Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.11/site-packages/sqlalchemy/orm/loading.py: 11%

Shortcuts on this page

r m x   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

658 statements  

1# orm/loading.py 

2# Copyright (C) 2005-2025 the SQLAlchemy authors and contributors 

3# <see AUTHORS file> 

4# 

5# This module is part of SQLAlchemy and is released under 

6# the MIT License: https://www.opensource.org/licenses/mit-license.php 

7# mypy: ignore-errors 

8 

9 

10"""private module containing functions used to convert database 

11rows into object instances and associated state. 

12 

13the functions here are called primarily by Query, Mapper, 

14as well as some of the attribute loading strategies. 

15 

16""" 

17 

18from __future__ import annotations 

19 

20from typing import Any 

21from typing import Dict 

22from typing import Iterable 

23from typing import List 

24from typing import Mapping 

25from typing import Optional 

26from typing import Sequence 

27from typing import Tuple 

28from typing import TYPE_CHECKING 

29from typing import TypeVar 

30from typing import Union 

31 

32from . import attributes 

33from . import exc as orm_exc 

34from . import path_registry 

35from .base import _DEFER_FOR_STATE 

36from .base import _RAISE_FOR_STATE 

37from .base import _SET_DEFERRED_EXPIRED 

38from .base import PassiveFlag 

39from .context import _ORMCompileState 

40from .context import FromStatement 

41from .context import QueryContext 

42from .util import _none_set 

43from .util import state_str 

44from .. import exc as sa_exc 

45from .. import util 

46from ..engine import result_tuple 

47from ..engine.result import ChunkedIteratorResult 

48from ..engine.result import FrozenResult 

49from ..engine.result import SimpleResultMetaData 

50from ..sql import select 

51from ..sql import util as sql_util 

52from ..sql.selectable import ForUpdateArg 

53from ..sql.selectable import LABEL_STYLE_TABLENAME_PLUS_COL 

54from ..sql.selectable import SelectState 

55from ..util import EMPTY_DICT 

56from ..util.typing import TupleAny 

57from ..util.typing import Unpack 

58 

59if TYPE_CHECKING: 

60 from ._typing import _IdentityKeyType 

61 from .base import LoaderCallableStatus 

62 from .interfaces import ORMOption 

63 from .mapper import Mapper 

64 from .query import Query 

65 from .session import Session 

66 from .state import InstanceState 

67 from ..engine.cursor import CursorResult 

68 from ..engine.interfaces import _ExecuteOptions 

69 from ..engine.result import Result 

70 from ..sql import Select 

71 

72_T = TypeVar("_T", bound=Any) 

73_O = TypeVar("_O", bound=object) 

74_new_runid = util.counter() 

75 

76 

77_PopulatorDict = Dict[str, List[Tuple[str, Any]]] 

78 

79 

80def instances( 

81 cursor: CursorResult[Unpack[TupleAny]], context: QueryContext 

82) -> Result[Unpack[TupleAny]]: 

83 """Return a :class:`.Result` given an ORM query context. 

84 

85 :param cursor: a :class:`.CursorResult`, generated by a statement 

86 which came from :class:`.ORMCompileState` 

87 

88 :param context: a :class:`.QueryContext` object 

89 

90 :return: a :class:`.Result` object representing ORM results 

91 

92 .. versionchanged:: 1.4 The instances() function now uses 

93 :class:`.Result` objects and has an all new interface. 

94 

95 """ 

96 

97 context.runid = _new_runid() 

98 

99 if context.top_level_context: 

100 is_top_level = False 

101 context.post_load_paths = context.top_level_context.post_load_paths 

102 else: 

103 is_top_level = True 

104 context.post_load_paths = {} 

105 

106 compile_state = context.compile_state 

107 filtered = compile_state._has_mapper_entities 

108 single_entity = ( 

109 not context.load_options._only_return_tuples 

110 and len(compile_state._entities) == 1 

111 and compile_state._entities[0].supports_single_entity 

112 ) 

113 

114 try: 

115 (process, labels, extra) = list( 

116 zip( 

117 *[ 

118 query_entity.row_processor(context, cursor) 

119 for query_entity in context.compile_state._entities 

120 ] 

121 ) 

122 ) 

123 

124 if context.yield_per and ( 

125 context.loaders_require_buffering 

126 or context.loaders_require_uniquing 

127 ): 

128 raise sa_exc.InvalidRequestError( 

129 "Can't use yield_per with eager loaders that require uniquing " 

130 "or row buffering, e.g. joinedload() against collections " 

131 "or subqueryload(). Consider the selectinload() strategy " 

132 "for better flexibility in loading objects." 

133 ) 

134 

135 except Exception: 

136 with util.safe_reraise(): 

137 cursor.close() 

138 

139 def _no_unique(entry): 

140 raise sa_exc.InvalidRequestError( 

141 "Can't use the ORM yield_per feature in conjunction with unique()" 

142 ) 

143 

144 def _not_hashable(datatype, *, legacy=False, uncertain=False): 

145 if not legacy: 

146 

147 def go(obj): 

148 if uncertain: 

149 try: 

150 return hash(obj) 

151 except: 

152 pass 

153 

154 raise sa_exc.InvalidRequestError( 

155 "Can't apply uniqueness to row tuple containing value of " 

156 f"""type {datatype!r}; { 

157 'the values returned appear to be' 

158 if uncertain 

159 else 'this datatype produces' 

160 } non-hashable values""" 

161 ) 

162 

163 return go 

164 elif not uncertain: 

165 return id 

166 else: 

167 _use_id = False 

168 

169 def go(obj): 

170 nonlocal _use_id 

171 

172 if not _use_id: 

173 try: 

174 return hash(obj) 

175 except: 

176 pass 

177 

178 # in #10459, we considered using a warning here, however 

179 # as legacy query uses result.unique() in all cases, this 

180 # would lead to too many warning cases. 

