uidgen 1.0.2

An adjective-noun-number identifier generator.

Generate "unique" identifiers of the form adjective-noun-number.

Usage

pip install uidgen

Then in python:

>>> import uidgen
>>> uidgen.generate()
spotless-leave-6
>>> uidgen.generate()
wet-diamond-0
>>> uidgen.generate()
dull-court-2

Or via CLI:

$ python -m uidgen
wild-sport-8

How "unique" are these IDs?

The total number of possible identifiers is:

>>> from uidgen.data import adjectives, nouns
>>> len(adjectives) * len(nouns) * 10
19996990

That's roughly 20 million. The uidgen.generate() function samples an ID uniformly from these possibilities. So, if you call uidgen.generate() a total of N times, the probability that all N ID are unique is:

>>> import math
>>> total = len(adjectives) * len(nouns) * 10
>>> for N in [10, 100, 1000, 10000, 100000]:
>>>     print(N, math.prod(total-i for i in range(N)) / total**N)
10 0.9999977496634996
100 0.9997524929700011
1000 0.9753302225997922
10000 0.08204043013318953
100000 1.696918121141932e-109

The takeaway: If you're generating under 200 IDs, its very unlikely you'll see a collision. Beyond 200 IDs, you're really risking a collision.

So what's the use of this package? Other ID specifications such as uuid4 have much stronger probabilistic guarantees. However, imagine naming a training run of a deep learning model d986a320-46e1-477e-8b0c-0f53352a8025. It doesn't exactly roll off the tongue when you're in a meeting discussing the takeways of the technique used in run d986a320-46e1-477e-8b0c-0f53352a8025 vs run f1b4666f-ab0d-409c-a54f-271ac4e33578. You really want something that's easy to say, is memorable, and is unique, which uidgen provides. This is the same technique that wandb.ai uses to name training runs by default.

You should NOT use this library if:

1. You're application cannot handle name collisions (i.e. the application doesn't detect collisions to regenerate an ID).
2. You need to generate a lot of IDs (more than 10000), in which case, even if you regenerate on a collision, you'll be waiting a while until you get a unique ID.