QuerySet API reference

This document describes the details of the QuerySet API. It builds on the material presented in the model and database query guides, so you’ll probably want to read and understand those documents before reading this one.

Throughout this reference we’ll use the example Weblog models presented in the database query guide.

When QuerySets are evaluated

Internally, a QuerySet can be constructed, filtered, sliced, and generally passed around without actually hitting the database. No database activity actually occurs until you do something to evaluate the queryset.

You can evaluate a QuerySet in the following ways:

  • Iteration. A QuerySet is iterable, and it executes its database query the first time you iterate over it. For example, this will print the headline of all entries in the database:

    for e in Entry.objects.all():

    Note: Don’t use this if all you want to do is determine if at least one result exists. It’s more efficient to use exists().

  • Slicing. As explained in Limiting QuerySets, a QuerySet can be sliced, using Python’s array-slicing syntax. Slicing an unevaluated QuerySet usually returns another unevaluated QuerySet, but Django will execute the database query if you use the “step” parameter of slice syntax, and will return a list. Slicing a QuerySet that has been evaluated (partially or fully) also returns a list.

  • Pickling/Caching. See the following section for details of what is involved when pickling QuerySets. The important thing for the purposes of this section is that the results are read from the database.

  • repr(). A QuerySet is evaluated when you call repr() on it. This is for convenience in the Python interactive interpreter, so you can immediately see your results when using the API interactively.

  • len(). A QuerySet is evaluated when you call len() on it. This, as you might expect, returns the length of the result list.

    Note: Don’t use len() on QuerySets if all you want to do is determine the number of records in the set. It’s much more efficient to handle a count at the database level, using SQL’s SELECT COUNT(*), and Django provides a count() method for precisely this reason. See count() below.

  • list(). Force evaluation of a QuerySet by calling list() on it. For example:

    entry_list = list(Entry.objects.all())

    Be warned, though, that this could have a large memory overhead, because Django will load each element of the list into memory. In contrast, iterating over a QuerySet will take advantage of your database to load data and instantiate objects only as you need them.

  • bool(). Testing a QuerySet in a boolean context, such as using bool(), or, and or an if statement, will cause the query to be executed. If there is at least one result, the QuerySet is True, otherwise False. For example:

    if Entry.objects.filter(headline="Test"):
       print("There is at least one Entry with the headline Test")

    Note: Don’t use this if all you want to do is determine if at least one result exists, and don’t need the actual objects. It’s more efficient to use exists() (see below).

Pickling QuerySets

If you pickle a QuerySet, this will force all the results to be loaded into memory prior to pickling. Pickling is usually used as a precursor to caching and when the cached queryset is reloaded, you want the results to already be present and ready for use (reading from the database can take some time, defeating the purpose of caching). This means that when you unpickle a QuerySet, it contains the results at the moment it was pickled, rather than the results that are currently in the database.

If you only want to pickle the necessary information to recreate the QuerySet from the database at a later time, pickle the query attribute of the QuerySet. You can then recreate the original QuerySet (without any results loaded) using some code like this:

>>> import pickle
>>> query = pickle.loads(s)     # Assuming 's' is the pickled string.
>>> qs = MyModel.objects.all()
>>> qs.query = query            # Restore the original 'query'.

The query attribute is an opaque object. It represents the internals of the query construction and is not part of the public API. However, it is safe (and fully supported) to pickle and unpickle the attribute’s contents as described here.

You can’t share pickles between versions

Pickles of QuerySets are only valid for the version of Django that was used to generate them. If you generate a pickle using Django version N, there is no guarantee that pickle will be readable with Django version N+1. Pickles should not be used as part of a long-term archival strategy.

QuerySet API

Though you usually won’t create one manually — you’ll go through a Manager — here’s the formal declaration of a QuerySet:

class QuerySet([model=None, query=None, using=None])

Usually when you’ll interact with a QuerySet you’ll use it by chaining filters. To make this work, most QuerySet methods return new querysets. These methods are covered in detail later in this section.

The QuerySet class has two public attributes you can use for introspection:


True if the QuerySet is ordered — i.e. has an order_by() clause or a default ordering on the model. False otherwise.


The database that will be used if this query is executed now.


The query parameter to QuerySet exists so that specialized query subclasses such as GeoQuerySet can reconstruct internal query state. The value of the parameter is an opaque representation of that query state and is not part of a public API. To put it simply: if you need to ask, you don’t need to use it.

Methods that return new QuerySets

Django provides a range of QuerySet refinement methods that modify either the types of results returned by the QuerySet or the way its SQL query is executed.



Returns a new QuerySet containing objects that match the given lookup parameters.

The lookup parameters (**kwargs) should be in the format described in Field lookups below. Multiple parameters are joined via AND in the underlying SQL statement.



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

The lookup parameters (**kwargs) should be in the format described in Field lookups below. Multiple parameters are joined via AND in the underlying SQL statement, and the whole thing is enclosed in a NOT().

This example excludes all entries whose pub_date is later than 2005-1-3 AND whose headline is “Hello”:

Entry.objects.exclude(pub_date__gt=datetime.date(2005, 1, 3), headline='Hello')

In SQL terms, that evaluates to:

WHERE NOT (pub_date > '2005-1-3' AND headline = 'Hello')

This example excludes all entries whose pub_date is later than 2005-1-3 OR whose headline is “Hello”:

Entry.objects.exclude(pub_date__gt=datetime.date(2005, 1, 3)).exclude(headline='Hello')

In SQL terms, that evaluates to:

WHERE NOT pub_date > '2005-1-3'
AND NOT headline = 'Hello'

Note the second example is more restrictive.


annotate(*args, **kwargs)

Annotates each object in the QuerySet with the provided list of aggregate values (averages, sums, etc) that have been computed over the objects that are related to the objects in the QuerySet. Each argument to annotate() is an annotation that will be added to each object in the QuerySet that is returned.

The aggregation functions that are provided by Django are described in Aggregation Functions below.

Annotations specified using keyword arguments will use the keyword as the alias for the annotation. Anonymous arguments will have an alias generated for them based upon the name of the aggregate function and the model field that is being aggregated.

