feat: add UUIDv7 documentation and customization

This commit is contained in:
Ben Ramsey
2022-09-14 18:21:10 -05:00
parent 28f509bd91
commit 4f4deb1dd6
17 changed files with 348 additions and 31 deletions
+5 -3
View File
@@ -4,10 +4,12 @@
Ordered-time Codec
==================
.. hint::
.. attention::
:ref:`Version 6, reordered time UUIDs <rfc4122.version6>` are a
new version of UUID that take the place of ordered-time UUIDs.
:ref:`Version 6, reordered time UUIDs <rfc4122.version6>` are a new version
of UUID that eliminate the need for the ordered-time codec. If you aren't
currently using the ordered-time codec, and you need time-based, sortable
UUIDs, consider using version 6 UUIDs.
UUIDs arrange their bytes according to the standard recommended by `RFC 4122`_.
Unfortunately, this means the bytes aren't in an arrangement that supports
@@ -4,6 +4,13 @@
Timestamp-first COMB Codec
==========================
.. attention::
:ref:`Version 7, Unix Epoch time UUIDs <rfc4122.version7>` are a new version
of UUID that eliminate the need for the timestamp-first COMB codec. If you
aren't currently using the timestamp-first COMB codec, and you need
time-based, sortable UUIDs, consider using version 7 UUIDs.
:ref:`Version 4, random UUIDs <rfc4122.version4>` are doubly problematic when it
comes to sorting and storing to databases (see :ref:`database.order`), since
their values are random, and there is no timestamp associated with them that may
+11 -12
View File
@@ -229,29 +229,28 @@ small.
Insertion Order and Sorting
###########################
UUIDs are not *monotonically increasing*. Even time-based UUIDs are not. If
using UUIDs as primary keys, the inserts will be random, and the data will be
UUID versions 1, 2, 3, 4, and 5 are not *monotonically increasing*. If using
these versions as primary keys, the inserts will be random, and the data will be
scattered on disk (for InnoDB). Over time, as the database size grows, lookups
will become slower and slower.
.. note::
.. tip::
See Percona's "`Storing UUID Values in MySQL`_" post, for more details on
the performance of UUIDs as primary keys.
To minimize these problems, two solutions have been devised:
1. Timestamp first COMBs
2. Ordered-time UUIDs
1. :ref:`rfc4122.version6` UUIDs
2. :ref:`rfc4122.version7` UUIDs
:ref:`customize.timestamp-first-comb-codec` explains the first solution and how
to use ramsey/uuid to implement it, while :ref:`customize.ordered-time-codec`
explains how to use ramsey/uuid to implement the second solution.
.. note::
.. hint::
:ref:`Version 6, reordered time UUIDs <rfc4122.version6>` are a
new version of UUID that take the place of ordered-time UUIDs.
We previously recommended the use of the :ref:`timestamp-first COMB
<customize.timestamp-first-comb-codec>` or :ref:`ordered-time
<customize.ordered-time-codec>` codecs to solve these problems. However,
UUID versions 6 and 7 were defined to provide these solutions in a
standardized way.
.. _ramsey/uuid-doctrine: https://github.com/ramsey/uuid-doctrine
+3 -4
View File
@@ -5,9 +5,9 @@ Introduction
============
ramsey/uuid is a PHP library for generating and working with `RFC 4122`_ version
1, 2, 3, 4, and 5 universally unique identifiers (UUID). ramsey/uuid also
supports optional and non-standard features, such as `version 6 UUIDs`_,
GUIDs, and other approaches for encoding/decoding UUIDs.
1, 2, 3, 4, 5, 6, and 7 universally unique identifiers (UUID). ramsey/uuid also
supports optional and non-standard features, such as GUIDs and other approaches
for encoding/decoding UUIDs.
What Is a UUID?
###############
@@ -29,4 +29,3 @@ UUIDs can also be stored in binary format, as a string of 16 bytes.
.. _RFC 4122: https://tools.ietf.org/html/rfc4122
.. _version 6 UUIDs: http://gh.peabody.io/uuidv6/
+2
View File
@@ -97,6 +97,8 @@ library.
- This generates a :ref:`rfc4122.version5` UUID.
* - :php:meth:`Uuid::uuid6() <Ramsey\\Uuid\\Uuid::uuid6>`
- This generates a :ref:`rfc4122.version6` UUID.
* - :php:meth:`Uuid::uuid7() <Ramsey\\Uuid\\Uuid::uuid7>`
- This generates a :ref:`rfc4122.version7` UUID.
