Making queries

Once you’ve created your data models, Django automatically gives you a database-abstraction API that lets you create, retrieve, update and delete objects. This document explains how to use this API. Refer to the data model reference for full details of all the various model lookup options.

Throughout this guide (and in the reference), we’ll refer to the following models, which comprise a blog application:

from datetime import date

from django.db import models

class Blog(models.Model):
    name = models.CharField(max_length=100)
    tagline = models.TextField()

    def __str__(self):

class Author(models.Model):
    name = models.CharField(max_length=200)
    email = models.EmailField()

    def __str__(self):

class Entry(models.Model):
    blog = models.ForeignKey(Blog, on_delete=models.CASCADE)
    headline = models.CharField(max_length=255)
    body_text = models.TextField()
    pub_date = models.DateField()
    mod_date = models.DateField(
    authors = models.ManyToManyField(Author)
    number_of_comments = models.IntegerField(default=0)
    number_of_pingbacks = models.IntegerField(default=0)
    rating = models.IntegerField(default=5)

    def __str__(self):
        return self.headline

Creating objects

To represent database-table data in Python objects, Django uses an intuitive system: A model class represents a database table, and an instance of that class represents a particular record in the database table.

To create an object, instantiate it using keyword arguments to the model class, then call save() to save it to the database.

Assuming models live in a file mysite/blog/, here’s an example:

>>> from blog.models import Blog
>>> b = Blog(name='Beatles Blog', tagline='All the latest Beatles news.')

This performs an INSERT SQL statement behind the scenes. Django doesn’t hit the database until you explicitly call save().

The save() method has no return value.

See also

save() takes a number of advanced options not described here. See the documentation for save() for complete details.

To create and save an object in a single step, use the create() method.

Saving changes to objects

To save changes to an object that’s already in the database, use save().

Given a Blog instance b5 that has already been saved to the database, this example changes its name and updates its record in the database:

>>> = 'New name'

This performs an UPDATE SQL statement behind the scenes. Django doesn’t hit the database until you explicitly call save().

Saving ForeignKey and ManyToManyField fields

Updating a ForeignKey field works exactly the same way as saving a normal field – assign an object of the right type to the field in question. This example updates the blog attribute of an Entry instance entry, assuming appropriate instances of Entry and Blog are already saved to the database (so we can retrieve them below):

>>> from blog.models import Blog, Entry
>>> entry = Entry.objects.get(pk=1)
>>> cheese_blog = Blog.objects.get(name="Cheddar Talk")
>>> = cheese_blog

Updating a ManyToManyField works a little differently – use the add() method on the field to add a record to the relation. This example adds the Author instance joe to the entry object:

>>> from blog.models import Author
>>> joe = Author.objects.create(name="Joe")
>>> entry.authors.add(joe)

To add multiple records to a ManyToManyField in one go, include multiple arguments in the call to add(), like this:

>>> john = Author.objects.create(name="John")
>>> paul = Author.objects.create(name="Paul")
>>> george = Author.objects.create(name="George")
>>> ringo = Author.objects.create(name="Ringo")
>>> entry.authors.add(john, paul, george, ringo)

Django will complain if you try to assign or add an object of the wrong type.

Retrieving objects

To retrieve objects from your database, construct a QuerySet via a Manager on your model class.

A QuerySet represents a collection of objects from your database. It can have zero, one or many filters. Filters narrow down the query results based on the given parameters. In SQL terms, a QuerySet equates to a SELECT statement, and a filter is a limiting clause such as WHERE or LIMIT.

You get a QuerySet by using your model’s Manager. Each model has at least one Manager, and it’s called objects by default. Access it directly via the model class, like so:

>>> Blog.objects
<django.db.models.manager.Manager object at ...>
>>> b = Blog(name='Foo', tagline='Bar')
>>> b.objects
AttributeError: "Manager isn't accessible via Blog instances."


Managers are accessible only via model classes, rather than from model instances, to enforce a separation between “table-level” operations and “record-level” operations.

The Manager is the main source of QuerySets for a model. For example, Blog.objects.all() returns a QuerySet that contains all Blog objects in the database.

Retrieving all objects

The simplest way to retrieve objects from a table is to get all of them. To do this, use the all() method on a Manager:

>>> all_entries = Entry.objects.all()

The all() method returns a QuerySet of all the objects in the database.

Retrieving specific objects with filters

The QuerySet returned by all() describes all objects in the database table. Usually, though, you’ll need to select only a subset of the complete set of objects.

To create such a subset, you refine the initial QuerySet, adding filter conditions. The two most common ways to refine a QuerySet are:

Returns a new QuerySet containing objects that match the given lookup parameters.
Returns a new QuerySet containing objects that do not match the given lookup parameters.

The lookup parameters (**kwargs in the above function definitions) should be in the format described in Field lookups below.

