django-picklefield 0.1

About

django-picklefield provides an implementation of a pickled object field. Such fields can contain any picklable objects.

The implementation is taken and adopted from Django snippet #1694 by taavi223, which is in turn based on Django snippet #513 by obeattie.

Usage

First of all, you need to have django-picklefield installed; for your convenience, recent versions should be available from PyPI.

To use, just define a field in your model:

from picklefield import PickledObjectField

class SomeObject(models.Model):

args = PickledObjectField()

and assign whatever you like (as long as it’s picklable) to the field:

obj = SomeObject() obj.args = [‘fancy’, {‘objects’: ‘inside’}] obj.save()

Notes

Here are the notes by taavi223, the original author:

Incredibly useful for storing just about anything in the database (provided it is Pickle-able, of course) when there isn’t a ‘proper’ field for the job.

PickledObjectField is database-agnostic, and should work with any database backend you can throw at it. You can pass in any Python object and it will automagically be converted behind the scenes. You never have to manually pickle or unpickle anything. Also works fine when querying; supports exact, in, and isnull lookups. It should be noted, however, that calling QuerySet.values() will only return the encoded data, not the original Python object.

Please note that this is supposed to be two files, one fields.py and one tests.py (if you don’t care about the unit tests, just use fields.py).

This PickledObjectField has a few improvements over the one in snippet #513.

This one solves the DjangoUnicodeDecodeError problem when saving an object containing non-ASCII data by base64 encoding the pickled output stream. This ensures that all stored data is ASCII, eliminating the problem.

PickledObjectField will now optionally use zlib to compress (and uncompress) pickled objects on the fly. This can be set per-field using the keyword argument “compress=True”. For most items this is probably not worth the small performance penalty, but for Models with larger objects, it can be a real space saver.

You can also now specify the pickle protocol per-field, using the protocol keyword argument. The default of 2 should always work, unless you are trying to access the data from outside of the Django ORM.

Worked around a rare issue when using the cPickle and performing lookups of complex data types. In short, cPickle would sometimes output different streams for the same object depending on how it was referenced. This of course could cause lookups for complex objects to fail, even when a matching object exists. See the docstrings and tests for more information.

You can now use the isnull lookup and have it function as expected. A consequence of this is that by default, PickledObjectField has null=True set (you can of course pass null=False if you want to change that). If null=False is set (the default for fields), then you wouldn’t be able to store a Python None value, since None values aren’t pickled or encoded (this in turn is what makes the isnull lookup possible).

You can now pass in an object as the default argument for the field without it being converted to a unicode string first. If you pass in a callable though, the field will still call it. It will not try to pickle and encode it.

You can manually import dbsafe_encode and dbsafe_decode from fields.py if you want to encode and decode objects yourself. This is mostly useful for decoding values returned from calling QuerySet.values(), which are still encoded strings.

The tests have been updated to match the added features, but if you find any bugs, please post them in the comments. My goal is to make this an error-proof implementation.

Note: If you are trying to store other django models in the PickledObjectField, please see the comments for a discussion on the problems associated with doing that. The easy solution is to put django models into a list or tuple before assigning them to the PickledObjectField.

Update 9/2/09: Fixed the value_to_string method so that serialization should now work as expected. Also added deepcopy back into dbsafe_encode, fixing #4 above, since deepcopy had somehow managed to remove itself. This means that lookups should once again work as expected in all situations. Also made the field editable=False by default (which I swear I already did once before!) since it is never a good idea to have a PickledObjectField be user editable.

Feedback

There is a home page with instructions on how to access the code repository.

Send feedback and suggestions to team@shrubberysoft.com.