* - :php:meth:`Uuid::isValid() <Ramsey\\Uuid\\Uuid::isValid>`
- Checks whether a string is a valid UUID.
* - :php:meth:`Uuid::fromString() <Ramsey\\Uuid\\Uuid::fromString>`
+1
View File
@@ -18,6 +18,7 @@ Reference
reference/rfc4122-uuidv4
reference/rfc4122-uuidv5
reference/rfc4122-uuidv6
reference/rfc4122-uuidv7
reference/guid-fields
reference/guid-guid
reference/nonstandard-fields
+20
View File
@@ -0,0 +1,20 @@
.. _reference.rfc4122.uuidv7:
===============
Rfc4122\\UuidV7
===============
.. php:namespace:: Ramsey\Uuid\Rfc4122
.. php:class:: UuidV7
Implements :php:interface:`Ramsey\\Uuid\\Rfc4122\\UuidInterface`.
UuidV7 represents a :ref:`version 7, Unix Epoch time UUID <rfc4122.version7>`.
In addition to providing the methods defined on the interface, this class
additionally provides the following methods.
.. php:method:: getDateTime()
:returns: A date object representing the timestamp associated with the UUID.
:returntype: ``\DateTimeInterface``
+12 -4
View File
@@ -14,11 +14,12 @@ RFC 4122 UUIDs
rfc4122/version4
rfc4122/version5
rfc4122/version6
rfc4122/version7
`RFC 4122`_ defines five versions of UUID. Each version has different generation
algorithms and properties. Which one you choose to use depends on your use-case.
You can find out more about their applications on the specific page for that
version.
`RFC 4122`_ defines five versions of UUID, while a `new Internet-Draft under
review`_ defines three new versions. Each version has different generation
algorithms and properties. Which one you choose depends on your use-case. You
can find out more about their applications on the specific page for that version.
Version 1: Gregorian Time
This version of UUID combines a timestamp, node value (in the form of a MAC
@@ -50,5 +51,12 @@ Version 6: Reordered Time
:ref:`version 1 UUID <rfc4122.version1>` with a *monotonically increasing*
UUID. For more details, see :ref:`rfc4122.version6`.
Version 7: Unix Epoch Time
This version of UUID combines a timestamp--based on milliseconds elapsed
since the Unix Epoch--and random bytes to create a monotonically increasing,
sortable UUID without the privacy and entropy concerns associated with
version 1 and version 6 UUIDs. For more details, see :ref:`rfc4122.version7`.
.. _RFC 4122: https://tools.ietf.org/html/rfc4122
.. _new Internet-Draft under review: https://datatracker.ietf.org/doc/html/draft-peabody-dispatch-new-uuid-format-04
+6
View File
@@ -4,6 +4,12 @@
Version 1: Gregorian Time
=========================
.. attention::
If you need a time-based UUID, and you don't need the other features
included in version 1 UUIDs, we recommend using
:ref:`version 7 UUIDs <rfc4122.version7>`.
A version 1 UUID uses the current time, along with the MAC address (or *node*)
for a network interface on the local machine. This serves two purposes:
+7 -1
View File
@@ -11,6 +11,12 @@ Version 6: Reordered Time
through the IETF process, the version 6 format is not expected to change
in any way that breaks compatibility.
.. attention::
If you need a time-based UUID, and you don't need the other features
included in version 6 UUIDs, we recommend using
:ref:`version 7 UUIDs <rfc4122.version7>`.
Version 6 UUIDs solve `two problems that have long existed`_ with the use of
:ref:`version 1 <rfc4122.version1>` UUIDs:
@@ -199,7 +205,7 @@ machines, see :ref:`rfc4122.version6.nodes`.
If you do not need an identifier with a node value embedded in it, but you still
need the benefit of a monotonically increasing unique identifier, see
:ref:`customize.timestamp-first-comb-codec`.
:ref:`rfc4122.version7`.
.. _Internet-Draft under review: https://datatracker.ietf.org/doc/draft-peabody-dispatch-new-uuid-format/
+173
View File
@@ -0,0 +1,173 @@
.. _rfc4122.version7:
==========================
Version 7: Unix Epoch Time
==========================
.. note::
Version 7, Unix Epoch time UUIDs are a new format of UUID, proposed in an
`Internet-Draft under review`_ at the IETF. While the draft is still going
through the IETF process, the version 7 format is not expected to change
in any way that breaks compatibility.
.. admonition:: ULIDs and Version 7 UUIDs
:class: hint
Version 7 UUIDs are binary-compatible with `ULIDs`_ (universally unique
lexicographically-sortable identifiers).
