caution

Note, this article is not finished! You can help by editing this doc page.

Overview

Simplified (since C++98)

template< class Key, /* ... */ >
class unordered_set;

Detailed

Regular (since C++98)

template<
    class Key,
    class Hash = std::hash<Key>,
    class Compare = std::equal_to<Key>,
    class Allocator = std::allocator<Key>
>
class unordered_set;

Polymorphic (since C++17)

namespace pmr {
    template <class Key,
              class Hash = std::hash<Key>,
              class Compare = std::equal_to<Key>>
    using unordered_set = std::unordered_set<Key, Hash, Compare,
                                             std::pmr::polymorphic_allocator<Key>>;
}

The std::unordered_set is an associative container for storing unique objects.

Example usage

The examples in this section are very simple ones. Navigate to examples section at the bottom for more.

Create

C++17

Using CTAD

#include <unordered_set>

int main() {
  std::unordered_set ints{1, 2, 3, 2, 1, 4, 5, 5};

 // Content: 1, 2, 3, 4, 5
}

until C++17

The elements will stay unique even if we repeat them at the point of construction.

#include <unordered_set>

int main() {
  std::unordered_set<int> ints{1, 2, 3, 2, 1, 4, 5, 5};

 // Content: 1, 2, 3, 4, 5
}

From range

We can construct almost every container from other containers using iterators.

#include <unordered_set>
#include <array>

int main() {
  std::array<int, 5> array{1, 4, 3, 3, 2};
  std::unordered_set<int> ints(array.cbegin(), array.cend());

 // Content: 1, 2, 3, 4
}

Order of elements

It's important to notice, that although there's an arbitrary order chosen in the examples' comments, the elements inside remain unordered (as the name of the container suggests), which means we aren't guaranteed to have any particular order while iterating over the unordered set.

Insert

Ordinary

No matter what, the set will always remain unique.

#include <unordered_set>

int main() {
  std::unordered_set<int> ints;
  
  ints.insert(1);
  ints.insert(2);
  ints.insert(1);
  ints.insert(2);

  // Content: 1 2
}

Many at once

We can insert several values at once.

#include <unordered_set>

int main() {
  std::unordered_set<int> ints;

  ints.insert({4, 3, 2, 1, 1});

  // Content: 1, 2, 3, 4
}

From range

We can also insert from other ranges, like in the construction.

#include <unordered_set>
#include <array>

int main() {
  std::array<int, 5> array = {1, 2, 3, 3, 4};
  std::unordered_set<int> ints;

  ints.insert(array.cbegin(), array.cend());

  // Content: 1, 2, 3, 4
}

Search

Contains (since C++20)

Since C++20 we can simply use .contains() on the set to determine if some arbitrary value exists in it.

#include <unordered_set>
#include <iostream>

int main() {
  std::unordered_set<int> ints{1, 2, 3, 4, 5};

  if(ints.contains(3)) {
    std::cout << "3 exists in the set.";
  } else {
    std::cout << "3 doesn't exist in the set.";
  }
}

Count

.count() is an another method of determining the existence of an element in the set.

#include <unordered_set>
#include <iostream>

int main() {
  std::unordered_set<int> ints{1, 2, 3, 4, 5};

  if(ints.count(3) == 1) { // or  simply if(ints.count(3)) , same effect
    std::cout << "3 exists in the set.";
  } else {
    std::cout << "3 doesn't exist in the set.";
  }
}

Find

We can use the .find() method to look for a specific value in the set. If the value isn't in the set, the returned iterator will be equal to the end iterator (obtained by .end() or .cend()).

#include <unordered_set>
#include <iostream>

int main() {
  std::unordered_set<int> ints{1, 2, 3, 4, 5};

  const auto it = ints.find(3); // we can use auto to deduce the type of the iterator returned

  if(it != ints.cend()) {
    // if the iterator is not equal to ints.cend(), the value is in the set
    std::cout << "3 exists in the set.";
  } else {
    std::cout << "3 doesn't exist in the set.";
  }
}

Technical details

Technical definition of an unordered_set

std::unordered_set is an associative container that contains a set of unique objects of type Key.

Search, insertion, and removal have average constant-time complexity.

Internally, the elements are not sorted in any particular order, but organized into buckets. Which bucket an element is placed into depends entirely on the hash of its value. This allows fast access to individual elements, since once a hash is computed, it refers to the exact bucket the element is placed into.

Container elements may not be modified (even by non const iterators) since modification could change an element's hash and corrupt the container.

std::unordered_set meets the requirements of Container, AllocatorAwareContainer, UnorderedAssociativeContainer.

std::unordered_set

Defined in

unordered_set

Template parameters

pub	Key	The type of the stored keys.
pub	Hash	A function object that perfoms hashing on two objects of type Key.
pub	Compare	A comparator type satisfying Compare.
pub	Allocator	An allocator type responsible for allocating and deallocating memory. Must satisfy Allocator.