181 _use_id = True 

182 

183 return id(obj) 

184 

185 return go 

186 

187 unique_filters = [ 

188 ( 

189 _no_unique 

190 if context.yield_per 

191 else ( 

192 _not_hashable( 

193 ent.column.type, # type: ignore 

194 legacy=context.load_options._legacy_uniquing, 

195 uncertain=ent._null_column_type, 

196 ) 

197 if ( 

198 not ent.use_id_for_hash 

199 and (ent._non_hashable_value or ent._null_column_type) 

200 ) 

201 else id if ent.use_id_for_hash else None 

202 ) 

203 ) 

204 for ent in context.compile_state._entities 

205 ] 

206 

207 row_metadata = SimpleResultMetaData( 

208 labels, extra, _unique_filters=unique_filters 

209 ) 

210 

211 def chunks(size): # type: ignore 

212 while True: 

213 yield_per = size 

214 

215 context.partials = {} 

216 

217 if yield_per: 

218 fetch = cursor.fetchmany(yield_per) 

219 

220 if not fetch: 

221 break 

222 else: 

223 fetch = cursor._raw_all_rows() 

224 

225 if single_entity: 

226 proc = process[0] 

227 rows = [proc(row) for row in fetch] 

228 else: 

229 rows = [ 

230 tuple([proc(row) for proc in process]) for row in fetch 

231 ] 

232 

233 # if we are the originating load from a query, meaning we 

234 # aren't being called as a result of a nested "post load", 

235 # iterate through all the collected post loaders and fire them 

236 # off. Previously this used to work recursively, however that 

237 # prevented deeply nested structures from being loadable 

238 if is_top_level: 

239 if yield_per: 

240 # if using yield per, memoize the state of the 

241 # collection so that it can be restored 

242 top_level_post_loads = list( 

243 context.post_load_paths.items() 

244 ) 

245 

246 while context.post_load_paths: 

247 post_loads = list(context.post_load_paths.items()) 

248 context.post_load_paths.clear() 

249 for path, post_load in post_loads: 

250 post_load.invoke(context, path) 

251 

252 if yield_per: 

253 context.post_load_paths.clear() 

254 context.post_load_paths.update(top_level_post_loads) 

255 

256 yield rows 

257 

258 if not yield_per: 

259 break 

260 

261 if context.execution_options.get("prebuffer_rows", False): 

262 # this is a bit of a hack at the moment. 

263 # I would rather have some option in the result to pre-buffer 

264 # internally. 

265 _prebuffered = list(chunks(None)) 

266 

267 def chunks(size): 

268 return iter(_prebuffered) 

269 

270 result = ChunkedIteratorResult( 

271 row_metadata, 

272 chunks, 

273 source_supports_scalars=single_entity, 

274 raw=cursor, 

275 dynamic_yield_per=cursor.context._is_server_side, 

276 ) 

277 

278 # filtered and single_entity are used to indicate to legacy Query that the 

279 # query has ORM entities, so legacy deduping and scalars should be called 

280 # on the result. 

281 result._attributes = result._attributes.union( 

282 dict(filtered=filtered, is_single_entity=single_entity) 

283 ) 

284 

285 # multi_row_eager_loaders OTOH is specific to joinedload. 

286 if context.compile_state.multi_row_eager_loaders: 

287 

288 def require_unique(obj): 

289 raise sa_exc.InvalidRequestError( 

290 "The unique() method must be invoked on this Result, " 

291 "as it contains results that include joined eager loads " 

292 "against collections" 

293 ) 

294 

295 result._unique_filter_state = (None, require_unique) 

296 

297 if context.yield_per: 

298 result.yield_per(context.yield_per) 

299 

300 return result 

301 

302 

303@util.preload_module("sqlalchemy.orm.context") 

304def merge_frozen_result(session, statement, frozen_result, load=True): 

305 """Merge a :class:`_engine.FrozenResult` back into a :class:`_orm.Session`, 

306 returning a new :class:`_engine.Result` object with :term:`persistent` 

307 objects. 

308 

309 See the section :ref:`do_orm_execute_re_executing` for an example. 

310 

311 .. seealso:: 

312 

313 :ref:`do_orm_execute_re_executing` 

314 

315 :meth:`_engine.Result.freeze` 

316 

317 :class:`_engine.FrozenResult` 

318 

319 """ 

320 querycontext = util.preloaded.orm_context 

321 

322 if load: 

323 # flush current contents if we expect to load data 

324 session._autoflush() 

325 

326 ctx = querycontext._ORMSelectCompileState._create_entities_collection( 

327 statement, legacy=False 

328 ) 

329 

330 with session.no_autoflush: 

331 mapped_entities = [ 

332 i 

333 for i, e in enumerate(ctx._entities) 

334 if isinstance(e, querycontext._MapperEntity) 

335 ] 

336 keys = [ent._label_name for ent in ctx._entities] 

337 

338 keyed_tuple = result_tuple( 

339 keys, [ent._extra_entities for ent in ctx._entities] 

340 ) 

341 

342 result = [] 

343 for newrow in frozen_result.rewrite_rows(): 

344 for i in mapped_entities: 

345 if newrow[i] is not None: 

346 newrow[i] = session._merge( 

347 attributes.instance_state(newrow[i]), 

348 attributes.instance_dict(newrow[i]), 

349 load=load, 

350 _recursive={}, 

351 _resolve_conflict_map={}, 

352 ) 

353 

354 result.append(keyed_tuple(newrow)) 

355 

356 return frozen_result.with_new_rows(result) 

357 

358 

359@util.became_legacy_20( 

360 ":func:`_orm.merge_result`", 

361 alternative="The function as well as the method on :class:`_orm.Query` " 

362 "is superseded by the :func:`_orm.merge_frozen_result` function.", 

363) 

364@util.preload_module("sqlalchemy.orm.context") 

365def merge_result( 

366 query: Query[Any], 

367 iterator: Union[FrozenResult, Iterable[Sequence[Any]], Iterable[object]], 

368 load: bool = True, 

369) -> Union[FrozenResult, Iterable[Any]]: 

370 """Merge a result into the given :class:`.Query` object's Session. 

371 

372 See :meth:`_orm.Query.merge_result` for top-level documentation on this 

373 function. 

374 

375 """ 

376 

377 querycontext = util.preloaded.orm_context 

378 

379 session = query.session 

380 if load: 

381 # flush current contents if we expect to load data 

382 session._autoflush() 

383 

384 # TODO: need test coverage and documentation for the FrozenResult 

385 # use case. 

386 if isinstance(iterator, FrozenResult): 

387 frozen_result = iterator 

388 iterator = iter(frozen_result.data) 

389 else: 

390 frozen_result = None 

391 

392 ctx = querycontext._ORMSelectCompileState._create_entities_collection( 

393 query, legacy=True 

394 ) 

395 

396 autoflush = session.autoflush 

397 try: 

398 session.autoflush = False 

399 single_entity = not frozen_result and len(ctx._entities) == 1 

400 

401 if single_entity: 

402 if isinstance(ctx._entities[0], querycontext._MapperEntity): 

403 result = [ 

404 session._merge( 

405 attributes.instance_state(instance), 

406 attributes.instance_dict(instance), 

407 load=load, 

408 _recursive={}, 

409 _resolve_conflict_map={}, 

410 ) 

411 for instance in iterator 

412 ] 

413 else: 

414 result = list(iterator) 

415 else: 

416 mapped_entities = [ 

417 i 

418 for i, e in enumerate(ctx._entities) 

419 if isinstance(e, querycontext._MapperEntity) 

