Refs #29444 -- Allowed returning multiple fields from INSERT statements on PostgreSQL.

Thanks Florian Apolloner, Tim Graham, Simon Charette, Nick Pope, and
Mariusz Felisiak for reviews.

django/db/backends/base/features.py

@@ -23,6 +23,7 @@ class BaseDatabaseFeatures:
 
     can_use_chunked_reads = True
     can_return_columns_from_insert = False
+    can_return_multiple_columns_from_insert = False
     can_return_rows_from_bulk_insert = False
     has_bulk_insert = True
     uses_savepoints = True

django/db/backends/base/operations.py

@@ -176,13 +176,12 @@ class BaseDatabaseOperations:
         else:
             return ['DISTINCT'], []
 
-    def fetch_returned_insert_id(self, cursor):
+    def fetch_returned_insert_columns(self, cursor):
         """
         Given a cursor object that has just performed an INSERT...RETURNING
-        statement into a table that has an auto-incrementing ID, return the
-        newly created ID.
+        statement into a table, return the newly created data.
         """
-        return cursor.fetchone()[0]
+        return cursor.fetchone()
 
     def field_cast_sql(self, db_type, internal_type):
         """
@@ -314,12 +313,11 @@ class BaseDatabaseOperations:
         """
         return value
 
-    def return_insert_id(self, field):
+    def return_insert_columns(self, fields):
         """
-        For backends that support returning the last insert ID as part of an
-        insert query, return the SQL and params to append to the INSERT query.
-        The returned fragment should contain a format string to hold the
-        appropriate column.
+        For backends that support returning columns as part of an insert query,
+        return the SQL and params to append to the INSERT query. The returned
+        fragment should contain a format string to hold the appropriate column.
         """
         pass
 

django/db/backends/oracle/operations.py

@@ -248,7 +248,7 @@ END;
     def deferrable_sql(self):
         return " DEFERRABLE INITIALLY DEFERRED"
 
-    def fetch_returned_insert_id(self, cursor):
+    def fetch_returned_insert_columns(self, cursor):
         value = cursor._insert_id_var.getvalue()
         if value is None or value == []:
             # cx_Oracle < 6.3 returns None, >= 6.3 returns empty list.
@@ -258,7 +258,7 @@ END;
                 'Oracle OCI library (see https://code.djangoproject.com/ticket/28859).'
             )
         # cx_Oracle < 7 returns value, >= 7 returns list with single value.
-        return value[0] if isinstance(value, list) else value
+        return value if isinstance(value, list) else [value]
 
     def field_cast_sql(self, db_type, internal_type):
         if db_type and db_type.endswith('LOB'):
@@ -341,8 +341,14 @@ END;
             match_option = "'i'"
         return 'REGEXP_LIKE(%%s, %%s, %s)' % match_option
 
-    def return_insert_id(self, field):
-        return 'RETURNING %s INTO %%s', (InsertVar(field),)
+    def return_insert_columns(self, fields):
+        if not fields:
+            return '', ()
+        sql = 'RETURNING %s.%s INTO %%s' % (
+            self.quote_name(fields[0].model._meta.db_table),
+            self.quote_name(fields[0].column),
+        )
+        return sql, (InsertVar(fields[0]),)
 
     def __foreign_key_constraints(self, table_name, recursive):
         with self.connection.cursor() as cursor:

django/db/backends/postgresql/features.py

@@ -8,6 +8,7 @@ from django.utils.functional import cached_property
 class DatabaseFeatures(BaseDatabaseFeatures):
     allows_group_by_selected_pks = True
     can_return_columns_from_insert = True
+    can_return_multiple_columns_from_insert = True
     can_return_rows_from_bulk_insert = True
     has_real_datatype = True
     has_native_uuid_field = True

django/db/backends/postgresql/operations.py

@@ -76,13 +76,12 @@ class DatabaseOperations(BaseDatabaseOperations):
     def deferrable_sql(self):
         return " DEFERRABLE INITIALLY DEFERRED"
 