Both use a 48-bit timestamp in milliseconds since the Unix Epoch, filling
the rest with random data. Version 7 UUIDs then add the version and variant
bits required by the UUID specification, which reduces the randomness from
80 bits to 74. Otherwise, they are identical.
You may even convert a version 7 UUID to a ULID.
:ref:`See below for an example. <rfc4122.version7.ulid>`
Version 7 UUIDs solve `two problems that have long existed`_ with the use of
:ref:`version 1 <rfc4122.version1>` UUIDs:
1. Scattered database records
2. Inability to sort by an identifier in a meaningful way (i.e., insert order)
To overcome these issues, we need the ability to generate UUIDs that are
*monotonically increasing*.
:ref:`Version 6 UUIDs <rfc4122.version6>` provide an excellent solution for
those who need monotonically increasing, sortable UUIDs with the features of
version 1 UUIDs (MAC address and clock sequence), but if those features aren't
necessary for your application, using a version 6 UUID might be overkill.
Version 7 UUIDs combine random data (like version 4 UUIDs) with a timestamp (in
milliseconds since the Unix Epoch, i.e., 1970-01-01 00:00:00 UTC) to create a
monotonically increasing, sortable UUID that doesn't have any privacy concerns,
since it doesn't include a MAC address.
For this reason, implementations should use version 7 UUIDs over versions 1 and
6, if possible.
.. code-block:: php
:caption: Generate a version 7, Unix Epoch time UUID
:name: rfc4122.version7.example
use Ramsey\Uuid\Uuid;
$uuid = Uuid::uuid7();
printf(
"UUID: %s\nVersion: %d\nDate: %s\n",
$uuid->toString(),
$uuid->getFields()->getVersion(),
$uuid->getDateTime()->format('r'),
);
This will generate a version 7 UUID and print out its string representation and
the time it was created.
It will look something like this:
.. code-block:: text
UUID: 01833ce0-3486-7bfd-84a1-ad157cf64005
Version: 7
Date: Wed, 14 Sep 2022 16:41:10 +0000
To use an existing date and time to generate a version 7 UUID, you may pass a
``\DateTimeInterface`` instance to the ``uuid7()`` method.
.. code-block:: php
:caption: Generate a version 7 UUID from an existing date and time
:name: rfc4122.version7.example-datetime
use DateTimeImmutable;
use Ramsey\Uuid\Uuid;
$dateTime = new DateTimeImmutable('@281474976710.655');
$uuid = Uuid::uuid7($dateTime);
printf(
"UUID: %s\nVersion: %d\nDate: %s\n",
$uuid->toString(),
$uuid->getFields()->getVersion(),
$uuid->getDateTime()->format('r'),
);
Which will print something like this:
.. code-block:: text
UUID: ffffffff-ffff-7964-a8f6-001336ac20cb
Version: 7
Date: Tue, 02 Aug 10889 05:31:50 +0000
.. tip::
Version 7 UUIDs generated in ramsey/uuid are instances of UuidV7. Check out
the :php:class:`Ramsey\\Uuid\\Rfc4122\\UuidV7` API documentation to learn
more about what you can do with a UuidV7 instance.
.. _rfc4122.version7.ulid:
Convert a Version 7 UUID to a ULID
##################################
As mentioned in the callout above, version 7 UUIDs are binary-compatible with
`ULIDs`_. This means you can encode a version 7 UUID using `Crockford's Base 32
algorithm`_ and it will be a valid ULID, timestamp and all.
Using the third-party library `tuupola/base32`_, here's how we can encode a
version 7 UUID as a ULID. Note that there's a little bit of work to perform the
conversion, since you're working with different bases.
.. code-block:: php
:caption: Encode a version 7, Unix Epoch time UUID as a ULID
:name: rfc4122.version7.example-ulid
use Ramsey\Uuid\Uuid;
use Tuupola\Base32;
$crockford = new Base32([
'characters' => Base32::CROCKFORD,
'padding' => false,
'crockford' => true,
]);
$uuid = Uuid::uuid7();
// First, we must pad the 16-byte string to 20 bytes
// for proper conversion without data loss.
$bytes = str_pad($uuid->getBytes(), 20, "\x00", STR_PAD_LEFT);
// Use Crockford's Base 32 encoding algorithm.
$encoded = $crockford->encode($bytes);
// That 20-byte string was encoded to 32 characters to avoid loss
// of data. We must strip off the first 6 characters--which are
// all zeros--to get a valid 26-character ULID string.
