Awesome
ulid
Universally Unique Lexicographically Sortable Identifier in Python 3.
Status
This project is actively maintained.
Installation
To install ulid from pip:
$ pip install ulid-py
To install ulid from source:
$ git clone git@github.com:ahawker/ulid.git
$ cd ulid && python setup.py install
Usage
Create a brand new ULID.
The timestamp value (48-bits) is from time.time() with millisecond precision.
The randomness value (80-bits) is from os.urandom().
>>> import ulid
>>> ulid.new()
<ULID('01BJQE4QTHMFP0S5J153XCFSP9')>
Create a new ULID from an existing 128-bit value, such as a UUID.
Supports ULID values as int
, bytes
, str
, and UUID
types.
>>> import ulid, uuid
>>> value = uuid.uuid4()
>>> value
UUID('0983d0a2-ff15-4d83-8f37-7dd945b5aa39')
>>> ulid.from_uuid(value)
<ULID('09GF8A5ZRN9P1RYDVXV52VBAHS')>
Create a new ULID from an existing timestamp value, such as a datetime object.
Supports timestamp values as int
, float
, str
, bytes
, bytearray
, memoryview
, datetime
, Timestamp
, and ULID
types.
>>> import datetime, ulid
>>> ulid.from_timestamp(datetime.datetime(1999, 1, 1))
<ULID('00TM9HX0008S220A3PWSFVNFEH')>
Create a new ULID from an existing randomness value.
Supports randomness values as int
, float
, str
, bytes
, bytearray
, memoryview
, Randomness
, and ULID
types.
>>> import os, ulid
>>> randomness = os.urandom(10)
>>> ulid.from_randomness(randomness)
>>> <ULID('01BJQHX2XEDK0VN0GMYWT9JN8S')>
For cases when you don't necessarily control the data type (input from external system), you can use the parse
method
which will attempt to make the correct determination for you. Please note that this will be slightly slower than creating
the instance from the respective from_*
method as it needs to make a number of type/conditional checks.
Supports values as int
, float
, str
, bytes
, bytearray
, memoryview
, uuid.UUID
, and ULID
types.
>>> import ulid
>>> value = db.model.get_id() ## Unsure about datatype -- Could be int, UUID, or string?
>>> ulid.parse(value)
>>> <ULID('0K0EDFETFM8SH912DBBD4ABXSZ')>
Once you have a ULID object, there are a number of ways to interact with it.
The timestamp
method will give you a snapshot view of the first 48-bits of the ULID while the randomness
method
will give you a snapshot of the last 80-bits.
>>> import ulid
>>> u = ulid.new()
>>> u
<ULID('01BJQM7SC7D5VVTG3J68ABFQ3N')>
>>> u.timestamp()
<Timestamp('01BJQM7SC7')>
>>> u.randomness()
<Randomness('D5VVTG3J68ABFQ3N')>
The ULID
, Timestamp
, and Randomness
classes all derive from the same base class, a MemoryView
.
A MemoryView
provides the bin
, bytes
, hex
, int
, oct
, and str
, methods for changing any values representation.
>>> import ulid
>>> u = ulid.new()
>>> u
<ULID('01BJQMF54D093DXEAWZ6JYRPAQ')>
>>> u.timestamp()
<Timestamp('01BJQMF54D')>
>>> u.timestamp().int
1497589322893
>>> u.timestamp().bytes
b'\x01\\\xafG\x94\x8d'
>>> u.timestamp().datetime
datetime.datetime(2017, 6, 16, 5, 2, 2, 893000, tzinfo=datetime.timezone.utc)
>>> u.randomness().bytes
b'\x02F\xde\xb9\\\xf9\xa5\xecYW'
>>> u.bytes[6:] == u.randomness().bytes
True
>>> u.str
'01BJQMF54D093DXEAWZ6JYRPAQ'
>>> u.int
1810474399624548315999517391436142935
>>> u.bin
'0b1010111001010111101000111100101001000110100000010010001101101111010111001010111001111100110100101111011000101100101010111'
>>> u.hex
'0x015caf47948d0246deb95cf9a5ec5957'
>>> u.oct
'0o12712750745106402215572712717464573054527'
A MemoryView
also provides rich comparison functionality.