420 ] 

421 result = [] 

422 keys = [ent._label_name for ent in ctx._entities] 

423 

424 keyed_tuple = result_tuple( 

425 keys, [ent._extra_entities for ent in ctx._entities] 

426 ) 

427 

428 for row in iterator: 

429 newrow = list(row) 

430 for i in mapped_entities: 

431 if newrow[i] is not None: 

432 newrow[i] = session._merge( 

433 attributes.instance_state(newrow[i]), 

434 attributes.instance_dict(newrow[i]), 

435 load=load, 

436 _recursive={}, 

437 _resolve_conflict_map={}, 

438 ) 

439 result.append(keyed_tuple(newrow)) 

440 

441 if frozen_result: 

442 return frozen_result.with_new_rows(result) 

443 else: 

444 return iter(result) 

445 finally: 

446 session.autoflush = autoflush 

447 

448 

449def get_from_identity( 

450 session: Session, 

451 mapper: Mapper[_O], 

452 key: _IdentityKeyType[_O], 

453 passive: PassiveFlag, 

454) -> Union[LoaderCallableStatus, Optional[_O]]: 

455 """Look up the given key in the given session's identity map, 

456 check the object for expired state if found. 

457 

458 """ 

459 instance = session.identity_map.get(key) 

460 if instance is not None: 

461 state = attributes.instance_state(instance) 

462 

463 if mapper.inherits and not state.mapper.isa(mapper): 

464 return attributes.PASSIVE_CLASS_MISMATCH 

465 

466 # expired - ensure it still exists 

467 if state.expired: 

468 if not passive & attributes.SQL_OK: 

469 # TODO: no coverage here 

470 return attributes.PASSIVE_NO_RESULT 

471 elif not passive & attributes.RELATED_OBJECT_OK: 

472 # this mode is used within a flush and the instance's 

473 # expired state will be checked soon enough, if necessary. 

474 # also used by immediateloader for a mutually-dependent 

475 # o2m->m2m load, :ticket:`6301` 

476 return instance 

477 try: 

478 state._load_expired(state, passive) 

479 except orm_exc.ObjectDeletedError: 

480 session._remove_newly_deleted([state]) 

481 return None 

482 return instance 

483 else: 

484 return None 

485 

486 

487def _load_on_ident( 

488 session: Session, 

489 statement: Union[Select, FromStatement], 

490 key: Optional[_IdentityKeyType], 

491 *, 

492 load_options: Optional[Sequence[ORMOption]] = None, 

493 refresh_state: Optional[InstanceState[Any]] = None, 

494 with_for_update: Optional[ForUpdateArg] = None, 

495 only_load_props: Optional[Iterable[str]] = None, 

496 no_autoflush: bool = False, 

497 bind_arguments: Mapping[str, Any] = util.EMPTY_DICT, 

498 execution_options: _ExecuteOptions = util.EMPTY_DICT, 

499 require_pk_cols: bool = False, 

500 is_user_refresh: bool = False, 

501): 

502 """Load the given identity key from the database.""" 

503 if key is not None: 

504 ident = key[1] 

505 identity_token = key[2] 

506 else: 

507 ident = identity_token = None 

508 

509 return _load_on_pk_identity( 

510 session, 

511 statement, 

512 ident, 

513 load_options=load_options, 

514 refresh_state=refresh_state, 

515 with_for_update=with_for_update, 

516 only_load_props=only_load_props, 

517 identity_token=identity_token, 

518 no_autoflush=no_autoflush, 

519 bind_arguments=bind_arguments, 

520 execution_options=execution_options, 

521 require_pk_cols=require_pk_cols, 

522 is_user_refresh=is_user_refresh, 

523 ) 

524 

525 

526def _load_on_pk_identity( 

527 session: Session, 

528 statement: Union[Select, FromStatement], 

529 primary_key_identity: Optional[Tuple[Any, ...]], 

530 *, 

531 load_options: Optional[Sequence[ORMOption]] = None, 

532 refresh_state: Optional[InstanceState[Any]] = None, 

533 with_for_update: Optional[ForUpdateArg] = None, 

534 only_load_props: Optional[Iterable[str]] = None, 

535 identity_token: Optional[Any] = None, 

536 no_autoflush: bool = False, 

537 bind_arguments: Mapping[str, Any] = util.EMPTY_DICT, 

538 execution_options: _ExecuteOptions = util.EMPTY_DICT, 

539 require_pk_cols: bool = False, 

540 is_user_refresh: bool = False, 

541): 

542 """Load the given primary key identity from the database.""" 

543 

544 query = statement 

545 q = query._clone() 

546 

547 assert not q._is_lambda_element 

548 

549 if load_options is None: 

550 load_options = QueryContext.default_load_options 

551 

552 if ( 

553 statement._compile_options 

554 is SelectState.default_select_compile_options 

555 ): 

556 compile_options = _ORMCompileState.default_compile_options 

557 else: 

558 compile_options = statement._compile_options 

559 

560 if primary_key_identity is not None: 

561 mapper = query._propagate_attrs["plugin_subject"] 

562 

563 (_get_clause, _get_params) = mapper._get_clause 

564 

565 # None present in ident - turn those comparisons 

566 # into "IS NULL" 

567 if None in primary_key_identity: 

568 nones = { 

569 _get_params[col].key 

570 for col, value in zip(mapper.primary_key, primary_key_identity) 

571 if value is None 

572 } 

573 

574 _get_clause = sql_util.adapt_criterion_to_null(_get_clause, nones) 

575 

576 if len(nones) == len(primary_key_identity): 

577 util.warn( 

578 "fully NULL primary key identity cannot load any " 

579 "object. This condition may raise an error in a future " 

580 "release." 

581 ) 

582 

583 q._where_criteria = ( 

584 sql_util._deep_annotate(_get_clause, {"_orm_adapt": True}), 

585 ) 

586 

587 params = { 

588 _get_params[primary_key].key: id_val 

589 for id_val, primary_key in zip( 

590 primary_key_identity, mapper.primary_key 

591 ) 

592 } 

593 else: 

594 params = None 

595 

596 if with_for_update is not None: 

597 version_check = True 

598 q._for_update_arg = with_for_update 

599 elif query._for_update_arg is not None: 

600 version_check = True 

601 q._for_update_arg = query._for_update_arg 

602 else: 

603 version_check = False 

604 

605 if require_pk_cols and only_load_props: 

606 if not refresh_state: 

607 raise sa_exc.ArgumentError( 

608 "refresh_state is required when require_pk_cols is present" 

609 ) 

610 

611 refresh_state_prokeys = refresh_state.mapper._primary_key_propkeys 

612 has_changes = { 

613 key 

614 for key in refresh_state_prokeys.difference(only_load_props) 

615 if refresh_state.attrs[key].history.has_changes() 

616 } 

617 if has_changes: 

618 # raise if pending pk changes are present. 

