Fixed #30913 -- Added support for covering indexes on PostgreSQL 11+.

docs/ref/models/constraints.txt

@@ -73,7 +73,7 @@ constraint.
 ``UniqueConstraint``
 ====================
 
-.. class:: UniqueConstraint(*, fields, name, condition=None, deferrable=None)
+.. class:: UniqueConstraint(*, fields, name, condition=None, deferrable=None, include=None)
 
     Creates a unique constraint in the database.
 
@@ -145,3 +145,26 @@ enforced immediately after every command.
 
     Deferred unique constraints may lead to a `performance penalty
     <https://www.postgresql.org/docs/current/sql-createtable.html#id-1.9.3.85.9.4>`_.
+
+``include``
+-----------
+
+.. attribute:: UniqueConstraint.include
+
+.. versionadded:: 3.2
+
+A list or tuple of the names of the fields to be included in the covering
+unique index as non-key columns. This allows index-only scans to be used for
+queries that select only included fields (:attr:`~UniqueConstraint.include`)
+and filter only by unique fields (:attr:`~UniqueConstraint.fields`).
+
+For example::
+
+    UniqueConstraint(name='unique_booking', fields=['room', 'date'], include=['full_name'])
+
+will allow filtering on ``room`` and ``date``, also selecting ``full_name``,
+while fetching data only from the index.
+
+``include`` is supported only on PostgreSQL.
+
+Non-key columns have the same database restrictions as :attr:`Index.include`.

docs/ref/models/indexes.txt

@@ -21,7 +21,7 @@ options`_.
 ``Index`` options
 =================
 
-.. class:: Index(fields=(), name=None, db_tablespace=None, opclasses=(), condition=None)
+.. class:: Index(fields=(), name=None, db_tablespace=None, opclasses=(), condition=None, include=None)
 
     Creates an index (B-Tree) in the database.
 
@@ -137,3 +137,40 @@ indexes records with more than 400 pages.
 
     The ``condition`` argument is ignored with MySQL and MariaDB as neither
     supports conditional indexes.
+
+``include``
+-----------
+
+.. attribute:: Index.include
+
+.. versionadded:: 3.2
+
+A list or tuple of the names of the fields to be included in the covering index
+as non-key columns. This allows index-only scans to be used for queries that
+select only included fields (:attr:`~Index.include`) and filter only by indexed
+fields (:attr:`~Index.fields`).
+
+For example::
+
+    Index(name='covering_index', fields=['headline'], include=['pub_date'])
+
+will allow filtering on ``headline``, also selecting ``pub_date``, while
+fetching data only from the index.
+
+Using ``include`` will produce a smaller index than using a multiple column
+index but with the drawback that non-key columns can not be used for sorting or
+filtering.
+
+``include`` is ignored for databases besides PostgreSQL.
+
+:attr:`Index.name` is required when using ``include``.
+
+See the PostgreSQL documentation for more details about `covering indexes`_.
+
+.. admonition:: Restrictions on PostgreSQL
+
+    PostgreSQL 11+ only supports covering B-Tree indexes, and PostgreSQL 12+
+    also supports covering :class:`GiST indexes
+    <django.contrib.postgres.indexes.GistIndex>`.
+
+.. _covering indexes: https://www.postgresql.org/docs/current/indexes-index-only-scans.html

docs/releases/3.2.txt

@@ -185,6 +185,10 @@ Models
 * :class:`When() <django.db.models.expressions.When>` expression now allows
   using the ``condition`` argument with ``lookups``.
 
+* The new :attr:`.Index.include` and :attr:`.UniqueConstraint.include`
+  attributes allow creating covering indexes and covering unique constraints on
+  PostgreSQL 11+.
+
 Requests and Responses
 ~~~~~~~~~~~~~~~~~~~~~~
 
@@ -263,6 +267,10 @@ backends.
   * ``introspected_small_auto_field_type``
   * ``introspected_boolean_field_type``
 
+* To enable support for covering indexes (:attr:`.Index.include`) and covering
+  unique constraints (:attr:`.UniqueConstraint.include`), set
+  ``DatabaseFeatures.supports_covering_indexes`` to ``True``.
+
 :mod:`django.contrib.gis`
 -------------------------