Pydantic & Marshmallow Integration

This document outlines how to integrate Datason with popular schema validation libraries Pydantic and Marshmallow. The goal is to keep validation and schema generation in their respective libraries while leveraging Datason for serialization and deserialization.

Background

Many users validate input data with Pydantic or Marshmallow before working with ML/AI workflows. They want an easy way to serialize these validated objects using Datason without losing type fidelity.

Key Points

• Validation remains the responsibility of Pydantic or Marshmallow.
• Serialization is handled by Datason.
• Integration helpers are optional and incur no new default dependencies.

Serializing Pydantic Models

from pydantic import BaseModel
import datason

class MyModel(BaseModel):
    a: int
    b: str

model = MyModel(a=1, b="foo")
json_data = datason.serialize(model)  # Or datason.serialize_pydantic(model)

Datason automatically extracts fields from BaseModel instances, including nested models. Type information is preserved so the data can be deserialized back to Python primitives or dictionaries.

Serializing Marshmallow Objects

from marshmallow import Schema, fields
import datason

class UserSchema(Schema):
    id = fields.Int()
    name = fields.Str()

schema = UserSchema()
user = schema.load({"id": 1, "name": "Alice"})
json_data = datason.serialize(user)  # Or datason.serialize_marshmallow(user)

The helpers work with the results of .load() or .dump(), enabling seamless round-tripping through Datason without rewriting validation logic.

Optional Dependencies

Datason’s core package does not require Pydantic or Marshmallow. If you use these helpers without installing the corresponding library, Datason raises a helpful ImportError explaining the missing dependency.

Documentation & Examples

• Using Datason with Pydantic
• Using Datason with Marshmallow

Include real-world examples in the docs (e.g., FastAPI with Pydantic, Flask with Marshmallow) to demonstrate how Datason plugs into existing validation flows.

Limitations

• Datason does not perform schema validation.
• Custom Pydantic or Marshmallow fields may require user-defined type handlers for perfect fidelity.