>>> import datetime, time, ulid
>>> u1 = ulid.new()
>>> time.sleep(5)
>>> u2 = ulid.new()
>>> u1 < u2
True
>>> u3 = ulid.from_timestamp(datetime.datetime(2039, 1, 1))
>>> u1 < u2 < u3
True
>>> [u.timestamp().datetime for u in sorted([u2, u3, u1])]
[datetime.datetime(2017, 6, 16, 5, 7, 14, 847000, tzinfo=datetime.timezone.utc), datetime.datetime(2017, 6, 16, 5, 7, 26, 775000, tzinfo=datetime.timezone.utc), datetime.datetime(2039, 1, 1, 8, 0, tzinfo=datetime.timezone.utc)]
Monotonic Support
This library supports two implementations for stronger guarantees of monotonically increasing randomness.
To use these implementations, simply import and alias it as ulid
. They supports an identical interface as ulid
, so no additional changes should be necessary.
Thread lock
The "thread lock" implementation is a simple implementation that follows that of the ulid/spec
. When two or more identifiers are created with the same millisecond, the subsequent identifiers use the previous identifiers randomness value + 1. See PR 473 for more details.
>>> import time
>>> from ulid import monotonic as ulid
>>> ts = time.time()
>>> ulid.from_timestamp(ts)
<ULID('01EFZ62V7VTEQR4Q788PSBBQP8')>
>>> ulid.from_timestamp(ts)
<ULID('01EFZ62V7VTEQR4Q788PSBBQP9')>
>>> ulid.from_timestamp(ts)
<ULID('01EFZ62V7VTEQR4Q788PSBBQPA')>
Microsecond
The "microsecond" implementation is not defined in the ulid/spec
. It uses a microsecond clock and uses those additional 10-bits into the first two bytes of the randomness value. This means that two identifiers generated within the same millisecond will be monotonically ordered. If two identifiers are generated within the same microsecond, they are ordered entirely by the randomness bytes. See PR 476 for more details.
>>> from ulid import microsecond as ulid
>>> ulid.new()
<ULID('01EH0VVVEC0BKJHF0370TNGQ4Z')>
>>> ulid.new()
<ULID('01EH0VVWPG0C6VDD0529CAHPNJ')>
>>> ulid.new()
<ULID('01EH0VVX8R0AN45DBYZZYMXVKT')>
>>> ulid.new()
<ULID('01EH0VVYA406BDKKRVCDJQZHYQ')>
Contributing
If you would like to contribute, simply fork the repository, push your changes and send a pull request.
Pull requests will be brought into the master
branch via a rebase and fast-forward merge with the goal of having a linear branch history with no merge commits.
License
Why not UUID?
UUID can be suboptimal for many uses-cases because:
- It isn't the most character efficient way of encoding 128 bits of randomness
- UUID v1/v2 is impractical in many environments, as it requires access to a unique, stable MAC address
- UUID v3/v5 requires a unique seed and produces randomly distributed IDs, which can cause fragmentation in many data structures
- UUID v4 provides no other information than randomness which can cause fragmentation in many data structures
ULID provides:
- 128-bit compatibility with UUID
- 1.21e+24 unique ULIDs per millisecond
- Lexicographically sortable!
- Canonically encoded as a 26 character string, as opposed to the 36 character UUID
- Uses Crockford's base32 for better efficiency and readability (5 bits per character)
- Case insensitive
- No special characters (URL safe)
Specification
Below is the current specification of ULID as implemented in this repository.
The binary format is implemented.
01AN4Z07BY 79KA1307SR9X4MV3
|----------| |----------------|
Timestamp Randomness
10chars 16chars
48bits 80bits
Components
Timestamp
- 48 bit integer
- UNIX-time in milliseconds
- Won't run out of space till the year 10895 AD.
Randomness
- 80 bits
- Cryptographically secure source of randomness, if possible
Sorting
The left-most character must be sorted first, and the right-most character sorted last (lexical order). The default ASCII character set must be used. Within the same millisecond, sort order is not guaranteed
Encoding
Crockford's Base32 is used as shown. This alphabet excludes the letters I, L, O, and U to avoid confusion and abuse.
0123456789ABCDEFGHJKMNPQRSTVWXYZ
Binary Layout and Byte Order
The components are encoded as 16 octets. Each component is encoded with the Most Significant Byte first (network byte order).
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 32_bit_uint_time_high |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 16_bit_uint_time_low | 16_bit_uint_random |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 32_bit_uint_random |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 32_bit_uint_random |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
String Representation
ttttttttttrrrrrrrrrrrrrrrr
where
t is Timestamp
r is Randomness