test-data-management

Define a factory class extending factory.Factory with a Meta model, then declare fields using Sequence, Faker, LazyAttribute, and SubFactory. Use create_batch for multiple instances and post_generation hooks for related objects.

Instantiate Faker and call providers like user_name, email, address, and date_of_birth to produce realistic values. Combine providers in generator functions to build complete dictionaries for users, posts, or other entities.

Factory Boy suits complex models with relationships and varied attributes, while plain pytest fixtures work for simple static setup. Factories reduce duplication when tests need many object variations.

Wrap each test in a database transaction using an autouse fixture and roll it back after the test completes. Alternatively, truncate all tables in a teardown fixture to guarantee a clean state.

Avoid factories for trivial single-object setups where a simple fixture suffices, and never use them with production data. They also add overhead when tests require only static, unchanging values.