Django doesn't properly set the unicode encoding on the psycopg2 driver so when you save unicode strings into an attribute, you get raw strings back which you have to manually decode into proper unicode strings.
Here's what Django does now:
In [1]: from travel.models import *
In [2]: t = Trip.objects.get(pk=1)
In [3]: t.name
Out[3]: '\xc3\x90'
In [4]: t.name = u'\xd0'
In [5]: t.name
Out[5]: u'\xd0'
In [6]: t.save()
In [7]: t.name
Out[7]: u'\xd0'
In [8]: t = Trip.objects.get(pk=1)
In [9]: t.name
Out[9]: '\xc3\x90'
In [10]: t.name.decode("utf8")
Out[10]: u'\xd0'
Here's what it does after the patch:
In [1]: from travel.models import *
In [2]: t = Trip.objects.get(pk=1)
In [3]: t.name
Out[3]: u'Taste of Maple'
In [4]: t.name = u'\xd0'
In [5]: t.name
Out[5]: u'\xd0'
In [6]: t.save()
In [7]: t.name
Out[7]: u'\xd0'
In [8]: t = Trip.objects.get(pk=1)
In [9]: t.name
Out[9]: u'\xd0'
In [10]:
patch:
Index: base.py
===================================================================
--- base.py (revision 3548)
+++ base.py (working copy)
@@ -7,6 +7,9 @@
from django.db.backends import util
try:
import psycopg2 as Database
+ # Register unicode conversions
+ import psycopg2.extensions
+ psycopg2.extensions.register_type(psycopg2.extensions.UNICODE)
except ImportError, e:
from django.core.exceptions import ImproperlyConfigured
raise ImproperlyConfigured, "Error loading psycopg2 module: %s" % e
