Nested Serializer Scenario: What is the best?

[giuseppenovielli]

Hello everyone,
in a complex scenario, we need to save or update nested objects.

Please consider the following two models.

class Car(models.Model):
    name = models.CharField(max_length=200, verbose_name='Name', validators=[MinLengthValidator(10)])
    engine = models.ForeignKey(Engine, on_delete=models.CASCADE, verbose_name='Engine')

    class Meta:
        ordering = ['-id']
        verbose_name = 'Car'
        verbose_name_plural = 'Cars'


class CarUser(models.Model):
    car = models.ForeignKey(Car, on_delete=models.CASCADE, verbose_name='Car', related_name='car_user_car')
    user = models.ForeignKey(User, on_delete=models.CASCADE, verbose_name='User', related_name='car_user_user')
    number_plate = models.CharField(max_length=10, verbose_name='Number plate',
                                    validators=[RegexValidator(regex=r'^[0-9a-zA-Z]*$')],
                                    )
    
    user_created = models.ForeignKey(User, on_delete=models.CASCADE, 
                                     verbose_name='User created', 
                                     related_name='car_user_created', 
                                     validators=[user_write_authorization_validator, user_is_active_validator])
    datetime_created = models.DateTimeField(default=timezone.now, verbose_name='Datetime')
 

    class Meta:
        ordering = ['-id']
        verbose_name = 'Car bouth by user'
        verbose_name_plural = 'Cars bouth by users'
        unique_together = ['car', 'user', 'number_plate']
    
        
    def _number_plate_user_validator(self):
        user = self.user
        last_name = user.last_name
        first_name = user.first_name
        fullname = '{} {}'.format(last_name, first_name)
        
        for number_plate_char in list(self.number_plate.lower()):
            if number_plate_char in list(fullname.lower()):
                return
        raise ValidationError(_('The name of the user must be included in the number plate'))
    
    
    def _car_user_validator(self):
        user = self.user
        last_name = user.last_name
        first_name = user.first_name
        fullname = '{} {}'.format(last_name, first_name)
        
        for car_name_char in list(self.car.name.lower()):
            if car_name_char in list(fullname.lower()):
                return
        raise ValidationError(_('At least one char of the car name must be included in the user fullname'))
        

    #https://stackoverflow.com/questions/61507845/model-clean-vs-model-clean-fields
    def clean(self):
        """
        Validate rules for multiple fields before save it
        https://www.django-rest-framework.org/api-guide/serializers/#object-level-validation
        https://docs.djangoproject.com/en/3.2/ref/models/instances/#django.db.models.Model.clean
        """
        super().clean()
        self._number_plate_user_validator()
        self._car_user_validator()

How you can read into the model: CarUser there are some custom business logic into clean() method.
The method def _car_user_validator(self) required Car instance to continue the validation.

Now, when i want to save 'nested serializers', considering call POST rest api, this mean that i save a new object Car and this Car instance must be associate to CarUser at field car.

The following json is more clear what i mean:

{
    "name": "electric_123456789",
    "engine" : 1,
    "car_user_car" : [
        {
            "user" : 1,
            "number_plate" : "a123"
        }
    ]
}

1. Ok, let's do as Django Rest Framework doc says:

class CarUser_CarExcluded_Serializer(FullCleanModelSerializer):
        
    class Meta:
        model = CarUser
        fields = ('id', 'user', 'number_plate',)
        
    def validate(self, attrs):
        attrs = super().validate(attrs)
        
        #Set the read_only fields values
        attrs['user_created'] = self.context['request'].user
        return attrs
    
        
class Car_FullCleanSerializer(FullCleanModelSerializer):
    car_user_car = CarUser_CarExcluded_Serializer()

    class Meta:
        model = Car
        fields = '__all__'

    def create(self, validated_data):
        car_user_data = validated_data.pop('car_user_car')
        car = Car.objects.create(**validated_data)
        CarUser.objects.create(car=car, **car_user_data)
        return car

s = CarFullCleanSerializer(data=request.POST)
s.is_valid(raise_exception=True)
s.save()

This code got an Exception Exception: Nested serializers are not supported.
FAIL

2. Let's try manually create nested instances:

class CarUser_CarExcluded_Serializer(FullCleanModelSerializer):
        
    class Meta:
        model = CarUser
        fields = ('id', 'user', 'number_plate',)

def validate(self, attrs):
   attrs = super().validate(attrs)
        
   #Set the read_only fields values
   attrs['user_created'] = self.context['request'].user
   return attrs

class Car_FullCleanSerializer(FullCleanModelSerializer):

    class Meta:
        model = Car
        fields = '__all__'

s_car = Car_FullCleanSerializer(data=request.data, context={'request': request})
s_car.is_valid(raise_exception=True)
car = s_car.save()
        
s = CarUser_CarExcluded_Serializer(data=request.data.get('car_user_car'), context={'request': request})
s.is_valid(raise_exception=True)
s.save(car=car)

This code got an Exception CarUser.car.RelatedObjectDoesNotExist: CarUser has no car.
CarUser model, into def _car_user_validator(self) method need to have a Car instance, if you don't provide it, the CarUser instance can't be saved into db, because we don't know if the data trust own business logic!

FAIL

3. Use is_valid(extra_include=None) parameter:

s_car = Car_FullCleanSerializer(data=request.data, context={'request': request})
s_car.is_valid(raise_exception=True)
car = s_car.save()
        
s = CarUser_CarExcluded_Serializer(data=request.data.get('car_user_car'), context={'request': request})
s.is_valid(raise_exception=True, extra_include={'car' : car})
s.save(car=car)

YES now CarUser is validated our data before save, but if CarUser clean() fail or ModelSerializer.validate() method fail?
A Car instance is already save!! But our endpoint get 400 Bad Request!

FAIL

4. Use is_valid(extra_include=None) parameter using partial instance:

s_car = Car_FullCleanSerializer(data=request.data, context={'request': request})
s_car.is_valid(raise_exception=True)
car = Car(**s_car.validate_data)
        
s = CarUser_CarExcluded_Serializer(data=request.data.get('car_user_car'), context={'request': request})
s.is_valid(raise_exception=True, extra_include={'car' : car})

car = s_car.save()
s.save(car=car)

Ok, now we don't save Car, if all our instance are validated!

Warning!
If use this hack you must design CarUser.clean() method for example, without perform filter, because Car instance isn't saved into db.

This is not the solution, but is a good hack to prevent the bad save, as described above at point 3.

Still investigating when there are Unique constraint.

What do you think?
I need your opinion.

Thanks!