Data::UUID::MT - Fast random UUID generator using the Mersenne Twister algorithm
my $ug1 = Data::UUID::MT->new( version => 4 ); # "1", "4" or "4s"
my $ug2 = Data::UUID::MT->new(); # default is "4"
# method interface
my $uuid1 = $ug->create(); # 16 byte binary string
my $uuid2 = $ug->create_hex();
my $uuid3 = $ug->create_string();
# iterator -- avoids some method call overhead
my $next = $ug->iterator;
my $uuid4 = $next->();
This UUID generator uses the excellent Math::Random::MT::Auto module as a source of fast, high-quality (pseudo) random numbers.
Three different types of UUIDs are supported. Two are consistent with RFC 4122 and one is a custom variant that provides a 'sequential UUID' that can be advantageous when used as a primary database key.
Note: The Mersenne Twister pseudo-random number generator has excellent statistical properties, but it is not considered cryptographically secure. Pseudo-random UUIDs are not recommended for use as security authentication tokens in cookies or other user-visible session identifiers.
The UUID generally follows the "version 1" spec from the RFC, however the clock sequence and MAC address are randomly generated each time. (This is permissible within the spec of the RFC.) The generated MAC address has the the multicast bit set as mandated by the RFC to ensure it does not conflict with real MAC addresses. This UUID has 60 bits of timestamp data, 61 bits of pseudo-random data and 7 mandated bits (multicast bit, "variant" field and "version" field).
The UUID follows the "version 4" spec, with 122 pseudo-random bits and 6 mandated bits ("variant" field and "version" field).
This is a custom UUID form that resembles "version 4" form, but that overlays the first 60 bits with a timestamp akin to "version 1", Unlike "version 1", this custom version preserves the ordering of bits from high to low, whereas "version 1" puts the low 32 bits of the timestamp first, then the middle 16 bits, then multiplexes the high bits with version field. This "4s" variant provides a "sequential UUID" with the timestamp providing order and the remaining random bits making collision with other UUIDs created at the exact same microsecond highly unlikely. This UUID has 60 timestamp bits, 62 pseudo-random bits and 6 mandated bits ("variant" field and "version" field).
This module focuses on generation of UUIDs with random elements and does not support UUID versions 2, 3 and 5.
my $ug = Data::UUID::MT->new( version => 4 );
Creates a UUID generator object. The only allowed versions are "1", "4" and "4s". If no version is specified, it defaults to "4".
my $uuid = $ug->create;
Returns a UUID packed into a 16 byte string.
my $uuid = $ug->create_hex();
Returns a UUID as a lowercase hex string, prefixed with "0x", e.g. 0xb0470602a64b11da863293ebf1c0e05a
my $uuid = $ug->create_string(); #
Returns UUID as a lowercase string in "standard" format, e.g. b0470602-a64b-11da-8632-93ebf1c0e05a
my $next = $ug->iterator;
my $uuid = $next->();
Returns a reference to the internal UUID generator function. Because this avoids method call overhead, it is slightly faster than calling create.
Reseeds the internal pseudo-random number generator. This happens automatically after a fork or thread creation (assuming Scalar::Util::weaken), but may be called manually if desired for some reason.
Any arguments provided are passed to Math::Random::MT::Auto::srand() for custom seeding.
$ug->reseed('hotbits' => 250, '/dev/random');
A UUID contains 16 bytes. A hex string representation looks like 0xb0470602a64b11da863293ebf1c0e05a. A "standard" representation looks like b0470602-a64b-11da-8632-93ebf1c0e05a. Sometimes these are seen in upper case and on Windows the standard format is often seen wrapped in parentheses.
Converting back and forth is easy with pack and unpack.
# string to 16 bytes
$string =~ s/^0x//i; # remove leading "0x"
$string =~ tr/()-//d; # strip '-' and parentheses
$binary = pack("H*", $string);
# 16 bytes to uppercase string formats
$hex = "0x" . uc unpack("H*", $binary);
$std = uc join "-", unpack("H8H4H4H4H12", $binary);
If you need a module that provides these conversions for you, consider UUID::Tiny.
At the time of writing, there are five other general purpose UUID generators on CPAN that I consider potential alternatives. Data::UUID::MT is included in the discussion below for comparison.