For example, to get a QuerySet of blog entries from the year 2006, use filter() like so:


With the default manager class, it is the same as:


Chaining filters

The result of refining a QuerySet is itself a QuerySet, so it’s possible to chain refinements together. For example:

>>> Entry.objects.filter(
...     headline__startswith='What'
... ).exclude(
... ).filter(
..., 1, 30)
... )

This takes the initial QuerySet of all entries in the database, adds a filter, then an exclusion, then another filter. The final result is a QuerySet containing all entries with a headline that starts with “What”, that were published between January 30, 2005, and the current day.

Filtered QuerySets are unique

Each time you refine a QuerySet, you get a brand-new QuerySet that is in no way bound to the previous QuerySet. Each refinement creates a separate and distinct QuerySet that can be stored, used and reused.


>>> q1 = Entry.objects.filter(headline__startswith="What")
>>> q2 = q1.exclude(
>>> q3 = q1.filter(

These three QuerySets are separate. The first is a base QuerySet containing all entries that contain a headline starting with “What”. The second is a subset of the first, with an additional criteria that excludes records whose pub_date is today or in the future. The third is a subset of the first, with an additional criteria that selects only the records whose pub_date is today or in the future. The initial QuerySet (q1) is unaffected by the refinement process.

QuerySets are lazy

QuerySets are lazy – the act of creating a QuerySet doesn’t involve any database activity. You can stack filters together all day long, and Django won’t actually run the query until the QuerySet is evaluated. Take a look at this example:

>>> q = Entry.objects.filter(headline__startswith="What")
>>> q = q.filter(
>>> q = q.exclude(body_text__icontains="food")
>>> print(q)

Though this looks like three database hits, in fact it hits the database only once, at the last line (print(q)). In general, the results of a QuerySet aren’t fetched from the database until you “ask” for them. When you do, the QuerySet is evaluated by accessing the database. For more details on exactly when evaluation takes place, see When QuerySets are evaluated.

Retrieving a single object with get()

filter() will always give you a QuerySet, even if only a single object matches the query - in this case, it will be a QuerySet containing a single element.

If you know there is only one object that matches your query, you can use the get() method on a Manager which returns the object directly:

>>> one_entry = Entry.objects.get(pk=1)

You can use any query expression with get(), just like with filter() - again, see Field lookups below.

Note that there is a difference between using get(), and using filter() with a slice of [0]. If there are no results that match the query, get() will raise a DoesNotExist exception. This exception is an attribute of the model class that the query is being performed on - so in the code above, if there is no Entry object with a primary key of 1, Django will raise Entry.DoesNotExist.

Similarly, Django will complain if more than one item matches the get() query. In this case, it will raise MultipleObjectsReturned, which again is an attribute of the model class itself.

Other QuerySet methods

Most of the time you’ll use all(), get(), filter() and exclude() when you need to look up objects from the database. However, that’s far from all there is; see the QuerySet API Reference for a complete list of all the various QuerySet methods.

Limiting QuerySets

Use a subset of Python’s array-slicing syntax to limit your QuerySet to a certain number of results. This is the equivalent of SQL’s LIMIT and OFFSET clauses.

For example, this returns the first 5 objects (LIMIT 5):

>>> Entry.objects.all()[:5]

This returns the sixth through tenth objects (OFFSET 5 LIMIT 5):

>>> Entry.objects.all()[5:10]

Negative indexing (i.e. Entry.objects.all()[-1]) is not supported.

Generally, slicing a QuerySet returns a new QuerySet – it doesn’t evaluate the query. An exception is if you use the “step” parameter of Python slice syntax. For example, this would actually execute the query in order to return a list of every second object of the first 10:

>>> Entry.objects.all()[:10:2]

Further filtering or ordering of a sliced queryset is prohibited due to the ambiguous nature of how that might work.

To retrieve a single object rather than a list (e.g. SELECT foo FROM bar LIMIT 1), use an index instead of a slice. For example, this returns the first Entry in the database, after ordering entries alphabetically by headline:

>>> Entry.objects.order_by('headline')[0]

This is roughly equivalent to:

>>> Entry.objects.order_by('headline')[0:1].get()

Note, however, that the first of these will raise IndexError while the second will raise DoesNotExist if no objects match the given criteria. See get() for more details.

Field lookups

Field lookups are how you specify the meat of an SQL WHERE clause. They’re specified as keyword arguments to the QuerySet methods filter(), exclude() and get().

Basic lookups keyword arguments take the form field__lookuptype=value. (That’s a double-underscore). For example:

>>> Entry.objects.filter(pub_date__lte='2006-01-01')

translates (roughly) into the following SQL:

SELECT * FROM blog_entry WHERE pub_date <= '2006-01-01';

How this is possible

Python has the ability to define functions that accept arbitrary name-value arguments whose names and values are evaluated at runtime. For more information, see Keyword Arguments in the official Python tutorial.

The field specified in a lookup has to be the name of a model field. There’s one exception though, in case of a ForeignKey you can specify the field name suffixed with _id. In this case, the value parameter is expected to contain the raw value of the foreign model’s primary key. For example:

>>> Entry.objects.filter(blog_id=4)

If you pass an invalid keyword argument, a lookup function will raise TypeError.

