모델을 위한 스키마 정의하기

A Schema is a Python class that lists the fields your data should have. It mirrors a model but lives separately, just for input and output.

Every schema inherits from Marshmallow's base Schema class. Naming it after your model, like UserSchema, keeps the intent obvious at a glance.

Inside the class you add attributes from the fields module. Each one names a key that will appear in your serialized JSON output.

Pick a field type that matches the data: Int, Str, Bool, Float, or DateTime. The type controls how values are formatted and validated.

A DateTime field serializes Python datetimes into clean ISO strings automatically. No more manual formatting scattered across your views.

Pass required=True to a field so loading fails if it is missing. This is your first taste of validation baked right into the schema.

Only list fields you want to expose. By simply not declaring password_hash, the schema can never leak it into a response.

Want camelCase JSON from a snake_case attribute? Use data_key to map an internal name to a different external key.

Some fields, like an id, should appear in output but never be accepted as input. Mark them dump_only to keep them read-only.

You create one instance to reuse: user_schema equals UserSchema(). Pass many=True later when you serialize a whole list.

This single class will both dump models to JSON and load JSON back into checked data. You define the shape once and use it both directions.

You want id readable in output but never accepted as input.

A Schema subclass declares typed fields that map to your model. Options like required, dump_only, and data_key shape exactly what goes in and out.