Note: This is the reference implementation of the IntToUuid specification.
Bidirectionally encodes a non-negative 64-bit unsigned "id" integer and optional 32-bit "namespace" integer into a valid RFC 9562 Version 8 UUID. The id and namespace integers are encoded to obscure their value and produce non-sequential UUIDs, while guaranteeing uniqueness and reproducibility.
This could be used to present an auto-incrementing integer "database id" as a UUID (proxy ID) in a public context, where you would not want to expose an enumerable, sequential value directly tied to your database structure/data. Since the encoded UUID can be converted back into integer namespace and id values at runtime, the UUID does not need to be persisted in the database or otherwise indexed to the ID it represents.
Note: The integer ID and namespace values are only encoded in the UUID, not encrypted, and the value can be recovered by a third party with effort. This library is intended to support on-demand conversion between an integer and a UUID, while mitigating basic "user enumeration attacks". Securely encrypting a 64-bit integer in the 122 bits available in a UUID is currently outside the scope of this library.
$id = \WickedByte\IntToUuid\IntegerId::make(12);
$uuid = \WickedByte\IntToUuid\IntToUuid::encode($id);
echo $uuid->toString(); // c81f423b-2ca0-8963-aefa-f067a191123f$id = \WickedByte\IntToUuid\IntegerId::make(42, 12);
$uuid = \WickedByte\IntToUuid\IntToUuid::encode($id);
echo $uuid->toString(); // dee5e9d2-c3e4-8273-b0d5-b3b5307bf749$uuid = \Ramsey\Uuid\Uuid::fromString('dee5e9d2-c3e4-8273-b0d5-b3b5307bf749');
$id = \WickedByte\IntToUuid\IntToUuid::decode($uuid);
echo $id->value; // 42
echo $id->namespace; // 12Encoding an integer uses a deterministic seed based on the xxHash (xxh3) hash
of the concatenated binary strings packed from the id and namespace values. The
first 32-bits of the hash are used as the contiguous time_hi_and_version,
clock_seq_hi_and_reserved, and clock_seq_low fields. To comply with the RFC
9562 Version 8, the seed is multiplexed with the required Version and Variant
bits, leaving 26 bits of deterministic "pseudo-randomness". The encoded id is
the id integer packed as a 64-bit binary string XOR the xxHash hash of the
namespace and seed. The encoded namespace is the namespace integer packed as a
32-bit binary string XOR the xxHash hash of the seed. The resulting octets are
arranged into a valid UUID and a new UuidInterface (from the ramsey/uuid
library) is returned.
Decoding is the reverse of the encoding process: the UUID octets are split into the encoded id, encoded namespace, and seed binary strings, XOR is applied to the encoded values and corresponding hashes, and a "checksum" seed is produced from the decoded binary strings, which are then unpacked into integer values. If the seed value from the UUID does not match the checksum seed, then UUID does not encode valid information, and an exception is thrown. An exception is also thrown if the UUID passed into the decode function is not a valid Version 8 UUID.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| time_low |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| time_mid | time_hi_and_version |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|clk_seq_hi_res | clk_seq_low | node (0-1) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| node (2-5) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| namespace |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| id (0-1) | seed (0-1) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| seed (2-3) | id (2-3) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| id (4-7) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Note: RFC 9562 incorporates and obsoletes the more well-known UUID specification, RFC 9562. It is common to see "on spec" UUIDs referred by either RFC 9562 or RFC 4122.
The other UUID versions defined by RFC 9562 have distinct generation algorithms and properties. Versions 1, 2, 6, and 7 are based on the current timestamp. Version 3 (Name-Based MD5) and Version 5 (Name-Based SHA1) are deterministic for a string "name" and "namespace" values, but are unidirectional because they are based on hash functions. Version 4 (Random) comes the closest to fulfilling our needs: 122 of the 128 bits are randomly/pseudo-randomly generated. The same algorithm used here could be used to generate encoded UUIDs that look like Version 4 UUIDs, but they would not be technically compatible with the RFC definition, or have the expected universal uniqueness property.
Version 8 defines an RFC-compatible format for experimental or vendor-defined UUIDs. The definition allows for both implementation-specific uniqueness and for the embedding of arbitrary information, both of which are key to this particular use case.