Django: Annotating price to row, then grouping rows by field, and annotating price sums to groups - python-3.x

Question

My situation: Working on a delivery thingie. I have a Customer, which has one price sheet. The price sheet has pricing zones that are based on postal codes. The customer has DeliveriesOrders, which can have multiple Deliveries (pickups, drop offs) and the price is calculated according the postal code.
The way I calculate prices to Delivery rows is:
sheet_zones = customer.price_sheet.price_sheet_zones.filter(postal_codes__code=OuterRef("postal_code"))
delivery_orders = DeliveryOrder.objects.select_related("customer") \
.prefetch_related(
Prefetch(
'deliveries',
queryset=Delivery.objects.annotate(price=Subquery(sheet_zones.values("price"))).order_by("-is_pickup"),
to_attr="priced_deliveries"
)
)
This enables me to have a price annotated for each Delivery row.
I however cannot annotate a Prefetch-field, as it results in an error, so .annotate(Sum("priced_deliveries")) doesn't work.
I am scratching my head quite hard to get a Sum of all deliveries in a delivery_order. But even more I am scratching my head how to group all deliveries by a field called "reference", and Sum all delivery prices per reference.
Pertinent models:
class DeliveryOrder(models.Model):
customer = models.ForeignKey(Customer, on_delete=models.SET_NULL, null=True, related_name="delivery_orders")
class Delivery(models.Model):
delivery_order = models.ForeignKey(DeliveryOrder, on_delete=models.SET_NULL, null=True, related_name="deliveries")
is_pickup = models.BooleanField(default=True)
reference = models.CharField(max_length=64, blank=True, null=True)
postal_code = models.IntegerField()
Customer related models:
class PriceSheet(models.Model):
name = models.CharField(max_length=255)
class Customer(models.Model):
name = models.CharField(max_length=255)
price_sheet = models.ForeignKey(PriceSheet,
on_delete=models.SET_NULL, blank=True, null=True, related_name="customers")
class PriceSheetZoneItem(models.Model):
price_sheet = models.ForeignKey(PriceSheet, on_delete=models.SET_NULL, null=True, related_name="price_sheet_zones")
zone = models.ForeignKey(Zone, on_delete=models.SET_NULL, null=True, related_name="price_sheet_zones")
postal_codes = models.ManyToManyField(PostalCode, related_name="price_sheet_postal_codes")
price = models.DecimalField(decimal_places=2, max_digits=9)
Postal code related:
class Town(models.Model):
name = models.CharField(max_length=64)
class Zone(models.Model):
name = models.CharField(max_length=32)
class PostalCode(models.Model):
code = models.IntegerField()
town = models.ForeignKey(Town, on_delete=models.SET_NULL, null=True, related_name="postal_codes")
I am open to all and any suggestions and help. I may very well be trying to do this the wrong way.
Using Django 3.2
Thank you in advance!

Answer 1

First, In the Delivery model, should postal_code be a ForeignKey to the PostalCode model? In the below I am going to assume that it is.
I think you want to annotate using a subquery. In SQL this looks something like:
SELECT
delivery_order.*
(SELECT SUM(<price computation from a single delivery>)
FROM delivery
JOIN <join to price sheet table so you can calculate price>
WHERE delivery.delivery_order_id=delivery_order.id
) AS total_price;
FROM
delivery_order;
And your ORM code is something like:
delivery_orders = DeliveryOrder.objects.annotate(
total_price=Subquery(
Delivery.objects.values(
'delivery_order' # This is necessary to get the proper group by
).annotate(
'customer_id'=OuterRef('id')
).annotate(
price=Sum(postal_code__pricesheetzoneitem__price)
).filter(
delivery_id=OuterRef('id'),
postal_code__pricesheetzoneitem__pricesheet__customer_id=OuterRef('id')
).values(
'price' # This is necessary to select only one value
)
)
)
This might be simpler using the django-sql-utils package
from sql_util.aggregates import SubquerySum
delivery_orders = DeliveryOrder.objects.annotate(
total_price=SubquerySum('delivery__postal_code__pricesheetzoneitem__price',
filter=Q(pricesheet__customer_id=OuterRef('id'))
)
)
I can't test this exact code, so I'm not sure it's 100% correct.
Edit: For postal_code not being a ForeignKey
First if you are storing the actual digits of the postal code, e.g. 90210, in the postal_code field, it should be a string, not an integer. Some postal codes start with 0.
You have a couple options. One is to make two fields, postal_code_string that stores the raw postal code info in case there is an error and postal_code which is a ForeignKey to PostalCode.
Another option is to use a JoinField. This allows you to keep everything the same, but the ORM can do the join on the postal_code field. JoinField is a little off topic for this question