-    def fetch_returned_insert_ids(self, cursor):
+    def fetch_returned_insert_rows(self, cursor):
         """
         Given a cursor object that has just performed an INSERT...RETURNING
-        statement into a table that has an auto-incrementing ID, return the
-        list of newly created IDs.
+        statement into a table, return the tuple of returned data.
         """
-        return [item[0] for item in cursor.fetchall()]
+        return cursor.fetchall()
 
     def lookup_cast(self, lookup_type, internal_type=None):
         lookup = '%s'
@@ -236,8 +235,16 @@ class DatabaseOperations(BaseDatabaseOperations):
             return cursor.query.decode()
         return None
 
-    def return_insert_id(self, field):
-        return "RETURNING %s", ()
+    def return_insert_columns(self, fields):
+        if not fields:
+            return '', ()
+        columns = [
+            '%s.%s' % (
+                self.quote_name(field.model._meta.db_table),
+                self.quote_name(field.column),
+            ) for field in fields
+        ]
+        return 'RETURNING %s' % ', '.join(columns), ()
 
     def bulk_insert_sql(self, fields, placeholder_rows):
         placeholder_rows_sql = (", ".join(row) for row in placeholder_rows)

django/db/models/base.py

@@ -876,10 +876,10 @@ class Model(metaclass=ModelBase):
             if not pk_set:
                 fields = [f for f in fields if f is not meta.auto_field]
 
-            update_pk = meta.auto_field and not pk_set
-            result = self._do_insert(cls._base_manager, using, fields, update_pk, raw)
-            if update_pk:
-                setattr(self, meta.pk.attname, result)
+            returning_fields = meta.db_returning_fields
+            results = self._do_insert(cls._base_manager, using, fields, returning_fields, raw)
+            for result, field in zip(results, returning_fields):
+                setattr(self, field.attname, result)
         return updated
 
     def _do_update(self, base_qs, using, pk_val, values, update_fields, forced_update):
@@ -909,13 +909,15 @@ class Model(metaclass=ModelBase):
             )
         return filtered._update(values) > 0
 
-    def _do_insert(self, manager, using, fields, update_pk, raw):
+    def _do_insert(self, manager, using, fields, returning_fields, raw):
         """
-        Do an INSERT. If update_pk is defined then this method should return
-        the new pk for the model.
+        Do an INSERT. If returning_fields is defined then this method should
+        return the newly created data for the model.
         """
-        return manager._insert([self], fields=fields, return_id=update_pk,
-                               using=using, raw=raw)
+        return manager._insert(
+            [self], fields=fields, returning_fields=returning_fields,
+            using=using, raw=raw,
+        )
 
     def delete(self, using=None, keep_parents=False):
         using = using or router.db_for_write(self.__class__, instance=self)

django/db/models/fields/__init__.py

@@ -735,6 +735,14 @@ class Field(RegisterLookupMixin):
     def db_tablespace(self):
         return self._db_tablespace or settings.DEFAULT_INDEX_TABLESPACE
 
+    @property
+    def db_returning(self):
+        """
+        Private API intended only to be used by Django itself. Currently only
+        the PostgreSQL backend supports returning multiple fields on a model.
+        """
+        return False
+
     def set_attributes_from_name(self, name):
         self.name = self.name or name
         self.attname, self.column = self.get_attname_column()
@@ -2311,6 +2319,7 @@ class UUIDField(Field):
 
 
 class AutoFieldMixin:
+    db_returning = True
 
     def __init__(self, *args, **kwargs):
         kwargs['blank'] = True

django/db/models/options.py

@@ -842,3 +842,14 @@ class Options:
             if isinstance(attr, property):
                 names.append(name)
         return frozenset(names)
+
+    @cached_property
+    def db_returning_fields(self):
+        """
+        Private API intended only to be used by Django itself.
+        Fields to be returned after a database insert.
+        """
+        return [
+            field for field in self._get_fields(forward=True, reverse=False, include_parents=PROXY_PARENTS)
+            if getattr(field, 'db_returning', False)
+        ]

