Thanks Malcome,
I was thinking of a kind of dispatch.connect "related-pre-save" and "related-post-save" for the manytomany related data. A bit like the existing "pre-save" and "post-save". The idea to look at the manipulator is not bad too. I'll investigate it, thanks. Malcolm Tredinnick wrote: > On Sat, 2006-07-08 at 17:49 +0000, william wrote: > > I've been trough the save() as documented here: > > http://code.djangoproject.com/wiki/RemovingTheMagic#Overridingsaveanddeletemodelmethods > > > > But, this does not work for fields having a ManyToMany relation. > > If you take the example with "publication" and "article" as described > > here: http://www.djangoproject.com/documentation/models/many_to_many/ > > > > If you put the following into the "Article" class: > > def save(self): > > print "Before",self.publications.all() > > super(Article, self).save() > > print "after",self.publications.all() > > > > You'll see that you'll always see the same result before and after the > > save. > > (standard fields are yet working correctly) > > It seems that the manytomany related data are saved after the complete > > execution of save(). > > I'm using revision 3275 > > In my case this is annoying. > > Related items (the things being saved into the many-to-many relation) > are not saved as part of a model's save method, as you have discovered. > Instead, the Add- and ChangeManipulators save the many-to-many items > later. In fact, for adding a new item, this is basically required, > because you need to know the new instance's primary key value before you > can save a reference to it in the m2m join table -- and that value does > not necessarily exist before it is saved to the database. > > At the moment, any workaround is going to involve custom manipulators, I > suspect (although I may be missing something obvious). We might be able > to come up with something a bit nicer with the upcoming manipulator > refactoring that is on the table. > > This does come up from time to time when somebody wants to take action > based on a new relationship being created. So you're not on the fringe > here by wanting this. > > Sorry, not much encouragement there except to say it's a known problem > (well, at least, I consider it a "problem"). Maybe somebody smarter than > me can suggest an easy solution at the moment *shrug*. > > Regards, > Malcolm --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Django users" group. To post to this group, send email to django-users@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/django-users -~----------~----~----~----~------~----~------~--~---