Data::GUID - version 1 UUIDs (wrapper around Data::UUID)
Data::UUID - version 1 or 3 UUIDs (derived from RFC 4122 code)
Data::UUID::LibUUID - version 1 or 4 UUIDs (libuuid)
UUID - version 1 or 4 UUIDs (libuuid)
UUID::Tiny - versions 1, 3, 4, or 5 (pure perl)
Data::UUID::MT - version 1 or 4 (or custom sequential "4s")
libuuid based UUIDs may generally be either version 4 (preferred) or version 1 (fallback), depending on the availability of a good random bit source (e.g. /dev/random). libuuid version 1 UUIDs could also be provided by the uuidd daemon if available.
UUID.pm leaves the choice of version up to libuuid. Data::UUID::LibUUID does so by default, but also allows specifying a specific version. Note that Data::UUID::LibUUID incorrectly refers to version 1 UUIDs as version 2 UUIDs. For example, to get a version 1 binary UUID explicitly, you would call Data::UUID::LibUUID::new_uuid_binary(2).
In addition to differences mentioned below, there are additional slight difference in how the modules (or libuuid) treat the "clock sequence" field and otherwise attempt to keep state between calls, but this is generally immaterial.
Version 1 UUID generators differ in whether they include the Ethernet MAC address as a "node identifier" as specified in RFC 4122. Including the MAC has security implications as Version 1 UUIDs can then be traced to a particular machine at a particular time.
For libuuid based modules, Version 1 UUIDs will include the actual MAC address, if available, or will substitute a random MAC (with multicast bit set).
Data::UUID version 1 UUIDs do not contain the MAC address, but replace it with an MD5 hash of data including the hostname and host id (possibly just the IP address), modified with the multicast bit.
Both UUID::Tiny and Data::UUID::MT version 1 UUIDs do not contain the actual MAC address, but replace it with a random multicast MAC address.
All the modules differ in the source of random bits.
libuuid based modules get random bits from /dev/random or /dev/urandom or fall back to a pseudo-random number generator.
Data::UUID only uses random data to see the clock sequence and gets bits from the C rand() function.
UUID::Tiny uses Perl's rand() function.
Data::UUID::MT gets random bits from Math::Random::MT::Auto, which uses the Mersenne Twister algorithm. Math::Random::MT::Auto seeds from system sources (including Win32 specific ones on that platform) if available and falls back to other less ideal sources if not.
Pseudo-random number generators used in generating UUIDs should be reseeded if the process forks or if threads are created.
Data::UUID::MT checks if the process ID has changed before generating a UUID and reseeds if necessary. If Scalar::Util is installed and provides weaken(), Data::UUID::MT will also reseed its objects on thread creation.
Data::UUID::LibUUID will reseed on fork on Mac OSX.
I have not explored further whether other UUID generators are fork/thread safe.
The examples/bench.pl program included with this module does some simple benchmarking of UUID generation speeds. Here is the output from my desktop system (AMD Phenom II X6 1045T CPU). Note that "v?" is used where the choice is left to libuuid -- which will result in version 4 UUIDs on my system.
Benchmark on Perl v5.14.0 for x86_64-linux with 8 byte integers.
U => UUID 0.02
UT => UUID::Tiny 1.03
DG => Data::GUID 0.046
DU => Data::UUID 1.217
DULU => Data::UUID::LibUUID 0.05
DUMT => Data::UUID::MT 0.001
Benchmarks are marked as to which UUID version is generated.
Some modules offer method ('meth') and function ('func') interfaces.
RFC 4122 A Universally Unique IDentifier (UUID) URN Namespace
Please report any bugs or feature requests through the issue tracker at https://github.com/dagolden/data-uuid-mt/issues. You will be notified automatically of any progress on your issue.
This is open source software. The code repository is available for public review and contribution under the terms of the license.
git clone git://github.com/dagolden/data-uuid-mt.git
David Golden <email@example.com>
Matt Koscica <firstname.lastname@example.org>
This software is Copyright (c) 2011 by David Golden.
This is free software, licensed under:
The Apache License, Version 2.0, January 2004
To install Data::UUID::MT, copy and paste the appropriate command in to your terminal.
perl -MCPAN -e shell
For more information on module installation, please visit the detailed CPAN module installation guide.