Abseil Containers

Abseil provides a number of containers as alternatives to STL containers. These containers generally adhere to the properties of STL containers, though there are often some associated API differences and/or implementation details which differ from the standard library.

The Abseil containers are designed to be more efficient in the general case; in some cases, however, the STL containers may be more efficient. Unlike some other abstractions that Abseil provides, these containers should not be considered drop-in replacements for their STL counterparts, as there are API and/or contract differences between the two sets of containers. For example, the Abseil containers often do not guarantee pointer stability1 after insertions or deletions.

The Abseil container library defines the following sets of containers:

  • Swiss table unordered containers
  • B-tree ordered containers

See below for more information about each of these container types.

Hash Tables

The Abseil container library contains a number of useful hash tables generally adhering to the STL container API contract:

  • absl::flat_hash_map
  • absl::flat_hash_set
  • absl::node_hash_map
  • absl::node_hash_set

Collectively, these hash tables are known as “Swiss tables” and are designed to be replacements for std::unordered_map and std::unordered_set They provide several advantages over the std::unordered_* containers:

  • Provides C++14 support for C++17 mechanisms such as try_emplace().
  • Supports heterogeneous lookup.
  • Allows optimizations for emplace({key, value}) to avoid allocating a pair in most common cases.
  • Supports a heterogeneous std::initializer_list to avoid extra copies for construction and insertion.
  • Guarantees an O(1) erase method by returning void instead of an iterator.

Construction

The set of Swiss table containers support the same overload set as std::unordered_map for construction and assignment:

// Examples using node_hash_set and node_hash_map are equivalent

// Default constructor
// No allocation for the table's elements is made.
absl::flat_hash_set<std::string> set1;

absl::flat_hash_map<int, std::string> map1;

// Initializer List constructor
absl::flat_hash_set<std::string> set2 = {{"huey"}, {"dewey"}, {"louie"},};

absl::flat_hash_map<int, std::string> map2 =
    {{1, "huey"}, {2, "dewey"}, {3, "louie"},};

// Copy constructor
absl::flat_hash_set<std::string> set3(set2);

absl::flat_hash_map<int, std::string> map3(map2);

// Copy assignment operator
// Hash functor and Comparator are copied as well
absl::flat_hash_set<std::string> set4;
set4 = set3;

absl::flat_hash_map<int, std::string> map4;
map4 = map3;

// Move constructor
// Move is guaranteed efficient
absl::flat_hash_set<std::string> set5(std::move(set4));

absl::flat_hash_map<int, std::string> map5(std::move(map4));

// Move assignment operator
// May be efficient if allocators are compatible
absl::flat_hash_set<std::string> set6;
set6 = std::move(set5);

absl::flat_hash_map<int, std::string> map6;
map6 = std::move(map5);

// Range constructor
std::vector<std::string> v = {"a", "b"};
absl::flat_hash_set<std::string> set7(v.begin(), v.end());

std::vector<std::pair<int, std::string>> v = {{1, "a"}, {2, "b"}};
absl::flat_hash_map<int, std::string> map7(v.begin(), v.end());

absl::flat_hash_map and absl::flat_hash_set

absl::flat_hash_map and absl::flat_hash_set are the recommended unordered containers for general use. These are flat data structures, which store their value_type directly in the slot array.

Guarantees

  • Keys and values are stored inline.
  • Iterators, references, and pointers to elements are invalidated on rehash.
  • Move operations do not invalidate iterators or pointers.

Memory Usage

Flat Hash Map Memory Layout

The container uses O((sizeof(std::pair<const K, V>) + 1) * bucket_count()) bytes. The max load factor is 87.5%, after which the table doubles in size (making load factor go down by 2x). Thus size() is usually between 0.4375*bucket_count() and 0.875*bucket_count(). For tables that have never rehashed the load factor can be even lower, but these numbers are sufficient for our estimates.

Recommendation

Use absl::flat_hash_map most of the time. If pointer stability of values (but not keys) is needed, use absl::flat_hash_map<Key, std::unique_ptr<Value>>.

absl::node_hash_map and absl::node_hash_set

These are near drop-in replacement for std::unordered_map and std::unordered_set. They are useful:

  • When pointer stability is required for both key and value.
  • For automatic migrations from std::unordered_map, std::unordered_set, hash_map or hash_set where it’s difficult to figure out whether the code relies on pointer stability.

These are node-based data structures in the STL standard sense: each value_type is allocated in a separate node and the main table contains pointers to those nodes.

Guarantees

  • Nodes have stable addresses.
  • Iterators are invalidated on rehash.
  • Move operations do not invalidate iterators.

Memory Usage

Node Hash Map Memory Layout

The slot array requires (sizeof(void*) + 1) * bucket_count() bytes and the nodes themselves require sizeof(value_type) * size() bytes. Together, this is O(9*bucket_count() + sizeof(std::pair<const K, V>)*size()) on most platforms.

