Option to not load initial fixtures during syncdb

[msiedlarek]

There are some situation in which you may want to prevent automatic loading of initial fixtures by syncdb management command. One example is when you want to create a database schema and fill it with data from the other database. If there were initial fixtures loaded in both databases there probably would be some primary key conflicts raised.

There is a stealth option for that, called load_initial_data, and used by testing setup, by I believe it's not fair to take away from user such a useful one.

[msiedlarek]

Patch

[Aymeric Augustin]

Why not, but it sounds like you're misusing the initial_data fixtures.

In my understanding, they exist for data that will always be in your database and will not change. For example, you have some fixed categories used by M2M fields in other models, so you create a model for them and hardcode the list in a fixture.

[msiedlarek]

I see your point, but as I said -- when you're using django-admin fixtures and syncdb everything should be ok, but when using some Django-independent database management utilities existence of initial data in freshly created database schema can cost you some trouble. In fact -- I'm facing this problem right now. Besides -- this is nice, simple feature which doesn't complicate anything that much.

[Aymeric Augustin]

#15985 was a duplicate.

[Aymeric Augustin]

#9549 is related.

[Jonathan Paugh]

What about the flush command? This change should be included there, too. I'm marking the patch as 'needs improvement.' By the way, I think this feature can be very useful, for development anyway. Whenever I change my models and need to resync, I often want to quickly verify the model without having to update my initial data---sometimes this is necessary anyway, but not always.

[Jonathan Paugh]

Reincarnation of patch supporting the flush command as well

[Jonathan Paugh]

Documentation for the (with-flush) patch

[Jonathan Paugh]

I think with the new patch and the documentation patch, this is ready for review by a core dev. Sorry for stealing your thunder, msiedlarek.

[Jonathan Paugh]

I made my patches against the SVN (git) tip, so I changed the version. I'm guessing that's the protocol.

[Aymeric Augustin]

Could you:

• consolidate all the changes in a single diff file?
• ensure you're indenting consistently with the surrounding code?

Also, this still needs tests.

[vlinhart]

I've consolidated the three pathes into one and are now up-todate with the current head. There is another patch which contains tests.

[Honza Kral <honza.kral@…>]

In [fedac99c859c626544d38fc0ea07207ef3fd101a]:

Fixed #15926 -- Added option --no-initial-data to syncdb and flush.

Thanks msiedlarek, jpaugh64 and vlinhart!

[Claude Paroz <claude@…>]

In [f0664dc8ae541b21e3cf421725e7933a9b3a799e]:

Made TestNoInitialDataLoading pass with MySQL (Refs #15926)

[Claude Paroz <claude@…>]

In [2c57809a560cb67c79b9e8a77cc713e8a2424c8e]:

Prevented TestNoInitialDataLoading to pollute other tests (Refs #15926)

Tests were still failing with MySQL. It seems a rollback is solving
the issue.