Type names

pub	key_type	Key
pub	value_type	Key
pub	size_type	Unsigned integer type (usually std::size_t).
pub	difference_type	Signed integer type (usually std::ptrdiff_t).
pub	hasher	Hash
pub	key_equal	KeyEqual
pub	allocator_type	Allocator
pub	reference	value_type&
pub	const_reference	const value_type&
pub	pointer	std::allocator_traits<Allocator>::pointer
pub	const_pointer	std::allocator_traits<Allocator>::const_pointer
pub	iterator ❗	Constant LegacyBidirectionalIterator to value_type
pub	const_iterator ❗	LegacyBidirectionalIterator to const value_type
pub	local_iterator	Like iterator but only to iterate across a single bucket.
pub	const_local_iterator	Like const_iterator but only to iterate across a single bucket.
pub	node_type (since C++17)	A specialization of node handle representing a container node.
pub	insert_return_type (since C++17)	Type describing the result of inserting a node_type, a specialization of template < class Iter, class NodeType > struct /* unspecified */ { Iter position; bool inserted; NodeType node; }; instantiated with template arguments iterator and node_type.

Unordered set's iterators

iterator is said to be constant because it is impossible to modify the values it's pointing to. This is because modyfing a value the iterator is pointing to could possibly invalidate the internal structure of the unordered set, which could then lead to all kinds of incorrect and unexpected behaviours.

The member type aliases iterator and const_iterator may be the same. This is left entirely to the implementation. This means that any overloaded functions/specializations differing only in the iterator types may be incorrect and may lead to ODR violation.

Iterator invalidation

pub	Operation	Invalidated
pub	All read only operations, swap, std::swap	Never
pub	Clear, rehash, reverse, operator=	Always
pub	Insert, emplace, emplace_hiny	Only if causes rehash
pub	Erase	Only to the element erased

Member functions

pub	(constructors)	Constructs a new unordered set.
pub	(destructor)	Destructs an unordered set.
pub	operator=	Assigns one unordered set to another.
pub	get_allocator	Returns an associated allocator.

Iterators

pub	begin cbegin	Returns an iterator to the beginning.
pub	end cend	Returns an iterator to the end.

Capacity

pub	empty	Returns true if a set is empty, otherwise false.
pub	size	Returns the number of elements in an unordered set.
pub	max_size	Returns the maximum possible number of elements.

Modifiers

pub	clear	Clears the contents of an unordered set.
pub	insert	Inserts elements (or nodes (extracted with .extract()) (since C++17)).
pub	emplace	Constructs a new element in place.
pub	emplace_hint	Constructs elements in-place using a hint (iterator).
pub	erase	Erases elements.
pub	swap	Swaps two unordered sets.
pub	extract (since C++17)	Extracts nodes from an unordered set (can be later inserted somewhere else).
pub	merge (since C++17)	Merges two unordered sets together.

Lookup

pub	count	Returns the number of elements matching a specific key.
pub	find	Searches for an element and returns an iterator to it, or an end iterator if not found.
pub	contains (since C++20)	Returns true if an element is inside a set, otherwise false.
pub	equal_range	Returns a range of elements matching a specific key.

Bucket interface

pub	begin (size_type) cbegin (size_type)	Returns an iterator to the beginning of a specified bucket.
pub	end (size_type) cend (size_type)	Returns an iterator to the end of a specified bucket.
pub	bucket_count	Returns a number of buckets.
pub	max_bucket_count	Returns a maximum number of buckets.
pub	bucket_size	Returns a number of elements in a specific bucket.
pub	bucket	Returns a bucket for a specific key.

Hash policy

pub	load_factor	Returns an average number of elements per bucket.
pub	max_load_factor	Manages a maximum average number of elements per bucket.
pub	rehash	Reserves at least the specified number of buckets and regenerates the hash table.
pub	reserve	Reserves space for at least a specified number of elements and regenerates the hash table.

Observers

pub	hash_function	Returns an internal function object that hashes keys.
pub	key_eq	Returns an internal function object that compares keys.

Non-member functions

pub	operator== operator!= (removed in C++20)	Compares values in an unordered_map.
pub	std::swap (std::unordered_set)	An overload for a std::swap algorithm.
pub	std::erase_if (std::unordered_set) (since C++20)	Overload for a std::erase_if algorithm.