django/db/models/query.py

@@ -470,23 +470,33 @@ class QuerySet:
             return objs
         self._for_write = True
         connection = connections[self.db]
-        fields = self.model._meta.concrete_fields
+        opts = self.model._meta
+        fields = opts.concrete_fields
         objs = list(objs)
         self._populate_pk_values(objs)
         with transaction.atomic(using=self.db, savepoint=False):
             objs_with_pk, objs_without_pk = partition(lambda o: o.pk is None, objs)
             if objs_with_pk:
-                self._batched_insert(objs_with_pk, fields, batch_size, ignore_conflicts=ignore_conflicts)
+                returned_columns = self._batched_insert(
+                    objs_with_pk, fields, batch_size, ignore_conflicts=ignore_conflicts,
+                )
+                for obj_with_pk, results in zip(objs_with_pk, returned_columns):
+                    for result, field in zip(results, opts.db_returning_fields):
+                        if field != opts.pk:
+                            setattr(obj_with_pk, field.attname, result)
                 for obj_with_pk in objs_with_pk:
                     obj_with_pk._state.adding = False
                     obj_with_pk._state.db = self.db
             if objs_without_pk:
                 fields = [f for f in fields if not isinstance(f, AutoField)]
-                ids = self._batched_insert(objs_without_pk, fields, batch_size, ignore_conflicts=ignore_conflicts)
+                returned_columns = self._batched_insert(
+                    objs_without_pk, fields, batch_size, ignore_conflicts=ignore_conflicts,
+                )
                 if connection.features.can_return_rows_from_bulk_insert and not ignore_conflicts:
-                    assert len(ids) == len(objs_without_pk)
-                for obj_without_pk, pk in zip(objs_without_pk, ids):
-                    obj_without_pk.pk = pk
+                    assert len(returned_columns) == len(objs_without_pk)
+                for obj_without_pk, results in zip(objs_without_pk, returned_columns):
+                    for result, field in zip(results, opts.db_returning_fields):
+                        setattr(obj_without_pk, field.attname, result)
                     obj_without_pk._state.adding = False
                     obj_without_pk._state.db = self.db
 
@@ -1181,7 +1191,7 @@ class QuerySet:
     # PRIVATE METHODS #
     ###################
 
-    def _insert(self, objs, fields, return_id=False, raw=False, using=None, ignore_conflicts=False):
+    def _insert(self, objs, fields, returning_fields=None, raw=False, using=None, ignore_conflicts=False):
         """
         Insert a new record for the given model. This provides an interface to
         the InsertQuery class and is how Model.save() is implemented.
@@ -1191,7 +1201,7 @@ class QuerySet:
             using = self.db
         query = sql.InsertQuery(self.model, ignore_conflicts=ignore_conflicts)
         query.insert_values(fields, objs, raw=raw)
-        return query.get_compiler(using=using).execute_sql(return_id)
+        return query.get_compiler(using=using).execute_sql(returning_fields)
     _insert.alters_data = True
     _insert.queryset_only = False
 
@@ -1203,21 +1213,22 @@ class QuerySet:
             raise NotSupportedError('This database backend does not support ignoring conflicts.')
         ops = connections[self.db].ops
         batch_size = (batch_size or max(ops.bulk_batch_size(fields, objs), 1))
-        inserted_ids = []
+        inserted_rows = []
         bulk_return = connections[self.db].features.can_return_rows_from_bulk_insert
         for item in [objs[i:i + batch_size] for i in range(0, len(objs), batch_size)]:
             if bulk_return and not ignore_conflicts:
-                inserted_id = self._insert(
-                    item, fields=fields, using=self.db, return_id=True,
+                inserted_columns = self._insert(
+                    item, fields=fields, using=self.db,
+                    returning_fields=self.model._meta.db_returning_fields,
                     ignore_conflicts=ignore_conflicts,
                 )
-                if isinstance(inserted_id, list):
-                    inserted_ids.extend(inserted_id)
+                if isinstance(inserted_columns, list):
+                    inserted_rows.extend(inserted_columns)
                 else:
-                    inserted_ids.append(inserted_id)
+                    inserted_rows.append(inserted_columns)
             else:
                 self._insert(item, fields=fields, using=self.db, ignore_conflicts=ignore_conflicts)
-        return inserted_ids
+        return inserted_rows
 
