pypackd 0.1.1

Unified Python serializer with msgspec MessagePack backend

pypackd

Unified Python serializer with msgspec MessagePack backend. One API for all your types — Pydantic, msgspec Structs, dataclasses, and every stdlib type.

Installation

pip install pypackd

With Pydantic support:

pip install pypackd[pydantic]

Usage

from serializer import Serializer

# Serialize any supported type
data = Serializer.serialize(my_object)

# Deserialize with target type
obj = Serializer.deserialize(data, MyType)

Primitives & Collections

Serializer.deserialize(Serializer.serialize(42), int)           # 42
Serializer.deserialize(Serializer.serialize("hello"), str)      # "hello"
Serializer.deserialize(Serializer.serialize([1, 2, 3]), list)   # [1, 2, 3]
Serializer.deserialize(Serializer.serialize({10, 20}), set)     # {10, 20}

Dataclasses

from dataclasses import dataclass

@dataclass
class Point:
    x: float
    y: float

p = Point(3.14, 2.72)
data = Serializer.serialize(p)
Serializer.deserialize(data, Point)  # Point(x=3.14, y=2.72)

msgspec Structs

import msgspec

class Config(msgspec.Struct):
    host: str
    port: int
    debug: bool = False

cfg = Config(host="localhost", port=8080)
data = Serializer.serialize(cfg)
Serializer.deserialize(data, Config)  # Config(host='localhost', port=8080, debug=False)

Pydantic Models

from pydantic import BaseModel
from datetime import datetime
from uuid import UUID

class User(BaseModel):
    id: UUID
    name: str
    created_at: datetime

user = User(id=UUID("abcdef01-2345-6789-abcd-ef0123456789"), name="Georg", created_at=datetime.now())
data = Serializer.serialize(user)
Serializer.deserialize(data, User)  # User(id=UUID('abcdef01-...'), name='Georg', ...)

Nested Types

Pydantic models with dataclass fields, nested Pydantic models, or any combination — it just works.

@dataclass
class Address:
    street: str
    city: str

class Order(BaseModel):
    order_id: int
    customer: User
    shipping: Address
    items: list[str]

order = Order(order_id=1, customer=user, shipping=Address("Hauptstr. 1", "Berlin"), items=["Laptop"])
data = Serializer.serialize(order)
Serializer.deserialize(data, Order)  # Full roundtrip with all nested types

Supported Types

Category	Types
Primitives	str, int, float, bool, None, bytes
Collections	list, dict, tuple, set, frozenset
Datetime	datetime, date, time, timedelta
Stdlib	UUID, Decimal, Enum
Structured	dataclass, NamedTuple, TypedDict, msgspec.Struct
Pydantic	BaseModel, RootModel (incl. computed fields, aliases, excluded fields, custom serializers, model/field validators)
Typing	Optional, Union, Literal, generics

How It Works

pypackd is a thin adapter layer over msgspec, which handles all non-Pydantic types natively with zero overhead. For Pydantic models, pypackd auto-detects the model's features and chooses the fastest serialization path:

• Simple models → obj.__dict__ (2-5x faster than model_dump())
• RootModel, models with computed fields, aliases, excluded fields, or custom serializers → model_dump(mode='python') for correctness

Type classifications are cached, Encoder/Decoder instances are reused, and all cache writes are thread-safe. 292 tests, 99% code coverage.

Error Handling

All errors are wrapped in a unified exception hierarchy:

from serializer import SerializerError, SerializeError, DeserializeError

try:
    Serializer.deserialize(data, MyModel)
except DeserializeError as e:
    print(e)            # Descriptive message
    print(e.__cause__)  # Original pydantic/msgspec exception

Requirements

• Python 3.10+
• msgspec >= 0.18
• pydantic >= 2.0 (optional)

License

MIT