Skip to content

databind

Python versions Documentation

The databind package provides a (de)serialization framework that understands most native Python types as well as dataclasses, as well as an implementation for serialize to/from JSON-like nested data structures.

Databind is intended mostly for flexible and easy to use configuration loading. It does not try achieve high-performance; you should look towards e.g. mashumaro for this usecase.

Example

@dataclass
class Server:
    host: str
    port: int

@dataclass
class Config:
    server: Server

from databind.json import dump, load

dict_payload = {"server": {"host": "localhost", "port": 8080}}
loaded = Config(server=Server(host="localhost", port=8080))

assert load(dict_payload, Config) == loaded
assert dump(loaded, Config) == dict_payload

Features ✨

  • Support for a plethora of builtin types, including Enum, Decimal, UUID, Path, datetime, date, time, timedelta
  • Support for multiple union serialization modes (nested, flat, keyed, typing.Literal)
  • Support for generic types, e.g. load([{"name": "Jane Doe"}], list[Person])
  • Support for new-style type hints in older Python versions when using forward refererences (strings or __future__.annotations) thanks to typeapi
  • Support for customized serialization and deserialization of types
  • Support for flattening fields of a nested dataclass or collecting remaining fields in a dict
  • Full runtime type checking during serialization
  • Use "settings" to customize serialization behaviour
    • As global settings per load()/dump() call: load(..., settings=[ExtraKeys(True)])
    • As class-level settings using a decorator: @Union(style=Union.FLAT) or @ExtraKeys(True)
    • As type-hint level settings using typing.Annotated (or typing_extensions.Annotated): full_name: Annotated[str, Alias("fullName")] or FullNameField = Annotated[str, Alias("fullName")]

Notable release notes

4.5.0

  • Merged databind.core and databind.json packages into databind. The old PyPI packages will remain as proxies until the next minor version.
  • Dropped support for Python 3.6 and 3.7.

Copyright © 2022 – Niklas Rosenstein