For example, if you were manipulating a list of blogs, you may want to determine how many entries have been made in each blog:

>>> q = Blog.objects.annotate(Count('entry'))
# The name of the first blog
>>> q[0].name
# The number of entries on the first blog
>>> q[0].entry__count

The Blog model doesn’t define an entry__count attribute by itself, but by using a keyword argument to specify the aggregate function, you can control the name of the annotation:

>>> q = Blog.objects.annotate(number_of_entries=Count('entry'))
# The number of entries on the first blog, using the name provided
>>> q[0].number_of_entries

For an in-depth discussion of aggregation, see the topic guide on Aggregation.



By default, results returned by a QuerySet are ordered by the ordering tuple given by the ordering option in the model’s Meta. You can override this on a per-QuerySet basis by using the order_by method.


Entry.objects.filter(pub_date__year=2005).order_by('-pub_date', 'headline')

The result above will be ordered by pub_date descending, then by headline ascending. The negative sign in front of "-pub_date" indicates descending order. Ascending order is implied. To order randomly, use "?", like so:


Note: order_by('?') queries may be expensive and slow, depending on the database backend you’re using.

To order by a field in a different model, use the same syntax as when you are querying across model relations. That is, the name of the field, followed by a double underscore (__), followed by the name of the field in the new model, and so on for as many models as you want to join. For example:

Entry.objects.order_by('blog__name', 'headline')

If you try to order by a field that is a relation to another model, Django will use the default ordering on the related model (or order by the related model’s primary key if there is no Meta.ordering specified. For example:


...is identical to:


...since the Blog model has no default ordering specified.

Be cautious when ordering by fields in related models if you are also using distinct(). See the note in distinct() for an explanation of how related model ordering can change the expected results.

It is permissible to specify a multi-valued field to order the results by (for example, a ManyToManyField field). Normally this won’t be a sensible thing to do and it’s really an advanced usage feature. However, if you know that your queryset’s filtering or available data implies that there will only be one ordering piece of data for each of the main items you are selecting, the ordering may well be exactly what you want to do. Use ordering on multi-valued fields with care and make sure the results are what you expect.

There’s no way to specify whether ordering should be case sensitive. With respect to case-sensitivity, Django will order results however your database backend normally orders them.

If you don’t want any ordering to be applied to a query, not even the default ordering, call order_by() with no parameters.

You can tell if a query is ordered or not by checking the QuerySet.ordered attribute, which will be True if the QuerySet has been ordered in any way.



Use the reverse() method to reverse the order in which a queryset’s elements are returned. Calling reverse() a second time restores the ordering back to the normal direction.

To retrieve the ‘’last’’ five items in a queryset, you could do this:


Note that this is not quite the same as slicing from the end of a sequence in Python. The above example will return the last item first, then the penultimate item and so on. If we had a Python sequence and looked at seq[-5:], we would see the fifth-last item first. Django doesn’t support that mode of access (slicing from the end), because it’s not possible to do it efficiently in SQL.

Also, note that reverse() should generally only be called on a QuerySet which has a defined ordering (e.g., when querying against a model which defines a default ordering, or when using order_by()). If no such ordering is defined for a given QuerySet, calling reverse() on it has no real effect (the ordering was undefined prior to calling reverse(), and will remain undefined afterward).



Returns a new QuerySet that uses SELECT DISTINCT in its SQL query. This eliminates duplicate rows from the query results.

By default, a QuerySet will not eliminate duplicate rows. In practice, this is rarely a problem, because simple queries such as Blog.objects.all() don’t introduce the possibility of duplicate result rows. However, if your query spans multiple tables, it’s possible to get duplicate results when a QuerySet is evaluated. That’s when you’d use distinct().


Any fields used in an order_by() call are included in the SQL SELECT columns. This can sometimes lead to unexpected results when used in conjunction with distinct(). If you order by fields from a related model, those fields will be added to the selected columns and they may make otherwise duplicate rows appear to be distinct. Since the extra columns don’t appear in the returned results (they are only there to support ordering), it sometimes looks like non-distinct results are being returned.

Similarly, if you use a values() query to restrict the columns selected, the columns used in any order_by() (or default model ordering) will still be involved and may affect uniqueness of the results.

The moral here is that if you are using distinct() be careful about ordering by related models. Similarly, when using distinct() and values() together, be careful when ordering by fields not in the values() call.

As of Django 1.4, you can pass positional arguments (*fields) in order to specify the names of fields to which the DISTINCT should apply. This translates to a SELECT DISTINCT ON SQL query.

Here’s the difference. For a normal distinct() call, the database compares each field in each row when determining which rows are distinct. For a distinct() call with specified field names, the database will only compare the specified field names.


This ability to specify field names is only available in PostgreSQL.


When you specify field names, you must provide an order_by() in the QuerySet, and the fields in order_by() must start with the fields in distinct(), in the same order.

For example, SELECT DISTINCT ON (a) gives you the first row for each value in column a. If you don’t specify an order, you’ll get some arbitrary row.


>>> Author.objects.distinct()

>>> Entry.objects.order_by('pub_date').distinct('pub_date')

>>> Entry.objects.order_by('blog').distinct('blog')

>>> Entry.objects.order_by('author', 'pub_date').distinct('author', 'pub_date')

>>> Entry.objects.order_by('blog__name', 'mod_date').distinct('blog__name', 'mod_date')

>>> Entry.objects.order_by('author', 'pub_date').distinct('author')



Returns a ValuesQuerySet — a QuerySet subclass that returns dictionaries when used as an iterable, rather than model-instance objects.

Each of those dictionaries represents an object, with the keys corresponding to the attribute names of model objects.

This example compares the dictionaries of values() with the normal model objects:

# This list contains a Blog object.
>>> Blog.objects.filter(name__startswith='Beatles')
[<Blog: Beatles Blog>]

# This list contains a dictionary.
>>> Blog.objects.filter(name__startswith='Beatles').values()
[{'id': 1, 'name': 'Beatles Blog', 'tagline': 'All the latest Beatles news.'}]

The values() method takes optional positional arguments, *fields, which specify field names to which the SELECT should be limited. If you specify the fields, each dictionary will contain only the field keys/values for the fields you specify. If you don’t specify the fields, each dictionary will contain a key and value for every field in the database table.