The database API supports about two dozen lookup types; a complete reference can be found in the field lookup reference. To give you a taste of what’s available, here’s some of the more common lookups you’ll probably use:


An “exact” match. For example:

>>> Entry.objects.get(headline__exact="Cat bites dog")

Would generate SQL along these lines:

SELECT ... WHERE headline = 'Cat bites dog';

If you don’t provide a lookup type – that is, if your keyword argument doesn’t contain a double underscore – the lookup type is assumed to be exact.

For example, the following two statements are equivalent:

>>> Blog.objects.get(id__exact=14)  # Explicit form
>>> Blog.objects.get(id=14)         # __exact is implied

This is for convenience, because exact lookups are the common case.


A case-insensitive match. So, the query:

>>> Blog.objects.get(name__iexact="beatles blog")

Would match a Blog titled "Beatles Blog", "beatles blog", or even "BeAtlES blOG".


Case-sensitive containment test. For example:


Roughly translates to this SQL:

SELECT ... WHERE headline LIKE '%Lennon%';

Note this will match the headline 'Today Lennon honored' but not 'today lennon honored'.

There’s also a case-insensitive version, icontains.

startswith, endswith
Starts-with and ends-with search, respectively. There are also case-insensitive versions called istartswith and iendswith.

Again, this only scratches the surface. A complete reference can be found in the field lookup reference.

Lookups that span relationships

Django offers a powerful and intuitive way to “follow” relationships in lookups, taking care of the SQL JOINs for you automatically, behind the scenes. To span a relationship, use the field name of related fields across models, separated by double underscores, until you get to the field you want.

This example retrieves all Entry objects with a Blog whose name is 'Beatles Blog':

>>> Entry.objects.filter(blog__name='Beatles Blog')

This spanning can be as deep as you’d like.

It works backwards, too. While it can be customized, by default you refer to a “reverse” relationship in a lookup using the lowercase name of the model.

This example retrieves all Blog objects which have at least one Entry whose headline contains 'Lennon':

>>> Blog.objects.filter(entry__headline__contains='Lennon')

If you are filtering across multiple relationships and one of the intermediate models doesn’t have a value that meets the filter condition, Django will treat it as if there is an empty (all values are NULL), but valid, object there. All this means is that no error will be raised. For example, in this filter:


(if there was a related Author model), if there was no author associated with an entry, it would be treated as if there was also no name attached, rather than raising an error because of the missing author. Usually this is exactly what you want to have happen. The only case where it might be confusing is if you are using isnull. Thus:


will return Blog objects that have an empty name on the author and also those which have an empty author on the entry. If you don’t want those latter objects, you could write:

Blog.objects.filter(entry__authors__isnull=False, entry__authors__name__isnull=True)

Spanning multi-valued relationships

When spanning a ManyToManyField or a reverse ForeignKey (such as from Blog to Entry), filtering on multiple attributes raises the question of whether to require each attribute to coincide in the same related object. We might seek blogs that have an entry from 2008 with “Lennon” in its headline, or we might seek blogs that merely have any entry from 2008 as well as some newer or older entry with “Lennon” in its headline.

To select all blogs containing at least one entry from 2008 having “Lennon” in its headline (the same entry satisfying both conditions), we would write:

Blog.objects.filter(entry__headline__contains='Lennon', entry__pub_date__year=2008)

Otherwise, to perform a more permissive query selecting any blogs with merely some entry with “Lennon” in its headline and some entry from 2008, we would write:


Suppose there is only one blog that has both entries containing “Lennon” and entries from 2008, but that none of the entries from 2008 contained “Lennon”. The first query would not return any blogs, but the second query would return that one blog. (This is because the entries selected by the second filter may or may not be the same as the entries in the first filter. We are filtering the Blog items with each filter statement, not the Entry items.) In short, if each condition needs to match the same related object, then each should be contained in a single filter() call.


As the second (more permissive) query chains multiple filters, it performs multiple joins to the primary model, potentially yielding duplicates.

>>> from datetime import date
>>> beatles = Blog.objects.create(name='Beatles Blog')
>>> pop = Blog.objects.create(name='Pop Music Blog')
>>> Entry.objects.create(
...     blog=beatles,
...     headline='New Lennon Biography',
...     pub_date=date(2008, 6, 1),
... )
<Entry: New Lennon Biography>
>>> Entry.objects.create(
...     blog=beatles,
...     headline='New Lennon Biography in Paperback',
...     pub_date=date(2009, 6, 1),
... )
<Entry: New Lennon Biography in Paperback>
>>> Entry.objects.create(
...     blog=pop,
...     headline='Best Albums of 2008',
...     pub_date=date(2008, 12, 15),
... )
<Entry: Best Albums of 2008>
>>> Entry.objects.create(
...     blog=pop,
...     headline='Lennon Would Have Loved Hip Hop',
...     pub_date=date(2020, 4, 1),
... )
<Entry: Lennon Would Have Loved Hip Hop>
>>> Blog.objects.filter(
...     entry__headline__contains='Lennon',
...     entry__pub_date__year=2008,
... )
<QuerySet [<Blog: Beatles Blog>]>
>>> Blog.objects.filter(
...     entry__headline__contains='Lennon',
... ).filter(
...     entry__pub_date__year=2008,
... )
<QuerySet [<Blog: Beatles Blog>, <Blog: Beatles Blog>, <Blog: Pop Music Blog]>


The behavior of filter() for queries that span multi-value relationships, as described above, is not implemented equivalently for exclude(). Instead, the conditions in a single exclude() call will not necessarily refer to the same item.