619 # technically, this could be limited to the case where we have 

620 # relationships in the only_load_props collection to be refreshed 

621 # also (and only ones that have a secondary eager loader, at that). 

622 # however, the error is in place across the board so that behavior 

623 # here is easier to predict. The use case it prevents is one 

624 # of mutating PK attrs, leaving them unflushed, 

625 # calling session.refresh(), and expecting those attrs to remain 

626 # still unflushed. It seems likely someone doing all those 

627 # things would be better off having the PK attributes flushed 

628 # to the database before tinkering like that (session.refresh() is 

629 # tinkering). 

630 raise sa_exc.InvalidRequestError( 

631 f"Please flush pending primary key changes on " 

632 "attributes " 

633 f"{has_changes} for mapper {refresh_state.mapper} before " 

634 "proceeding with a refresh" 

635 ) 

636 

637 # overall, the ORM has no internal flow right now for "dont load the 

638 # primary row of an object at all, but fire off 

639 # selectinload/subqueryload/immediateload for some relationships". 

640 # It would probably be a pretty big effort to add such a flow. So 

641 # here, the case for #8703 is introduced; user asks to refresh some 

642 # relationship attributes only which are 

643 # selectinload/subqueryload/immediateload/ etc. (not joinedload). 

644 # ORM complains there's no columns in the primary row to load. 

645 # So here, we just add the PK cols if that 

646 # case is detected, so that there is a SELECT emitted for the primary 

647 # row. 

648 # 

649 # Let's just state right up front, for this one little case, 

650 # the ORM here is adding a whole extra SELECT just to satisfy 

651 # limitations in the internal flow. This is really not a thing 

652 # SQLAlchemy finds itself doing like, ever, obviously, we are 

653 # constantly working to *remove* SELECTs we don't need. We 

654 # rationalize this for now based on 1. session.refresh() is not 

655 # commonly used 2. session.refresh() with only relationship attrs is 

656 # even less commonly used 3. the SELECT in question is very low 

657 # latency. 

658 # 

659 # to add the flow to not include the SELECT, the quickest way 

660 # might be to just manufacture a single-row result set to send off to 

661 # instances(), but we'd have to weave that into context.py and all 

662 # that. For 2.0.0, we have enough big changes to navigate for now. 

663 # 

664 mp = refresh_state.mapper._props 

665 for p in only_load_props: 

666 if mp[p]._is_relationship: 

667 only_load_props = refresh_state_prokeys.union(only_load_props) 

668 break 

669 

670 if refresh_state and refresh_state.load_options: 

671 compile_options += {"_current_path": refresh_state.load_path.parent} 

672 q = q.options(*refresh_state.load_options) 

673 

674 new_compile_options, load_options = _set_get_options( 

675 compile_options, 

676 load_options, 

677 version_check=version_check, 

678 only_load_props=only_load_props, 

679 refresh_state=refresh_state, 

680 identity_token=identity_token, 

681 is_user_refresh=is_user_refresh, 

682 ) 

683 

684 q._compile_options = new_compile_options 

685 q._order_by = None 

686 

687 if no_autoflush: 

688 load_options += {"_autoflush": False} 

689 

690 execution_options = util.EMPTY_DICT.merge_with( 

691 execution_options, {"_sa_orm_load_options": load_options} 

692 ) 

693 result = ( 

694 session.execute( 

695 q, 

696 params=params, 

697 execution_options=execution_options, 

698 bind_arguments=bind_arguments, 

699 ) 

700 .unique() 

701 .scalars() 

702 ) 

703 

704 try: 

705 return result.one() 

706 except orm_exc.NoResultFound: 

707 return None 

708 

709 

710def _set_get_options( 

711 compile_opt, 

712 load_opt, 

713 populate_existing=None, 

714 version_check=None, 

715 only_load_props=None, 

716 refresh_state=None, 

717 identity_token=None, 

718 is_user_refresh=None, 

719): 

720 compile_options = {} 

721 load_options = {} 

722 if version_check: 

723 load_options["_version_check"] = version_check 

724 if populate_existing: 

725 load_options["_populate_existing"] = populate_existing 

726 if refresh_state: 

727 load_options["_refresh_state"] = refresh_state 

728 compile_options["_for_refresh_state"] = True 

729 if only_load_props: 

730 compile_options["_only_load_props"] = frozenset(only_load_props) 

731 if identity_token: 

732 load_options["_identity_token"] = identity_token 

733 

734 if is_user_refresh: 

735 load_options["_is_user_refresh"] = is_user_refresh 

736 if load_options: 

737 load_opt += load_options 

738 if compile_options: 

739 compile_opt += compile_options 

740 

741 return compile_opt, load_opt 

742 

743 

744def _setup_entity_query( 

745 compile_state, 

746 mapper, 

747 query_entity, 

748 path, 

749 adapter, 

750 column_collection, 

751 with_polymorphic=None, 

752 only_load_props=None, 

753 polymorphic_discriminator=None, 

754 **kw, 

755): 

756 if with_polymorphic: 

757 poly_properties = mapper._iterate_polymorphic_properties( 

758 with_polymorphic 

759 ) 

760 else: 

761 poly_properties = mapper._polymorphic_properties 

762 

763 quick_populators = {} 

764 

765 path.set(compile_state.attributes, "memoized_setups", quick_populators) 

766 

767 # for the lead entities in the path, e.g. not eager loads, and 

768 # assuming a user-passed aliased class, e.g. not a from_self() or any 

769 # implicit aliasing, don't add columns to the SELECT that aren't 

770 # in the thing that's aliased. 

771 check_for_adapt = adapter and len(path) == 1 and path[-1].is_aliased_class 

772 

773 for value in poly_properties: 

774 if only_load_props and value.key not in only_load_props: 

775 continue 

776 value.setup( 

777 compile_state, 

778 query_entity, 

779 path, 

780 adapter, 

781 only_load_props=only_load_props, 

782 column_collection=column_collection, 

783 memoized_populators=quick_populators, 

784 check_for_adapt=check_for_adapt, 

785 **kw, 

786 ) 

787 

788 if ( 

789 polymorphic_discriminator is not None 

790 and polymorphic_discriminator is not mapper.polymorphic_on 

791 ): 

792 if adapter: 

793 pd = adapter.columns[polymorphic_discriminator] 

794 else: 

795 pd = polymorphic_discriminator 

796 column_collection.append(pd) 

797 

798 

799def _warn_for_runid_changed(state): 

800 util.warn( 

801 "Loading context for %s has changed within a load/refresh " 

802 "handler, suggesting a row refresh operation took place. If this " 

803 "event handler is expected to be " 

804 "emitting row refresh operations within an existing load or refresh " 

805 "operation, set restore_load_context=True when establishing the " 

806 "listener to ensure the context remains unchanged when the event " 

807 "handler completes." % (state_str(state),) 

808 ) 

809 

810 

811def _instance_processor( 

812 query_entity, 

813 mapper, 

814 context, 

815 result, 

816 path, 

817 adapter, 

818 only_load_props=None, 

819 refresh_state=None, 

820 polymorphic_discriminator=None, 

821 _polymorphic_from=None, 

822): 

823 """Produce a mapper level row processor callable 

824 which processes rows into mapped instances.""" 

