Backend5 min read

Pydantic & Data Validation

Validate and serialize Python data with Pydantic - type-safe models, nested validation, custom validators, settings management, and FastAPI integration.

PydanticPythonData ValidationType SafetyFastAPI

skill: Pydantic & Data Validation category: Backend tags: [Pydantic, Python, Data Validation, Type Safety, FastAPI] readingTime: 5

Pydantic validates data at the boundary of your system - API inputs, config files, external API responses - so business logic sees clean, typed data. Catching malformed data at the boundary is cheaper than handling it deep in business logic.

Workflow

Define models with types and validationPydantic BaseModel parses and validates data on instantiation - wrong types raise ValidationError with field-level error messages.
Write custom validatorsField constraints (min_length, ge) cover simple rules; custom validators encode domain logic like uniqueness or cross-field constraints.
Use nested models and serializationNested Pydantic models validate complex structures recursively. model_dump() serializes to dict for JSON responses.
Manage settings with BaseSettingsBaseSettings reads from environment variables automatically - typed config with validation at startup, not at runtime.
Integrate with FastAPIFastAPI uses Pydantic for request/response validation and OpenAPI schema generation - no glue code needed.

Output Format

Pydantic models: Field constraints, custom validators, nested models, BaseSettings for config, FastAPI integration.

class MyModel(BaseModel):
    field: type = Field(..., constraint)

    @field_validator('field')
    @classmethod
    def validate_field(cls, v): ...

model_dump() / model_dump(mode='json')

Examples

Input

Output

Edge Cases & Gotchas

Strict vs lax mode - By default Pydantic coerces "1" → 1 for int fields. Use model_config = {"strict": True} to reject type mismatches instead of coercing.
Extra fields - Unknown fields are ignored by default. Use model_config = {"extra": "forbid"} to raise on unexpected fields - useful for strict API contracts.
Recursive models - Pydantic supports self-referential models (tree structures) via model_rebuild() - required after circular imports.
Pydantic v1 vs v2 - Pydantic v2 (released 2023) has breaking API changes - @validator → @field_validator, .dict() → .model_dump(). Check the migration guide before upgrading.

When to Break These Rules

Use attrs or dataclasses for simple data containers without validation - Pydantic adds startup overhead not justified for internal data classes.
Use marshmallow for complex serialization with custom field rendering logic - Pydantic is better for validation, marshmallow for flexible output formatting.

Reference Files

https://docs.pydantic.dev/latest/
https://docs.pydantic.dev/latest/concepts/pydantic_settings/