For example, the following query would exclude blogs that contain both entries with “Lennon” in the headline and entries published in 2008:


However, unlike the behavior when using filter(), this will not limit blogs based on entries that satisfy both conditions. In order to do that, i.e. to select all blogs that do not contain entries published with “Lennon” that were published in 2008, you need to make two queries:


Filters can reference fields on the model

In the examples given so far, we have constructed filters that compare the value of a model field with a constant. But what if you want to compare the value of a model field with another field on the same model?

Django provides F expressions to allow such comparisons. Instances of F() act as a reference to a model field within a query. These references can then be used in query filters to compare the values of two different fields on the same model instance.

For example, to find a list of all blog entries that have had more comments than pingbacks, we construct an F() object to reference the pingback count, and use that F() object in the query:

>>> from django.db.models import F
>>> Entry.objects.filter(number_of_comments__gt=F('number_of_pingbacks'))

Django supports the use of addition, subtraction, multiplication, division, modulo, and power arithmetic with F() objects, both with constants and with other F() objects. To find all the blog entries with more than twice as many comments as pingbacks, we modify the query:

>>> Entry.objects.filter(number_of_comments__gt=F('number_of_pingbacks') * 2)

To find all the entries where the rating of the entry is less than the sum of the pingback count and comment count, we would issue the query:

>>> Entry.objects.filter(rating__lt=F('number_of_comments') + F('number_of_pingbacks'))

You can also use the double underscore notation to span relationships in an F() object. An F() object with a double underscore will introduce any joins needed to access the related object. For example, to retrieve all the entries where the author’s name is the same as the blog name, we could issue the query:

>>> Entry.objects.filter(authors__name=F('blog__name'))

For date and date/time fields, you can add or subtract a timedelta object. The following would return all entries that were modified more than 3 days after they were published:

>>> from datetime import timedelta
>>> Entry.objects.filter(mod_date__gt=F('pub_date') + timedelta(days=3))

The F() objects support bitwise operations by .bitand(), .bitor(), .bitxor(), .bitrightshift(), and .bitleftshift(). For example:

>>> F('somefield').bitand(16)


Oracle doesn’t support bitwise XOR operation.

Expressions can reference transforms

Django supports using transforms in expressions.

For example, to find all Entry objects published in the same year as they were last modified:

>>> from django.db.models import F
>>> Entry.objects.filter(pub_date__year=F('mod_date__year'))

To find the earliest year an entry was published, we can issue the query:

>>> from django.db.models import Min
>>> Entry.objects.aggregate(first_published_year=Min('pub_date__year'))

This example finds the value of the highest rated entry and the total number of comments on all entries for each year:

>>> from django.db.models import OuterRef, Subquery, Sum
>>> Entry.objects.values('pub_date__year').annotate(
...     top_rating=Subquery(
...         Entry.objects.filter(
...             pub_date__year=OuterRef('pub_date__year'),
...         ).order_by('-rating').values('rating')[:1]
...     ),
...     total_comments=Sum('number_of_comments'),
... )

The pk lookup shortcut

For convenience, Django provides a pk lookup shortcut, which stands for “primary key”.

In the example Blog model, the primary key is the id field, so these three statements are equivalent:

>>> Blog.objects.get(id__exact=14) # Explicit form
>>> Blog.objects.get(id=14) # __exact is implied
>>> Blog.objects.get(pk=14) # pk implies id__exact

The use of pk isn’t limited to __exact queries – any query term can be combined with pk to perform a query on the primary key of a model:

# Get blogs entries with id 1, 4 and 7
>>> Blog.objects.filter(pk__in=[1,4,7])

# Get all blog entries with id > 14
>>> Blog.objects.filter(pk__gt=14)

pk lookups also work across joins. For example, these three statements are equivalent:

>>> Entry.objects.filter(blog__id__exact=3) # Explicit form
>>> Entry.objects.filter(blog__id=3)        # __exact is implied
>>> Entry.objects.filter(blog__pk=3)        # __pk implies __id__exact

Escaping percent signs and underscores in LIKE statements

The field lookups that equate to LIKE SQL statements (iexact, contains, icontains, startswith, istartswith, endswith and iendswith) will automatically escape the two special characters used in LIKE statements – the percent sign and the underscore. (In a LIKE statement, the percent sign signifies a multiple-character wildcard and the underscore signifies a single-character wildcard.)