825 

826 # note that this method, most of which exists in a closure 

827 # called _instance(), resists being broken out, as 

828 # attempts to do so tend to add significant function 

829 # call overhead. _instance() is the most 

830 # performance-critical section in the whole ORM. 

831 

832 identity_class = mapper._identity_class 

833 compile_state = context.compile_state 

834 

835 # look for "row getter" functions that have been assigned along 

836 # with the compile state that were cached from a previous load. 

837 # these are operator.itemgetter() objects that each will extract a 

838 # particular column from each row. 

839 

840 getter_key = ("getters", mapper) 

841 getters = path.get(compile_state.attributes, getter_key, None) 

842 

843 if getters is None: 

844 # no getters, so go through a list of attributes we are loading for, 

845 # and the ones that are column based will have already put information 

846 # for us in another collection "memoized_setups", which represents the 

847 # output of the LoaderStrategy.setup_query() method. We can just as 

848 # easily call LoaderStrategy.create_row_processor for each, but by 

849 # getting it all at once from setup_query we save another method call 

850 # per attribute. 

851 props = mapper._prop_set 

852 if only_load_props is not None: 

853 props = props.intersection( 

854 mapper._props[k] for k in only_load_props 

855 ) 

856 

857 quick_populators = path.get( 

858 context.attributes, "memoized_setups", EMPTY_DICT 

859 ) 

860 

861 todo = [] 

862 cached_populators = { 

863 "new": [], 

864 "quick": [], 

865 "deferred": [], 

866 "expire": [], 

867 "existing": [], 

868 "eager": [], 

869 } 

870 

871 if refresh_state is None: 

872 # we can also get the "primary key" tuple getter function 

873 pk_cols = mapper.primary_key 

874 

875 if adapter: 

876 pk_cols = [adapter.columns[c] for c in pk_cols] 

877 primary_key_getter = result._tuple_getter(pk_cols) 

878 else: 

879 primary_key_getter = None 

880 

881 getters = { 

882 "cached_populators": cached_populators, 

883 "todo": todo, 

884 "primary_key_getter": primary_key_getter, 

885 } 

886 for prop in props: 

887 if prop in quick_populators: 

888 # this is an inlined path just for column-based attributes. 

889 col = quick_populators[prop] 

890 if col is _DEFER_FOR_STATE: 

891 cached_populators["new"].append( 

892 (prop.key, prop._deferred_column_loader) 

893 ) 

894 elif col is _SET_DEFERRED_EXPIRED: 

895 # note that in this path, we are no longer 

896 # searching in the result to see if the column might 

897 # be present in some unexpected way. 

898 cached_populators["expire"].append((prop.key, False)) 

899 elif col is _RAISE_FOR_STATE: 

900 cached_populators["new"].append( 

901 (prop.key, prop._raise_column_loader) 

902 ) 

903 else: 

904 getter = None 

905 if adapter: 

906 # this logic had been removed for all 1.4 releases 

907 # up until 1.4.18; the adapter here is particularly 

908 # the compound eager adapter which isn't accommodated 

909 # in the quick_populators right now. The "fallback" 

910 # logic below instead took over in many more cases 

911 # until issue #6596 was identified. 

912 

913 # note there is still an issue where this codepath 

914 # produces no "getter" for cases where a joined-inh 

915 # mapping includes a labeled column property, meaning 

916 # KeyError is caught internally and we fall back to 

917 # _getter(col), which works anyway. The adapter 

918 # here for joined inh without any aliasing might not 

919 # be useful. Tests which see this include 

920 # test.orm.inheritance.test_basic -> 

921 # EagerTargetingTest.test_adapt_stringency 

922 # OptimizedLoadTest.test_column_expression_joined 

923 # PolymorphicOnNotLocalTest.test_polymorphic_on_column_prop # noqa: E501 

924 # 

925 

926 adapted_col = adapter.columns[col] 

927 if adapted_col is not None: 

928 getter = result._getter(adapted_col, False) 

929 if not getter: 

930 getter = result._getter(col, False) 

931 if getter: 

932 cached_populators["quick"].append((prop.key, getter)) 

933 else: 

934 # fall back to the ColumnProperty itself, which 

935 # will iterate through all of its columns 

936 # to see if one fits 

937 prop.create_row_processor( 

938 context, 

939 query_entity, 

940 path, 

941 mapper, 

942 result, 

943 adapter, 

944 cached_populators, 

945 ) 

946 else: 

947 # loader strategies like subqueryload, selectinload, 

948 # joinedload, basically relationships, these need to interact 

949 # with the context each time to work correctly. 

950 todo.append(prop) 

951 

952 path.set(compile_state.attributes, getter_key, getters) 

953 

954 cached_populators = getters["cached_populators"] 

955 

956 populators = {key: list(value) for key, value in cached_populators.items()} 

957 for prop in getters["todo"]: 

958 prop.create_row_processor( 

959 context, query_entity, path, mapper, result, adapter, populators 

960 ) 

961 

962 propagated_loader_options = context.propagated_loader_options 

963 load_path = ( 

964 context.compile_state.current_path + path 

965 if context.compile_state.current_path.path 

966 else path 

967 ) 

968 

969 session_identity_map = context.session.identity_map 

970 

971 populate_existing = context.populate_existing or mapper.always_refresh 

972 load_evt = bool(mapper.class_manager.dispatch.load) 

973 refresh_evt = bool(mapper.class_manager.dispatch.refresh) 

974 persistent_evt = bool(context.session.dispatch.loaded_as_persistent) 

975 if persistent_evt: 

976 loaded_as_persistent = context.session.dispatch.loaded_as_persistent 

977 instance_state = attributes.instance_state 

978 instance_dict = attributes.instance_dict 

979 session_id = context.session.hash_key 

980 runid = context.runid 

981 identity_token = context.identity_token 

982 

983 version_check = context.version_check 

984 if version_check: 

985 version_id_col = mapper.version_id_col 

986 if version_id_col is not None: 

987 if adapter: 

988 version_id_col = adapter.columns[version_id_col] 

