Dataclasses vs Pydantic vs attrs
Choose the right modeling tool.
Recipe
Quick-reference recipe card - copy-paste ready.
@dataclass
class Point:
x: int
y: intWhen to reach for this:
- Internal DTOs
- Validated API edges
- Immutable value objects
Working Example
class PointModel(BaseModel):
model_config = ConfigDict(frozen=True)
x: int
y: intWhat this demonstrates:
- dataclass simplicity
- Pydantic validation
- frozen models
Deep Dive
How It Works
- dataclasses: stdlib, light.
- Pydantic: validation + schema.
- attrs: powerful, less API schema focus.
Gotchas
- Boundary validation skipped - Invalid data reaches persistence layers.. Fix: Validate with Pydantic or framework forms at the edge..
- Leaking stack traces - Clients see internal errors.. Fix: Map exceptions to stable HTTP responses..
- Blocking async event loops - Workers stall under concurrent load.. Fix: Use async drivers or threadpool wrappers..
- Secrets in source control - Credentials leak via git history.. Fix: Load secrets from env or a vault at runtime..
- Missing observability - Incidents are hard to debug.. Fix: Add structured logs, metrics, and request IDs..
Alternatives
| Alternative | Use When | Don't Use When |
|---|---|---|
| Alternate framework in this cookbook | Team standard or existing monolith | Greenfield API with different constraints |
| Managed BaaS | CRUD-only MVP | Custom auth, workflows, or compliance needs |
| gRPC | Internal high-performance RPC | Public HTTP clients and browser access |
FAQs
When should I adopt modeling tool choice?
Use it when the patterns and trade-offs on this page match your API or data boundary.
What is the top production mistake with modeling tool choice?
Skipping validation, timeouts, or explicit error contracts at the HTTP edge.
How do I test modeling tool choice?
Use the framework test client, override dependencies, and assert status plus JSON shape.
Does modeling tool choice work with Python 3.14?
Yes - examples target Python 3.14 with pinned framework versions from the stack footer.
How does modeling tool choice relate to Pydantic 2?
Validate and serialize at boundaries; keep services working with typed domain objects.
Sync or async?
Prefer async routes when I/O dominates; keep CPU work small or offload to workers.
Where should business logic live?
Thin handlers; services own rules; repositories own queries.
How do I document APIs?
Publish OpenAPI or schema docs that match response models in code.
How do I handle versioning?
Explicit URL or header versioning with deprecation windows - avoid silent breaks.
What should I read next?
Follow the Related links for the next layer of depth in this section.
How do I stay secure?
Authenticate callers, authorize per resource, rate-limit, and never log secrets.
Performance first step?
Measure DB and upstream latency before swapping frameworks.
Related
- Pydantic Basics - Core models
- Validators - Custom rules
- Serialization - model_dump
- Settings Management - env config
Stack versions: This page was written for Python 3.14.0 (stable 3.14, maintenance 3.13), FastAPI 0.115+, Django 5.2, Flask 3.1, Pydantic 2, PyTorch 2.6+, pandas 2.2+, Polars 1.x, ruff 0.9+, and uv 0.6+.