Deduction guides (since C++17)

Click to expand

Deduction guides

// (1)
template< class InputIt,
          class Hash = std::hash<iter_val_t<InputIt>>,
          class Pred = std::equal_to<iter_val_t<InputIt>>,
          class Alloc = std::allocator<iter_val_t<InputIt>> >
unordered_set( InputIt, InputIt,
         typename /*see below*/::size_type = /*see below*/,
         Hash = Hash(), Pred = Pred(), Alloc = Alloc() )
  -> unordered_set<iter_val_t<InputIt>, Hash, Pred, Alloc>;

// (2)
template< class InputIt, class Alloc >
unordered_set( InputIt, InputIt, typename /*see below*/::size_type, Alloc )
  -> unordered_set<iter_val_t<InputIt>,
              std::hash<iter_val_t<InputIt>>,
              std::equal_to<iter_val_t<InputIt>>,
              Alloc>;

// (3)
template< class InputIt, class Hash, class Alloc >
unordered_set( InputIt, InputIt, typename /*see below*/::size_type, Hash, Alloc )
  -> unordered_set<iter_val_t<InputIt>, Hash,
             std::equal_to<iter_val_t<InputIt>>, Allocator>;

(1 - 3) allow deduction from an iterator range

// (4)
template< class T, class Allocator >
unordered_set( std::initializer_list<T>, typename /*see below*/::size_type, Allocator )
  -> unordered_set<T, std::hash<T>, std::equal_to<T>, Alloc>;

// (5)
template< class T, class Hash, class Alloc >
unordered_set( std::initializer_list<T>, typename /*see below*/::size_type, Hash, Alloc )
  -> unordered_set<T, Hash, std::equal_to<T>, Alloc>;

// (6)
template< class T,
          class Hash = std::hash<T>,
          class Pred = std::equal_to<T>,
          class Alloc = std::allocator<T> >
unordered_set( std::initializer_list<T>,
         typename /*see below*/::size_type = /*see below*/,
         Hash = Hash(), Pred = Pred(), Alloc = Alloc() )
  -> unordered_set<T, Hash, Pred, Alloc>;

(4- 6) allow deduction from a std::initializer_list

Alias iter_val_t is defined as if follows:

template< class InputIt >
using iter_val_t = typename std::iterator_traits<InputIt>::value_type;

note

Note that this alias isn't guaranteed to be found anywhere in the standard library. It is defined solely for presentation purposes of these deduction guides and weren't present in the standard library at the point of writing this document.

The size_type parameter type refers to the size_type member type of the type deduced by the deduction guide.

Overload resolution

In order for any of the deduction guides to participate in overload resolution, the folllowing requirements must be met:

• InputIt satisfies LegacyInputIterator
• Alloc satisfies Allocator
• Hash doesn't deduce to an integral type or doesn't satisfy Allocator
• Pred doesn't satisfy Allocator

note

The extent to which the library determines that a type does not satisfy LegacyInputIterator is unspecified, except that as a minimum:

• Integral types do not qualify as input iterators.

Likewise, the extent to which it determines that a type does not satisfy Allocator is unspecified, except that as a minimum:

• The member type Alloc::value_type must exist.
• The expression std::declval<Alloc&>().allocate(std::size_t{}) must be well-formed when treated as an unevaluated operand.

More examples

Creating a custom hash and comparator.

We can create our own comparators by defining function objects with operator() overloaded. To create a custom hash for an object, we can create a specialization for it in the std namespace.

#include <iostream>
#include <string>
#include <unordered_set>

struct student {
  std::string name;
  std::string surname;
  int age;

  bool operator==(const student&) const noexcept = default;
};

/*
 *	We need to define a specialization for std::hash for the unordered_set to work
 *  properly.
 */
namespace std {

  template <>
  struct hash<student> {
    std::size_t operator()(const student& obj) const noexcept {
      constexpr auto str_hash = std::hash<std::string>();

      return (str_hash(obj.name) * 31) ^ (str_hash(obj.surname) * 31) ^ obj.age;
    }
  }; 

}

struct student_comparator {
  bool operator()(const student& first, const student& second) const noexcept {
    return first == second; 
  }
};

auto main() -> int {
  using student_set = std::unordered_set<student, std::hash<student>, student_comparator>;
  
  const auto student1 = student("Philip", "Davies", 17);
  const auto student2 = student("Philip", "Davies", 17);
  const auto student3 = student("Philip", "Smith", 19);
  const auto student4 = student("Jacob", "Jones", 19);

  auto students = student_set{student1, student2, student3, student4};

  for(const auto& student : students) {
    std::cout << "Student: " << student.name << ' ' 
							<< student.surname << ", " 
							<< student.age << " years old.\n";
  }
}

Result

Student: Jacob Jones, 19 years old.
Student: Philip Smith, 19 years old.
Student: Philip Davies, 17 years old.

Custom hashes

Note that although we've added four students into the set, we only see information about three of them being displayed. This is because there are two students with exactly the same data. When the hashes of both of them are being compared, they turn out to be the same, which then leads to a comparision of the two objects themselves (using operator==, hence the it's presence inside student). If the comparision tests positive, the new object (that compared the same as the one already inserted) is not being added.