>>> Blog.objects.values()
[{'id': 1, 'name': 'Beatles Blog', 'tagline': 'All the latest Beatles news.'}],
>>> Blog.objects.values('id', 'name')
[{'id': 1, 'name': 'Beatles Blog'}]

A few subtleties that are worth mentioning:

  • If you have a field called foo that is a ForeignKey, the default values() call will return a dictionary key called foo_id, since this is the name of the hidden model attribute that stores the actual value (the foo attribute refers to the related model). When you are calling values() and passing in field names, you can pass in either foo or foo_id and you will get back the same thing (the dictionary key will match the field name you passed in).

    For example:

    >>> Entry.objects.values()
    [{'blog_id': 1, 'headline': u'First Entry', ...}, ...]
    >>> Entry.objects.values('blog')
    [{'blog': 1}, ...]
    >>> Entry.objects.values('blog_id')
    [{'blog_id': 1}, ...]
  • When using values() together with distinct(), be aware that ordering can affect the results. See the note in distinct() for details.

  • If you use a values() clause after an extra() call, any fields defined by a select argument in the extra() must be explicitly included in the values() call. Any extra() call made after a values() call will have its extra selected fields ignored.

A ValuesQuerySet is useful when you know you’re only going to need values from a small number of the available fields and you won’t need the functionality of a model instance object. It’s more efficient to select only the fields you need to use.

Finally, note a ValuesQuerySet is a subclass of QuerySet, so it has all methods of QuerySet. You can call filter() on it, or order_by(), or whatever. Yes, that means these two calls are identical:


The people who made Django prefer to put all the SQL-affecting methods first, followed (optionally) by any output-affecting methods (such as values()), but it doesn’t really matter. This is your chance to really flaunt your individualism.

You can also refer to fields on related models with reverse relations through OneToOneField, ForeignKey and ManyToManyField attributes:

Blog.objects.values('name', 'entry__headline')
[{'name': 'My blog', 'entry__headline': 'An entry'},
     {'name': 'My blog', 'entry__headline': 'Another entry'}, ...]


Because ManyToManyField attributes and reverse relations can have multiple related rows, including these can have a multiplier effect on the size of your result set. This will be especially pronounced if you include multiple such fields in your values() query, in which case all possible combinations will be returned.



This is similar to values() except that instead of returning dictionaries, it returns tuples when iterated over. Each tuple contains the value from the respective field passed into the values_list() call — so the first item is the first field, etc. For example:

>>> Entry.objects.values_list('id', 'headline')
[(1, u'First entry'), ...]

If you only pass in a single field, you can also pass in the flat parameter. If True, this will mean the returned results are single values, rather than one-tuples. An example should make the difference clearer:

>>> Entry.objects.values_list('id').order_by('id')
[(1,), (2,), (3,), ...]

>>> Entry.objects.values_list('id', flat=True).order_by('id')
[1, 2, 3, ...]

It is an error to pass in flat when there is more than one field.

If you don’t pass any values to values_list(), it will return all the fields in the model, in the order they were declared.


dates(field, kind, order='ASC')

Returns a DateQuerySet — a QuerySet that evaluates to a list of datetime.datetime objects representing all available dates of a particular kind within the contents of the QuerySet.

field should be the name of a DateField or DateTimeField of your model.

kind should be either "year", "month" or "day". Each datetime.datetime object in the result list is “truncated” to the given type.

  • "year" returns a list of all distinct year values for the field.
  • "month" returns a list of all distinct year/month values for the field.
  • "day" returns a list of all distinct year/month/day values for the field.

order, which defaults to 'ASC', should be either 'ASC' or 'DESC'. This specifies how to order the results.


>>> Entry.objects.dates('pub_date', 'year')
[datetime.datetime(2005, 1, 1)]
>>> Entry.objects.dates('pub_date', 'month')
[datetime.datetime(2005, 2, 1), datetime.datetime(2005, 3, 1)]
>>> Entry.objects.dates('pub_date', 'day')
[datetime.datetime(2005, 2, 20), datetime.datetime(2005, 3, 20)]
>>> Entry.objects.dates('pub_date', 'day', order='DESC')
[datetime.datetime(2005, 3, 20), datetime.datetime(2005, 2, 20)]
>>> Entry.objects.filter(headline__contains='Lennon').dates('pub_date', 'day')
[datetime.datetime(2005, 3, 20)]


When time zone support is enabled, Django uses UTC in the database connection, which means the aggregation is performed in UTC. This is a known limitation of the current implementation.



Returns an EmptyQuerySet — a QuerySet subclass that always evaluates to an empty list. This can be used in cases where you know that you should return an empty result set and your caller is expecting a QuerySet object (instead of returning an empty list, for example.)


>>> Entry.objects.none()



Returns a copy of the current QuerySet (or QuerySet subclass). This can be useful in situations where you might want to pass in either a model manager or a QuerySet and do further filtering on the result. After calling all() on either object, you’ll definitely have a QuerySet to work with.


extra(select=None, where=None, params=None, tables=None, order_by=None, select_params=None)

Sometimes, the Django query syntax by itself can’t easily express a complex WHERE clause. For these edge cases, Django provides the extra() QuerySet modifier — a hook for injecting specific clauses into the SQL generated by a QuerySet.

By definition, these extra lookups may not be portable to different database engines (because you’re explicitly writing SQL code) and violate the DRY principle, so you should avoid them if possible.

Specify one or more of params, select, where or tables. None of the arguments is required, but you should use at least one of them.

  • select

    The select argument lets you put extra fields in the SELECT clause. It should be a dictionary mapping attribute names to SQL clauses to use to calculate that attribute.


    Entry.objects.extra(select={'is_recent': "pub_date > '2006-01-01'"})

    As a result, each Entry object will have an extra attribute, is_recent, a boolean representing whether the entry’s pub_date is greater than Jan. 1, 2006.

