faker: Reproducibility of results broken between Faker-0.7.18 and (Faker-0.8.0, 0.8.7)
Faker-0.8.0 and Faker-0.7.18 generate different values when given the same random source.
Is the deterministic generation a goal of the project or not? I assumed that it is before… because the README specifically says that it’s useful for unit testing, which is how I use it.
About this issue
- Original URL
- State: closed
- Created 7 years ago
- Reactions: 1
- Comments: 23 (5 by maintainers)
Commits related to this issue
- add unit test for provider __init__ — committed to joke2k/faker by reverbc 7 years ago
I think it’s fine to use semver, just add a notice that the data inside the providers is subject to change between versions, and that the only guarantee of reproducibility is when using the same faker version.