Skip to content

Serialization and deserialization library for Python

License

Notifications You must be signed in to change notification settings

drhagen/serialite

Repository files navigation

Serialite

Serialite is a library serializing and deserializing arbitrarily complex objects in Python. You apply the @serializable decorator to a dataclass to automatically create to_data and from_data methods using the type annotations. Or for more control, inherit from the SerializableMixin and implement the class attribute __fields_serializer__. For even more control, inherit from the abstract base class Serializer and implement the to_data and from_data methods directly.

Basics

The abstract base class is Serializer:

class Serializer(Generic[Output]):
    def from_data(self, data: Json) -> DeserializationSuccess[Output]: ...
    def to_data(self, value: Output) -> Json: ...

The class is generic in the type of the object that it serializes. The two abstract methods from_data and to_data are the key to the whole design, which revolves around getting objects to and from JSON-serializable data, which are objects constructed entirely of bools, ints, floats, lists, and dicts. Such structures can be consumed by json.dumps to produce a string and produced by json.loads after consuming a string. By basing the serialization around JSON serializable data, complex structures can be built up or torn down piece by piece while alternatively building up complex error messages during deserialization which pinpoint the location in the structure where the bad data exist.

For new classes, it is recommended that the Serializer be implemented on the class itself. There is an abstract base class Serializable that classes can inherit from to indicate this. There is a mixin SerializableMixin that provides an implementation of from_data and to_data for any class that implements the __fields_serializer class attribute.

For dataclasses, it is even easier. There is a decorator serializable that inserts SerializableMixin into the list of base classes after the dataclass decorator has run and also generates __fields_serializer__ from the data class attributes.

Finding the correct serializer for each type can be a pain, so serializer(cls: type) -> Serializer is provided as a convenience function. This is a single dispatch function, which looks up the serializer registered for a particular type. For example, serializer(list[float]) will return ListSerializer(FloatSerializer).

About

Serialization and deserialization library for Python

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages