""" PostgreSQL database backend for Django. Requires psycopg2 >= 2.8.4 or psycopg >= 3.1.8 """ import asyncio import threading import warnings from contextlib import contextmanager from django.conf import settings from django.core.exceptions import ImproperlyConfigured from django.db import DatabaseError as WrappedDatabaseError from django.db import connections from django.db.backends.base.base import NO_DB_ALIAS, BaseDatabaseWrapper from django.db.backends.utils import CursorDebugWrapper as BaseCursorDebugWrapper from django.utils.asyncio import async_unsafe from django.utils.functional import cached_property from django.utils.safestring import SafeString from django.utils.version import get_version_tuple try: try: import psycopg as Database except ImportError: import psycopg2 as Database except ImportError: raise ImproperlyConfigured("Error loading psycopg2 or psycopg module") def psycopg_version(): version = Database.__version__.split(" ", 1)[0] return get_version_tuple(version) if psycopg_version() < (2, 8, 4): raise ImproperlyConfigured( f"psycopg2 version 2.8.4 or newer is required; you have {Database.__version__}" ) if (3,) <= psycopg_version() < (3, 1, 8): raise ImproperlyConfigured( f"psycopg version 3.1.8 or newer is required; you have {Database.__version__}" ) from .psycopg_any import IsolationLevel, is_psycopg3 # NOQA isort:skip if is_psycopg3: from psycopg import adapters, sql from psycopg.pq import Format from .psycopg_any import get_adapters_template, register_tzloader TIMESTAMPTZ_OID = adapters.types["timestamptz"].oid else: import psycopg2.extensions import psycopg2.extras psycopg2.extensions.register_adapter(SafeString, psycopg2.extensions.QuotedString) psycopg2.extras.register_uuid() # Register support for inet[] manually so we don't have to handle the Inet() # object on load all the time. INETARRAY_OID = 1041 INETARRAY = psycopg2.extensions.new_array_type( (INETARRAY_OID,), "INETARRAY", psycopg2.extensions.UNICODE, ) psycopg2.extensions.register_type(INETARRAY) # Some of these import psycopg, so import them after checking if it's installed. from .client import DatabaseClient # NOQA isort:skip from .creation import DatabaseCreation # NOQA isort:skip from .features import DatabaseFeatures # NOQA isort:skip from .introspection import DatabaseIntrospection # NOQA isort:skip from .operations import DatabaseOperations # NOQA isort:skip from .schema import DatabaseSchemaEditor # NOQA isort:skip def _get_varchar_column(data): if data["max_length"] is None: return "varchar" return "varchar(%(max_length)s)" % data class DatabaseWrapper(BaseDatabaseWrapper): vendor = "postgresql" display_name = "PostgreSQL" # This dictionary maps Field objects to their associated PostgreSQL column # types, as strings. Column-type strings can contain format strings; they'll # be interpolated against the values of Field.__dict__ before being output. # If a column type is set to None, it won't be included in the output. data_types = { "AutoField": "integer", "BigAutoField": "bigint", "BinaryField": "bytea", "BooleanField": "boolean", "CharField": _get_varchar_column, "DateField": "date", "DateTimeField": "timestamp with time zone", "DecimalField": "numeric(%(max_digits)s, %(decimal_places)s)", "DurationField": "interval", "FileField": "varchar(%(max_length)s)", "FilePathField": "varchar(%(max_length)s)", "FloatField": "double precision", "IntegerField": "integer", "BigIntegerField": "bigint", "IPAddressField": "inet", "GenericIPAddressField": "inet", "JSONField": "jsonb", "OneToOneField": "integer", "PositiveBigIntegerField": "bigint", "PositiveIntegerField": "integer", "PositiveSmallIntegerField": "smallint", "SlugField": "varchar(%(max_length)s)", "SmallAutoField": "smallint", "SmallIntegerField": "smallint", "TextField": "text", "TimeField": "time", "UUIDField": "uuid", } data_type_check_constraints = { "PositiveBigIntegerField": '"%(column)s" >= 0', "PositiveIntegerField": '"%(column)s" >= 0', "PositiveSmallIntegerField": '"%(column)s" >= 0', } data_types_suffix = { "AutoField": "GENERATED BY DEFAULT AS IDENTITY", "BigAutoField": "GENERATED BY DEFAULT AS IDENTITY", "SmallAutoField": "GENERATED BY DEFAULT AS IDENTITY", } operators = { "exact": "= %s", "iexact": "= UPPER(%s)", "contains": "LIKE %s", "icontains": "LIKE UPPER(%s)", "regex": "~ %s", "iregex": "~* %s", "gt": "> %s", "gte": ">= %s", "lt": "< %s", "lte": "<= %s", "startswith": "LIKE %s", "endswith": "LIKE %s", "istartswith": "LIKE UPPER(%s)", "iendswith": "LIKE UPPER(%s)", } # The patterns below are used to generate SQL pattern lookup