Skip to content

Mocks, fixtures, and snapshots

R-MOCK-1 - Mocks do not replace the behavior under test

Rule. A test may mock dependencies, providers, time, randomness, or network responses, but must not mock away the code path it claims to verify.

Why. Over-mocking creates tests that only prove the mock setup. This is especially easy when AI-generated tests mock every imported module.

vcqa. Flag test files with more mocked app modules than assertions, or tests that mock the subject under test and then assert mocked return values.

References.

R-FIXTURE-1 - Fixtures are realistic and maintained

Rule. Fixtures must represent valid and invalid real-world shapes, and must be updated when public contracts change.

Why. Stale or toy fixtures miss contract drift. Realistic fixtures catch serialization, validation, optional-field, and compatibility failures.

vcqa. Flag fixtures made only of empty objects, single-field examples, or names like foo/bar where the production schema is richer.

References.

R-SNAP-1 - Snapshots are narrow and reviewed

Rule. Snapshots must be small enough to review and paired with behavior assertions when the output has user or API semantics.

Why. Large snapshots are often accepted blindly. They detect change, not correctness.

vcqa. Flag large snapshot files, snapshot-only tests for interactive behavior, and snapshot updates with no nearby assertion or explanation.

References.