Session serialisers should document limitations on values

[Sasha Romijn]

The JSON session serialiser in Django (the default) does not allow non-UTF8 bytes to be stored in session values. This fact is not documented, although it may not be entirely surprising as Django expects utf-8 everywhere. However, for clarity it may still be better to mention this, probably under ​https://docs.djangoproject.com/en/dev/topics/http/sessions/#session-serialization

Here's a specific example of what you can't do, because '\xd9' is not valid unicode:

>>> from django.contrib.sessions.backends.db import SessionStore
>>> s = SessionStore()
>>> s['foo'] = '\xd9'
>>> s.save()
......
UnicodeDecodeError: 'utf8' codec can't decode byte 0xd9 in position 0: unexpected end of data

I'm considering this a documentation bug, as according to my research there is no way to encode this non-utf8 data into JSON.

[Claude Paroz]

A workaround could be to base64 encode, then decode at read, but still, arbitrary raw bytes aren't JSON transmittable as is (AFAIK).

[Tim Graham]

​PR

[Tim Graham <timograham@…>]

In 1f82b85:

Fixed #26831 -- Documented session data must be JSON encodable for JSONSerializer.

[Tim Graham <timograham@…>]

In e6db7271:

[1.9.x] Fixed #26831 -- Documented session data must be JSON encodable for JSONSerializer.

Backport of 1f82b857ceb75f2d7a68e79c6a00c30bfe7f1318 from master

[Tim Graham <timograham@…>]

In 5d8332aa:

[1.10.x] Fixed #26831 -- Documented session data must be JSON encodable for JSONSerializer.

Backport of 1f82b857ceb75f2d7a68e79c6a00c30bfe7f1318 from master