Django Model() vs Model.objects.create()

The differences between Model() and Model.objects.create() are the following:

1. INSERT vs UPDATE

   Model.save() does either INSERT or UPDATE of an object in a DB, while Model.objects.create() does only INSERT.

   Model.save() does

   • UPDATE If the object’s primary key attribute is set to a value that evaluates to True

   • INSERT If the object’s primary key attribute is not set or if the UPDATE didn’t update anything (e.g. if primary key is set to a value that doesn’t exist in the database).

2. Existing primary key

   If primary key attribute is set to a value and such primary key already exists, then Model.save() performs UPDATE, but Model.objects.create() raises IntegrityError.

   Consider the following models.py:

   class Subject(models.Model): subject_id = models.PositiveIntegerField(primary_key=True, db_column='subject_id') name = models.CharField(max_length=255) max_marks = models.PositiveIntegerField() 

   1. Insert/Update to db with Model.save()

      physics = Subject(subject_id=1, name='Physics', max_marks=100) physics.save() math = Subject(subject_id=1, name='Math', max_marks=50) # Case of update math.save() 

      Result:

      Subject.objects.all().values() <QuerySet [{'subject_id': 1, 'name': 'Math', 'max_marks': 50}]> 

   2. Insert to db with Model.objects.create()

      Subject.objects.create(subject_id=1, name='Chemistry', max_marks=100) IntegrityError: UNIQUE constraint failed: m****t.subject_id 

   Explanation: In the example, math.save() does an UPDATE (changes name from Physics to Math, and max_marks from 100 to 50), because subject_id is a primary key and subject_id=1 already exists in the DB. But Subject.objects.create() raises IntegrityError, because, again the primary key subject_id with the value 1 already exists.

3. Forced insert

   Model.save() can be made to behave as Model.objects.create() by using force_insert=True parameter: Model.save(force_insert=True).

4. Return value

   Model.save() return None where Model.objects.create() return model instance i.e. package_name.models.Model

Conclusion: Model.objects.create() does model initialization and performs save() with force_insert=True.

Excerpt from the source code of Model.objects.create()

def create(self, **kwargs): """ Create a new object with the given kwargs, saving it to the database and returning the created object. """ obj = self.model(**kwargs) self._for_write = True obj.save(force_insert=True, using=self.db) return obj 

For more details follow the links:

1. https://docs.djangoproject.com/en/stable/ref/models/querysets/#create

2. https://github.com/django/django/blob/2d8dcba03aae200aaa103ec1e69f0a0038ec2f85/django/db/models/query.py#L440

The two syntaxes are not equivalent and it can lead to unexpected errors. Here is a simple example showing the differences. If you have a model:

from django.db import models class Test(models.Model): added = models.DateTimeField(auto_now_add=True) 

And you create a first object:

foo = Test.objects.create(pk=1) 

Then you try to create an object with the same primary key:

foo_duplicate = Test.objects.create(pk=1) # returns the error: # django.db.utils.IntegrityError: (1062, "Duplicate entry '1' for key 'PRIMARY'") foo_duplicate = Test(pk=1).save() # returns the error: # django.db.utils.IntegrityError: (1048, "Column 'added' cannot be null") 

UPDATE 15.3.2017:

I have opened a Django-issue on this and it seems to be preliminary accepted here: https://code.djangoproject.com/ticket/27825

My experience is that when using the Constructor (ORM) class by references with Django 1.10.5 there might be some inconsistencies in the data (i.e. the attributes of the created object may get the type of the input data instead of the casted type of the ORM object property) example:

models

class Payment(models.Model): amount_cash = models.DecimalField() 

some_test.py - object.create

Class SomeTestCase: def generate_orm_obj(self, _constructor, base_data=None, modifiers=None): objs = [] if not base_data: base_data = {'amount_case': 123.00} for modifier in modifiers: actual_data = deepcopy(base_data) actual_data.update(modifier) # Hacky fix, _obj = _constructor.objects.create(**actual_data) print(type(_obj.amount_cash)) # Decimal assert created objs.append(_obj) return objs 

some_test.py - Constructor()

Class SomeTestCase: def generate_orm_obj(self, _constructor, base_data=None, modifiers=None): objs = [] if not base_data: base_data = {'amount_case': 123.00} for modifier in modifiers: actual_data = deepcopy(base_data) actual_data.update(modifier) # Hacky fix, _obj = _constructor(**actual_data) print(type(_obj.amount_cash)) # Float assert created objs.append(_obj) return objs 

Model.objects.create() creates a model instance and saves it. Model() only creates an in memory model instance. It's not saved to the database until you call the instance's save() method to save it. That's when validation happens also.

• Use Model() to create model instances you would want to use right away. You don’t want to store these in a backup store (database or file) probably because of latency or what not. You could cache though (as static variables ^{[Stackoverflow]} for example)
• Use Model.objects.create() to create something and store right away in some database somewhere.

Does the second one immediately create a BarModel in the database, while for FooModel, the save() method has to be called explicitly to add it to the database?

Exactly right!