Django

Code

Ticket #2698 (closed: fixed)

Opened 3 years ago

Last modified 11 months ago

Custom default managers interfere with delete operations

Reported by: anonymous Assigned to: mtredinnick
Milestone: 1.1 Component: Database layer (models, ORM)
Version: SVN Keywords:
Cc: ned@nedbatchelder.com Triage Stage: Accepted
Has patch: 0 Needs documentation: 0
Needs tests: 0 Patch needs improvement: 0

Description

In our model, we have a Story class, with a boolean 'deleted' attribute. We defined a custom default manager to return all of the stories that are not deleted, so that most of the code on the site will not have to deal with filtering out the deleted stories (there's another custom manager called with_deleted for those code paths that need to see the deleted stories as well).

The problem comes up when deleting a user. The deletion code iterates the user's stories to delete them, but uses the default manager, so stories with deleted=1 are not found. So (ironically) any stories marked with the 'deleted' field are not actually deleted when the user is removed. This prevents the user from being deleted, because the story table still has a foreign key reference to the user table, and MySQL prevents the user record from being deleted.

The docs for custom managers say: "it's generally a good idea for the first Manager to be relatively unfiltered". But it seems that any filtering in the default manager will interfere with cascading deletes.

If the default manager does any filtering at all, then deleting an object may fail because its related objects will not all be deleted, and the database's referential integrity will prevent the deletion. When looking for related obejcts to delete, you have to always get all of them. No filtering allowed.

Attachments

Change History

09/12/06 05:02:24 changed by anonymous

  • cc set to ned@nedbatchelder.com.

02/27/08 19:09:48 changed by jacob

  • stage changed from Unreviewed to Accepted.

Looks like the objects-to-delete gathererer needs to figure out how to use an unfiltered queryset.

03/13/09 03:33:02 changed by mtredinnick

  • owner changed from nobody to mtredinnick.
  • status changed from new to assigned.

03/14/09 22:41:33 changed by mtredinnick

(In [10056]) Use plain model.Manager, or suitable proxy, for model saving.

We can't use the default manager in Model.save_base(), since we need to retrieve existing objects which might be filtered out by that manager. We now always use a plain Manager instance at that point (or something that can replace it, such as a GeoManager?), making all existing rows in the database visible to the saving code.

The logic for detecting a "suitable replacement" plain base is the same as for related fields: if the use_for_related_fields is set on the manager subclass, we can use it. The general requirement here is that we want a base class that returns the appropriate QuerySet? subclass, but does not restrict the rows returned.

Fixed #8990, #9527.

Refs #2698 (which is not fixed by this change, but it's the first part of a larger change to fix that bug.)

03/14/09 22:45:32 changed by mtredinnick

(In [10059]) [1.0.X] Use plain model.Manager, or suitable proxy, for model saving.

We can't use the default manager in Model.save_base(), since we need to retrieve existing objects which might be filtered out by that manager. We now always use a plain Manager instance at that point (or something that can replace it, such as a GeoManager?), making all existing rows in the database visible to the saving code.

The logic for detecting a "suitable replacement" plain base is the same as for related fields: if the use_for_related_fields is set on the manager subclass, we can use it. The general requirement here is that we want a base class that returns the appropriate QuerySet? subclass, but does not restrict the rows returned.

Fixed #8990, #9527.

Refs #2698 (which is not fixed by this change, but it's the first part of a larger change to fix that bug.)

Backport of r10056 from trunk.

03/18/09 05:13:15 changed by mtredinnick

  • milestone set to 1.1.

Okay, I think I just worked out how to solve this without having to rewrite all the foreign relation fields for 1.1 (it's too big a job for the time we've got). So I can put this on the 1.1 milestone now.

03/19/09 22:57:12 changed by mtredinnick

  • status changed from assigned to closed.
  • resolution set to fixed.

(In [10104]) Fixed #2698 -- Fixed deleting in the presence of custom managers.

A custom manager on a related object that filtered away objects would prevent those objects being deleted via the relation. This is now fixed.

03/20/09 17:14:32 changed by mtredinnick

(In [10106]) [1.0.X] Fixed #2698 -- Fixed deleting in the presence of custom managers.

A custom manager on a related object that filtered away objects would prevent those objects being deleted via the relation. This is now fixed.

Backport of r10104 from trunk.

Add/Change #2698 (Custom default managers interfere with delete operations)




Change Properties
Action