Django - 用于字段计算的模型属性方法与信号

Question

I have a model that does some basic aggregation of some values on a m2m field. I can achieve this by doing the calculation in a Model method, in which case the results would be stored in a model attribute.

Or I can do the calculation in a pre_save signal function, and then the result would be stored in a model field.

I want to know are there are any cons to using signals? or in which cases it would be better to use a model property method?

I understand that the calculations will run at different times, with the signal function running when save() is called, and the model method running automatically.

class CartItem(models.Model):
    name = models.CharField(max_length=100)
    total_price = models.DecimalField(max_digits=10, decimal_places=2)

##Model Method Approach
    @property
    def price(self):
        return self.extras.all().aggregate(Sum('price'))['price__sum']

##Signal Approach
def set_total_price(sender, instance, **kwargs):
      instance.total_price= instance.extras.all().aggregate(Sum('price'))['price__sum'] 

pre_save.connect(set_total_price, sender=CartItem)

The signal approach has the benefit of resulting in a model field rather than a model attribute, which makes it easier use in Parent or related models for aggregation. So why would one ever use a Model method over a model signal?

Answer 1

The difference is that the @property will be calculated every time you access it, and the signal one will be calculated once and then stored, if you're not using the right indices they can have a big difference in performance

Another difference is that if you use CartItem.objects.update , signals won't run, so you would end up with invalid data, same thing if you updated the object outside of Django directly through SQL.

Answer 2

I'm VERY new to django, but I was asking a similar question. I had 2 time fields and needed the difference between them. My first try was with a property which worked reasonably well until I realized I needed to be able to easily get a total of that difference across a queryset.

I ended up implementing a pre_save signal to set the value of the field. In my use case, the data doesn't change frequently, so it made sense. Then I created a property on the queryset to aggregate a sum. This way I can filter first and get a sum across the subset.