type-serialize 2.1.0

Serialize with type annotations

type-serialize

Serialize with type annotations

Features

• Supported in Python 3.9 and later.
• Serialize classes without additional code.
  • Custom classes
  • @dataclas
  • NamedTuple
  • Enum
  • (Experimental) numpy.ndarray
• Deserialization using type annotations.
• No dependencies
• Compress serialization results: bz2, gzip, lzma, zlib - Removed to focus on 'serialization' itself.

Overview

To pass a custom object to the json.dumps and json.loads functions, there is the following method.

• Expand json.JSONEncoder and json.JSONDecoder.
• Convert to built-in Python object supported by json.JSONEncoder and json.JSONDecoder.

Both methods require additional code and have some problems.

• Problem of not checking symbols when manipulating strings.
• When adding/deleting/editing a property, all related codes must be changed together.
• Painful typecasting (as the biggest problem).

As a way to hide these problems with a library and use serialization and deserialization, I chose type annotations. (Although the library is complicated; haha...) There are some additional advantages to using this.

• Static type checking using mypy.
• Autocomplete in IDE like PyCharm.

Things to know

• All public fields are serialized.
• Methods are not serialized.
• Private fields that start with an underscore (_) are not serialized.
• Members specified with the @property decorator are not serialized.
• When deserializing, all fields must be type-annotated.
• A value of None is ignored by the serialization target.
• When deserializing, the __init__ function must have NO required arguments.
• Implement __serialize__ to override the serialization method.
• Implement __deserialize__ to override the deserialization method.

Installation

pip install type-serialize

Usage

Serializable python object

from dataclasses import dataclass
from type_serialize import deserialize, serialize


@dataclass
class Sample:
    field1: str
    field2: int


data = Sample(field1="a", field2=100)
obj = serialize(data)
assert isinstance(obj, dict)
assert obj["field1"] == "a"
assert obj["field2"] == 100
print(obj)

result = deserialize(obj, Sample)
assert isinstance(result, Sample)
assert data == result
print(result)

Override serialize and deserialize

from dataclasses import dataclass
from type_serialize import deserialize, serialize


@dataclass
class Sample:
    value: int

    def __serialize__(self):
        return {"a": self.value}

    def __deserialize__(self, data) -> None:
        self.value = data["a"]

    def __init__(self, value=100):
        self.value = value


test = Sample(value=200)
obj = serialize(test)
assert isinstance(obj, dict)
assert obj["a"] == 200
print(obj)

result = deserialize(obj, Sample)
assert isinstance(result, Sample)
assert test == result
print(result)

License

See the LICENSE file for details. In summary, type-serialize is licensed under the MIT license.