    Django inserts the given SQL snippet directly into the SELECT statement, so the resulting SQL of the above example would be something like:

    SELECT blog_entry.*, (pub_date > '2006-01-01') AS is_recent
    FROM blog_entry;

    The next example is more advanced; it does a subquery to give each resulting Blog object an entry_count attribute, an integer count of associated Entry objects:

            'entry_count': 'SELECT COUNT(*) FROM blog_entry WHERE blog_entry.blog_id = blog_blog.id'

    In this particular case, we’re exploiting the fact that the query will already contain the blog_blog table in its FROM clause.

    The resulting SQL of the above example would be:

    SELECT blog_blog.*, (SELECT COUNT(*) FROM blog_entry WHERE blog_entry.blog_id = blog_blog.id) AS entry_count
    FROM blog_blog;

    Note that the parentheses required by most database engines around subqueries are not required in Django’s select clauses. Also note that some database backends, such as some MySQL versions, don’t support subqueries.

    In some rare cases, you might wish to pass parameters to the SQL fragments in extra(select=...). For this purpose, use the select_params parameter. Since select_params is a sequence and the select attribute is a dictionary, some care is required so that the parameters are matched up correctly with the extra select pieces. In this situation, you should use a django.utils.datastructures.SortedDict for the select value, not just a normal Python dictionary.

    This will work, for example:

        select=SortedDict([('a', '%s'), ('b', '%s')]),
        select_params=('one', 'two'))

    The only thing to be careful about when using select parameters in extra() is to avoid using the substring "%%s" (that’s two percent characters before the s) in the select strings. Django’s tracking of parameters looks for %s and an escaped % character like this isn’t detected. That will lead to incorrect results.

  • where / tables

    You can define explicit SQL WHERE clauses — perhaps to perform non-explicit joins — by using where. You can manually add tables to the SQL FROM clause by using tables.

    where and tables both take a list of strings. All where parameters are “AND”ed to any other search criteria.


    Entry.objects.extra(where=["foo='a' OR bar = 'a'", "baz = 'a'"])

    ...translates (roughly) into the following SQL:

    SELECT * FROM blog_entry WHERE (foo='a' OR bar='a') AND (baz='a')

    Be careful when using the tables parameter if you’re specifying tables that are already used in the query. When you add extra tables via the tables parameter, Django assumes you want that table included an extra time, if it is already included. That creates a problem, since the table name will then be given an alias. If a table appears multiple times in an SQL statement, the second and subsequent occurrences must use aliases so the database can tell them apart. If you’re referring to the extra table you added in the extra where parameter this is going to cause errors.

    Normally you’ll only be adding extra tables that don’t already appear in the query. However, if the case outlined above does occur, there are a few solutions. First, see if you can get by without including the extra table and use the one already in the query. If that isn’t possible, put your extra() call at the front of the queryset construction so that your table is the first use of that table. Finally, if all else fails, look at the query produced and rewrite your where addition to use the alias given to your extra table. The alias will be the same each time you construct the queryset in the same way, so you can rely upon the alias name to not change.

  • order_by

    If you need to order the resulting queryset using some of the new fields or tables you have included via extra() use the order_by parameter to extra() and pass in a sequence of strings. These strings should either be model fields (as in the normal order_by() method on querysets), of the form table_name.column_name or an alias for a column that you specified in the select parameter to extra().

    For example:

    q = Entry.objects.extra(select={'is_recent': "pub_date > '2006-01-01'"})
    q = q.extra(order_by = ['-is_recent'])

    This would sort all the items for which is_recent is true to the front of the result set (True sorts before False in a descending ordering).

    This shows, by the way, that you can make multiple calls to extra() and it will behave as you expect (adding new constraints each time).

  • params

    The where parameter described above may use standard Python database string placeholders — '%s' to indicate parameters the database engine should automatically quote. The params argument is a list of any extra parameters to be substituted.


    Entry.objects.extra(where=['headline=%s'], params=['Lennon'])

    Always use params instead of embedding values directly into where because params will ensure values are quoted correctly according to your particular backend. For example, quotes will be escaped correctly.




    Entry.objects.extra(where=['headline=%s'], params=['Lennon'])



In some complex data-modeling situations, your models might contain a lot of fields, some of which could contain a lot of data (for example, text fields), or require expensive processing to convert them to Python objects. If you are using the results of a queryset in some situation where you don’t know if you need those particular fields when you initially fetch the data, you can tell Django not to retrieve them from the database.

This is done by passing the names of the fields to not load to defer():

Entry.objects.defer("headline", "body")

A queryset that has deferred fields will still return model instances. Each deferred field will be retrieved from the database if you access that field (one at a time, not all the deferred fields at once).

You can make multiple calls to defer(). Each call adds new fields to the deferred set:

# Defers both the body and headline fields.

The order in which fields are added to the deferred set does not matter. Calling defer() with a field name that has already been deferred is harmless (the field will still be deferred).

You can defer loading of fields in related models (if the related models are loading via select_related()) by using the standard double-underscore notation to separate related fields:

Blog.objects.select_related().defer("entry__headline", "entry__body")

If you want to clear the set of deferred fields, pass None as a parameter to defer():

# Load all fields immediately.

Some fields in a model won’t be deferred, even if you ask for them. You can never defer the loading of the primary key. If you are using select_related() to retrieve related models, you shouldn’t defer the loading of the field that connects from the primary model to the related one, doing so will result in an error.


The defer() method (and its cousin, only(), below) are only for advanced use-cases. They provide an optimization for when you have analyzed your queries closely and understand exactly what information you need and have measured that the difference between returning the fields you need and the full set of fields for the model will be significant.

Even if you think you are in the advanced use-case situation, only use defer() when you cannot, at queryset load time, determine if you will need the extra fields or not. If you are frequently loading and using a particular subset of your data, the best choice you can make is to normalize your models and put the non-loaded data into a separate model (and database table). If the columns must stay in the one table for some reason, create a model with Meta.managed = False (see the managed attribute documentation) containing just the fields you normally need to load and use that where you might otherwise call defer(). This makes your code more explicit to the reader, is slightly faster and consumes a little less memory in the Python process.


