Using Pickle

Pickle Example

   1 # Save a dictionary into a pickle file.
   2 import pickle
   3 
   4 favorite_color = { "lion": "yellow", "kitty": "red" }
   5 
   6 pickle.dump( favorite_color, open( "save.p", "wb" ) )

   1 # Load the dictionary back from the pickle file.
   2 import pickle
   3 
   4 favorite_color = pickle.load( open( "save.p", "rb" ) )
   5 # favorite_color is now { "lion": "yellow", "kitty": "red" }

For a more complex example, see the official Pickle example, and for API details, see the official Pickle use documentation.

cPickle (Python 2.x only)

In Python 2, you can speed up your pickle access with cPickle. (In Python3, importing pickle will automatically use the accelerated version if it is available.)

The only difference is that you write...

   1 import cPickle as pickle

...at the top of your file, and now all your pickle calls are much faster. ("...up to 1000 times faster.")

Why would you ever use native pickle, rather than cPickle?

Flying Pickle Alert!

Pickle files can be hacked. If you receive a raw pickle file over the network, don't trust it! It could have malicious code in it, that would run arbitrary python when you try to de-pickle it.

However, if you are doing your own pickle writing and reading, you're safe. (Provided no one else has access to the pickle file, of course.)

You might also check out http://trustedpickle.sourceforge.net/ if you want to generate pickles, and later verify that you were the one that generated the pickle (e.g., if you are putting a pickle in a hidden field).

What can you Pickle?

Generally you can pickle any object if you can pickle every attribute of that object. Classes, functions, and methods cannot be pickled -- if you pickle an object, the object's class is not pickled, just a string that identifies what class it belongs to. This works fine for most pickles (but note the discussion about long-term storage of pickles).

With pickle protocol v1, you cannot pickle open file objects, network connections, or database connections. When you think about it, it makes sense -- pickle cannot will the connection for file object to exist when you unpickle your object, and the process of creating that connection goes beyond what pickle can automatically do for you. If you really want to pickle something that has an attribute that is causing problems, look at the pickle documentation for __getstate__, __setstate__, and __getinitargs__ -- using these you can exclude problematic attributes.

With pickle protocol v2, you are able to pickle open file objects. This will change in a future version of Python. See this bug report for more information.

Contributors

LionKimbro, IanBicking, lwickjr

Discussion

Pickles can cause problems if you save a pickle, then update your code and read the pickle in. Attribute added to your __init__ may not be present in the unpickled object; also, if pickle can't find your class and module (e.g., if you renamed the module) you will get errors. However you can handle renaming modules and classes as described in UsingPickle/RenamingModules

For this reason, you should be wary of using pickles for long-term storage where the underlying code is not highly stable.

[lwickjr]: Another possibility re unpicklable objects is to register pickling and unpickling functions with copy_reg. Regarding renaming modules/classes/functions: I've had to deal with this repeatedly in my own code. I have developed an ad-hock procedure that works for me:

A more robust approach would be to perform step one above, and just leave it at that, in case you missed a pickle or two. If desired, you can then perform step 3 after you judge normal processing to have performed step 2 for you, say, a couple years later. ;)

Awkward, but it works. Anyone have any ideas for a better way to do this? --lwickjr

UsingPickle (last edited 2014-06-04 13:32:28 by EricYe)

Unable to edit the page? See the FrontPage for instructions.