This means things should work intuitively, so the abstraction doesn’t leak. For example, to retrieve all the entries that contain a percent sign, use the percent sign as any other character:

>>> Entry.objects.filter(headline__contains='%')

Django takes care of the quoting for you; the resulting SQL will look something like this:

SELECT ... WHERE headline LIKE '%\%%';

Same goes for underscores. Both percentage signs and underscores are handled for you transparently.

Caching and QuerySets

Each QuerySet contains a cache to minimize database access. Understanding how it works will allow you to write the most efficient code.

In a newly created QuerySet, the cache is empty. The first time a QuerySet is evaluated – and, hence, a database query happens – Django saves the query results in the QuerySet’s cache and returns the results that have been explicitly requested (e.g., the next element, if the QuerySet is being iterated over). Subsequent evaluations of the QuerySet reuse the cached results.

Keep this caching behavior in mind, because it may bite you if you don’t use your QuerySets correctly. For example, the following will create two QuerySets, evaluate them, and throw them away:

>>> print([e.headline for e in Entry.objects.all()])
>>> print([e.pub_date for e in Entry.objects.all()])

That means the same database query will be executed twice, effectively doubling your database load. Also, there’s a possibility the two lists may not include the same database records, because an Entry may have been added or deleted in the split second between the two requests.

To avoid this problem, save the QuerySet and reuse it:

>>> queryset = Entry.objects.all()
>>> print([p.headline for p in queryset]) # Evaluate the query set.
>>> print([p.pub_date for p in queryset]) # Reuse the cache from the evaluation.

When QuerySets are not cached

Querysets do not always cache their results. When evaluating only part of the queryset, the cache is checked, but if it is not populated then the items returned by the subsequent query are not cached. Specifically, this means that limiting the queryset using an array slice or an index will not populate the cache.

For example, repeatedly getting a certain index in a queryset object will query the database each time:

>>> queryset = Entry.objects.all()
>>> print(queryset[5]) # Queries the database
>>> print(queryset[5]) # Queries the database again

However, if the entire queryset has already been evaluated, the cache will be checked instead:

>>> queryset = Entry.objects.all()
>>> [entry for entry in queryset] # Queries the database
>>> print(queryset[5]) # Uses cache
>>> print(queryset[5]) # Uses cache

Here are some examples of other actions that will result in the entire queryset being evaluated and therefore populate the cache:

>>> [entry for entry in queryset]
>>> bool(queryset)
>>> entry in queryset
>>> list(queryset)


Simply printing the queryset will not populate the cache. This is because the call to __repr__() only returns a slice of the entire queryset.

Asynchronous queries

New in Django 4.1.

If you are writing asynchronous views or code, you cannot use the ORM for queries in quite the way we have described above, as you cannot call blocking synchronous code from asynchronous code - it will block up the event loop (or, more likely, Django will notice and raise a SynchronousOnlyOperation to stop that from happening).

Fortunately, you can do many queries using Django’s asynchronous query APIs. Every method that might block - such as get() or delete() - has an asynchronous variant (aget() or adelete()), and when you iterate over results, you can use asynchronous iteration (async for) instead.

Query iteration

New in Django 4.1.

The default way of iterating over a query - with for - will result in a blocking database query behind the scenes as Django loads the results at iteration time. To fix this, you can swap to async for:

async for entry in Authors.objects.filter(name__startswith="A"):

Be aware that you also can’t do other things that might iterate over the queryset, such as wrapping list() around it to force its evaluation (you can use async for in a comprehension, if you want it).

Because QuerySet methods like filter() and exclude() do not actually run the query - they set up the queryset to run when it’s iterated over - you can use those freely in asynchronous code. For a guide to which methods can keep being used like this, and which have asynchronous versions, read the next section.

QuerySet and manager methods

New in Django 4.1.

Some methods on managers and querysets - like get() and first() - force execution of the queryset and are blocking. Some, like filter() and exclude(), don’t force execution and so are safe to run from asynchronous code. But how are you supposed to tell the difference?

While you could poke around and see if there is an a-prefixed version of the method (for example, we have aget() but not afilter()), there is a more logical way - look up what kind of method it is in the QuerySet reference.

In there, you’ll find the methods on QuerySets grouped into two sections:

  • Methods that return new querysets: These are the non-blocking ones, and don’t have asynchronous versions. You’re free to use these in any situation, though read the notes on defer() and only() before you use them.
  • Methods that do not return querysets: These are the blocking ones, and have asynchronous versions - the asynchronous name for each is noted in its documentation, though our standard pattern is to add an a prefix.

Using this distinction, you can work out when you need to use asynchronous versions, and when you don’t. For example, here’s a valid asynchronous query:

user = await User.objects.filter(username=my_input).afirst()

filter() returns a queryset, and so it’s fine to keep chaining it inside an asynchronous environment, whereas first() evaluates and returns a model instance - thus, we change to afirst(), and use await at the front of the whole expression in order to call it in an asynchronous-friendly way.


