Easy-Serialize

Turn custom python objects into strings and vice-versa.

Strengths

• no extra code to write in most cases
• supports json
• supports nested objects
• no dependencies outside of python's library (only pytest for developing)
• ability to add methods to control serializing and deserializing
• type hints and works with mypy

Drawbacks

• objects like tuples are automatically turned into lists
• only supports json right now
• can't handle circular references
• if an object has references in many places, it will be present in the json repeatedly and after deserializing will be separate objects
• Because of this, it cannot handle linked lists or similar interconnected data structures well
• stores __dict__ of the object, so classes using __slots__ will likely require overriding the serialize method and deserialize classmethod.

How to use

Extend Serializable

class A(Serializable):
    ...

or use the make_serializable decorator

@make_serializable
class A:
    ...

serialize using

serialize(object_to_serialize)

and deserialize using

deserialize(stringified_object)

Custom serializing

In the case that the standard method doesn't cut it, it is possible to add custom serializing and deserializing methods to a class

@make_serializable
class A:
    def __init__(self, x):
        self.x = x

    def serialize(self) -> dict:
        return {'x': self.x}
    
    @classmethod
    def deserialize(cls, data: dict) -> 'A':
        return A(data['x'])

Notes

• __init__ is by default not called for deserialized objects
  • Because it needs to create the object and then copy over all the values of __dict__, it doesn't know the arguments used in __init__. Thus if the class requires arguments in __init__, the object will be created with __new__ which avoids calling __init__. However, if __init__ can take zero arguments, it will be called. A warning will be emitted if __init__ is not called.

Future

Things to possibly add:

• other data formats like xml or the like
• support for objects with circular references