     def _chain(self, **kwargs):
         """

django/db/models/sql/compiler.py

@@ -1159,7 +1159,7 @@ class SQLCompiler:
 
 
 class SQLInsertCompiler(SQLCompiler):
-    return_id = False
+    returning_fields = None
 
     def field_as_sql(self, field, val):
         """
@@ -1290,14 +1290,14 @@ class SQLInsertCompiler(SQLCompiler):
         # queries and generate their own placeholders. Doing that isn't
         # necessary and it should be possible to use placeholders and
         # expressions in bulk inserts too.
-        can_bulk = (not self.return_id and self.connection.features.has_bulk_insert)
+        can_bulk = (not self.returning_fields and self.connection.features.has_bulk_insert)
 
         placeholder_rows, param_rows = self.assemble_as_sql(fields, value_rows)
 
         ignore_conflicts_suffix_sql = self.connection.ops.ignore_conflicts_suffix_sql(
             ignore_conflicts=self.query.ignore_conflicts
         )
-        if self.return_id and self.connection.features.can_return_columns_from_insert:
+        if self.returning_fields and self.connection.features.can_return_columns_from_insert:
             if self.connection.features.can_return_rows_from_bulk_insert:
                 result.append(self.connection.ops.bulk_insert_sql(fields, placeholder_rows))
                 params = param_rows
@@ -1306,12 +1306,11 @@ class SQLInsertCompiler(SQLCompiler):
                 params = [param_rows[0]]
             if ignore_conflicts_suffix_sql:
                 result.append(ignore_conflicts_suffix_sql)
-            col = "%s.%s" % (qn(opts.db_table), qn(opts.pk.column))
-            r_fmt, r_params = self.connection.ops.return_insert_id(opts.pk)
-            # Skip empty r_fmt to allow subclasses to customize behavior for
+            # Skip empty r_sql to allow subclasses to customize behavior for
             # 3rd party backends. Refs #19096.
-            if r_fmt:
-                result.append(r_fmt % col)
+            r_sql, r_params = self.connection.ops.return_insert_columns(self.returning_fields)
+            if r_sql:
+                result.append(r_sql)
                 params += [r_params]
             return [(" ".join(result), tuple(chain.from_iterable(params)))]
 
@@ -1328,25 +1327,33 @@ class SQLInsertCompiler(SQLCompiler):
                 for p, vals in zip(placeholder_rows, param_rows)
             ]
 
-    def execute_sql(self, return_id=False):
+    def execute_sql(self, returning_fields=None):
         assert not (
-            return_id and len(self.query.objs) != 1 and
+            returning_fields and len(self.query.objs) != 1 and
             not self.connection.features.can_return_rows_from_bulk_insert
         )
-        self.return_id = return_id
+        self.returning_fields = returning_fields
         with self.connection.cursor() as cursor:
             for sql, params in self.as_sql():
                 cursor.execute(sql, params)
-            if not return_id:
-                return
+            if not self.returning_fields:
+                return []
             if self.connection.features.can_return_rows_from_bulk_insert and len(self.query.objs) > 1:
-                return self.connection.ops.fetch_returned_insert_ids(cursor)
+                return self.connection.ops.fetch_returned_insert_rows(cursor)
             if self.connection.features.can_return_columns_from_insert:
-                assert len(self.query.objs) == 1
-                return self.connection.ops.fetch_returned_insert_id(cursor)
-            return self.connection.ops.last_insert_id(
-                cursor, self.query.get_meta().db_table, self.query.get_meta().pk.column
+                if (
+                    len(self.returning_fields) > 1 and
+                    not self.connection.features.can_return_multiple_columns_from_insert
+                ):
+                    raise NotSupportedError(
+                        'Returning multiple columns from INSERT statements is '
+                        'not supported on this database backend.'
                     )
+                assert len(self.query.objs) == 1
+                return self.connection.ops.fetch_returned_insert_columns(cursor)
+            return [self.connection.ops.last_insert_id(
+                cursor, self.query.get_meta().db_table, self.query.get_meta().pk.column
+            )]
 
