_load_factor()`), then the new number of slots
|
|
// will be at least `size()` / `max_load_factor()`.
|
|
//
|
|
// To force a rehash, pass rehash(0).
|
|
//
|
|
// NOTE: unlike behavior in `std::unordered_set`, references are also
|
|
// invalidated upon a `rehash()`.
|
|
using Base::rehash;
|
|
|
|
// node_hash_set::reserve(count)
|
|
//
|
|
// Sets the number of slots in the `node_hash_set` to the number needed to
|
|
// accommodate at least `count` total elements without exceeding the current
|
|
// maximum load factor, and may rehash the container if needed.
|
|
using Base::reserve;
|
|
|
|
// node_hash_set::contains()
|
|
//
|
|
// Determines whether an element comparing equal to the given `key` exists
|
|
// within the `node_hash_set`, returning `true` if so or `false` otherwise.
|
|
using Base::contains;
|
|
|
|
// node_hash_set::count(const Key& key) const
|
|
//
|
|
// Returns the number of elements comparing equal to the given `key` within
|
|
// the `node_hash_set`. note that this function will return either `1` or `0`
|
|
// since duplicate elements are not allowed within a `node_hash_set`.
|
|
using Base::count;
|
|
|
|
// node_hash_set::equal_range()
|
|
//
|
|
// Returns a closed range [first, last], defined by a `std::pair` of two
|
|
// iterators, containing all elements with the passed key in the
|
|
// `node_hash_set`.
|
|
using Base::equal_range;
|
|
|
|
// node_hash_set::find()
|
|
//
|
|
// Finds an element with the passed `key` within the `node_hash_set`.
|
|
using Base::find;
|
|
|
|
// node_hash_set::bucket_count()
|
|
//
|
|
// Returns the number of "buckets" within the `node_hash_set`. Note that
|
|
// because a flat hash map contains all elements within its internal storage,
|
|
// this value simply equals the current capacity of the `node_hash_set`.
|
|
using Base::bucket_count;
|
|
|
|
// node_hash_set::load_factor()
|
|
//
|
|
// Returns the current load factor of the `node_hash_set` (the average number
|
|
// of slots occupied with a value within the hash map).
|
|
using Base::load_factor;
|
|
|
|
// node_hash_set::max_load_factor()
|
|
//
|
|
// Manages the maximum load factor of the `node_hash_set`. Overloads are
|
|
// listed below.
|
|
//
|
|
// float node_hash_set::max_load_factor()
|
|
//
|
|
// Returns the current maximum load factor of the `node_hash_set`.
|
|
//
|
|
// void node_hash_set::max_load_factor(float ml)
|
|
//
|
|
// Sets the maximum load factor of the `node_hash_set` to the passed value.
|
|
//
|
|
// NOTE: This overload is provided only for API compatibility with the STL;
|
|
// `node_hash_set` will ignore any set load factor and manage its rehashing
|
|
// internally as an implementation detail.
|
|
using Base::max_load_factor;
|
|
|
|
// node_hash_set::get_allocator()
|
|
//
|
|
// Returns the allocator function associated with this `node_hash_set`.
|
|
using Base::get_allocator;
|
|
|
|
// node_hash_set::hash_function()
|
|
//
|
|
// Returns the hashing function used to hash the keys within this
|
|
// `node_hash_set`.
|
|
using Base::hash_function;
|
|
|
|
// node_hash_set::key_eq()
|
|
//
|
|
// Returns the function used for comparing keys equality.
|
|
using Base::key_eq;
|
|
|
|
ABSL_DEPRECATED("Call `hash_function()` instead.")
|
|
typename Base::hasher hash_funct() { return this->hash_function(); }
|
|
|
|
ABSL_DEPRECATED("Call `rehash()` instead.")
|
|
void resize(typename Base::size_type hint) { this->rehash(hint); }
|
|
};
|
|
|
|
// erase_if(node_hash_set<>, Pred)
|
|
//
|
|
// Erases all elements that satisfy the predicate `pred` from the container `c`.
|
|
template <typename T, typename H, typename E, typename A, typename Predicate>
|
|
void erase_if(node_hash_set<T, H, E, A>& c, Predicate pred) {
|
|
container_internal::EraseIf(pred, &c);
|
|
}
|
|
|
|
namespace container_internal {
|
|
|
|
template <class T>
|
|
struct NodeHashSetPolicy
|
|
: absl::container_internal::node_hash_policy<T&, NodeHashSetPolicy<T>> {
|
|
using key_type = T;
|
|
using init_type = T;
|
|
using constant_iterators = std::true_type;
|
|
|
|
template <class Allocator, class... Args>
|
|
static T* new_element(Allocator* alloc, Args&&... args) {
|
|
using ValueAlloc =
|
|
typename absl::allocator_traits<Allocator>::template rebind_alloc<T>;
|
|
ValueAlloc value_alloc(*alloc);
|
|
T* res = absl::allocator_traits<ValueAlloc>::allocate(value_alloc, 1);
|
|
absl::allocator_traits<ValueAlloc>::construct(value_alloc, res,
|
|
std::forward<Args>(args)...);
|
|
return res;
|
|
}
|
|
|
|
template <class Allocator>
|
|
static void delete_element(Allocator* alloc, T* elem) {
|
|
using ValueAlloc =
|
|
typename absl::allocator_traits<Allocator>::template rebind_alloc<T>;
|
|
ValueAlloc value_alloc(*alloc);
|
|
absl::allocator_traits<ValueAlloc>::destroy(value_alloc, elem);
|
|
absl::allocator_traits<ValueAlloc>::deallocate(value_alloc, elem, 1);
|
|
}
|
|
|
|
template <class F, class... Args>
|
|
static decltype(absl::container_internal::DecomposeValue(
|
|
std::declval<F>(), std::declval<Args>()...))
|
|
apply(F&& f, Args&&... args) {
|
|
return absl::container_internal::DecomposeValue(
|
|
std::forward<F>(f), std::forward<Args>(args)...);
|
|
}
|
|
|
|
static size_t element_space_used(const T*) { return sizeof(T); }
|
|
};
|
|
} // namespace container_internal
|
|
|
|
namespace container_algorithm_internal {
|
|
|
|
// Specialization of trait in absl/algorithm/container.h
|
|
template <class Key, class Hash, class KeyEqual, class Allocator>
|
|
struct IsUnorderedContainer<absl::node_hash_set<Key, Hash, KeyEqual, Allocator>>
|
|
: std::true_type {};
|
|
|
|
} // namespace container_algorithm_internal
|
|
ABSL_NAMESPACE_END
|
|
} // namespace absl
|
|
|
|
#endif // ABSL_CONTAINER_NODE_HASH_SET_H_
|