If you forget to put the await part in, you may see errors like “coroutine object has no attribute x” or “<coroutine …>” strings in place of your model instances. If you ever see these, you are missing an await somewhere to turn that coroutine into a real value.


New in Django 4.1.

Transactions are not currently supported with asynchronous queries and updates. You will find that trying to use one raises SynchronousOnlyOperation.

If you wish to use a transaction, we suggest you write your ORM code inside a separate, synchronous function and then call that using sync_to_async - see Asynchronous support for more.

Querying JSONField

Lookups implementation is different in JSONField, mainly due to the existence of key transformations. To demonstrate, we will use the following example model:

from django.db import models

class Dog(models.Model):
    name = models.CharField(max_length=200)
    data = models.JSONField(null=True)

    def __str__(self):

Storing and querying for None

As with other fields, storing None as the field’s value will store it as SQL NULL. While not recommended, it is possible to store JSON scalar null instead of SQL NULL by using Value('null').

Whichever of the values is stored, when retrieved from the database, the Python representation of the JSON scalar null is the same as SQL NULL, i.e. None. Therefore, it can be hard to distinguish between them.

This only applies to None as the top-level value of the field. If None is inside a list or dict, it will always be interpreted as JSON null.

When querying, None value will always be interpreted as JSON null. To query for SQL NULL, use isnull:

>>> Dog.objects.create(name='Max', data=None)  # SQL NULL.
<Dog: Max>
>>> Dog.objects.create(name='Archie', data=Value('null'))  # JSON null.
<Dog: Archie>
>>> Dog.objects.filter(data=None)
<QuerySet [<Dog: Archie>]>
>>> Dog.objects.filter(data=Value('null'))
<QuerySet [<Dog: Archie>]>
>>> Dog.objects.filter(data__isnull=True)
<QuerySet [<Dog: Max>]>
>>> Dog.objects.filter(data__isnull=False)
<QuerySet [<Dog: Archie>]>

Unless you are sure you wish to work with SQL NULL values, consider setting null=False and providing a suitable default for empty values, such as default=dict.


Storing JSON scalar null does not violate null=False.

Key, index, and path transforms

To query based on a given dictionary key, use that key as the lookup name:

>>> Dog.objects.create(name='Rufus', data={
...     'breed': 'labrador',
...     'owner': {
...         'name': 'Bob',
...         'other_pets': [{
...             'name': 'Fishy',
...         }],
...     },
... })
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'breed': 'collie', 'owner': None})
<Dog: Meg>
>>> Dog.objects.filter(data__breed='collie')
<QuerySet [<Dog: Meg>]>

Multiple keys can be chained together to form a path lookup:

>>> Dog.objects.filter(data__owner__name='Bob')
<QuerySet [<Dog: Rufus>]>

If the key is an integer, it will be interpreted as an index transform in an array:

>>> Dog.objects.filter(data__owner__other_pets__0__name='Fishy')
<QuerySet [<Dog: Rufus>]>

If the key you wish to query by clashes with the name of another lookup, use the contains lookup instead.

To query for missing keys, use the isnull lookup:

>>> Dog.objects.create(name='Shep', data={'breed': 'collie'})
<Dog: Shep>
>>> Dog.objects.filter(data__owner__isnull=True)
<QuerySet [<Dog: Shep>]>


The lookup examples given above implicitly use the exact lookup. Key, index, and path transforms can also be chained with: icontains, endswith, iendswith, iexact, regex, iregex, startswith, istartswith, lt, lte, gt, and gte, as well as with Containment and key lookups.


Due to the way in which key-path queries work, exclude() and filter() are not guaranteed to produce exhaustive sets. If you want to include objects that do not have the path, add the isnull lookup.


Since any string could be a key in a JSON object, any lookup other than those listed below will be interpreted as a key lookup. No errors are raised. Be extra careful for typing mistakes, and always check your queries work as you intend.

MariaDB and Oracle users

Using order_by() on key, index, or path transforms will sort the objects using the string representation of the values. This is because MariaDB and Oracle Database do not provide a function that converts JSON values into their equivalent SQL values.

Oracle users

On Oracle Database, using None as the lookup value in an exclude() query will return objects that do not have null as the value at the given path, including objects that do not have the path. On other database backends, the query will return objects that have the path and the value is not null.

PostgreSQL users

On PostgreSQL, if only one key or index is used, the SQL operator -> is used. If multiple operators are used then the #> operator is used.

SQLite users

On SQLite, "true", "false", and "null" string values will always be interpreted as True, False, and JSON null respectively.

Containment and key lookups


The contains lookup is overridden on JSONField. The returned objects are those where the given dict of key-value pairs are all contained in the top-level of the field. For example:

>>> Dog.objects.create(name='Rufus', data={'breed': 'labrador', 'owner': 'Bob'})
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'breed': 'collie', 'owner': 'Bob'})
<Dog: Meg>
>>> Dog.objects.create(name='Fred', data={})
<Dog: Fred>
>>> Dog.objects.filter(data__contains={'owner': 'Bob'})
<QuerySet [<Dog: Rufus>, <Dog: Meg>]>
>>> Dog.objects.filter(data__contains={'breed': 'collie'})
<QuerySet [<Dog: Meg>]>

Oracle and SQLite

contains is not supported on Oracle and SQLite.