When calling save() for instances with deferred fields, only the loaded fields will be saved. See save() for more details.



The only() method is more or less the opposite of defer(). You call it with the fields that should not be deferred when retrieving a model. If you have a model where almost all the fields need to be deferred, using only() to specify the complementary set of fields can result in simpler code.

Suppose you have a model with fields name, age and biography. The following two querysets are the same, in terms of deferred fields:

Person.objects.defer("age", "biography")

Whenever you call only() it replaces the set of fields to load immediately. The method’s name is mnemonic: only those fields are loaded immediately; the remainder are deferred. Thus, successive calls to only() result in only the final fields being considered:

# This will defer all fields except the headline.
Entry.objects.only("body", "rating").only("headline")

Since defer() acts incrementally (adding fields to the deferred list), you can combine calls to only() and defer() and things will behave logically:

# Final result is that everything except "headline" is deferred.
Entry.objects.only("headline", "body").defer("body")

# Final result loads headline and body immediately (only() replaces any
# existing set of fields).
Entry.objects.defer("body").only("headline", "body")

All of the cautions in the note for the defer() documentation apply to only() as well. Use it cautiously and only after exhausting your other options. Also note that using only() and omitting a field requested using select_related() is an error as well.


When calling save() for instances with deferred fields, only the loaded fields will be saved. See save() for more details.



This method is for controlling which database the QuerySet will be evaluated against if you are using more than one database. The only argument this method takes is the alias of a database, as defined in DATABASES.

For example:

# queries the database with the 'default' alias.
>>> Entry.objects.all()

# queries the database with the 'backup' alias
>>> Entry.objects.using('backup')



Returns a queryset that will lock rows until the end of the transaction, generating a SELECT ... FOR UPDATE SQL statement on supported databases.

For example:

entries = Entry.objects.select_for_update().filter(author=request.user)

All matched entries will be locked until the end of the transaction block, meaning that other transactions will be prevented from changing or acquiring locks on them.

Usually, if another transaction has already acquired a lock on one of the selected rows, the query will block until the lock is released. If this is not the behavior you want, call select_for_update(nowait=True). This will make the call non-blocking. If a conflicting lock is already acquired by another transaction, DatabaseError will be raised when the queryset is evaluated.

Note that using select_for_update() will cause the current transaction to be considered dirty, if under transaction management. This is to ensure that Django issues a COMMIT or ROLLBACK, releasing any locks held by the SELECT FOR UPDATE.

Currently, the postgresql_psycopg2, oracle, and mysql database backends support select_for_update(). However, MySQL has no support for the nowait argument. Obviously, users of external third-party backends should check with their backend’s documentation for specifics in those cases.

Passing nowait=True to select_for_update using database backends that do not support nowait, such as MySQL, will cause a DatabaseError to be raised. This is in order to prevent code unexpectedly blocking.

Using select_for_update on backends which do not support SELECT ... FOR UPDATE (such as SQLite) will have no effect.

Methods that do not return QuerySets

The following QuerySet methods evaluate the QuerySet and return something other than a QuerySet.

These methods do not use a cache (see Caching and QuerySets). Rather, they query the database each time they’re called.



Returns the object matching the given lookup parameters, which should be in the format described in Field lookups.

get() raises MultipleObjectsReturned if more than one object was found. The MultipleObjectsReturned exception is an attribute of the model class.

get() raises a DoesNotExist exception if an object wasn’t found for the given parameters. This exception is also an attribute of the model class. Example:

Entry.objects.get(id='foo') # raises Entry.DoesNotExist

The DoesNotExist exception inherits from django.core.exceptions.ObjectDoesNotExist, so you can target multiple DoesNotExist exceptions. Example:

from django.core.exceptions import ObjectDoesNotExist
    e = Entry.objects.get(id=3)
    b = Blog.objects.get(id=1)
except ObjectDoesNotExist:
    print("Either the entry or blog doesn't exist.")



A convenience method for creating an object and saving it all in one step. Thus:

p = Person.objects.create(first_name="Bruce", last_name="Springsteen")


p = Person(first_name="Bruce", last_name="Springsteen")

are equivalent.

The force_insert parameter is documented elsewhere, but all it means is that a new object will always be created. Normally you won’t need to worry about this. However, if your model contains a manual primary key value that you set and if that value already exists in the database, a call to create() will fail with an IntegrityError since primary keys must be unique. Be prepared to handle the exception if you are using manual primary keys.



A convenience method for looking up an object with the given kwargs, creating one if necessary.

Returns a tuple of (object, created), where object is the retrieved or created object and created is a boolean specifying whether a new object was created.

This is meant as a shortcut to boilerplatish code and is mostly useful for data-import scripts. For example:

    obj = Person.objects.get(first_name='John', last_name='Lennon')
except Person.DoesNotExist:
    obj = Person(first_name='John', last_name='Lennon', birthday=date(1940, 10, 9))

This pattern gets quite unwieldy as the number of fields in a model goes up. The above example can be rewritten using get_or_create() like so:

obj, created = Person.objects.get_or_create(first_name='John', last_name='Lennon',
                  defaults={'birthday': date(1940, 10, 9)})

Any keyword arguments passed to get_or_create()except an optional one called defaults — will be used in a get() call. If an object is found, get_or_create() returns a tuple of that object and False. If multiple objects are found, get_or_create raises MultipleObjectsReturned. If an object is not found, get_or_create() will instantiate and save a new object, returning a tuple of the new object and True. The new object will be created roughly according to this algorithm:

defaults = kwargs.pop('defaults', {})
params = dict([(k, v) for k, v in kwargs.items() if '__' not in k])
obj = self.model(**params)

In English, that means start with any non-'defaults' keyword argument that doesn’t contain a double underscore (which would indicate a non-exact lookup). Then add the contents of defaults, overriding any keys if necessary, and use the result as the keyword arguments to the model class. As hinted at above, this is a simplification of the algorithm that is used, but it contains all the pertinent details. The internal implementation has some more error-checking than this and handles some extra edge-conditions; if you’re interested, read the code.

