Boost C++ Libraries of the most highly regarded and expertly designed C++ library projects in the world. Herb Sutter and Andrei Alexandrescu, C++ Coding Standards



  1. Introduction
  2. Configuration
  3. Examples
  4. Reference
  5. Concepts
  6. Design notes
  7. History and Acknowledgements


A UUID, or Universally unique identifier, is intended to uniquely identify information in a distributed environment without significant central coordination. It can be used to tag objects with very short lifetimes, or to reliably identify very persistent objects across a network.

A formal definition for UUID can be found in RFC 4122.

UUIDs have many applications. Some examples follow: Databases may use UUIDs to identify rows or records in order to ensure that they are unique across different databases, or for publication/subscription services. Network messages may be identified with a UUID to ensure that different parts of a message are put back together again. Distributed computing may use UUIDs to identify a remote procedure call. Transactions and classes involved in serialization may be identified by UUIDs. Microsoft's component object model (COM) uses UUIDs to distinguish different software component interfaces. UUIDs are inserted into documents from Microsoft Office programs. UUIDs identify audio or video streams in the Advanced Systems Format (ASF). UUIDs are also a basis for OIDs (object identifiers), and URNs (uniform resource name).

An attractive feature of UUIDs when compared to alternatives is their relative small size, of 128-bits, or 16-bytes. Another is that the creation of UUIDs does not require a centralized authority.

When UUIDs are generated by one of the defined mechanisms, they are either guaranteed to be unique, different from all other generated UUIDs (that is, it has never been generated before and it will never be generated again), or it is extremely likely to be unique (depending on the mechanism).


The library does not require building or any special configuration to be used. However, there are a few options that can be enabled by defining macros prior to including library headers. These macros are summarized in the following table.

BOOST_UUID_NO_SIMDIf defined, disables any optimizations for SIMD-enabled processors. Generic versions of algorithms will be used instead. This may result in suboptimal performance. By default, optimized algorithms are used, when the library is able to detect the availability of SIMD extensions at compile time.
BOOST_UUID_USE_SSE2If defined, enables optimizations for SSE2 exstensions available in modern x86 processors.
BOOST_UUID_USE_SSE3If defined, enables optimizations for SSE3 exstensions available in modern x86 processors.
BOOST_UUID_USE_SSE41If defined, enables optimizations for SSE4.1 exstensions available in modern x86 processors.

By default the library attempts to detect the availability of SIMD extensions in the target CPU at compile time and automatically defines the appropriate macros if succeeded. The BOOST_UUID_USE_SSE* macros can be defined by users, if auto-detection fails and it is known that the target CPU will have the extension. Do not enable these extensions unless you're certain that they will always be available on any machine that will run your program. The library performs no run time checks, so if an extension is missing, the program will likely crash. Note that enabling more advanced extensions implies that more basic ones are also available.



// example of tagging an object with a uuid
// see boost/libs/uuid/test/test_tagging.cpp

#include <boost/uuid/uuid.hpp>
#include <boost/uuid/uuid_generators.hpp>

class object
        : tag(boost::uuids::random_generator()())
        , state(0)

    explicit object(int state)
        : tag(boost::uuids::random_generator()())
        , state(state)

    object(object const& rhs)
        : tag(rhs.tag)
        , state(rhs.state)

    bool operator==(object const& rhs) const {
        return tag == rhs.tag;

    object& operator=(object const& rhs) {
        tag = rhs.tag;
        state = rhs.state;

    int get_state() const { return state; }
    void set_state(int new_state) { state = new_state; }

    boost::uuids::uuid tag;

    int state;

object o1(1);
object o2 = o1;
assert(o1 == o2);

object o3(3);
assert(o1 != o3);
assert(o2 != o3);

POD Efficiencies

This library implements a UUID as a POD allowing a UUID to be used in the most efficient ways, including using memcpy, and aggregate initializers. A drawback is that a POD can not have any constructors, and thus declaring a UUID will not initialize it to a value generated by one of the defined mechanisms. But a class based on a UUID can be defined that does initialize itself to a value generated by one of the defined mechanisms.

