Skip to content
A Python 3 library for better timestamped UUIDs
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


uuid0 - better timestamped UUIDs in Python

V4 UUIDs are often used as primary keys or as part of database indexes. However, their random and non-sequential nature can cause locality and performance issues.

This Python 3 package mitigates these issues by using encoding a UNIX timestamp in the first 6 bytes of the UUID, and filling the remaining bytes with random data. The uuid0.UUID class inherits from the standard uuid.UUID class so it's safe to use all the usual properties (hex, int, bytes, etc.)

The package also contains uuid0.django which provides Django form and model fields based on the UUID0 type, as well as an abstract UUID0Model.


This package is in alpha status, use at your own risk.


Install by running:

pip install uuid0

Example usage:

>>> import uuid0

# make a UUID based on the current time
>>> uuid0.generate()

# make a UUID from a string of hex digits (braces and hyphens ignored)
>>> x = uuid0.UUID('{0b7dd8d0-7e40-8360-9322-4a361d7b573f}')

# extract the datetime from a UUID
>>> str(x.datetime)
'2010-01-15 00:00:36'

# convert a UUID to a string of hex digits in standard form
>>> str(x)

# convert a UUID to a base62 string
>>> x.base62

# make a UUID from a base62 string
>>> uuid0.UUID(base62=x.base62)

# get the raw 16 bytes of the UUID
>>> x.bytes

# make a UUID from a 16-byte string
>>> uuid0.UUID(bytes=x.bytes)

What about collisions?

Collisions are a concern, of course, but only if you are generating really large amounts of UUIDs in a short timespan.

By default, uuid0 only uses 6 bytes to encode the time and does not encode any version bits, leaving 10 bytes (80 bits) of random data.

The probability of collision is near zero for most use cases since for each 1/10000s period, there are 280 possible UUIDs.


Generating a UUID0 is slower than generating a UUID v4, but faster than a UUID v1.

Below are results for 500k iterations on an i7-6700HQ and Python 3.6.2 (generated using the benchmarks/ script):

method it/s ?s/it % slower
uuid.uuid4() 453043 2.207 0.0%
uuid0.generate() 311184 3.214 31.31%
uuid.uuid1() 230163 4.345 49.2%

In terms of database performance, indexes are approximately 18% smaller and inserts about 23% faster on PostgreSQL with a uuid type column, but YMMV depending on your use case.


This project is licensed under the MIT License. See the LICENSE file for details.

You can’t perform that action at this time.