testing-patterns

Trait-based dependency injection abstracts I/O and external dependencies as traits, enabling you to inject real implementations in production and fake or mock implementations in tests. Define an algebra trait (e.g., Clock, Database), create a RealClock adapter for production, and a FixedClock fake for testing. Pass the trait object or generic parameter to your domain logic, keeping business rules pure and testable.

Hand-rolled fakes are best for simple, deterministic test scenarios where you control exact behavior—like a FixedClock returning a preset time. Expectation-based mocks (mockall, wiremock) suit complex interactions where you verify call sequences, argument counts, or specific invocations. Fakes are lighter and faster; mocks provide stricter verification but add overhead.

Property-based testing uses proptest to generate random inputs and verify that invariants and business rules hold across many cases, not just hand-written examples. It catches edge cases and off-by-one errors that unit tests miss, ensuring your domain logic is robust and correct under diverse conditions beyond typical scenarios.

Model async I/O (HTTP, database, timers) as async trait methods on your effect algebra. Implement a real adapter for production and a fake async implementation for tests. Use #[tokio::test] or similar to run async test functions, inject the fake, and verify behavior without touching real external systems.

Yes. Generics (monomorphic approach) offer zero-cost abstraction and compile-time verification but require generic parameters throughout your call stack. Trait objects provide runtime polymorphism and simpler signatures but add dynamic dispatch overhead. Choose generics for performance-critical code, trait objects for flexibility and simpler APIs.

Pure domain logic—free of I/O and side effects—is testable, deterministic, and easy to reason about. By pushing all effects to injected traits, you separate business rules from infrastructure concerns, enabling fast, isolated unit tests without mocks and reducing test fragility when external systems change.