How to pickle yourself?

问题:

I want my class to implement Save and Load functions which simply do a pickle of the class. But apparently you cannot use 'self' in the fashion below. How can you do this?

self = cPickle.load(f)

cPickle.dump(self,f,2)

回答1:

This is what I ended up doing. Updating the __dict__ means we keep any new member variables I add to the class and just update the ones that were there when the object was last pickle'd. It seems the simplest while maintaining the saving and loading code inside the class itself so calling code just does an object.save().

def load(self):
    f = open(self.filename, 'rb')
    tmp_dict = cPickle.load(f)
    f.close()          

    self.__dict__.update(tmp_dict) 


def save(self):
    f = open(self.filename, 'wb')
    cPickle.dump(self.__dict__, f, 2)
    f.close()

回答2:

The dump part should work as you suggested. for the loading part, you can define a @classmethod that loads an instance from a given file and returns it.

@classmethod
def loader(cls,f):
    return cPickle.load(f)

then the caller would do something like:

class_instance = ClassName.loader(f)

回答3:

If you want to have your class update itself from a saved pickle… you pretty much have to use __dict__.update, as you have in your own answer. It's kind of like a cat chasing it's tail, however… as you are asking the instance to essentially "reset" itself with prior state.

There's a slight tweak to your answer. You can actually pickle self.

>>> import dill
>>> class Thing(object):
...   def save(self):
...     return dill.dumps(self)
...   def load(self, obj):
...     self.__dict__.update(dill.loads(obj).__dict__)
... 
>>> t = Thing()
>>> t.x = 1
>>> _t = t.save()
>>> t.x = 2
>>> t.x
2
>>> t.load(_t)
>>> t.x
1

I used loads and dumps instead of load and dump because I wanted the pickle to save to a string. Using load and dump to a file also works. And, actually, I can use dill to pickle an class instance to a file, for later use… even if the class is defined interactively. Continuing from above...

>>> with open('self.pik', 'w') as f:
...   dill.dump(t, f)
... 
>>> 

then stopping and restarting...

Python 2.7.10 (default, May 25 2015, 13:16:30) 
[GCC 4.2.1 Compatible Apple LLVM 5.1 (clang-503.0.40)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import dill
>>> with open('self.pik', 'r') as f:
...   t = dill.load(f)
... 
>>> t.x
1
>>> print dill.source.getsource(t.__class__)
class Thing(object):
  def save(self):
    return dill.dumps(self)
  def load(self, obj):
    self.__dict__.update(dill.loads(obj).__dict__)

>>> 

I'm using dill, which is available here: https://github.com/uqfoundation

回答4:

There is an example of how to pickle an instance here, in the docs. (Search down for the "TextReader" example). The idea is to define __getstate__ and __setstate__ methods, which allow you to define what data needs to be pickled, and how to use that data to re-instantiate the object.

回答5:

How about writing a class called Serializable that would implement dump and load and make your class inherit from it?