• Language: en

Model index reference

Index classes ease creating database indexes. They can be added using the Meta.indexes option. This document explains the API references of Index which includes the index options.

Referencing built-in indexes

Indexes are defined in django.db.models.indexes, but for convenience they’re imported into django.db.models. The standard convention is to use from django.db import models and refer to the indexes as models.<IndexClass>.

Index options

class Index(fields=(), name=None, db_tablespace=None, opclasses=(), condition=None)[source]

Creates an index (B-Tree) in the database.



A list or tuple of the name of the fields on which the index is desired.

By default, indexes are created with an ascending order for each column. To define an index with a descending order for a column, add a hyphen before the field’s name.

For example Index(fields=['headline', '-pub_date']) would create SQL with (headline, pub_date DESC). Index ordering isn’t supported on MySQL. In that case, a descending index is created as a normal index.

Changed in Django 2.1:

Older versions don’t accept a tuple.



The name of the index. If name isn’t provided Django will auto-generate a name. For compatibility with different databases, index names cannot be longer than 30 characters and shouldn’t start with a number (0-9) or underscore (_).



The name of the database tablespace to use for this index. For single field indexes, if db_tablespace isn’t provided, the index is created in the db_tablespace of the field.

If Field.db_tablespace isn’t specified (or if the index uses multiple fields), the index is created in tablespace specified in the db_tablespace option inside the model’s class Meta. If neither of those tablespaces are set, the index is created in the same tablespace as the table.

See also

For a list of PostgreSQL-specific indexes, see django.contrib.postgres.indexes.


New in Django Development version:

The names of the PostgreSQL operator classes to use for this index. If you require a custom operator class, you must provide one for each field in the index.

For example, GinIndex(name='json_index', fields=['jsonfield'], opclasses=['jsonb_path_ops']) creates a gin index on jsonfield using jsonb_path_ops.

opclasses are ignored for databases besides PostgreSQL.

Index.name is required when using opclasses.


New in Django Development version:

If the table is very large and your queries mostly target a subset of rows, it may be useful to restrict an index to that subset. Specify a condition as a Q. For example, condition=Q(pages__gt=400) indexes records with more than 400 pages.

Index.name is required when using condition.

Restrictions on PostgreSQL

PostgreSQL requires functions referenced in the condition to be be marked as IMMUTABLE. Django doesn’t validate this but PostgreSQL will error. This means that functions such as Date functions and Concat aren’t accepted. If you store dates in DateTimeField, comparison to datetime objects may require the tzinfo argument to be provided because otherwise the comparison could result in a mutable function due to the casting Django does for lookups.

Restrictions on SQLite

SQLite imposes restrictions on how a partial index can be constructed.


Oracle does not support partial indexes. Instead, partial indexes can be emulated using functional indexes. Use a migration to add the index using RunSQL.

MySQL and MariaDB

The condition argument is ignored with MySQL and MariaDB as neither supports conditional indexes.

Back to Top