xattrs

xattrs is eXtensible toolkits of your attrs and dataclasses types with sensible default behaviors.

Table of Contents

• Motivation
• How to contribute
• License

Motivation

pydantic is a great library to validate and serialize data, but it's not designed to work with attrs or dataclasses as first class support.

attrs is a great library to define classes with declarative and immutable data, but it's not designed to serialize or deserialize instances directly and validate data value by given schema.

attrs incubated dataclasses to define classes without boilerplate in a standard library way. And

...

• Data Classes are intentionally less powerful than attrs.
• ...
• dataclasses can and will be more conservative. We are not bound to any release schedules and we have a clear deprecation policy.

...

-- attrs Documentaion - Why not ...

As I understand, attrs is more powerful and flexible as a third-party library, and has better situation than dataclass to develop and test bleeding-edge ideas and unstable features. And finally, these could also feedback to dataclass.

Unfortunately, while attrs itself is widely adopted, its ecosystem is not as prosperous as pydantic's. Very few libraries are designed to enrich validators, filters or converters for attrs. cattrs attempted to do so, but it hasn't gained the popularity that was expected.

So xattrs is here born to fill the gap between attrs and pydantic, and (try) to provide a better way to work with attrs and dataclasses in a more elegant and powerful way.

Architecture

xattrs has layered architecture, which is designed to be extensible and customizable.

• xattrs is the fundamental layer, which provides the core functionalities to serialize and deserialize instances of attrs and dataclasses types.
  • Just use it like you thought what attrs-serde or dataclass-serde will do.
  • Similar alternatives
    • yukinarit/pyserde
    • Fatal1ty/mashumaro
    • lidatong/dataclasses-json
• xattrs.schema is the schema layer, which provides the schema functionalities to validate annotated schema for instances of attrs and dataclasses.
  • If you're familar with pydantic, you could consider it as a better (?) TypeAdapter for attrs and dataclass.
  • If you're familar with zod (TypeScript), its API is inspired from zod.
  • Try to import xattrs.schema as xs and see what you could do with it.
  • Similar alternatives
    • pydantic/pydantic
    • marshmallow-code/marshmallow
    • python-jsonschema/jsonschema
    • utilmeta/utype

Project scope

Let dataclasses or attrs just store declaretive and immutable 'data'.

Goals:

• Easy to use, easy to extend, easy to customize.
• Extending upstream project as more as its recommended ways, see also attrs Extending.
• Pythonic and powerful API for (de)serialization and schema validation
• Well-tested and well-documented
• Runable under pyodide (WebAssembly)

Installation

pip install xattrs

How to contribute

The first and the best way to contribute to xattrs is to use it in your project and give feedback on your experience about it both good and bad. You could participate our community by submitting issues or even pull requests, not limited to bugs, but also proposals, documentation, use cases and best practices.

Then you could also try to help us to improve the project by promoting it to your teams, writing blog posts, or even giving a talk in your local community.

If you have did all above, but still thought you could do more. Consider to sponsor the project for long-term maintenance and development.

The most important thing is the community, and we are looking forward to how people enjoy using xattrs and how it could help them to solve their problems.

Development Guide

Major upstream dependencies

• attrs
• dataclasses
• annotated

License

xattrs is distributed under the terms of the MIT license.