Question

I have a pretty generic Article model, with m2m relation to Tag model. I want to keep count of each tag usage, i think the best way would be to denormalise count field on Tag model and update it each time Article being saved. How can i accomplish this, or maybe there’s a better way?

Answer 1

You can do this by creating an intermediate model for the M2M relationship and use it as your hook for the post_save and post_delete signals to update the denormalised column in the Article table.

For example, I do this for favourited Question counts in soclone, where Users have a M2M relationship with Questions:

from django.contrib.auth.models import User from django.db import connection, models, transaction from django.db.models.signals import post_delete, post_save  class Question(models.Model):     # ...     favourite_count = models.PositiveIntegerField(default=0)  class FavouriteQuestion(models.Model):     question = models.ForeignKey(Question)     user     = models.ForeignKey(User)  def update_question_favourite_count(instance, **kwargs):     '''     Updates the favourite count for the Question related to the given     FavouriteQuestion.     '''     if kwargs.get('raw', False):         return     cursor = connection.cursor()     cursor.execute(         'UPDATE soclone_question SET favourite_count = ('             'SELECT COUNT(*) from soclone_favouritequestion '             'WHERE soclone_favouritequestion.question_id = soclone_question.id'         ') '         'WHERE id = %s', [instance.question_id])     transaction.commit_unless_managed()  post_save.connect(update_question_favourite_count, sender=FavouriteQuestion) post_delete.connect(update_question_favourite_count, sender=FavouriteQuestion)  # Very, very naughty User.add_to_class('favourite_questions',                   models.ManyToManyField(Question, through=FavouriteQuestion,                                          related_name='favourited_by')) 

There’s been a bit of discussion on the django-developers mailing list about implementing a means of declaratively declaring denormalisations to avoid having to write code like the above:

• Denormalisation, magic, and is it really that useful?
• Denormalisation Magic, Round Two