python pickle dump and load private variables

Question

I need to protect a class variable. But what to do if the class supports save and load options?

import numpy as np
import pickle
class data(object):
    def __init__(self):
        self.__a = range(100)

    @property    
    def a(self):
         return self.__a

    def save(self, path):
        pickle.dump(self,open(path, 'wb'), protocol=2)

    def load(self, path):
        obj = pickle.load(open(path, 'wb'))
        self.__a = obj.a

This is simple but __aattribute is no more protected because calling instance.a returns the exact instance.__a list and it can changed from the outside which is dangerous in my case.

is there any way around this?

in python there is no protection. You always can change `__a` even from outside. — Daniel, May 10 '14 at 16:10
yes but you don't have any access to it. try instance.__a and you get rejected with an AttributeError — Cobry, May 10 '14 at 16:14
@Cobry: It is mangled to help separate the namespace from subclasses. It's there as `instance._data__a`. — Martijn Pieters, May 10 '14 at 16:14
obviously you need to load the save object attributes. and you cannot do self = obj and its not safe to do self.__dict__ = obj.__dict__ — Cobry, May 10 '14 at 16:14
@Cobry: [Python name mangling: When in doubt, do what?](http://stackoverflow.com/q/7456807) — Martijn Pieters, May 10 '14 at 16:15
so, when it is dangerous to change `a`, then simply don't do it. — Daniel, May 10 '14 at 16:19

score 1 · Answer 1 · answered May 10 '14 at 16:13

1

To protect lists from being changed, you can return a copy of the list by your property:

@property
def a(self):
    return list(self.__a)

answered May 10 '14 at 16:13

Daniel

42,087
4
55
81

this is a copy and i'm dealing with huge data. – Cobry May 10 '14 at 16:17

score 0 · Answer 2 · answered May 10 '14 at 18:11

Instead of non-standard save/load methods, stick with the standard pythonic way of pickling objects, i.e. using pickle.dump and pickle.load directly.

The data members will be as protected after loading as they were before dumping, i.e. your object behaves the same.

class data(object):
    def __init__(self):
        self.__a = range(100)
    @property    
    def a(self):
         return self.__a

obj = data()
# when you want to save, do:
x = pickle.dumps(obj)
# and for loading, do:
obj = pickle.loads(x)
obj.__dict__
=> {'_data__a': [0,
  1,
  2,
  3,
  ...
]}

This approach has many advantages, e.g. you can safely pickle objects which reference instances of your class data.

python pickle dump and load private variables

2 Answers2