I like Django. A lot.

But there's some things about it that bug me, particularly the settings module. The settings module, as you may or may not already know, is the central configuration file for your django project. It handles everything about your project, including: database host, login and password; email host (for emailing errors and whatnot), authentication; applications installed; etc... Well, you get the idea.

Fine, so what's the problem? Well, there isn't one. Unless you use version control like Subversion. And want to have separate environments for testing, staging, production. And the settings module that is version controlled specifies a production database and you need it to point to a test database for testing -- you know, like a real-life development environment!

There's a couple solutions to this dilemma: one involves leaving the settings module out of version control altogether and manually editing it in each environment you'll be running in. Or you could have it in source control and hack on it in each environment; then deal with merging changes, fixing conflicts, etc. after every update.

Or you can use the following approach that requires a little manual editing at the very bottom of your settings module:

# settings.py
# Django settings for mccormac.org project.

DEBUG = False
TEMPLATE_DEBUG = DEBUG

DATABASE_ENGINE = "postgresql"
#... etc...
# all the settings you want at default values

# import local settings overriding the defaults
try:
    from local_settings import *
except ImportError:
    try:
        from mod_python import apache
        apache.log_error( "local_settings.py not set; using default settings", apache.APLOG_NOTICE )
    except ImportError:
        import sys
        sys.stderr.write( "local_settings.py not set; using default settings\n" )

If you look at the example above, you can see we attempt to import from a module called local_settings; if we don't find it we log the issue and move on. This particular implementation attempts to use the apache log, but only works when running on apache.

Indeed, the mod_python module won't be available unless the settings module is actually running on apache, so you have to catch the exception otherwise the embedded webserver in development will crash on the line: from mod_python import apache.

Our local_settings module can now look like this:

# local_settings.py
# Override default Django settings module
#
DEBUG = True
DATABASE_ENGINE = "sqlite3"
DATABASE_NAME = "/home/mike/test.db"

TEMPLATE_DIRS = (
    "/home/mike/dev/templates"
)

With this configuration, the local_settings module doesn't even have to exist -- what I do is set default production values in the settings module and override necessary settings for staging and development in their corresponding local_settings module. Now you can add the main settings module to version control, create a 'svn:ignore' property for the local_settings module and you have a pretty flexible setup.