clauses when # the right-hand side of the lookup isn't a raw string (it might be an expression # or the result of a bilateral transformation). # In those cases, special characters for LIKE operators (e.g. \, *, _) should be # escaped on database side. # # Note: we use str.format() here for readability as '%' is used as a wildcard for # the LIKE operator. pattern_esc = ( r"REPLACE(REPLACE(REPLACE({}, E'\\', E'\\\\'), E'%%', E'\\%%'), E'_', E'\\_')" ) pattern_ops = { "contains": "LIKE '%%' || {} || '%%'", "icontains": "LIKE '%%' || UPPER({}) || '%%'", "startswith": "LIKE {} || '%%'", "istartswith": "LIKE UPPER({}) || '%%'", "endswith": "LIKE '%%' || {}", "iendswith": "LIKE '%%' || UPPER({})", } Database = Database SchemaEditorClass = DatabaseSchemaEditor # Classes instantiated in __init__(). client_class = DatabaseClient creation_class = DatabaseCreation features_class = DatabaseFeatures introspection_class = DatabaseIntrospection ops_class = DatabaseOperations # PostgreSQL backend-specific attributes. _named_cursor_idx = 0 _connection_pools = {} @property def pool(self): pool_options = self.settings_dict["OPTIONS"].get("pool") if self.alias == NO_DB_ALIAS or not pool_options: return None if self.alias not in self._connection_pools: if self.settings_dict.get("CONN_MAX_AGE", 0) != 0: raise ImproperlyConfigured( "Pooling doesn't support persistent connections." ) # Set the default options. if pool_options is True: pool_options = {} try: from psycopg_pool import ConnectionPool except ImportError as err: raise ImproperlyConfigured( "Error loading psycopg_pool module.\nDid you install psycopg[pool]?" ) from err connect_kwargs = self.get_connection_params() # Ensure we run in autocommit, Django properly sets it later on. connect_kwargs["autocommit"] = True enable_checks = self.settings_dict["CONN_HEALTH_CHECKS"] pool = ConnectionPool( kwargs=connect_kwargs, open=False, # Do not open the pool during startup. configure=self._configure_connection, check=ConnectionPool.check_connection if enable_checks else None, **pool_options, ) # setdefault() ensures that multiple threads don't set this in # parallel. Since we do not open the pool during it's init above, # this means that at worst during startup multiple threads generate # pool objects and the first to set it wins. self._connection_pools.setdefault(self.alias, pool) return self._connection_pools[self.alias] def close_pool(self): if self.pool: self.pool.close() del self._connection_pools[self.alias] def get_database_version(self): """ Return a tuple of the database's version. E.g. for pg_version 120004, return (12, 4). """ return divmod(self.pg_version, 10000) def get_connection_params(self): settings_dict = self.settings_dict # None may be used to connect to the default 'postgres' db if settings_dict["NAME"] == "" and not settings_dict["OPTIONS"].get("service"): raise ImproperlyConfigured( "settings.DATABASES is improperly configured. " "Please supply the NAME or OPTIONS['service'] value." ) if len(settings_dict["NAME"] or "") > self.ops.max_name_length(): raise ImproperlyConfigured( "The database name '%s' (%d characters) is longer than " "PostgreSQL's limit of %d characters. Supply a shorter NAME " "in settings.DATABASES." % ( settings_dict["NAME"], len(settings_dict["NAME"]), self.ops.max_name_length(), ) ) if settings_dict["NAME"]: conn_params = { "dbname": settings_dict["NAME"], **settings_dict["OPTIONS"], } elif settings_dict["NAME"] is None: # Connect to the default 'postgres' db. settings_dict["OPTIONS"].pop("service", None) conn_params = {"dbname": "postgres", **settings_dict["OPTIONS"]} else: conn_params = {**settings_dict["OPTIONS"]} conn_params["client_encoding"] = "UTF8" conn_params.pop("assume_role", None) conn_params.pop("isolation_level", None) pool_options = conn_params.pop("pool", None) if pool_options and not is_psycopg3: raise ImproperlyConfigured("Database pooling requires psycopg >= 3") server_side_binding = conn_params.pop("server_side_binding", None) conn_params.setdefault( "cursor_factory", ( ServerBindingCursor if is_psycopg3 and server_side_binding is True else Cursor ), ) if settings_dict["USER"]: conn_params["user"] = settings_dict["USER"] if settings_dict["PASSWORD"]: conn_params["password"] = settings_dict["PASSWORD"] if settings_dict["HOST"]: conn_params["host"] = settings_dict["HOST"] if settings_dict["PORT"]: conn_params["port"] = settings_dict["PORT"] if is_psycopg3: conn_params["context"] = get_adapters_template( settings.USE_TZ, self.timezone ) # Disable prepared statements by default to keep connection