Note that boost::is_pod is specialized for boost::uuids::uuid and depends on Boost.TypeTraits. Define BOOST_UUID_NO_TYPE_TRAITS before including boost/uuid/uuid.hpp to remove the dependency on Boost.TypeTraits.

// example using memcpy and aggregate initializers
// example of a class uuid see boost/libs/uuid/test/test_uuid_class.cpp

#include <boost/uuid/uuid.hpp>
#include <boost/uuid/uuid_generators.hpp>

{ // example using memcpy
    unsigned char uuid_data[16];
    // fill uuid_data

    boost::uuids::uuid u;

    memcpy(&u, uuid_data, 16);

{ // example using aggregate initializers
    boost::uuids::uuid u =
    { 0x12 ,0x34, 0x56, 0x78
    , 0x90, 0xab
    , 0xcd, 0xef
    , 0x12, 0x34
    , 0x56, 0x78, 0x90, 0xab, 0xcd, 0xef

// example of creating a uuid class that
// initializes the uuid in the constructor
// using a defined mechanism

class uuid_class : public boost::uuids::uuid
        : boost::uuids::uuid(boost::uuids::random_generator()())

    explicit uuid_class(boost::uuids::uuid const& u)
        : boost::uuids::uuid(u)

    operator boost::uuids::uuid() {
        return static_cast<boost::uuids::uuid&>(*this);

    operator boost::uuids::uuid() const {
        return static_cast<boost::uuids::uuid const&>(*this);

uuid_class u1;
uuid_class u2;

assert(u1 != u2);

Byte Extraction

It is sometimes useful to get at the 16 bytes of a uuid directly. Typical use is as follows:

boost::uuids::uuid u;
std::vector<uint8_t> v(u.size());
std::copy(u.begin(), u.end(), v.begin());

Note: boost::uuids::uuid::size() always returns 16.




namespace boost {
namespace uuids {

class uuid {
    typedef uint8_t value_type;
    typedef uint8_t& reference;
    typedef uint8_t const& const_reference;
    typedef uint8_t* iterator;
    typedef uint8_t const* const_iterator;
    typedef std::size_t size_type;
    typedef std::ptrdiff_t difference_type;

    static constexpr size_type static_size() noexcept;

    // iteration
    iterator begin() noexcept;
    iterator end() noexcept;
    const_iterator begin() const noexcept;
    const_iterator end() const noexcept;

    constexpr size_type size() const noexcept;

    bool is_nil() const noexcept;

    enum variant_type {
        variant_ncs, // NCS backward compatibility
        variant_rfc_4122, // defined in RFC 4122 document
        variant_microsoft, // Microsoft Corporation backward compatibility
        variant_future // future definition
    variant_type variant() const noexcept;

    enum version_type {
        version_unknown = -1,
        version_time_based = 1,
        version_dce_security = 2,
        version_name_based_md5 = 3,
        version_random_number_based = 4,
        version_name_based_sha1 = 5
    version_type version() const noexcept;

    // Swap function
    void swap(uuid& rhs) noexcept;

    uint8_t data[static_size()];

// standard operators
bool operator==(uuid const& lhs, uuid const& rhs) noexcept;
bool operator!=(uuid const& lhs, uuid const& rhs) noexcept;
bool operator<(uuid const& lhs, uuid const& rhs) noexcept;
bool operator>(uuid const& lhs, uuid const& rhs) noexcept;
bool operator<=(uuid const& lhs, uuid const& rhs) noexcept;
bool operator>=(uuid const& lhs, uuid const& rhs) noexcept;

void swap(uuid& lhs, uuid& rhs) noexcept;

std::size_t hash_value(uuid const& u) noexcept;