Recommendation

Prefer absl::flat_hash_map or absl::flat_hash_set in most new code (see above).

Use absl::node_hash_map or absl::node_hash_set when pointer stability of both keys and values is required (rare), or for code migrations from other containers with this property. Note: Do not use popularity as a guide. You will see the “node” containers used a lot, but only because it was safe to migrate code to them from other containers.

Construction and Usage

absl::flat_hash_map<int, string> numbers =
    {{1, "one"}, {2, "two"}, {3, "three"}};
numbers.try_emplace(4, "four");

absl::flat_hash_map<string, std::unique_ptr<string>> strings;
strings.try_emplace("foo", absl::make_unique<string>("bar"));

Heterogeneous Lookup

Inserting into or looking up an element within an associative container requires a key. In general, containers require the keys to be of a specific type, which can lead to inefficiencies at call sites that need to convert between near-equivalent types (such as std::string and absl::string_view).

std::map<std::string, int> m = ...;
absl::string_view some_key = ...;
// Construct a temporary `std::string` to do the query.
// The allocation + copy + deallocation might dominate the find() call.
auto it = m.find(std::string(some_key));

To avoid this unnecessary work, the Swiss tables provide heterogeneous lookup for conversions to string types (allowing absl::string_view in the lookup, for example), and for conversions to smart pointer types (std::unique_ptr, std::shared_ptr), through the absl::Hash hashing framework. (The supporting comparators are built into absl::Hash.)

absl::flat_hash_map<std::string, int> m = ...;
absl::string_view some_key = ...;
// We can use string_view directly as the key search.
auto it = m.find(some_key);

Iteration Order Instability

While std::unordered_map makes no guarantees about iteration order, many implementations happen to have a deterministic order based on the keys and their insert order. This is not true of absl::flat_hash_map and absl::node_hash_map. Thus, converting from std::unordered_map to absl::flat_hash_map can expose latent bugs where the code incorrectly depends on iteration order.

A special case which can create a subtle bug is summing float values in an unordered container. While mathematical sums do not depend on order, floating point sums do, and it can be the case that a sum is deterministic with std::unordered_set but non-deterministic with absl::flat_hash_set.

B-tree Ordered Containers

The Abseil container library contains ordered containers generally adhering to the STL container API contract, but implemented using (generally more efficient) B-trees rather than binary trees (as used in std::map et al):

  • absl::btree_map
  • absl::btree_set
  • absl::btree_multimap
  • absl::btree_multiset

These ordered containers are designed to be more efficient replacements for std::map and std::set in most cases. Specifically, they provide several advantages over the ordered std:: containers:

  • Provide lower memory overhead in most cases than their STL equivalents.
  • Are generally more cache friendly (and hence faster) than their STL equivalents.
  • Provide C++14 support for C++17 mechanisms such as try_emplace().
  • Support heterogeneous lookup.

Construction

The set of B-tree containers support the same overload set as std::map for construction and assignment:

// Examples using btree_multimap and btree_multiset are equivalent

// Default constructor
// No allocation for the B-tree's elements is made.
absl::btree_set<std::string> set1;

absl::btree_map<int, std::string> map1;

// Initializer List constructor
absl::btree_set<std::string> set2 = {{"huey"}, {"dewey"}, {"louie"},};

absl::btree_map<int, std::string> map2 =
    {{1, "huey"}, {2, "dewey"}, {3, "louie"},};

// Copy constructor
absl::btree_set<std::string> set3(set2);

absl::btree_map<int, std::string> map3(map2);

// Copy assignment operator
// Hash functor and Comparator are copied as well
absl::btree_set<std::string> set4;
set4 = set3;

absl::btree_map<int, std::string> map4;
map4 = map3;

// Move constructor
// Move is guaranteed efficient
absl::btree_set<std::string> set5(std::move(set4));

absl::btree_map<int, std::string> map5(std::move(map4));

// Move assignment operator
// May be efficient if allocators are compatible
absl::btree_set<std::string> set6;
set6 = std::move(set5);

absl::btree_map<int, std::string> map6;
map6 = std::move(map5);

// Range constructor
std::vector<std::string> v = {"a", "b"};
absl::btree_set<std::string> set7(v.begin(), v.end());

std::vector<std::pair<int, std::string>> v = {{1, "a"}, {2, "b"}};
absl::btree_map<int, std::string> map7(v.begin(), v.end());

Notes

  1. “Pointer stability” means that a pointer to an element remains valid (is not invalidated) so long as the element is present, allowing code to cache pointers to elements even when the underlying container is mutated. Saying that a container has pointer stability is the same as saying that it doesn’t move elements in memory; their addresses do not change. Pointer stability/invalidation is the same as reference stability/invalidation.