Hey Carl,

On Friday, November 21, 2014 7:57:47 PM UTC+1, Carl Meyer wrote:
>
> Hi Markus, 
>
> On 11/21/2014 05:35 AM, Markus Holtermann wrote: 
> > As of now (Django 1.7.x and master) the migration writer serializes 
> > classes like the CreateModel operation with kwargs and not args. 
> > Which is the right way to do. However, as soon as Django changes some 
> > of these operations' constructor signatures (as I plan to do in 
> > https://code.djangoproject.com/ticket/23822) newly created migration 
> > files become backwards incompatible for 3rd party apps. 
> > 
> > As a concrete example: CreateModel will get a "managers" argument in 
> > 1.8 (if my pull request will be accepted). Hence all migration files 
> > created with Django 1.8 will contain a managers argument as well. 
> > Third party apps that want to provide compatibility for 1.7 and 1.8 
> > will have a problem as managers is an unknown kwarg to CreateModel in 
> > 1.7. 
> > 
> > I therefore propose to add **kwargs to all operations and would like 
> > that to be backported to 1.7 although its neither a bug nor security 
> > fix. 
>
> This makes sense to me. 
>

Thanks, I opened a ticket (https://code.djangoproject.com/ticket/23892) and 
added a pull-request (https://github.com/django/django/pull/3602)
 

> > I haven't looked into other classes that are serialized by the 
> > migration writer if they would profit from the same patch, but from 
> > my perspective model fields don't need that change as their signature 
> > didn't change over "a decade". 
> > 
> > Another possible problem I want to address but that I haven't looked 
> > into yet, is the serialization of args in general. I think having 
> > **kwargs for every deconstructible class and always serialize with 
> > kwargs and not having any args in the migration classes will prevent 
> > problems in the future, too. 
>
> I'm not sure how you're proposing to address this? It seems like it's up 
> to the implementor of `deconstruct()` for any particular class to decide 
> which things to record as args and which as kwargs, and to take into 
> account the backwards-compatibility considerations. It's not possible in 
> the general case to "always serialize with kwargs", since some objects 
> might accept some things only via `*args`. 
>
> If you just mean looking through the implementations of `deconstruct()` 
> in Django itself and using more kwargs and fewer args there, I think 
> that's a good idea. 
>

Yes, that's my idea. To keep track of the issue I opened 
https://code.djangoproject.com/ticket/23894.

I went through the code and from what I I've seen so far except for the 
PostgreSQL ArrayField and the migration operations all classes having an 
explicit deconstruct() method are using kwargs already. Those classes 
relying on the deconstructible decorator fall back to to whatever way they 
are used due to the underlying argument catching. But that should be fine. 
I we realize at some point that the positional arguments are bad there we 
can still make the deconstructible decorator more intelligent to merge the 
positional argument into keyword arguments.

Markus

-- 
You received this message because you are subscribed to the Google Groups 
"Django developers  (Contributions to Django itself)" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/django-developers.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/django-developers/0e3d5c01-0fe0-40b3-a02c-6fa9c87bed0a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to