}} // namespace boost::uuids


The size of a uuid (in bytes) can be obtained either by calling the function boost::uuids::uuid::size() or by calling the static function boost::uuids::uuid::static_size(), both always return 16.

    boost::uuids::uuid u;
    assert(16 == u.size());
    assert(16 == boost::uuids::uuid::static_size());


Both iterators and constant iterators are provided.

    boost::uuids::uuid u;
    for (boost::uuids::uuid::const_iterator it=u.begin(); it!=u.end(); ++it) {
        boost::uuids::uuid::value_type v = *it;
    for (boost::uuids::uuid::iterator it=u.begin(); it!=u.end(); ++it) {
        *it = 0;

Nil uuid

The function, boost::uuids::uuid::is_nil() returns true if and only if the uuid is equal to {00000000-0000-0000-0000-000000000000}.


Three bits of a uuid determine the variant.

    boost::uuids::uuid u;
    boost::uuids::uuid::variant_type v = u.variant();


Four bits of a uuid determine the variant, that is the mechanism used to generate the uuid.

    boost::uuids::uuid u;
    boost::uuids::uuid::version_type v = u.version();


Both boost::uuids::uuid::swap() and boost::uuids::swap() are provided.

    boost::uuids::uuid u1, u2;
    swap(u1, u2);


All of the standard numeric operators are defined for the uuid class. These include:


Hash Function

This function allows uuids to be used with boost::hash

boost::hash<boost::uuids::uuid> uuid_hasher;
std::size_t uuid_hash_value = uuid_hasher(boost::uuids::uuid());



This file includes all the uuid generators for convenience.
#include <boost/uuid/nil_generator.hpp>
#include <boost/uuid/string_generator.hpp>
#include <boost/uuid/name_generator.hpp>
#include <boost/uuid/random_generator.hpp>



namespace boost {
namespace uuids {

struct nil_generator {
    typedef uuid result_type;

    uuid operator()() const;
uuid nil_uuid();

}} //namespace boost::uuids

Nil Generator

The boost::uuids::nil_generator class always generates a nil uuid.

boost::uuids::nil_generator gen;
boost::uuids::uuid u = gen();
assert(u.is_nil() == true);

// or for convenience
boost::uuids::uuid u = boost::uuids::nil_uuid();
assert(u.is_nil() == true);



namespace boost {
namespace uuids {

struct string_generator {
    typedef uuid result_type;

    template <typename ch, typename char_traits, typename alloc>
        uuid operator()(std::basic_string<ch, char_traits, alloc> const& s) const;

}} //namespace boost::uuids

String Generator

The boost::uuids::string_generator class generates a uuid from a string. In addition to the standards-defined string format in RFC 4122 (p. 3), the string generator accepts a few variants. Valid strings match the following PCRE regular expression:

Or more generally, the following formats are accepted as valid string formats, where h is hexadecimal, without case sensitivity, and without any leading or trailing whitespace:
For example:
boost::uuids::string_generator gen;
boost::uuids::uuid u1 = gen("{01234567-89ab-cdef-0123-456789abcdef}");
boost::uuids::uuid u2 = gen(L"01234567-89ab-cdef-0123-456789abcdef");
boost::uuids::uuid u3 = gen(std::string("0123456789abcdef0123456789abcdef"));
boost::uuids::uuid u4 = gen(std::wstring(L"01234567-89AB-CDEF-0123-456789ABCDEF"));
Invalid input will generate a std::runtime_error exception.



namespace boost {
namespace uuids {

template<class NameHashProvider>
class basic_name_generator {
    typedef uuid result_type;

    explicit basic_name_generator(uuid const& namespace_uuid);

    uuid operator()(const char* name) const;
    uuid operator()(const wchar_t* name) const;
    tempate <typename ch, typename char_traits, typename alloc>
        uuid operator()(std::basic_string<ch, char_traits, alloc> const& name) const;
    uuid operator()(void const* buffer, std::size_t byte_count) const;

typedef basic_name_generator<detail::md5> name_generator_md5;
typedef basic_name_generator<detail::sha1> name_generator_sha1;
typedef name_generator_sha1 name_generator; // deprecated
typedef name_generator_sha1 name_generator_latest;

}} //namespace boost::uuids

Name Generator

RFC 4122 specifies that a name-based uuid is derived from content in a namespace. A uuid with identical content in an identical namespace shall yield the same uuid as long as the same hasing algorithm is used.