 
 class SQLDeleteCompiler(SQLCompiler):

docs/releases/3.0.txt

@@ -448,14 +448,20 @@ backends.
   :class:`~django.db.models.DateTimeField` in ``datetime_cast_date_sql()``,
   ``datetime_extract_sql()``, etc.
 
-* ``DatabaseOperations.return_insert_id()`` now requires an additional
-  ``field`` argument with the model field.
-
 * Entries for ``AutoField``, ``BigAutoField``, and ``SmallAutoField`` are added
   to  ``DatabaseOperations.integer_field_ranges`` to support the integer range
   validators on these field types. Third-party backends may need to customize
   the default entries.
 
+* ``DatabaseOperations.fetch_returned_insert_id()`` is replaced by
+  ``fetch_returned_insert_columns()`` which returns a list of values returned
+  by the ``INSERT … RETURNING`` statement, instead of a single value.
+
+* ``DatabaseOperations.return_insert_id()`` is replaced by
+  ``return_insert_columns()`` that accepts a ``fields``
+  argument, which is an iterable of fields to be returned after insert. Usually
+  this is only the auto-generated primary key.
+
 :mod:`django.contrib.admin`
 ---------------------------
 

tests/backends/models.py

@@ -5,10 +5,6 @@ from django.contrib.contenttypes.models import ContentType
 from django.db import models
 
 
-class NonIntegerAutoField(models.Model):
-    creation_datetime = models.DateTimeField(primary_key=True)
-
-
 class Square(models.Model):
     root = models.IntegerField()
     square = models.PositiveIntegerField()

tests/backends/oracle/tests.py

@@ -1,4 +1,3 @@
-import datetime
 import unittest
 
 from django.db import connection
@@ -6,7 +5,7 @@ from django.db.models.fields import BooleanField, NullBooleanField
 from django.db.utils import DatabaseError
 from django.test import TransactionTestCase
 
-from ..models import NonIntegerAutoField, Square
+from ..models import Square
 
 
 @unittest.skipUnless(connection.vendor == 'oracle', 'Oracle tests')
@@ -96,23 +95,3 @@ class TransactionalTests(TransactionTestCase):
             self.assertIn('ORA-01017', context.exception.args[0].message)
         finally:
             connection.settings_dict['PASSWORD'] = old_password
-
-    def test_non_integer_auto_field(self):
-        with connection.cursor() as cursor:
-            # Create trigger that fill non-integer auto field.
-            cursor.execute("""
-                CREATE OR REPLACE TRIGGER "TRG_FILL_CREATION_DATETIME"
-                BEFORE INSERT ON "BACKENDS_NONINTEGERAUTOFIELD"
-                FOR EACH ROW
-                BEGIN
-                    :NEW.CREATION_DATETIME := SYSTIMESTAMP;
-                END;
-            """)
-        try:
-            NonIntegerAutoField._meta.auto_field = NonIntegerAutoField.creation_datetime
-            obj = NonIntegerAutoField.objects.create()
-            self.assertIsNotNone(obj.creation_datetime)
-            self.assertIsInstance(obj.creation_datetime, datetime.datetime)
-        finally:
-            with connection.cursor() as cursor:
-                cursor.execute('DROP TRIGGER "TRG_FILL_CREATION_DATETIME"')