989 version_id_getter = result._getter(version_id_col) 

990 else: 

991 version_id_getter = None 

992 

993 if not refresh_state and _polymorphic_from is not None: 

994 key = ("loader", path.path) 

995 

996 if key in context.attributes and context.attributes[key].strategy == ( 

997 ("selectinload_polymorphic", True), 

998 ): 

999 option_entities = context.attributes[key].local_opts["entities"] 

1000 else: 

1001 option_entities = None 

1002 selectin_load_via = mapper._should_selectin_load( 

1003 option_entities, 

1004 _polymorphic_from, 

1005 ) 

1006 

1007 if selectin_load_via and selectin_load_via is not _polymorphic_from: 

1008 # only_load_props goes w/ refresh_state only, and in a refresh 

1009 # we are a single row query for the exact entity; polymorphic 

1010 # loading does not apply 

1011 assert only_load_props is None 

1012 

1013 if selectin_load_via.is_mapper: 

1014 _load_supers = [] 

1015 _endmost_mapper = selectin_load_via 

1016 while ( 

1017 _endmost_mapper 

1018 and _endmost_mapper is not _polymorphic_from 

1019 ): 

1020 _load_supers.append(_endmost_mapper) 

1021 _endmost_mapper = _endmost_mapper.inherits 

1022 else: 

1023 _load_supers = [selectin_load_via] 

1024 

1025 for _selectinload_entity in _load_supers: 

1026 if _PostLoad.path_exists( 

1027 context, load_path, _selectinload_entity 

1028 ): 

1029 continue 

1030 callable_ = _load_subclass_via_in( 

1031 context, 

1032 path, 

1033 _selectinload_entity, 

1034 _polymorphic_from, 

1035 option_entities, 

1036 ) 

1037 _PostLoad.callable_for_path( 

1038 context, 

1039 load_path, 

1040 _selectinload_entity.mapper, 

1041 _selectinload_entity, 

1042 callable_, 

1043 _selectinload_entity, 

1044 ) 

1045 

1046 post_load = _PostLoad.for_context(context, load_path, only_load_props) 

1047 

1048 if refresh_state: 

1049 refresh_identity_key = refresh_state.key 

1050 if refresh_identity_key is None: 

1051 # super-rare condition; a refresh is being called 

1052 # on a non-instance-key instance; this is meant to only 

1053 # occur within a flush() 

1054 refresh_identity_key = mapper._identity_key_from_state( 

1055 refresh_state 

1056 ) 

1057 else: 

1058 refresh_identity_key = None 

1059 

1060 primary_key_getter = getters["primary_key_getter"] 

1061 

1062 if mapper.allow_partial_pks: 

1063 is_not_primary_key = _none_set.issuperset 

1064 else: 

1065 is_not_primary_key = _none_set.intersection 

1066 

1067 def _instance(row): 

1068 # determine the state that we'll be populating 

1069 if refresh_identity_key: 

1070 # fixed state that we're refreshing 

1071 state = refresh_state 

1072 instance = state.obj() 

1073 dict_ = instance_dict(instance) 

1074 isnew = state.runid != runid 

1075 currentload = True 

1076 loaded_instance = False 

1077 else: 

1078 # look at the row, see if that identity is in the 

1079 # session, or we have to create a new one 

1080 identitykey = ( 

1081 identity_class, 

1082 primary_key_getter(row), 

1083 identity_token, 

1084 ) 

1085 

1086 instance = session_identity_map.get(identitykey) 

1087 

1088 if instance is not None: 

1089 # existing instance 

1090 state = instance_state(instance) 

1091 dict_ = instance_dict(instance) 

1092 

1093 isnew = state.runid != runid 

1094 currentload = not isnew 

1095 loaded_instance = False 

1096 

1097 if version_check and version_id_getter and not currentload: 

1098 _validate_version_id( 

1099 mapper, state, dict_, row, version_id_getter 

1100 ) 

1101 

1102 else: 

1103 # create a new instance 

1104 

1105 # check for non-NULL values in the primary key columns, 

1106 # else no entity is returned for the row 

1107 if is_not_primary_key(identitykey[1]): 

1108 return None 

1109 

1110 isnew = True 

1111 currentload = True 

1112 loaded_instance = True 

1113 

1114 instance = mapper.class_manager.new_instance() 

1115 

1116 dict_ = instance_dict(instance) 

1117 state = instance_state(instance) 

1118 state.key = identitykey 

1119 state.identity_token = identity_token 

1120 

1121 # attach instance to session. 

1122 state.session_id = session_id 

1123 session_identity_map._add_unpresent(state, identitykey) 

1124 

1125 effective_populate_existing = populate_existing 

1126 if refresh_state is state: 

1127 effective_populate_existing = True 

1128 

1129 # populate. this looks at whether this state is new 

1130 # for this load or was existing, and whether or not this 

1131 # row is the first row with this identity. 

1132 if currentload or effective_populate_existing: 

1133 # full population routines. Objects here are either 

1134 # just created, or we are doing a populate_existing 

1135 

1136 # be conservative about setting load_path when populate_existing 

1137 # is in effect; want to maintain options from the original 

1138 # load. see test_expire->test_refresh_maintains_deferred_options 

1139 if isnew and ( 

1140 propagated_loader_options or not effective_populate_existing 

1141 ): 

1142 state.load_options = propagated_loader_options 

1143 state.load_path = load_path 

1144 

1145 _populate_full( 

1146 context, 

1147 row, 

1148 state, 

1149 dict_, 

1150 isnew, 

1151 load_path, 

1152 loaded_instance, 

1153 effective_populate_existing, 

1154 populators, 

1155 ) 

1156 

1157 if isnew: 

1158 # state.runid should be equal to context.runid / runid 

1159 # here, however for event checks we are being more conservative 

1160 # and checking against existing run id 

1161 # assert state.runid == runid 

1162 

1163 existing_runid = state.runid 

1164 

1165 if loaded_instance: 

1166 if load_evt: 

1167 state.manager.dispatch.load(state, context) 

1168 if state.runid != existing_runid: 

1169 _warn_for_runid_changed(state) 

1170 if persistent_evt: 

1171 loaded_as_persistent(context.session, state) 

1172 if state.runid != existing_runid: 

1173 _warn_for_runid_changed(state) 

1174 elif refresh_evt: 

1175 state.manager.dispatch.refresh( 

1176 state, context, only_load_props 

1177 ) 

1178 if state.runid != runid: 

1179 _warn_for_runid_changed(state) 

1180 

1181 if effective_populate_existing or state.modified: 

1182 if refresh_state and only_load_props: 

1183 state._commit(dict_, only_load_props) 

1184 else: 

1185 state._commit_all(dict_, session_identity_map) 

1186 

1187 if post_load: 

1188 post_load.add_state(state, True) 

1189 

1190 else: 

1191 # partial population routines, for objects that were already 

1192 # in the Session, but a row matches them; apply eager loaders 

1193 # on existing objects, etc. 

