Serialization + JSON

Synopsis

This library is intended to serialize objects to and from Python primitives. That is, objects will be represented as dict, list, int, float, bool, str, and None. The resulting primitives may be easily serialized to JSON using python.json or this library.

Because the library focuses on representation with native primitives, it could be useful for serialization to other formats.

Usage

The library uses decorators to mark methods in charge of serialization. There's a serializer decorator that associates a function with serializing a specific type, and there's a corresponding deserializer:

@serializer(datetime)
def dthandler(dt: datetime):
    return dt.isoformat()


@deserializer(datetime)
def str2dt(datestr: str):
    return datetime.fromisoformat(datestr)

Finally, the library implements a serialization decorator to make a class in charge of serializing itself. Please see the example to illustrate.

The xtelligent_serial.json namespace includes two convenience methods for reading and writing to and from JSON. The from_json and to_json functions are documented here. The functions serialize types that your code supports with the serialization decorators.

Automatic dataclass support

Python's dataclass with the frozen option creates the rough equivalent of a NamedTuple, but you may still add methods and properties. Immutability is a great practice for creating testable, readable code. A frozen dataclass does not acquire the "infinite state machine" behaviors common to object-oriented class design. It is beyond the scope of this document to create full justification of immutability, and there is an abundance of material on this subject. It is obvious that a frozen dataclass maps very well to JSON documents. The consistent structure of dataclass types make it straightforward for this library to support these classes automatically, without decorators. The only caveat is that all member attribute types must be supported by decorators, or they must also be primitive or dataclass types. Again, please see the example to illustrate.

Documentation

API Reference

Example

Source Code

Roadmap

• Integration with json.JSONDecoder and json.JSONEncoder. For now, this module is an alternative to the json module.
• Serialization convenience methods on the decorators.
• Support for automatic deserialization. Right now, it is required to pass a parameter indicating the type to deserialize to.
• Research automatic support for NamedTuple.