Hashing Algorithms
RFC 4122 defines two hashing mechanisms for uuid generation: Since RFC 4122 was written, both of these hash algorithms have been rendered insecure. In anticipation of a new RFC for uuid arriving, name_generator has been deprecated, and replaced with: name_generator, while deprecated, remains a type alias for name_generator_sha1 so the behavior is identical to previous releases. When the successor to SHA1 is chosen, it will be implemented under a new name similar to those above. If your application does not rely on stable hashing over time, you can use name_generator_latest to always use the latest hashing algorithm available.

Consumers are free to use their own hash provider as long as it satisfies the NameHashProvider concept.

There are four well-known namespaces defined in RFC 4122, Appendix C which are defined as:
boost::uuids::ns::dns()     // == {6ba7b810-9dad-11d1-80b4-00c04fd430c8}
boost::uuids::ns::url()     // == {6ba7b811-9dad-11d1-80b4-00c04fd430c8}
boost::uuids::ns::oid()     // == {6ba7b812-9dad-11d1-80b4-00c04fd430c8}
boost::uuids::ns::x500dn()  // == {6ba7b814-9dad-11d1-80b4-00c04fd430c8}
of course, you are free to use your own namespace if you prefer. Here is an example of name generator usage:
boost::uuids::name_generator_sha1 gen(boost::uuids::ns::dns());
boost::uuids::uuid udoc = gen("");
std::cout << " uuid in dns namespace, sha1 version: " << udoc << std::endl;
produces the output: uuid in dns namespace, sha1 version: 0043f363-bbb4-5369-840a-322df6ec1926



namespace boost {
namespace uuids {

class random_generator {
    typedef uuid result_type;
    result_type operator()();

template <typename UniformRandomNumberGenerator>
class basic_random_generator {
    typedef uuid result_type;

    explicit basic_random_generator(UniformRandomNumberGenerator& gen);
    explicit basic_random_generator(UniformRandomNumberGenerator* pGen);

    result_type operator()();

typedef basic_random_generator<boost::mt19937> random_generator_mt19937;

}} // namespace boost::uuids

Random Generator

boost::uuids::random_generator class generates uuids using operating system provided entropy. For the majority of use cases this should be sufficient, as this generator has very low startup overhead when compared to a generator with seeding requirements.

boost::uuids::basic_random_generator class generates a random number based uuid from a random number generator (one that conforms to the UniformRandomNumberGenerator concept). The default constructor will properly seed the random number generator if it requires the behavior for proper operation. Additional constructors allow you to provide your own UniformRandomNumberGenerator and you are responsible for properly seeding it if necessary.

boost::uuids::random_generator_mt19937 is a type definition for: boost::uuids::basic_random_generator<mt19937> and is provided for convenience.


The following platforms are supported and entropy selection logic explained:

Preprocessor Macros


In most cases random_generator will be optimal. A benchmark can be found in the tests (test_bench_random) that will determine the cutoff point where random_generator_mt19937 outperforms random_generator in wall time.

On Windows when using the wincrypt entropy provider, a measurable delay may occur the first time a random_generator is constructed within a running instance. This has been observed using test_bench_random and was so significant that the benchmark was changed to throw out the first loop of measurements.


The exception boost::uuids::entropy_error is thrown if there is an error getting entropy from the operating system.

