renew

Semi-text-pickling in pure python. If you meet just a few restrictions, you can store classes state into a python file and import or evaluate it somewhere else or later on. You can even use it as a database unless the amount of data is huge.

1 minute - example:

import renew 

class ThatNiceClass(renew.Mold):
    # manual implementation of __init__ is needed. Constructor_arguments
    # have to be actual names of this class attributes
    def __init__(self, f_a, f_b, *f_c, **f_d):
        self.f_a, self.f_b, self.f_c, self.f_d = f_a, f_b, f_c, f_d

c = ThatNiceClass(1, 2, 3, 4, five=5, six=6)

assert repr(c) == "ThatNiceClass(1, 2, 3, 4, five=5, six=6)"
assert c == eval(repr(c))  # __eq__ implemented 
assert repr(c) == repr(eval(repr(c)))  # pure reproduction, instance "survives" eval

class SecondClass(renew.Mold):
    _cls_namespace = "foo_pkg"

    def __init__(self, one, two="number two", three=None):
        self.one, self.two, self.three = one, two, three

s1 = SecondClass(1)
s2 = SecondClass(3.14159, "non default")
s3 = SecondClass("Lorem ipsum dolor sit amet, consectetur adipiscing elit")
s4 = SecondClass(4, three=ThatNiceClass(1, 2, 3, 4, five=5, six=6))

d = ThatNiceClass(s1, s2, lorem=s3, im_nesting=s4)

assert repr(d) == """\
ThatNiceClass(
    foo_pkg.SecondClass(1),
    foo_pkg.SecondClass(3.14159, 'non default'),
    im_nesting=foo_pkg.SecondClass(4, three=ThatNiceClass(1, 2, 3, 4, five=5, six=6)),
    lorem=foo_pkg.SecondClass('Lorem ipsum dolor sit amet, consectetur adipiscing elit'),
)"""

The __repr__ story - repr(object)

Does repr stand for "representation" or "reproduction"? According to python documentation __repr__ functionality has two separate approaches. From https://docs.python.org/3/library/functions.html#repr (v 3.7.2)

repr(object) Return a string containing a printable representation of an object. For many types, this function makes an attempt to return a string that would yield an object with the same value when passed to eval(), otherwise the representation is a string enclosed in angle brackets that contains the name of the type of the object together with additional information often including the name and address of the object. A class can control what this function returns for its instances by defining a __repr__() method.

1. reproducible repr:

For several native objects it returns a string that can be used to reproduce given object, i.e. to create a copy of given object.

a = [1, 3.141559, None, "string"]
statement_str = repr(a)
assert statement_str == '[1, 3.141559, None, "string"]'

You may tell that repr of an object is reproducible if this is meet:

a = [1, 3.14159, None, "string"]
statement_str = repr(a)
assert repr(eval(statement_str)) == statement_str
# if the object implements __eq__ this should be also true:
assert eval(statement_str) == a

2. descriptive repr:

Unfortunately python does not serve the "reproducible repr" out of the box for types defined by user:

class Car(object):
    def __init__(self, body_type, engine_power):
        self.body_type = body_type
        self.engine_power = engine_power

car = Car("coupe", 124.0)
# repr(car) == '<__main__.Car object at 0x7f0ff6313290>'
# but using renew:

import renew

class ReproducibleCar(renew.Mold):
    _cls_namespace = "bar"
    def __init__(self, body_type, engine_power):
        self.body_type = body_type
        self.engine_power = engine_power

car2 = ReproducibleCar("sedan", 110.0)
assert repr(car2) == 'bar.ReproducibleCar("sedan", 110.0)'

The method above is implemented as a decorator, but you can also use a inheritance to get the same result.

import renew

class Car(renew.Mold):
    _cls_namespace = "cars"
    _cls_dependency = "that.things"

    def __init__(self, body_type, engine_power, fuel, seats, color=None):
        self.body_type = body_type
        self.engine_power = engine_power
        self.fuel = fuel
        self.seats = seats
        self.color = color

class Driver(renew.Mold):
    _cls_namespace = "persons"

    def __init__(self, first_name, last_name, *cars):
        self.first_name = first_name
        self.last_name = last_name
        self.cars = cars

car_1 = Car("Truck", 120.0, "diesel", 2)
car_2 = Car("Van", 145.0, "diesel", seats=7, color="silver")
car_3 = Car("Roadster", 210.0, "gasoline", seats=2)

driver_1 = Driver("Blenda", "Klapa", car_1)
driver_2 = Driver("Trytka", "Blotnick", car_2, car_3)

assert repr(driver_1) == ".Driver('Blenda', 'Klapa', cars.Car('Truck', 120.0, 'diesel', 2))"
assert repr(driver_2) == """\
persons.Driver(
    'Trytka',
    'Blotnick',
    cars.Car('Van', 145.0, 'diesel', 7, 'silver'),
    cars.Car('Roadster', 210.0, 'gasoline', 2),
)"""

renew.serialize("/tmp/target.py", blenda=driver_1, trytka=driver_2)

The created file looks like this:

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# This file has been created with renew.
# A py-pickling tool: https://pypi.org/project/renew/

from living.things import persons
from that.things import cars

blenda = persons.Driver('Blenda', 'Klapa', cars.Car('Truck', 120.0, 'diesel', 2))

trytka = persons.Driver(
    'Trytka',
    'Blotnick',
    cars.Car('Van', 145.0, 'diesel', 7, 'silver'),
    cars.Car('Roadster', 210.0, 'gasoline', 2),
)

How it works?

Note that ReproducibleCar does not explicitly implement the __repr__, but the renew.reproducible decorator supplements it (overrides it if any has been defined before). renew.reproduction inspects constructor's argument specification of decorated class and yields a string that tries to be a call statement composed of

• namespace, e.g. your package name (according to desired importing convention)
• given class name
• given class' attributes values, that have the same names and order as constructor arguments

That forms the only one usage restriction:

The class has to store all the constructor arguments in its attributes with the same name (as in ReproducibleCar definition above).

Variadic args have to be stored in the instance either as list, tuple (no cast needed), set or OrderedDict (set is rendered with sorting). Keyword args have to be stored in the instance as a dict or OrderedDict.

from collections import OrderedDict
import renew

class ThatClass(renew.Mold):
    def __init__(self, x=1, *others, **kw_args):
        self.x = x
        self.others = OrderedDict(others)
        self.kw_args = kw_args

that = ThatClass(3.14159, ("a", "A"), ("b", "B"), one=1, two=2, many=666)

assert repr(that) == "ThatClass(3.14159, many=666, one=1, two=2)"
assert that.x == 3.14159
assert that.others == OrderedDict([("a", "A"), ("b", "B")])
assert that.kw_args == dict(one=1, two=2, many=666)
 

Limitations

• keys of plain dict being "complex" objects get a bit ugly layout if repr of given key spans multiple lines.

• renew does not cross-reference objects while serializing. Although neither pickle nor marshal does cross-reference, renew most probably could do it but it's hard to tell how to let renew know where and how a chain of objects have to be cross-referenced.

• For ultra-capable meta programming MacroPy: https://pypi.org/project/MacroPy/ would be a better choice.

For full list of features and usage examples, please refer to unit tests, especially tests/test_renew.py.