This is the inverse of the contains lookup - the objects returned will be those where the key-value pairs on the object are a subset of those in the value passed. For example:

>>> Dog.objects.create(name='Rufus', data={'breed': 'labrador', 'owner': 'Bob'})
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'breed': 'collie', 'owner': 'Bob'})
<Dog: Meg>
>>> Dog.objects.create(name='Fred', data={})
<Dog: Fred>
>>> Dog.objects.filter(data__contained_by={'breed': 'collie', 'owner': 'Bob'})
<QuerySet [<Dog: Meg>, <Dog: Fred>]>
>>> Dog.objects.filter(data__contained_by={'breed': 'collie'})
<QuerySet [<Dog: Fred>]>

Oracle and SQLite

contained_by is not supported on Oracle and SQLite.


Returns objects where the given key is in the top-level of the data. For example:

>>> Dog.objects.create(name='Rufus', data={'breed': 'labrador'})
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'breed': 'collie', 'owner': 'Bob'})
<Dog: Meg>
>>> Dog.objects.filter(data__has_key='owner')
<QuerySet [<Dog: Meg>]>


Returns objects where all of the given keys are in the top-level of the data. For example:

>>> Dog.objects.create(name='Rufus', data={'breed': 'labrador'})
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'breed': 'collie', 'owner': 'Bob'})
<Dog: Meg>
>>> Dog.objects.filter(data__has_keys=['breed', 'owner'])
<QuerySet [<Dog: Meg>]>


Returns objects where any of the given keys are in the top-level of the data. For example:

>>> Dog.objects.create(name='Rufus', data={'breed': 'labrador'})
<Dog: Rufus>
>>> Dog.objects.create(name='Meg', data={'owner': 'Bob'})
<Dog: Meg>
>>> Dog.objects.filter(data__has_any_keys=['owner', 'breed'])
<QuerySet [<Dog: Rufus>, <Dog: Meg>]>

Complex lookups with Q objects

Keyword argument queries – in filter(), etc. – are “AND”ed together. If you need to execute more complex queries (for example, queries with OR statements), you can use Q objects.

A Q object (django.db.models.Q) is an object used to encapsulate a collection of keyword arguments. These keyword arguments are specified as in “Field lookups” above.

For example, this Q object encapsulates a single LIKE query:

from django.db.models import Q

Q objects can be combined using the &, |, and ^ operators. When an operator is used on two Q objects, it yields a new Q object.

For example, this statement yields a single Q object that represents the “OR” of two "question__startswith" queries:

Q(question__startswith='Who') | Q(question__startswith='What')

This is equivalent to the following SQL WHERE clause:

WHERE question LIKE 'Who%' OR question LIKE 'What%'

You can compose statements of arbitrary complexity by combining Q objects with the &, |, and ^ operators and use parenthetical grouping. Also, Q objects can be negated using the ~ operator, allowing for combined lookups that combine both a normal query and a negated (NOT) query:

Q(question__startswith='Who') | ~Q(pub_date__year=2005)

Each lookup function that takes keyword-arguments (e.g. filter(), exclude(), get()) can also be passed one or more Q objects as positional (not-named) arguments. If you provide multiple Q object arguments to a lookup function, the arguments will be “AND”ed together. For example:

    Q(pub_date=date(2005, 5, 2)) | Q(pub_date=date(2005, 5, 6))

… roughly translates into the SQL:

SELECT * from polls WHERE question LIKE 'Who%'
    AND (pub_date = '2005-05-02' OR pub_date = '2005-05-06')

Lookup functions can mix the use of Q objects and keyword arguments. All arguments provided to a lookup function (be they keyword arguments or Q objects) are “AND”ed together. However, if a Q object is provided, it must precede the definition of any keyword arguments. For example:

    Q(pub_date=date(2005, 5, 2)) | Q(pub_date=date(2005, 5, 6)),

… would be a valid query, equivalent to the previous example; but:

    Q(pub_date=date(2005, 5, 2)) | Q(pub_date=date(2005, 5, 6))

… would not be valid.

See also

The OR lookups examples in Django’s unit tests show some possible uses of Q.

Changed in Django 4.1:

Support for the ^ (XOR) operator was added.

Comparing objects

To compare two model instances, use the standard Python comparison operator, the double equals sign: ==. Behind the scenes, that compares the primary key values of two models.

Using the Entry example above, the following two statements are equivalent:

>>> some_entry == other_entry
>>> ==

If a model’s primary key isn’t called id, no problem. Comparisons will always use the primary key, whatever it’s called. For example, if a model’s primary key field is called name, these two statements are equivalent:

>>> some_obj == other_obj
>>> ==

Deleting objects

The delete method, conveniently, is named delete(). This method immediately deletes the object and returns the number of objects deleted and a dictionary with the number of deletions per object type. Example:

>>> e.delete()
(1, {'blog.Entry': 1})

You can also delete objects in bulk. Every QuerySet has a delete() method, which deletes all members of that QuerySet.