1194 unloaded = state.unloaded 

1195 isnew = state not in context.partials 

1196 

1197 if not isnew or unloaded or populators["eager"]: 

1198 # state is having a partial set of its attributes 

1199 # refreshed. Populate those attributes, 

1200 # and add to the "context.partials" collection. 

1201 

1202 to_load = _populate_partial( 

1203 context, 

1204 row, 

1205 state, 

1206 dict_, 

1207 isnew, 

1208 load_path, 

1209 unloaded, 

1210 populators, 

1211 ) 

1212 

1213 if isnew: 

1214 if refresh_evt: 

1215 existing_runid = state.runid 

1216 state.manager.dispatch.refresh(state, context, to_load) 

1217 if state.runid != existing_runid: 

1218 _warn_for_runid_changed(state) 

1219 

1220 state._commit(dict_, to_load) 

1221 

1222 if post_load and context.invoke_all_eagers: 

1223 post_load.add_state(state, False) 

1224 

1225 return instance 

1226 

1227 if mapper.polymorphic_map and not _polymorphic_from and not refresh_state: 

1228 # if we are doing polymorphic, dispatch to a different _instance() 

1229 # method specific to the subclass mapper 

1230 def ensure_no_pk(row): 

1231 identitykey = ( 

1232 identity_class, 

1233 primary_key_getter(row), 

1234 identity_token, 

1235 ) 

1236 if not is_not_primary_key(identitykey[1]): 

1237 return identitykey 

1238 else: 

1239 return None 

1240 

1241 _instance = _decorate_polymorphic_switch( 

1242 _instance, 

1243 context, 

1244 query_entity, 

1245 mapper, 

1246 result, 

1247 path, 

1248 polymorphic_discriminator, 

1249 adapter, 

1250 ensure_no_pk, 

1251 ) 

1252 

1253 return _instance 

1254 

1255 

1256def _load_subclass_via_in( 

1257 context, path, entity, polymorphic_from, option_entities 

1258): 

1259 mapper = entity.mapper 

1260 

1261 # TODO: polymorphic_from seems to be a Mapper in all cases. 

1262 # this is likely not needed, but as we dont have typing in loading.py 

1263 # yet, err on the safe side 

1264 polymorphic_from_mapper = polymorphic_from.mapper 

1265 not_against_basemost = polymorphic_from_mapper.inherits is not None 

1266 

1267 zero_idx = len(mapper.base_mapper.primary_key) == 1 

1268 

1269 if entity.is_aliased_class or not_against_basemost: 

1270 q, enable_opt, disable_opt = mapper._subclass_load_via_in( 

1271 entity, polymorphic_from 

1272 ) 

1273 else: 

1274 q, enable_opt, disable_opt = mapper._subclass_load_via_in_mapper 

1275 

1276 def do_load(context, path, states, load_only, effective_entity): 

1277 if not option_entities: 

1278 # filter out states for those that would have selectinloaded 

1279 # from another loader 

1280 # TODO: we are currently ignoring the case where the 

1281 # "selectin_polymorphic" option is used, as this is much more 

1282 # complex / specific / very uncommon API use 

1283 states = [ 

1284 (s, v) 

1285 for s, v in states 

1286 if s.mapper._would_selectin_load_only_from_given_mapper(mapper) 

1287 ] 

1288 

1289 if not states: 

1290 return 

1291 

1292 orig_query = context.query 

1293 

1294 if path.parent: 

1295 enable_opt_lcl = enable_opt._prepend_path(path) 

1296 disable_opt_lcl = disable_opt._prepend_path(path) 

1297 else: 

1298 enable_opt_lcl = enable_opt 

1299 disable_opt_lcl = disable_opt 

1300 options = ( 

1301 (enable_opt_lcl,) + orig_query._with_options + (disable_opt_lcl,) 

1302 ) 

1303 

1304 q2 = q.options(*options) 

1305 

1306 q2._compile_options = context.compile_state.default_compile_options 

1307 q2._compile_options += {"_current_path": path.parent} 

1308 

1309 if context.populate_existing: 

1310 q2 = q2.execution_options(populate_existing=True) 

1311 

1312 context.session.execute( 

1313 q2, 

1314 dict( 

1315 primary_keys=[ 

1316 state.key[1][0] if zero_idx else state.key[1] 

1317 for state, load_attrs in states 

1318 ] 

1319 ), 

1320 ).unique().scalars().all() 

1321 

1322 return do_load 

1323 

1324 

1325def _populate_full( 

1326 context, 

1327 row, 

1328 state, 

1329 dict_, 

1330 isnew, 

1331 load_path, 

1332 loaded_instance, 

1333 populate_existing, 

1334 populators, 

1335): 

1336 if isnew: 

1337 # first time we are seeing a row with this identity. 

1338 state.runid = context.runid 

1339 

1340 for key, getter in populators["quick"]: 

1341 dict_[key] = getter(row) 

1342 if populate_existing: 

1343 for key, set_callable in populators["expire"]: 

1344 dict_.pop(key, None) 

1345 if set_callable: 

1346 state.expired_attributes.add(key) 

1347 else: 

1348 for key, set_callable in populators["expire"]: 

1349 if set_callable: 

1350 state.expired_attributes.add(key) 

1351 

1352 for key, populator in populators["new"]: 

1353 populator(state, dict_, row) 

1354 

1355 elif load_path != state.load_path: 

1356 # new load path, e.g. object is present in more than one 

1357 # column position in a series of rows 

1358 state.load_path = load_path 

1359 

1360 # if we have data, and the data isn't in the dict, OK, let's put 

1361 # it in. 

1362 for key, getter in populators["quick"]: 

1363 if key not in dict_: 

1364 dict_[key] = getter(row) 

1365 

1366 # otherwise treat like an "already seen" row 

1367 for key, populator in populators["existing"]: 

1368 populator(state, dict_, row) 

1369 # TODO: allow "existing" populator to know this is 

1370 # a new path for the state: 

1371 # populator(state, dict_, row, new_path=True) 

1372 

1373 else: 

1374 # have already seen rows with this identity in this same path. 