poolers # working. Can be reenabled via OPTIONS in the settings dict. conn_params["prepare_threshold"] = conn_params.pop( "prepare_threshold", None ) return conn_params @async_unsafe def get_new_connection(self, conn_params): # self.isolation_level must be set: # - after connecting to the database in order to obtain the database's # default when no value is explicitly specified in options. # - before calling _set_autocommit() because if autocommit is on, that # will set connection.isolation_level to ISOLATION_LEVEL_AUTOCOMMIT. options = self.settings_dict["OPTIONS"] set_isolation_level = False try: isolation_level_value = options["isolation_level"] except KeyError: self.isolation_level = IsolationLevel.READ_COMMITTED else: # Set the isolation level to the value from OPTIONS. try: self.isolation_level = IsolationLevel(isolation_level_value) set_isolation_level = True except ValueError: raise ImproperlyConfigured( f"Invalid transaction isolation level {isolation_level_value} " f"specified. Use one of the psycopg.IsolationLevel values." ) if self.pool: # If nothing else has opened the pool, open it now. self.pool.open() connection = self.pool.getconn() else: connection = self.Database.connect(**conn_params) if set_isolation_level: connection.isolation_level = self.isolation_level if not is_psycopg3: # Register dummy loads() to avoid a round trip from psycopg2's # decode to json.dumps() to json.loads(), when using a custom # decoder in JSONField. psycopg2.extras.register_default_jsonb( conn_or_curs=connection, loads=lambda x: x ) return connection def ensure_timezone(self): # Close the pool so new connections pick up the correct timezone. self.close_pool() if self.connection is None: return False return self._configure_timezone(self.connection) def _configure_timezone(self, connection): conn_timezone_name = connection.info.parameter_status("TimeZone") timezone_name = self.timezone_name if timezone_name and conn_timezone_name != timezone_name: with connection.cursor() as cursor: cursor.execute(self.ops.set_time_zone_sql(), [timezone_name]) return True return False def _configure_role(self, connection): if new_role := self.settings_dict["OPTIONS"].get("assume_role"): with connection.cursor() as cursor: sql = self.ops.compose_sql("SET ROLE %s", [new_role]) cursor.execute(sql) return True return False def _configure_connection(self, connection): # This function is called from init_connection_state and from the # psycopg pool itself after a connection is opened. # Commit after setting the time zone. commit_tz = self._configure_timezone(connection) # Set the role on the connection. This is useful if the credential used # to login is not the same as the role that owns database resources. As # can be the case when using temporary or ephemeral credentials. commit_role = self._configure_role(connection) return commit_role or commit_tz def _close(self): if self.connection is not None: # `wrap_database_errors` only works for `putconn` as long as there # is no `reset` function set in the pool because it is deferred # into a thread and not directly executed. with self.wrap_database_errors: if self.pool: # Ensure the correct pool is returned. This is a workaround # for tests so a pool can be changed on setting changes # (e.g. USE_TZ, TIME_ZONE). self.connection._pool.putconn(self.connection) # Connection can no longer be used. self.connection = None else: return self.connection.close() def init_connection_state(self): super().init_connection_state() if self.connection is not None and not self.pool: commit = self._configure_connection(self.connection) if commit and not self.get_autocommit(): self.connection.commit() @async_unsafe def create_cursor(self, name=None): if name: if is_psycopg3 and ( self.settings_dict["OPTIONS"].get("server_side_binding") is not True ): # psycopg >= 3 forces the usage of server-side bindings for # named cursors so a specialized class that implements # server-side cursors while performing client-side bindings # must be used if `server_side_binding` is disabled (default). cursor = ServerSideCursor( self.connection, name=name, scrollable=False, withhold=self.connection.autocommit, ) else: # In autocommit mode, the cursor will be used outside of a # transaction, hence use a holdable cursor. cursor = self.connection.cursor( name, scrollable=False, withhold=self.connection.autocommit ) else: cursor = self.connection.cursor() if is_psycopg3: # Register the cursor timezone only if the connection disagrees, to # avoid copying the adapter map. tzloader = self.connection.adapters.get_loader(TIMESTAMPTZ_OID, Format.TEXT) if