$ulid = substr($encoded, 6);
printf("ULID: %s\n", $ulid);
This will print something like this:
.. code-block:: text
ULID: 01GCZ05N3JFRKBRWKNGCQZGP44
.. caution::
Be aware that all version 7 UUIDs may be converted to ULIDs but not all
ULIDs may be converted to UUIDs.
For that matter, all UUIDs of any version may be encoded as ULIDs, but they
will not be monotonically increasing and sortable unless they are version 7
UUIDs. You will also not be able to extract a meaningful timestamp from the
ULID, unless it was converted from a version 7 UUID.
.. _ULIDs: https://github.com/ulid/spec
.. _Internet-Draft under review: https://datatracker.ietf.org/doc/html/draft-peabody-dispatch-new-uuid-format-04#section-5.2
.. _two problems that have long existed: https://www.percona.com/blog/2014/12/19/store-uuid-optimized-way/
.. _Crockford's Base 32 algorithm: https://www.crockford.com/base32.html
.. _tuupola/base32: https://packagist.org/packages/tuupola/base32
+6 -2
View File
@@ -683,16 +683,20 @@ class Uuid implements UuidInterface
/**
* Returns a version 7 (Unix Epoch time) UUID
*
* @param DateTimeInterface|null $dateTime An optional date/time from which
* to create the version 7 UUID. If not provided, the UUID is generated
* using the current date/time.
*
* @return UuidInterface A UuidInterface instance that represents a
* version 7 UUID
*/
public static function uuid7(): UuidInterface
public static function uuid7(?DateTimeInterface $dateTime = null): UuidInterface
{
$factory = self::getFactory();
if (method_exists($factory, 'uuid7')) {
/** @var UuidInterface */
return $factory->uuid7();
return $factory->uuid7($dateTime);
}
throw new UnsupportedOperationException(
+25 -2
View File
@@ -18,13 +18,16 @@ use DateTimeInterface;
use Ramsey\Uuid\Builder\UuidBuilderInterface;
use Ramsey\Uuid\Codec\CodecInterface;
use Ramsey\Uuid\Converter\NumberConverterInterface;
use Ramsey\Uuid\Converter\Time\UnixTimeConverter;
use Ramsey\Uuid\Converter\TimeConverterInterface;
use Ramsey\Uuid\Generator\DceSecurityGeneratorInterface;
use Ramsey\Uuid\Generator\DefaultTimeGenerator;
use Ramsey\Uuid\Generator\NameGeneratorInterface;
use Ramsey\Uuid\Generator\RandomGeneratorInterface;
use Ramsey\Uuid\Generator\TimeGeneratorInterface;
use Ramsey\Uuid\Generator\UnixTimeGenerator;
use Ramsey\Uuid\Lazy\LazyUuidFromString;
use Ramsey\Uuid\Math\BrickMathCalculator;
use Ramsey\Uuid\Provider\NodeProviderInterface;
use Ramsey\Uuid\Provider\Time\FixedTimeProvider;
use Ramsey\Uuid\Type\Hexadecimal;
@@ -421,12 +424,32 @@ class UuidFactory implements UuidFactoryInterface
/**
* Returns a version 7 (Unix Epoch time) UUID
*
* @param DateTimeInterface|null $dateTime An optional date/time from which
* to create the version 7 UUID. If not provided, the UUID is generated
* using the current date/time.
*
* @return UuidInterface A UuidInterface instance that represents a
* version 7 UUID
*/
public function uuid7(): UuidInterface
public function uuid7(?DateTimeInterface $dateTime = null): UuidInterface
{
return $this->uuidFromBytesAndVersion($this->unixTimeGenerator->generate(), Uuid::UUID_TYPE_UNIX_TIME);
if ($dateTime !== null) {
$timeProvider = new FixedTimeProvider(
new Time($dateTime->format('U'), $dateTime->format('u'))
);
$timeGenerator = new UnixTimeGenerator(
new UnixTimeConverter(new BrickMathCalculator()),
$timeProvider,
$this->randomGenerator,
);
$bytes = $timeGenerator->generate();
} else {
$bytes = $this->unixTimeGenerator->generate();
}
return $this->uuidFromBytesAndVersion($bytes, Uuid::UUID_TYPE_UNIX_TIME);
}
/**
+15
View File
@@ -15,6 +15,7 @@ declare(strict_types=1);
namespace Ramsey\Uuid;
use DateTimeInterface;
use Ramsey\Uuid\Type\Hexadecimal;
use Ramsey\Uuid\Type\Integer as IntegerObject;
@@ -121,3 +122,17 @@ function v6(?Hexadecimal $node = null, ?int $clockSeq = null): string
{
return Uuid::uuid6($node, $clockSeq)->toString();
}
/**
* Returns a version 7 (Unix Epoch time) UUID
*
* @param DateTimeInterface|null $dateTime An optional date/time from which
* to create the version 7 UUID. If not provided, the UUID is generated
* using the current date/time.
