marshmallow-FastOneOfSchema

An extension to marshmallow to support fast schema (de)multiplexing.

marshmallow is a fantastic library for serialization and deserialization of data. For more on that project see its GitHub page or its Documentation.

This library adds a special kind of schema that actually multiplexes other schemas based on object type. When serializing values, it uses get_obj_type() method to get object type name. Then it uses type_schemas name-to-Schema mapping to get schema for that particular object type, serializes object using that schema and adds an extra field with name of object type. Deserialization is reverse.

Installing

This fork is designed to be a drop‑in replacement for marshmallow-oneofschema.

Using uv:

$ uv add marshmallow-fastoneofschema

Using poetry:

$ poetry add marshmallow-fastoneofschema

pip:

$ pip install marshmallow-fastoneofschema

Note: If marshmallow-oneofschema remains installed alongside this fork, the library logs an error and uses the fast fork. For reliability, uninstall the upstream package to avoid any packaging/order surprises. This wheel also includes a small startup hook (.pth) that redirects imports of marshmallow_oneofschema to the fast fork, even if the upstream was installed later. You can disable this behavior by setting FASTONEOFSCHEMA_DISABLE_PTH=1.

Example

The code below demonstrates how to set up a polymorphic schema. For the full context check out the tests. Once setup the schema should act like any other schema. If it does not then please file an Issue.

import marshmallow
import marshmallow.fields
from marshmallow_fastoneofschema import OneOfSchema


class Foo:
    def __init__(self, foo):
        self.foo = foo


class Bar:
    def __init__(self, bar):
        self.bar = bar


class FooSchema(marshmallow.Schema):
    foo = marshmallow.fields.String(required=True)

    @marshmallow.post_load
    def make_foo(self, data, **kwargs):
        return Foo(**data)


class BarSchema(marshmallow.Schema):
    bar = marshmallow.fields.Integer(required=True)

    @marshmallow.post_load
    def make_bar(self, data, **kwargs):
        return Bar(**data)


class MyUberSchema(OneOfSchema):
    type_schemas = {"foo": FooSchema, "bar": BarSchema}

    def get_obj_type(self, obj):
        if isinstance(obj, Foo):
            return "foo"
        elif isinstance(obj, Bar):
            return "bar"
        else:
            raise Exception("Unknown object type: {}".format(obj.__class__.__name__))


MyUberSchema().dump([Foo(foo="hello"), Bar(bar=123)], many=True)
# => [{'type': 'foo', 'foo': 'hello'}, {'type': 'bar', 'bar': 123}]

MyUberSchema().load(
    [{"type": "foo", "foo": "hello"}, {"type": "bar", "bar": 123}], many=True
)
# => [Foo('hello'), Bar(123)]

By default get_obj_type() returns obj.__class__.__name__, so you can just reuse that to save some typing:

class MyUberSchema(OneOfSchema):
    type_schemas = {"Foo": FooSchema, "Bar": BarSchema}

You can customize type field with type_field class property:

class MyUberSchema(OneOfSchema):
    type_field = "object_type"
    type_schemas = {"Foo": FooSchema, "Bar": BarSchema}


MyUberSchema().dump([Foo(foo="hello"), Bar(bar=123)], many=True)
# => [{'object_type': 'Foo', 'foo': 'hello'}, {'object_type': 'Bar', 'bar': 123}]

You can use resulting schema everywhere marshmallow.Schema can be used, e.g.

import marshmallow as m
import marshmallow.fields as f


class MyOtherSchema(m.Schema):
    items = f.List(f.Nested(MyUberSchema))

License

MIT licensed. See the bundled LICENSE file for more details.

Performance & Compatibility Notes

• This fork preserves the public API and error shapes of the upstream package.
• Optimizations avoid unnecessary copies when type_field_remove is False and reduce overhead for many=True in default configurations.
• Aggressive instance caching can be disabled via FOO_DISABLE_AGGRESSIVE_MODE=1 if needed.
• Supported Python versions: 3.11+.

DeepFriedMarshmallow Plugin

This fork ships a DFM plugin that can enable JIT inlining for Nested(OneOfSchema) fields.

• Discovery: via entry point group deepfriedmarshmallow.plugins or env DFM_PLUGINS.
• Initial scope: Only engages when get_data_type/get_obj_type are default and all type_schemas keys are strings.
• Fallback: If conditions aren't met, DFM falls back to its standard generation.

Per‑Schema Controls & Flags

Migration

Basic usage stays the same (plug‑and‑play). For projects that want to migrate explicitly to the new names:

• Replace dependency marshmallow-oneofschema with this fork.
• Optionally switch imports to the new package/class names: - from marshmallow_fastoneofschema import FastOneOfSchema - Or keep from marshmallow_oneofschema import OneOfSchema (alias provided). - Existing code continues to work; new names are recommended for clarity.
• Env flags: - FOO_DISABLE_AGGRESSIVE_MODE=1: disable instance caching. - FOO_CONTEXT_ISOLATION=1: enable per-task schema instance caching (ContextVars).
• Per-schema overrides via class Meta: fastoneof = {...}: - aggressive_mode: bool - context_isolation: bool
• Per-request overrides via ContextVars (advanced): - OneOfSchema._foo_aggressive_ctx.set(True|False) - OneOfSchema._foo_isolation_ctx.set(True|False)