If you have a field named defaults and want to use it as an exact lookup in get_or_create(), just use 'defaults__exact', like so:

Foo.objects.get_or_create(defaults__exact='bar', defaults={'defaults': 'baz'})

The get_or_create() method has similar error behavior to create() when you’re using manually specified primary keys. If an object needs to be created and the key already exists in the database, an IntegrityError will be raised.

Finally, a word on using get_or_create() in Django views. As mentioned earlier, get_or_create() is mostly useful in scripts that need to parse data and create new records if existing ones aren’t available. But if you need to use get_or_create() in a view, please make sure to use it only in POST requests unless you have a good reason not to. GET requests shouldn’t have any effect on data; use POST whenever a request to a page has a side effect on your data. For more, see Safe methods in the HTTP spec.


bulk_create(objs, batch_size=None)

This method inserts the provided list of objects into the database in an efficient manner (generally only 1 query, no matter how many objects there are):

>>> Entry.objects.bulk_create([
...     Entry(headline="Django 1.0 Released"),
...     Entry(headline="Django 1.1 Announced"),
...     Entry(headline="Breaking: Django is awesome")
... ])

This has a number of caveats though:

  • The model’s save() method will not be called, and the pre_save and post_save signals will not be sent.
  • It does not work with child models in a multi-table inheritance scenario.
  • If the model’s primary key is an AutoField it does not retrieve and set the primary key attribute, as save() does.

The batch_size parameter controls how many objects are created in single query. The default is to create all objects in one batch, except for SQLite where the default is such that at maximum 999 variables per query is used.

The batch_size parameter was added in version 1.5.



Returns an integer representing the number of objects in the database matching the QuerySet. The count() method never raises exceptions.


# Returns the total number of entries in the database.

# Returns the number of entries whose headline contains 'Lennon'

A count() call performs a SELECT COUNT(*) behind the scenes, so you should always use count() rather than loading all of the record into Python objects and calling len() on the result (unless you need to load the objects into memory anyway, in which case len() will be faster).

Depending on which database you’re using (e.g. PostgreSQL vs. MySQL), count() may return a long integer instead of a normal Python integer. This is an underlying implementation quirk that shouldn’t pose any real-world problems.



Takes a list of primary-key values and returns a dictionary mapping each primary-key value to an instance of the object with the given ID.


>>> Blog.objects.in_bulk([1])
{1: <Blog: Beatles Blog>}
>>> Blog.objects.in_bulk([1, 2])
{1: <Blog: Beatles Blog>, 2: <Blog: Cheddar Talk>}
>>> Blog.objects.in_bulk([])

If you pass in_bulk() an empty list, you’ll get an empty dictionary.



Evaluates the QuerySet (by performing the query) and returns an iterator (see PEP 234) over the results. A QuerySet typically caches its results internally so that repeated evaluations do not result in additional queries. In contrast, iterator() will read results directly, without doing any caching at the QuerySet level (internally, the default iterator calls iterator() and caches the return value). For a QuerySet which returns a large number of objects that you only need to access once, this can results in better performance and a significant reduction in memory.

Note that using iterator() on a QuerySet which has already been evaluated will force it to evaluate again, repeating the query.

Also, use of iterator() causes previous prefetch_related() calls to be ignored since these two optimizations do not make sense together.


Some Python database drivers like psycopg2 perform caching if using client side cursors (instantiated with connection.cursor() and what Django’s ORM uses). Using iterator() does not affect caching at the database driver level. To disable this caching, look at server side cursors.



Returns the latest object in the table, by date, using the field_name provided as the date field.

This example returns the latest Entry in the table, according to the pub_date field:


If your model’s Meta specifies get_latest_by, you can leave off the field_name argument to latest(). Django will use the field specified in get_latest_by by default.

Like get(), latest() raises DoesNotExist if there is no object with the given parameters.

Note latest() exists purely for convenience and readability.


aggregate(*args, **kwargs)

Returns a dictionary of aggregate values (averages, sums, etc) calculated over the QuerySet. Each argument to aggregate() specifies a value that will be included in the dictionary that is returned.

The aggregation functions that are provided by Django are described in Aggregation Functions below.

Aggregates specified using keyword arguments will use the keyword as the name for the annotation. Anonymous arguments will have a name generated for them based upon the name of the aggregate function and the model field that is being aggregated.

For example, when you are working with blog entries, you may want to know the number of authors that have contributed blog entries:

>>> q = Blog.objects.aggregate(Count('entry'))
{'entry__count': 16}

By using a keyword argument to specify the aggregate function, you can control the name of the aggregation value that is returned:

>>> q = Blog.objects.aggregate(number_of_entries=Count('entry'))
{'number_of_entries': 16}

For an in-depth discussion of aggregation, see the topic guide on Aggregation.



Returns True if the QuerySet contains any results, and False if not. This tries to perform the query in the simplest and fastest way possible, but it does execute nearly the same query as a normal QuerySet query.

exists() is useful for searches relating to both object membership in a QuerySet and to the existence of any objects in a QuerySet, particularly in the context of a large QuerySet.

The most efficient method of finding whether a model with a unique field (e.g. primary_key) is a member of a QuerySet is:

entry = Entry.objects.get(pk=123)
if some_query_set.filter(pk=entry.pk).exists():
    print("Entry contained in queryset")

Which will be faster than the following which requires evaluating and iterating through the entire queryset:

if entry in some_query_set:
   print("Entry contained in QuerySet")

And to find whether a queryset contains any items:

if some_query_set.exists():
    print("There is at least one object in some_query_set")

Which will be faster than:

if some_query_set:
    print("There is at least one object in some_query_set")

... but not by a large degree (hence needing a large queryset for efficiency gains).

Additionally, if a some_query_set has not yet been evaluated, but you know that it will be at some point, then using some_query_set.exists() will do more overall work (one query for the existence check plus an extra one to later retrieve the results) than simply using bool(some_query_set), which retrieves the results and then checks if any were returned.



Performs an SQL update query for the specified fields, and returns the number of rows matched (which may not be equal to the number of rows updated if some rows already have the new value).