// Depending on the platform there may be a setup cost in
// initializing the generator so plan to reuse it if you can.
boost::uuids::random_generator gen;
boost::uuids::uuid id = gen();
std::cout << id << std::endl;
boost::uuids::uuid id2 = gen();
std::cout << id2 << std::endl;
assert(id != id2);

// You can still use a PseudoRandomNumberGenerator to create
// UUIDs, however this is not the preferred mechanism.
boost::uuids::random_generator_mt19937 bulkgen;
for (size_t i = 0; i < 1000; ++i)
    boost::uuids::uuid u = bulkgen();
    // do something with u



namespace boost {
namespace uuids {

template <typename ch, typename char_traits>
    std::basic_ostream<ch, char_traits>& operator<<(std::basic_ostream<ch, char_traits> &os, uuid const& u);

template <typename ch, typename char_traits>
    std::basic_istream<ch, char_traits>& operator>>(std::basic_istream<ch, char_traits> &is, uuid &u);

template<class OutputIterator>
    OutputIterator to_chars(uuid const& u, OutputIterator out);

bool to_chars(uuid const& u, char* first, char* last);

std::string to_string(uuid const& u);
std::wstring to_wstring(uuid const& u);

}} // namespace boost::uuids

Stream Operators

The standard input and output stream operators << and >> are provided by including boost/uuid/uuid_io.hpp. The string representation of a uuid is hhhhhhhh-hhhh-hhhh-hhhh-hhhhhhhhhhhh where h is hexadecimal.

boost::uuids::uuid u1; // initialize uuid

std::stringstream ss;
ss << u1;

boost::uuids::uuid u2;
ss >> u2;

assert(u1, u2);

One can also use boost::lexical_cast.

boost::uuids::uuid u1; // initialize uuid

std::string s = boost::lexical_cast<std::string>(u);
boost::uuids::uuid u2 = boost::lexical_cast<boost::uuids::uuid>(s);

assert(u1 == u2);

To Chars

The function to_chars is a fast non-allocating (and non-throwing if the iterator does not throw) function to write a uuid to a string buffer. It writes exactly 36 characters to the iterator on success (not null-terminated).

boost::uuids::uuid u; // initialize uuid

char buf[36];
boost::uuids::to_chars(u, buf);
// OR
bool ret = boost::uuids::to_chars(u, std::begin(buf), std::end(buf));

To String

The functions to_string and to_wstring are provided as a convenience to convert a uuid to a string. They are also likely faster than the stream operators or using boost::lexical_cast.

boost::uuids::uuid u; // initialize uuid

std::string s1 = to_string(u);

std::wstring s2 = to_wstring(u);



namespace boost {
namespace uuids {

BOOST_CLASS_IMPLEMENTATION(boost::uuids::uuid, boost::serialization::primitive_type)

}} // namespace boost::uuids


Serialization is accomplished with the Boost Serialization library. A uuid is serialized as a primitive type, thus only the uuid value will be saved to/loaded from an archive.

Include boost/uuid/uuid_serialize.hpp to enable serialization for uuids.


This section describes all of the concepts defined by the library.


A NameHashProvder type is supplied as a template argument to the basic_name_generator class. It provides the hashing function that the name generator uses to generate a uuid.


In this table, G is a type meeting the requirements of NameHashProvider:

Expression Semantics, Pre/Post-conditions
typedef G::digest_type A type definition of contiguous raw storage at least 16 bytes in length.
void G::process_byte(unsigned char)
void G::process_bytes(void const*, std::size_t)
void G::get_digest(typename G::digest_type&) Copies the digest into the supplied parameter. Called once.
unsigned char G::get_version() const Returns the RFC 4122 version for the hashing algorithm (4 bits) in 0x07.

Design notes

The document,, was used to design and implement the boost::uuids::uuid struct.

All functions are re-entrant. Classes are as thread-safe as an int. That is an instance can not be shared between threads without proper synchronization.

History and Acknowledgements

A number of people on the mailing list provided useful comments and greatly helped to shape the library.

Revised November 8, 2017

© Copyright Andy Tompkins, 2006

© Copyright 2017 James E. King III

Distributed under the Boost Software License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at