py2dataclasses 3.14.115

PEP-557 compatible dataclasses backport for Python 2.7+

py2dataclasses

This project is a PEP-557 compatible dataclass implementation for Python 2.7.

Usage

1. pip install py2dataclasses

2.  # We have to use `field` syntax because Python 2 doesn't support type annotations.
    from dataclasses import dataclass, field
    
    @dataclass
    class Point(object):
        x = field(int)
        y = field(int)
    
    p = Point(3, 4)
    print(p)        # Point(x=3, y=4)
    print(p.x)      # 3
    print(p == Point(3, 4))  # True
   

Big fat (working!) example

from __future__ import print_function
from dataclasses import (dataclass, field, fields, asdict, astuple,
                         replace, make_dataclass, is_dataclass, InitVar,)
from typing import ClassVar

@dataclass
class Address(object):
    street = field(str)
    city = field(str)
    zipcode = field(str, default="00000")

@dataclass
class Person(object):
    company = field(ClassVar[str], "Acme Corp")
    name = field(str)
    age = field(int)
    address = field(Address)
    salary_multiplier = field(InitVar[float], default=1.0)
    salary = field(int, default=50000)

    def __post_init__(self, salary_multiplier):
        self.salary = int(self.salary * salary_multiplier)

addr = Address("123 Main St", "Anytown")
person = Person("Alice", 30, addr, salary_multiplier=1.5)

print(person)
# Person(name='Alice', age=30, address=Address(...), salary=75000)

print(asdict(person))
# {'name': 'Alice', 'age': 30, 'address': {...}, 'salary': 75000}

person2 = replace(person, age=31)
assert person2.age == 31
assert person.age == 30

assert Person.company == "Acme Corp"
assert is_dataclass(person)

More examples here.

loads() / dumps() use json by default and can use any serializer module with loads() / dumps() functions:

import msgpack
from dataclasses import dumps, loads

payload = dumps(Point(3, 4), serializer=msgpack, use_bin_type=True)
point = loads(Point, payload, serializer=msgpack, raw=False)

load() / loads() reject unknown input keys by default. Pass unknown=EXCLUDE to ignore unknown, ClassVar, or init=False input keys. Scalar fields use marshmallow-style coercion by default; pass strict_types=True to require values to already match their annotated runtime types. dump() / dumps() serialize the current dataclass instance and do not run load-time validation.

Dataclass helper caches are enabled by default with @dataclass(cache=True). The repeated load() / loads() / validate() / validates() / dump() / dumps() path reuses resolved field metadata for better performance. Classes that intentionally update field definitions, field types, __annotations__, __dataclass_fields__, or ClassVar / init=False metadata at runtime can opt out per class with @dataclass(cache=False) or make_dataclass(..., cache=False).

Development

Several testenvs provided to run tests in different configurations:

• 

  • runs ported to py2.7 tests on backported dataclasses (runs with py27)

• 

  • runs ported to py2.7 tests on backported dataclasses (runs with py3)

• 

  • runs native python 3.14 tests on stdlib dataclasses (runs with py3)

• 

  • runs native python 3.14 tests on backported dataclasses (runs with py3)

• 

  • runs ported to py2.7 tests on stdlib dataclasses (runs with py3)

Fixtures

• tests._fixtures_py314 dir contains original python 3.14 tests for the dataclasses lib. Test themselves are in __init__.py
• tests._fixtures_py27 dir contains tests ported from python 3.14 to python 2.7 as closely as possible. Test themselves are in __init__.py

Helpers

• tests._xmlrunner is a helper to run coverage alongside tests
• tests._xmlprocessor is a helper for the CI

Adapters

• tests._fixtures_compat_backport_to_stdlib
• tests._fixtures_compat_stdlib_to_backport

Source

• dataclasses
• _py2dataclasses

⚠ WARNING

• Please, do not use Python 2.7 in 2026