For example, to turn comments off for all blog entries published in 2010, you could do this:

>>> Entry.objects.filter(pub_date__year=2010).update(comments_on=False)

(This assumes your Entry model has fields pub_date and comments_on.)

You can update multiple fields — there’s no limit on how many. For example, here we update the comments_on and headline fields:

>>> Entry.objects.filter(pub_date__year=2010).update(comments_on=False, headline='This is old')

The update() method is applied instantly, and the only restriction on the QuerySet that is updated is that it can only update columns in the model’s main table, not on related models. You can’t do this, for example:

>>> Entry.objects.update(blog__name='foo') # Won't work!

Filtering based on related fields is still possible, though:

>>> Entry.objects.filter(blog__id=1).update(comments_on=True)

You cannot call update() on a QuerySet that has had a slice taken or can otherwise no longer be filtered.

The update() method returns the number of affected rows:

>>> Entry.objects.filter(id=64).update(comments_on=True)

>>> Entry.objects.filter(slug='nonexistent-slug').update(comments_on=True)

>>> Entry.objects.filter(pub_date__year=2010).update(comments_on=False)

If you’re just updating a record and don’t need to do anything with the model object, the most efficient approach is to call update(), rather than loading the model object into memory. For example, instead of doing this:

e = Entry.objects.get(id=10)
e.comments_on = False

...do this:


Using update() also prevents a race condition wherein something might change in your database in the short period of time between loading the object and calling save().

Finally, realize that update() does an update at the SQL level and, thus, does not call any save() methods on your models, nor does it emit the pre_save or post_save signals (which are a consequence of calling Model.save()). If you want to update a bunch of records for a model that has a custom save() method, loop over them and call save(), like this:

for e in Entry.objects.filter(pub_date__year=2010):
    e.comments_on = False



Performs an SQL delete query on all rows in the QuerySet. The delete() is applied instantly. You cannot call delete() on a QuerySet that has had a slice taken or can otherwise no longer be filtered.

For example, to delete all the entries in a particular blog:

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

# Delete all the entries belonging to this Blog.
>>> Entry.objects.filter(blog=b).delete()

By default, Django’s ForeignKey emulates the SQL constraint ON DELETE CASCADE — in other words, any objects with foreign keys pointing at the objects to be deleted will be deleted along with them. For example:

blogs = Blog.objects.all()
# This will delete all Blogs and all of their Entry objects.

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

The delete() method does a bulk delete and does not call any delete() methods on your models. It does, however, emit the pre_delete and post_delete signals for all deleted objects (including cascaded deletions).

Allow fast-path deletion of objects

Django needs to fetch objects into memory to send signals and handle cascades. However, if there are no cascades and no signals, then Django may take a fast-path and delete objects without fetching into memory. For large deletes this can result in significantly reduced memory usage. The amount of executed queries can be reduced, too.

ForeignKeys which are set to on_delete DO_NOTHING do not prevent taking the fast-path in deletion.

Note that the queries generated in object deletion is an implementation detail subject to change.

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().

For an introduction, see models and database queries documentation.


Exact match. If the value provided for comparison is None, it will be interpreted as an SQL NULL (see isnull for more details).



SQL equivalents:

SELECT ... WHERE id = 14;

MySQL comparisons

In MySQL, a database table’s “collation” setting determines whether exact comparisons are case-sensitive. This is a database setting, not a Django setting. It’s possible to configure your MySQL tables to use case-sensitive comparisons, but some trade-offs are involved. For more information about this, see the collation section in the databases documentation.


Case-insensitive exact match.


Blog.objects.get(name__iexact='beatles blog')

SQL equivalent:

SELECT ... WHERE name ILIKE 'beatles blog';

Note this will match 'Beatles Blog', 'beatles blog', 'BeAtLes BLoG', etc.

SQLite users

When using the SQLite backend and Unicode (non-ASCII) strings, bear in mind the database note about string comparisons. SQLite does not do case-insensitive matching for Unicode strings.


Case-sensitive containment test.



SQL equivalent:

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

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

SQLite users

SQLite doesn’t support case-sensitive LIKE statements; contains acts like icontains for SQLite. See the database note for more information.


Case-insensitive containment test.



SQL equivalent:

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

SQLite users

When using the SQLite backend and Unicode (non-ASCII) strings, bear in mind the database note about string comparisons.


In a given list.


Entry.objects.filter(id__in=[1, 3, 4])

SQL equivalent:

SELECT ... WHERE id IN (1, 3, 4);

You can also use a queryset to dynamically evaluate the list of values instead of providing a list of literal values:

inner_qs = Blog.objects.filter(name__contains='Cheddar')
entries = Entry.objects.filter(blog__in=inner_qs)

This queryset will be evaluated as subselect statement:

SELECT ... WHERE blog.id IN (SELECT id FROM ... WHERE NAME LIKE '%Cheddar%')

If you pass in a ValuesQuerySet or ValuesListQuerySet (the result of calling values() or values_list() on a queryset) as the value to an __in lookup, you need to ensure you are only extracting one field in the result. For example, this will work (filtering on the blog names):

inner_qs = Blog.objects.filter(name__contains='Ch').values('name')
entries = Entry.objects.filter(blog__name__in=inner_qs)

This example will raise an exception, since the inner query is trying to extract two field values, where only one is expected:

# Bad code! Will raise a TypeError.
inner_qs = Blog.objects.filter(name__contains='Ch').values('name', 'id')
entries = Entry.objects.filter(blog__name__in=inner_qs)

Performance considerations

Be cautious about using nested queries and understand your database server’s performance characteristics (if in doubt, benchmark!). Some database backends, most notably MySQL, don’t optimize nested queries very well. It is more efficient, in those cases, to extract a list of values and then pass that into the second query. That is, execute two queries instead of one:

values = Blog.objects.filter(
        name__contains='Cheddar').values_list('pk', flat=True)
entries = Entry.objects.filter(blog__in=list(values))

Note the list() call around the Blog QuerySet to force execution of the first query. Without it, a nested query would be executed, because QuerySets are lazy.


Greater than.



SQL equivalent:

SELECT ... WHERE id > 4;


Greater than or equal to.


Less than.


Less than or equal to.