*
* @return non-empty-string Version 7 UUID as a string
*/
function v7(?DateTimeInterface $dateTime = null): string
{
return Uuid::uuid7($dateTime)->toString();
}
@@ -52,7 +52,7 @@ class UnixTimeConverterTest extends TestCase
'microseconds' => '999000',
],
// This is the last possible time supported by v7 UUIDs.
// This is the last possible time supported by v7 UUIDs (2 ^ 48 - 1).
// 10889-08-02 05:31:50.655 +00:00
[
'uuidTimestamp' => new Hexadecimal('ffffffffffff'),
@@ -155,7 +155,7 @@ class UnixTimeConverterTest extends TestCase
'expected' => '000000000000',
],
// This is the last possible time supported by v7 UUIDs:
// This is the last possible time supported by v7 UUIDs (2 ^ 48 - 1):
// 10889-08-02 05:31:50.655 +00:00
[
'seconds' => '281474976710',
+38 -1
View File
@@ -4,7 +4,10 @@ declare(strict_types=1);
namespace Ramsey\Uuid\Test;
use DateTimeImmutable;
use DateTimeInterface;
use Ramsey\Uuid\Rfc4122\FieldsInterface;
use Ramsey\Uuid\Rfc4122\UuidV7;
use Ramsey\Uuid\Type\Hexadecimal;
use Ramsey\Uuid\Type\Integer as IntegerObject;
use Ramsey\Uuid\Uuid;
@@ -15,6 +18,7 @@ use function Ramsey\Uuid\v3;
use function Ramsey\Uuid\v4;
use function Ramsey\Uuid\v5;
use function Ramsey\Uuid\v6;
use function Ramsey\Uuid\v7;
class FunctionsTest extends TestCase
{
@@ -79,6 +83,39 @@ class FunctionsTest extends TestCase
$fields = Uuid::fromString($v6)->getFields();
$this->assertIsString($v6);
$this->assertSame(Uuid::UUID_TYPE_PEABODY, $fields->getVersion());
$this->assertSame(Uuid::UUID_TYPE_REORDERED_TIME, $fields->getVersion());
}
public function testV7ReturnsVersion7UuidString(): void
{
$v7 = v7();
/** @var UuidV7 $uuid */
$uuid = Uuid::fromString($v7);
/** @var FieldsInterface $fields */
$fields = $uuid->getFields();
$this->assertIsString($v7);
$this->assertSame(Uuid::UUID_TYPE_UNIX_TIME, $fields->getVersion());
$this->assertInstanceOf(DateTimeInterface::class, $uuid->getDateTime());
}
public function testV7WithCustomDateTimeReturnsVersion7UuidString(): void
{
$dateTime = new DateTimeImmutable('2022-09-14T22:44:33+00:00');
$v7 = v7($dateTime);
/** @var UuidV7 $uuid */
$uuid = Uuid::fromString($v7);
/** @var FieldsInterface $fields */
$fields = $uuid->getFields();
$this->assertIsString($v7);
$this->assertSame(Uuid::UUID_TYPE_UNIX_TIME, $fields->getVersion());
$this->assertInstanceOf(DateTimeInterface::class, $uuid->getDateTime());
$this->assertSame(1663195473, $uuid->getDateTime()->getTimestamp());
}
}
+15
View File
@@ -6,6 +6,7 @@ namespace Ramsey\Uuid\Test;
use Brick\Math\BigDecimal;
use Brick\Math\RoundingMode;
use DateTimeImmutable;
use DateTimeInterface;
use Mockery;
use Mockery\MockInterface;
@@ -728,6 +729,20 @@ class UuidTest extends TestCase
Uuid::uuid7();
}
public function testUuid7WithDateTime(): void
{
$dateTime = new DateTimeImmutable('@281474976710.655');
$uuid = Uuid::uuid7($dateTime);
$this->assertInstanceOf(DateTimeInterface::class, $uuid->getDateTime());
$this->assertSame(2, $uuid->getVariant());
$this->assertSame(7, $uuid->getVersion());
$this->assertSame(
'10889-08-02T05:31:50.655+00:00',
$uuid->getDateTime()->format(DateTimeInterface::RFC3339_EXTENDED),
);
}
/**
* Tests known version-3 UUIDs
*