papyri.serde

A mini-implementation of an automatic serialiser-deserialiser for nested dataclass like class based on type annotations.

Example

In [14]: from dataclasses import dataclass
    ...: from typing import Optional, Union, List
    ...:

Note that Author and Reviewer are isomorphic even if totally unrelated.

In [15]: @dataclass
    ...: class Author:
    ...:     first: Optional[str]
    ...:     last: str
    ...:
    ...: @dataclass
    ...: class Reviewer:
    ...:     first: Optional[str]
    ...:     last: str
    ...:

Here, items can be heterogenous, or of ambiguous type based only on its fields values.

In [16]: @dataclass
    ...: class Book:
    ...:     author: List[Union[Author, Reviewer]]
    ...:     title: str
    ...:


In [17]: obj = Book([Author("Matthias", "B"), Reviewer("Tony", "Fast")], "pyshs")
    ...:
    ...: data = serialize(obj , Book)
    ...:
    ...: deserialize(Book, Book, data)

Out[17]: Book(author=[Author(first='Matthias', last='B'), Reviewer(first='Tony', last='Fast')], title='pyshs')

                      ^...................................^
                                        .
                                        .Note the conserved types.

Unlike other similar libraries that automatically serialise/deserialise it has the following properties:

• object do not need to have a give baseclass, they need to have an __init__ or _deserialise class method that takes each parameter as kwargs.

• Subclass or isomorphic classes are kept in the de-serialisation, in particular in Union and List of Unions. That is to say it will properly de-serialise and heterogenous list or dict, as long as those respect the type annotation.

Both Pydantic and Jetblack-serialize would have erased the types and returned either 2 Authors or 2 Reviewers.

• it is also compatible with Rust Serde with adjacently tagged Unions (not

  critical but nice to have)