django_db_sampler 0.21

This python module was written to help create database fixtures from an existing
database that is too large to simply dump it in its entirety.
Thanks to Jacob Rigby, Technical lead at Reya Group (http://www.reya.com)
for his unhesitating agreement to release this to the open source community.

The process copies the desired model instances to a separate database
along with their dependencies, and optionally children (instances dependent on
them). You can then dump the other database to get a minimal fixture
for your test case.

Quick Reference:
def db_sample(
db_obj_iterable, # Iterable containing Django model instances
child_depth=1, # Set this to 2+ to also include children of children
db_alias='fixture_maker' # If you really need to use a different alias you can.
)

First use:
1. Add a database 'fixture_maker' to your settings file pointing to a
database with a different name from your default database. Add
"'TEST_MIRROR': 'default'" to the fixture_maker database settings
to prevent the new database from being created during tests.

2. Create the database pointed to by alias fixture_maker (This step can be skipped
if you're using sqlite, as it will automatically create the database when
sync'd)

3. Run 'python manage.py syncdb --database=fixture_maker'.
*note I've run into some trouble with South, syncdb, and the --database
flag. If you're using south, you can work around this:
Move your default database, do a 'syncdb --migrate' then rename
the resulting new default database as
the fixture maker database then move your original default database back.

4. Pass an iterable containing django model instances to db_sample().
This Example will save all the Pizza instances with olive topping to the
fixture_maker database with their dependencies and by default children to
1 level deep:

from db_sampler_script import db_sample

p = Pizza.objects.filter(toppings__type='olive')
db_sample(p)
5. Run 'python manage.py dumpdata --database=fixture_maker > your_new_fixture'

Enjoy.

Subsequent use:
Skip steps 1 & 2, and replace step 3 with:
'python manage.py flush --database=fixture_maker'
Be careful not to run this command without the --database flag, lest
you erase your entire default database.

Assumptions:
Primary key for models is obj.id, if you've changed this for some models,
please let me know how it blows up.
Foreign keys don't form loops (an object reference an object which eventually
references the first). This is very rarely needed, and many databases
make it difficult, so if you're not sure don't worry about it.
If you have some loops like these and genuinely need them, let me know &
I'll update the code to deal with it.
Foreign keys that aren't NULL are assumed to be required for database
consistency. At worst this will add some model instances to the resulting
fixture that aren't really needed. Unless someone contacts me about
this causing a real problem, it will likely stay like this.