For example, this deletes all Entry objects with a pub_date year of 2005:

>>> Entry.objects.filter(pub_date__year=2005).delete()
(5, {'webapp.Entry': 5})

Keep in mind that this will, whenever possible, be executed purely in SQL, and so the delete() methods of individual object instances will not necessarily be called during the process. If you’ve provided a custom delete() method on a model class and want to ensure that it is called, you will need to “manually” delete instances of that model (e.g., by iterating over a QuerySet and calling delete() on each object individually) rather than using the bulk delete() method of a QuerySet.

When Django deletes an object, by default it emulates the behavior of the SQL constraint ON DELETE CASCADE – in other words, any objects which had foreign keys pointing at the object to be deleted will be deleted along with it. For example:

b = Blog.objects.get(pk=1)
# This will delete the Blog and all of its Entry objects.

This cascade behavior is customizable via the on_delete argument to the ForeignKey.

Note that delete() is the only QuerySet method that is not exposed on a Manager itself. This is a safety mechanism to prevent you from accidentally requesting Entry.objects.delete(), and deleting all the entries. If you do want to delete all the objects, then you have to explicitly request a complete query set:


Copying model instances

Although there is no built-in method for copying model instances, it is possible to easily create new instance with all fields’ values copied. In the simplest case, you can set pk to None and _state.adding to True. Using our blog example:

blog = Blog(name='My blog', tagline='Blogging is easy') # == 1 = None
blog._state.adding = True # == 2

Things get more complicated if you use inheritance. Consider a subclass of Blog:

class ThemeBlog(Blog):
    theme = models.CharField(max_length=200)

django_blog = ThemeBlog(name='Django', tagline='Django is easy', theme='python') # == 3

Due to how inheritance works, you have to set both pk and id to None, and _state.adding to True: = None = None
django_blog._state.adding = True # == 4

This process doesn’t copy relations that aren’t part of the model’s database table. For example, Entry has a ManyToManyField to Author. After duplicating an entry, you must set the many-to-many relations for the new entry:

entry = Entry.objects.all()[0] # some previous entry
old_authors = entry.authors.all() = None
entry._state.adding = True

For a OneToOneField, you must duplicate the related object and assign it to the new object’s field to avoid violating the one-to-one unique constraint. For example, assuming entry is already duplicated as above:

detail = EntryDetail.objects.all()[0] = None
detail._state.adding = True
detail.entry = entry

Updating multiple objects at once

Sometimes you want to set a field to a particular value for all the objects in a QuerySet. You can do this with the update() method. For example:

# Update all the headlines with pub_date in 2007.
Entry.objects.filter(pub_date__year=2007).update(headline='Everything is the same')

You can only set non-relation fields and ForeignKey fields using this method. To update a non-relation field, provide the new value as a constant. To update ForeignKey fields, set the new value to be the new model instance you want to point to. For example:

>>> b = Blog.objects.get(pk=1)

# Change every Entry so that it belongs to this Blog.
>>> Entry.objects.update(blog=b)

The update() method is applied instantly and returns the number of rows matched by the query (which may not be equal to the number of rows updated if some rows already have the new value). The only restriction on the QuerySet being updated is that it can only access one database table: the model’s main table. You can filter based on related fields, but you can only update columns in the model’s main table. Example:

>>> b = Blog.objects.get(pk=1)

# Update all the headlines belonging to this Blog.
>>> Entry.objects.filter(blog=b).update(headline='Everything is the same')

Be aware that the update() method is converted directly to an SQL statement. It is a bulk operation for direct updates. It doesn’t run any save() methods on your models, or emit the pre_save or post_save signals (which are a consequence of calling save()), or honor the auto_now field option. If you want to save every item in a QuerySet and make sure that the save() method is called on each instance, you don’t need any special function to handle that. Loop over them and call save():

for item in my_queryset:

Calls to update can also use F expressions to update one field based on the value of another field in the model. This is especially useful for incrementing counters based upon their current value. For example, to increment the pingback count for every entry in the blog:

>>> Entry.objects.update(number_of_pingbacks=F('number_of_pingbacks') + 1)

However, unlike F() objects in filter and exclude clauses, you can’t introduce joins when you use F() objects in an update – you can only reference fields local to the model being updated. If you attempt to introduce a join with an F() object, a FieldError will be raised:

# This will raise a FieldError
>>> Entry.objects.update(headline=F('blog__name'))

Falling back to raw SQL

If you find yourself needing to write an SQL query that is too complex for Django’s database-mapper to handle, you can fall back on writing SQL by hand. Django has a couple of options for writing raw SQL queries; see Performing raw SQL queries.

Finally, it’s important to note that the Django database layer is merely an interface to your database. You can access your database via other tools, programming languages or database frameworks; there’s nothing Django-specific about your database.

Back to Top