self.timezone != tzloader.timezone: register_tzloader(self.timezone, cursor) else: cursor.tzinfo_factory = self.tzinfo_factory if settings.USE_TZ else None return cursor def tzinfo_factory(self, offset): return self.timezone @async_unsafe def chunked_cursor(self): self._named_cursor_idx += 1 # Get the current async task # Note that right now this is behind @async_unsafe, so this is # unreachable, but in future we'll start loosening this restriction. # For now, it's here so that every use of "threading" is # also async-compatible. try: current_task = asyncio.current_task() except RuntimeError: current_task = None # Current task can be none even if the current_task call didn't error if current_task: task_ident = str(id(current_task)) else: task_ident = "sync" # Use that and the thread ident to get a unique name return self._cursor( name="_django_curs_%d_%s_%d" % ( # Avoid reusing name in other threads / tasks threading.current_thread().ident, task_ident, self._named_cursor_idx, ) ) def _set_autocommit(self, autocommit): with self.wrap_database_errors: self.connection.autocommit = autocommit def check_constraints(self, table_names=None): """ Check constraints by setting them to immediate. Return them to deferred afterward. """ with self.cursor() as cursor: cursor.execute("SET CONSTRAINTS ALL IMMEDIATE") cursor.execute("SET CONSTRAINTS ALL DEFERRED") def is_usable(self): if self.connection is None: return False try: # Use a psycopg cursor directly, bypassing Django's utilities. with self.connection.cursor() as cursor: cursor.execute("SELECT 1") except Database.Error: return False else: return True def close_if_health_check_failed(self): if self.pool: # The pool only returns healthy connections. return return super().close_if_health_check_failed() @contextmanager def _nodb_cursor(self): cursor = None try: with super()._nodb_cursor() as cursor: yield cursor except (Database.DatabaseError, WrappedDatabaseError): if cursor is not None: raise warnings.warn( "Normally Django will use a connection to the 'postgres' database " "to avoid running initialization queries against the production " "database when it's not needed (for example, when running tests). " "Django was unable to create a connection to the 'postgres' database " "and will use the first PostgreSQL database instead.", RuntimeWarning, ) for connection in connections.all(): if ( connection.vendor == "postgresql" and connection.settings_dict["NAME"] != "postgres" ): conn = self.__class__( { **self.settings_dict, "NAME": connection.settings_dict["NAME"], }, alias=self.alias, ) try: with conn.cursor() as cursor: yield cursor finally: conn.close() break else: raise @cached_property def pg_version(self): with self.temporary_connection(): return self.connection.info.server_version def make_debug_cursor(self, cursor): return CursorDebugWrapper(cursor, self) if is_psycopg3: class CursorMixin: """ A subclass of psycopg cursor implementing callproc. """ def callproc(self, name, args=None): if not isinstance(name, sql.Identifier): name = sql.Identifier(name) qparts = [sql.SQL("SELECT * FROM "), name, sql.SQL("(")] if args: for item in args: qparts.append(sql.Literal(item)) qparts.append(sql.SQL(",")) del qparts[-1] qparts.append(sql.SQL(")")) stmt = sql.Composed(qparts) self.execute(stmt) return args class ServerBindingCursor(CursorMixin, Database.Cursor): pass class Cursor(CursorMixin, Database.ClientCursor): pass class ServerSideCursor( CursorMixin, Database.client_cursor.ClientCursorMixin, Database.ServerCursor ): """ psycopg >= 3 forces the usage of server-side bindings when using named cursors but the ORM doesn't yet support the systematic generation of prepareable SQL (#20516). ClientCursorMixin forces the usage of client-side bindings while ServerCursor implements the logic required to declare and scroll through named cursors. Mixing ClientCursorMixin in wouldn't be necessary if Cursor allowed to specify how parameters should be bound instead, which ServerCursor would inherit, but that's not the case. """ class CursorDebugWrapper(BaseCursorDebugWrapper): def copy(self, statement): with self.debug_sql(statement): return self.cursor.copy(statement) else: Cursor = psycopg2.extensions.cursor class CursorDebugWrapper(BaseCursorDebugWrapper): def copy_expert(self, sql, file, *args): with self.debug_sql(sql): return self.cursor.copy_expert(sql, file, *args) def copy_to(self, file, table, *args, **kwargs): with self.debug_sql(sql="COPY %s TO STDOUT" % table): return self.cursor.copy_to(file, table, *args, **kwargs)