python：带有用户定义的类名和路径的pickle对象

Question

Here is the problem: I have a set of pickle files with objects and it takes large computing time (like ~2 months) to recreate those objects. I need to convert those objects ( class A stored in file Av1.py ) into a new class ( class A stored in file Av2.py ) objects but save them such that during loading pickle would search for class A in Av1.py such that later I could just rename Av2.py => Av1.py and have the new objects interpreted as having type of class A stored in the original file Av1.py . How can I do it?

Here is an example:

# Av1.py
class A:
    def __init__(self, x):
        self.x = x

# Av2.py
class A:
    def __init__(self, y):
        self.y = y

# python session 1
from Av1 import A as Av1
from Av2 import A as Av2

def convert(av1): return Av2(av1.x+1)

av1 = Av1(1)
av2 = convert(av1)

import pickle, gzip
with gzip.open('o.gzpkl', 'wb', compresslevel=9) as f:
    pickle.dump(av2, f, protocol=pickle.HIGHEST_PROTOCOL)

mv Av2.py Av1.py # Linux shell command to rename file Av2.py into Av1.py

# python session 2
import pickle, gzip
with gzip.open('o.gzpkl', 'rb') as f:
    av = pickle.load(f)

Expected result: python session 1 is modified such that python session 2 works and interprets the object in o.gzpkl as having type Av1.A .

Thank you very much for your help!

Answer 1

Here is the solution: I need to do the following before writing av2

av2.__class__ = Av1

ie python session 1 should be

from Av1 import A as Av1
from Av2 import A as Av2

def convert(av1): return Av2(av1.x+1)

av1 = Av1(1)
av2 = convert(av1)

av2.__class__ = Av1 # solution !!!!

import pickle, gzip
with gzip.open('o.gzpkl', 'wb', compresslevel=9) as f:
    pickle.dump(av2, f, protocol=pickle.HIGHEST_PROTOCOL)