Warn when unique=True is set on ForeignKey

[Curtis Maloney]

Long ago in the dark times before 1.0, there was a bug in Admin that meant using ForeignKey(unique=True) was preferable to using a OneToOneField.

Somehow, people (on IRC) are still doing this -- even new users-- , and I can't figure out why.

Unless I'm missing something, there is no case where a ForeignKey(unique=True) is preferable to a OneToOneField.

Shouldn't we, then, at least add a check/warn to say this?

[Daniele Procida]

I've marked this ticket as especially suitable for people following the ​Don't be afraid to commit tutorial at the DjangoCon US 2014 sprints.

If you're tackling this ticket, please don't hesitate to ask me for guidance if you'd like any, either at the sprints themselves, or here or on the Django IRC channels, where I can be found as EvilDMP.

[JLinden]

Pull request sent at ​https://github.com/django/django/pull/3196

[Diego Guimarães]

New pull request with tests and updated documentation at ​https://github.com/django/django/pull/3262

[Diego Guimarães]

The patch and test are basically working, but a design decision should be made how to handle the tests violating the unique check.

I think that would be helpful, if someone that knows better the System Checks could take a look. I detailed the problems in the PR ​https://github.com/django/django/pull/3262

[Tim Graham]

On the PR Carl says,

Is this check worth adding? A unique ForeignKey works just fine and doesn't hurt anyone. If ForeignKey(unique=True) makes more sense to someone, or if they just changed it to be unique and have a bunch of existing accessor code that they don't want to have to change to OneToOneField style, is there really any value in Django complaining at them about it?

I think if we do add this, the warning needs to be carefully worded so as not to imply that there's anything wrong or broken about ForeignKey(unique=True), because there isn't; it's just that they may find the OneToOneField accessor slightly more convenient if they weren't aware of that option. But to me this seems like a coding-style thing that doesn't warrant a check warning.

[Carl Meyer]

Thanks Tim, I should have posted that here instead.

[Tim Graham <timograham@…>]

In f39b0421b406b411c3bcb58e8aa415d885fea505:

Fixed #23338 -- Added warning when unique=True on ForeigKey

Thanks Jonathan Lindén for the initial patch, and Tim Graham
and Gabe Jackson for the suggestions.