Case-sensitive starts-with.



SQL equivalent:

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

SQLite doesn’t support case-sensitive LIKE statements; startswith acts like istartswith for SQLite.


Case-insensitive starts-with.



SQL equivalent:

SELECT ... WHERE headline ILIKE 'Will%';

SQLite users

When using the SQLite backend and Unicode (non-ASCII) strings, bear in mind the database note about string comparisons.


Case-sensitive ends-with.



SQL equivalent:

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

SQLite users

SQLite doesn’t support case-sensitive LIKE statements; endswith acts like iendswith for SQLite. Refer to the database note documentation for more.


Case-insensitive ends-with.



SQL equivalent:

SELECT ... WHERE headline ILIKE '%will'

SQLite users

When using the SQLite backend and Unicode (non-ASCII) strings, bear in mind the database note about string comparisons.


Range test (inclusive).


start_date = datetime.date(2005, 1, 1)
end_date = datetime.date(2005, 3, 31)
Entry.objects.filter(pub_date__range=(start_date, end_date))

SQL equivalent:

SELECT ... WHERE pub_date BETWEEN '2005-01-01' and '2005-03-31';

You can use range anywhere you can use BETWEEN in SQL — for dates, numbers and even characters.


Filtering a DateTimeField with dates won’t include items on the last day, because the bounds are interpreted as “0am on the given date”. If pub_date was a DateTimeField, the above expression would be turned into this SQL:

SELECT ... WHERE pub_date BETWEEN '2005-01-01 00:00:00' and '2005-03-31 00:00:00';

Generally speaking, you can’t mix dates and datetimes.


For date/datetime fields, exact year match. Takes a four-digit year.



SQL equivalent:

SELECT ... WHERE pub_date BETWEEN '2005-01-01' AND '2005-12-31';

(The exact SQL syntax varies for each database engine.)


For date and datetime fields, an exact month match. Takes an integer 1 (January) through 12 (December).



SQL equivalent:

SELECT ... WHERE EXTRACT('month' FROM pub_date) = '12';

(The exact SQL syntax varies for each database engine.)


For date and datetime fields, an exact day match.



SQL equivalent:

SELECT ... WHERE EXTRACT('day' FROM pub_date) = '3';

(The exact SQL syntax varies for each database engine.)

Note this will match any record with a pub_date on the third day of the month, such as January 3, July 3, etc.


For date and datetime fields, a ‘day of the week’ match.

Takes an integer value representing the day of week from 1 (Sunday) to 7 (Saturday).



(No equivalent SQL code fragment is included for this lookup because implementation of the relevant query varies among different database engines.)

Note this will match any record with a pub_date that falls on a Monday (day 2 of the week), regardless of the month or year in which it occurs. Week days are indexed with day 1 being Sunday and day 7 being Saturday.


When time zone support is enabled, Django uses UTC in the database connection, which means the year, month, day and week_day lookups are performed in UTC. This is a known limitation of the current implementation.


Takes either True or False, which correspond to SQL queries of IS NULL and IS NOT NULL, respectively.



SQL equivalent:

SELECT ... WHERE pub_date IS NULL;


Case-sensitive regular expression match.

The regular expression syntax is that of the database backend in use. In the case of SQLite, which has no built in regular expression support, this feature is provided by a (Python) user-defined REGEXP function, and the regular expression syntax is therefore that of Python’s re module.


Entry.objects.get(title__regex=r'^(An?|The) +')

SQL equivalents:

SELECT ... WHERE title REGEXP BINARY '^(An?|The) +'; -- MySQL

SELECT ... WHERE REGEXP_LIKE(title, '^(an?|the) +', 'c'); -- Oracle

SELECT ... WHERE title ~ '^(An?|The) +'; -- PostgreSQL

SELECT ... WHERE title REGEXP '^(An?|The) +'; -- SQLite

Using raw strings (e.g., r'foo' instead of 'foo') for passing in the regular expression syntax is recommended.


Case-insensitive regular expression match.


Entry.objects.get(title__iregex=r'^(an?|the) +')

SQL equivalents:

SELECT ... WHERE title REGEXP '^(an?|the) +'; -- MySQL

SELECT ... WHERE REGEXP_LIKE(title, '^(an?|the) +', 'i'); -- Oracle

SELECT ... WHERE title ~* '^(an?|the) +'; -- PostgreSQL

SELECT ... WHERE title REGEXP '(?i)^(an?|the) +'; -- SQLite

Aggregation functions

Django provides the following aggregation functions in the django.db.models module. For details on how to use these aggregate functions, see the topic guide on aggregation.


class Avg(field)

Returns the mean value of the given field, which must be numeric.

  • Default alias: <field>__avg
  • Return type: float


class Count(field, distinct=False)

Returns the number of objects that are related through the provided field.

  • Default alias: <field>__count
  • Return type: int

Has one optional argument:


If distinct=True, the count will only include unique instances. This is the SQL equivalent of COUNT(DISTINCT <field>). The default value is False.


class Max(field)

Returns the maximum value of the given field.

  • Default alias: <field>__max
  • Return type: same as input field


class Min(field)

Returns the minimum value of the given field.

  • Default alias: <field>__min
  • Return type: same as input field


class StdDev(field, sample=False)

Returns the standard deviation of the data in the provided field.

  • Default alias: <field>__stddev
  • Return type: float

Has one optional argument:


By default, StdDev returns the population standard deviation. However, if sample=True, the return value will be the sample standard deviation.


SQLite doesn’t provide StdDev out of the box. An implementation is available as an extension module for SQLite. Consult the SQlite documentation for instructions on obtaining and installing this extension.


class Sum(field)

Computes the sum of all values of the given field.

  • Default alias: <field>__sum
  • Return type: same as input field


class Variance(field, sample=False)

Returns the variance of the data in the provided field.

  • Default alias: <field>__variance
  • Return type: float

Has one optional argument:


By default, Variance returns the population variance. However, if sample=True, the return value will be the sample variance.


SQLite doesn’t provide Variance out of the box. An implementation is available as an extension module for SQLite. Consult the SQlite documentation for instructions on obtaining and installing this extension.