tests/model_meta/tests.py

@@ -279,3 +279,8 @@ class PropertyNamesTests(SimpleTestCase):
         # Instance only descriptors don't appear in _property_names.
         self.assertEqual(AbstractPerson().test_instance_only_descriptor, 1)
         self.assertEqual(AbstractPerson._meta._property_names, frozenset(['pk', 'test_property']))
+
+
+class ReturningFieldsTests(SimpleTestCase):
+    def test_pk(self):
+        self.assertEqual(Relation._meta.db_returning_fields, [Relation._meta.pk])

tests/queries/models.py

@@ -4,6 +4,7 @@ Various complex queries that have been problematic in the past.
 import threading
 
 from django.db import models
+from django.db.models.functions import Now
 
 
 class DumbCategory(models.Model):
@@ -730,3 +731,19 @@ class RelatedIndividual(models.Model):
 class CustomDbColumn(models.Model):
     custom_column = models.IntegerField(db_column='custom_name', null=True)
     ip_address = models.GenericIPAddressField(null=True)
+
+
+class CreatedField(models.DateTimeField):
+    db_returning = True
+
+    def __init__(self, *args, **kwargs):
+        kwargs.setdefault('default', Now)
+        super().__init__(*args, **kwargs)
+
+
+class ReturningModel(models.Model):
+    created = CreatedField(editable=False)
+
+
+class NonIntegerPKReturningModel(models.Model):
+    created = CreatedField(editable=False, primary_key=True)

tests/queries/test_db_returning.py

@@ -0,0 +1,64 @@
+import datetime
+
+from django.db import NotSupportedError, connection
+from django.test import TestCase, skipIfDBFeature, skipUnlessDBFeature
+from django.test.utils import CaptureQueriesContext
+
+from .models import DumbCategory, NonIntegerPKReturningModel, ReturningModel
+
+
+@skipUnlessDBFeature('can_return_columns_from_insert')
+class ReturningValuesTests(TestCase):
+    def test_insert_returning(self):
+        with CaptureQueriesContext(connection) as captured_queries:
+            DumbCategory.objects.create()
+        self.assertIn(
+            'RETURNING %s.%s' % (
+                connection.ops.quote_name(DumbCategory._meta.db_table),
+                connection.ops.quote_name(DumbCategory._meta.get_field('id').column),
+            ),
+            captured_queries[-1]['sql'],
+        )
+
+    def test_insert_returning_non_integer(self):
+        obj = NonIntegerPKReturningModel.objects.create()
+        self.assertTrue(obj.created)
+        self.assertIsInstance(obj.created, datetime.datetime)
+
+    @skipUnlessDBFeature('can_return_multiple_columns_from_insert')
+    def test_insert_returning_multiple(self):
+        with CaptureQueriesContext(connection) as captured_queries:
+            obj = ReturningModel.objects.create()
+        table_name = connection.ops.quote_name(ReturningModel._meta.db_table)
+        self.assertIn(
+            'RETURNING %s.%s, %s.%s' % (
+                table_name,
+                connection.ops.quote_name(ReturningModel._meta.get_field('id').column),
+                table_name,
+                connection.ops.quote_name(ReturningModel._meta.get_field('created').column),
+            ),
+            captured_queries[-1]['sql'],
+        )
+        self.assertTrue(obj.pk)
+        self.assertIsInstance(obj.created, datetime.datetime)
+
+    @skipIfDBFeature('can_return_multiple_columns_from_insert')
+    def test_insert_returning_multiple_not_supported(self):
+        msg = (
+            'Returning multiple columns from INSERT statements is '
+            'not supported on this database backend.'
+        )
+        with self.assertRaisesMessage(NotSupportedError, msg):
+            ReturningModel.objects.create()
+
+    @skipUnlessDBFeature(
+        'can_return_rows_from_bulk_insert',
+        'can_return_multiple_columns_from_insert',
+    )
+    def test_bulk_insert(self):
+        objs = [ReturningModel(), ReturningModel(pk=2 ** 11), ReturningModel()]
+        ReturningModel.objects.bulk_create(objs)
+        for obj in objs:
+            with self.subTest(obj=obj):
+                self.assertTrue(obj.pk)
+                self.assertIsInstance(obj.created, datetime.datetime)