By default, MySQL 4.1 (at least) uses a case-insensitive collation, utf8_general_ci. This means that checks for equality are, annoyingly, case-insensitive. You can set the collation for a database in the CREATE DATABASE command, or you can set it per-server with an option in the MySQL my.cnf file. You can also set it per-column if you want.
Sadly, using a collation like utf8_bin, which gives case-sensitive comparisons, causes unexpected results through the ORM.
To reproduce:
Run "python manage.py test" such that it tests this models.py:
from django.db import models
class Dummy(models.Model):
"""
>>> d = Dummy(ok='ok', notok='not ok')
>>> d.save()
>>> d = Dummy.objects.all()[0]
>>> d.ok, d.notok
('ok', 'not ok')
"""
ok = models.CharField(maxlength=100)
notok = models.TextField()
You'll get something like this (irrelevant info removed):
Failed example:
d.ok, d.notok
Expected:
('ok', 'not ok')
Got:
('ok', array('c', 'not ok'))
Without the collation-server line in my.cnf, the tests pass. Even without that line the tests fail if the CREATE DATABASE line specifies utf8_bin as the collation (and I imagine other binary collations), but the test framework uses the server defaults for those.
My own scrabbling around discovered that MySQLdb (version 1.2.0, at least) includes a suspicious class in init.py:
def Binary(x):
from array import array
return array('c', x)
... but that's as far as my limited knowledge goes - I have no idea how it gets used.
