Category: python

← Back to all posts

Types That Can't Lie: Safe Type Modeling in Python

pythontyping·by Adrian Benavides·September 2, 2026·6 min read

A string field accepts anything, and the crash lands far from the cause. Design Python types that can't lie: value objects, canonical enums, and tagged unions the type checker can rely on.

Immutability and Defensive State Design in Python

pythontyping·by Adrian Benavides·August 6, 2026·8 min read

Frozen dataclasses are the default for domain value objects, but they don't validate at runtime. msgspec.Struct closes that gap with a single type that decodes, validates, and stays immutable across the wire boundary and the domain layer. attrs and cattrs cover the same ground in two libraries, and pydantic is only worth it when the ecosystem integration matters.

Cross-Cutting Concerns Without Constructor Over-Injection

pythonarchitecture·by Adrian Benavides·August 3, 2026·6 min read

Use contextvars.ContextVar for cross-cutting concerns that almost every service needs, without threading them through every constructor. Covers per-task isolation, structlog and opentelemetry designs, sync-vs-async leaks, and the failure modes that show up in production.

Hexagonal Architecture in Python: Building Decoupled, Testable Systems

pythonarchitecture·by Adrian Benavides·July 10, 2026·9 min read

Build decoupled, testable Python systems with Hexagonal Architecture. Covers ports, adapters, driven/driving roles, structural typing with Protocol, and the pitfalls of port proliferation, primitive obsession, and infrastructure leakage.

Domain-Driven Design in Python: Preventing the Big Ball of Mud

pythonarchitecture·by Adrian Benavides·July 2, 2026·11 min read

Apply DDD tactical patterns in Python with immutable value objects, aggregate roots, and bounded contexts (the modular monolith default, the modern tooling landscape, and when DDD is overkill).