user667804 user667804 - 1 year ago 62
Python Question

Django save behaving randomly

I have a Story model with a M2M relationship to some Resource objects. Some of the Resource objects are missing a name so I want to copy the title of the Story to the assigned Resource objects.

Here is my code:

from collector import models
from django.core.paginator import Paginator

paginator = Paginator(models.Story.objects.all(), 1000)

def fix_issues():
for page in range(1, paginator.num_pages + 1):
for story in
name_story = story.title
for r in story.resources.select_subclasses():
if != name_story: = name_story
if len( == 0:
print("Something went wrong: " + name_story)
print("done processing page %s out of %s" % (page, paginator.num_pages))


I need to use a paginator because I'm dealing with a million objects. The weird part is that after calling fix_issues() about half of my resources that had no name, now have the correct name, while the other half still has no name. I can call fix_issues() again and again and every time more objects receive a name. This seems really weird to me, why would an object not be updated the first time but only the second time?

Additional information:

  • The "Something went wrong: " message is never printed.

  • I'm using select_subclasses from django-model-utils to iterate over all resources (any type).

  • The story.title is never empty.

  • No error message is printed, when I run these commands.

  • I did not override the save method of the Resource model (only the save method of the Story model).

  • I tried to use @transaction.atomic but the result was the same.

My Model:

class Resource(models.Model):
name = models.CharField(max_length=200)
# Important for retrieving the correct subtype.
objects = InheritanceManager()

def __str__(self):
return str(

class CustomResource(Resource):
homepage = models.CharField(max_length=3000, default="", blank=True, null=True)

class Story(models.Model):
url = models.URLField(max_length=3000)
resources = models.ManyToManyField(Resource)
popularity = models.FloatField()

def _update_popularity(self):
self.popularity = 3

def save(self, *args, **kwargs):
super(Story, self).save(*args, **kwargs)
super(Story, self).save(*args, **kwargs)

Documentation for the select_subclasses:

Further investigation:
I thought that maybe select_subclasses did not return all the objects. Right now every story has exactly one resource. So it was easy enough to check that select_subclasses always returns one item. This is the function I used:

def find_issues():
for page in range(1, paginator.num_pages + 1):
for story in
assert(len(story.resources.select_subclasses()) == 1)
print("done processing page %s out of %s" % (page, paginator.num_pages))

But again, this executes without any problems. So I don't thing the select_subclasses is to blame. I also checked if paginator.num_pages is right and it is. If i divide by 1000 (items per page) I get exactly the number of stories I have in my database.

Answer Source

I think I know what is happening:

The Paginator loads a Queryset and gives me the first n items. I process these and update some of the values. But for the next iteration the order of the items in the queryset changes (because I updated some of them and did not define an order). So I'm skipping over items that are now on the first page. I can avoid it by specifying an order (pk for example).

If you think I'm wrong, please let me know. Otherwise I will accept this as the correct answer. Thank you.

Recommended from our users: Dynamic Network Monitoring from WhatsUp Gold from IPSwitch. Free Download