1375 for key, populator in populators["existing"]: 

1376 populator(state, dict_, row) 

1377 

1378 # TODO: same path 

1379 # populator(state, dict_, row, new_path=False) 

1380 

1381 

1382def _populate_partial( 

1383 context, row, state, dict_, isnew, load_path, unloaded, populators 

1384): 

1385 if not isnew: 

1386 if unloaded: 

1387 # extra pass, see #8166 

1388 for key, getter in populators["quick"]: 

1389 if key in unloaded: 

1390 dict_[key] = getter(row) 

1391 

1392 to_load = context.partials[state] 

1393 for key, populator in populators["existing"]: 

1394 if key in to_load: 

1395 populator(state, dict_, row) 

1396 else: 

1397 to_load = unloaded 

1398 context.partials[state] = to_load 

1399 

1400 for key, getter in populators["quick"]: 

1401 if key in to_load: 

1402 dict_[key] = getter(row) 

1403 for key, set_callable in populators["expire"]: 

1404 if key in to_load: 

1405 dict_.pop(key, None) 

1406 if set_callable: 

1407 state.expired_attributes.add(key) 

1408 for key, populator in populators["new"]: 

1409 if key in to_load: 

1410 populator(state, dict_, row) 

1411 

1412 for key, populator in populators["eager"]: 

1413 if key not in unloaded: 

1414 populator(state, dict_, row) 

1415 

1416 return to_load 

1417 

1418 

1419def _validate_version_id(mapper, state, dict_, row, getter): 

1420 if mapper._get_state_attr_by_column( 

1421 state, dict_, mapper.version_id_col 

1422 ) != getter(row): 

1423 raise orm_exc.StaleDataError( 

1424 "Instance '%s' has version id '%s' which " 

1425 "does not match database-loaded version id '%s'." 

1426 % ( 

1427 state_str(state), 

1428 mapper._get_state_attr_by_column( 

1429 state, dict_, mapper.version_id_col 

1430 ), 

1431 getter(row), 

1432 ) 

1433 ) 

1434 

1435 

1436def _decorate_polymorphic_switch( 

1437 instance_fn, 

1438 context, 

1439 query_entity, 

1440 mapper, 

1441 result, 

1442 path, 

1443 polymorphic_discriminator, 

1444 adapter, 

1445 ensure_no_pk, 

1446): 

1447 if polymorphic_discriminator is not None: 

1448 polymorphic_on = polymorphic_discriminator 

1449 else: 

1450 polymorphic_on = mapper.polymorphic_on 

1451 if polymorphic_on is None: 

1452 return instance_fn 

1453 

1454 if adapter: 

1455 polymorphic_on = adapter.columns[polymorphic_on] 

1456 

1457 def configure_subclass_mapper(discriminator): 

1458 try: 

1459 sub_mapper = mapper.polymorphic_map[discriminator] 

1460 except KeyError: 

1461 raise AssertionError( 

1462 "No such polymorphic_identity %r is defined" % discriminator 

1463 ) 

1464 else: 

1465 if sub_mapper is mapper: 

1466 return None 

1467 elif not sub_mapper.isa(mapper): 

1468 return False 

1469 

1470 return _instance_processor( 

1471 query_entity, 

1472 sub_mapper, 

1473 context, 

1474 result, 

1475 path, 

1476 adapter, 

1477 _polymorphic_from=mapper, 

1478 ) 

1479 

1480 polymorphic_instances = util.PopulateDict(configure_subclass_mapper) 

1481 

1482 getter = result._getter(polymorphic_on) 

1483 

1484 def polymorphic_instance(row): 

1485 discriminator = getter(row) 

1486 if discriminator is not None: 

1487 _instance = polymorphic_instances[discriminator] 

1488 if _instance: 

1489 return _instance(row) 

1490 elif _instance is False: 

1491 identitykey = ensure_no_pk(row) 

1492 

1493 if identitykey: 

1494 raise sa_exc.InvalidRequestError( 

1495 "Row with identity key %s can't be loaded into an " 

1496 "object; the polymorphic discriminator column '%s' " 

1497 "refers to %s, which is not a sub-mapper of " 

1498 "the requested %s" 

1499 % ( 

1500 identitykey, 

1501 polymorphic_on, 

1502 mapper.polymorphic_map[discriminator], 

1503 mapper, 

1504 ) 

1505 ) 

1506 else: 

1507 return None 

1508 else: 

1509 return instance_fn(row) 

1510 else: 

1511 identitykey = ensure_no_pk(row) 

1512 

1513 if identitykey: 

1514 raise sa_exc.InvalidRequestError( 

1515 "Row with identity key %s can't be loaded into an " 

1516 "object; the polymorphic discriminator column '%s' is " 

1517 "NULL" % (identitykey, polymorphic_on) 

1518 ) 

1519 else: 

1520 return None 

1521 

1522 return polymorphic_instance 

1523 

1524 

1525class _PostLoad: 

1526 """Track loaders and states for "post load" operations.""" 

1527 

1528 __slots__ = "loaders", "states", "load_keys" 

1529 

1530 def __init__(self): 

1531 self.loaders = {} 

1532 self.states = util.OrderedDict() 

1533 self.load_keys = None 

1534 

1535 def add_state(self, state, overwrite): 

1536 # the states for a polymorphic load here are all shared 

1537 # within a single PostLoad object among multiple subtypes. 

1538 # Filtering of callables on a per-subclass basis needs to be done at 

1539 # the invocation level 

1540 self.states[state] = overwrite 

1541 

1542 def invoke(self, context, path): 

1543 if not self.states: 

1544 return 

1545 path = path_registry.PathRegistry.coerce(path) 

1546 for ( 

1547 effective_context, 

1548 token, 

1549 limit_to_mapper, 

1550 loader, 

1551 arg, 

1552 kw, 

1553 ) in self.loaders.values(): 

1554 states = [ 

1555 (state, overwrite) 

1556 for state, overwrite in self.states.items() 

1557 if state.manager.mapper.isa(limit_to_mapper) 

1558 ] 

1559 if states: 

1560 loader( 

1561 effective_context, path, states, self.load_keys, *arg, **kw 

1562 ) 

1563 self.states.clear() 

1564 

1565 @classmethod 

1566 def for_context(cls, context, path, only_load_props): 

1567 pl = context.post_load_paths.get(path.path) 

1568 if pl is not None and only_load_props: 

1569 pl.load_keys = only_load_props 

1570 return pl 

1571 

1572 @classmethod 

1573 def path_exists(self, context, path, key): 

1574 return ( 

1575 path.path in context.post_load_paths 

1576 and key in context.post_load_paths[path.path].loaders 

1577 ) 

1578 

1579 @classmethod 

1580 def callable_for_path( 

1581 cls, context, path, limit_to_mapper, token, loader_callable, *arg, **kw 

1582 ): 

1583 if path.path in context.post_load_paths: 

1584 pl = context.post_load_paths[path.path] 

1585 else: 

1586 pl = context.post_load_paths[path.path] = _PostLoad() 

1587 pl.loaders[token] = ( 

1588 context, 

1589 token, 

1590 limit_to_mapper, 

1591 loader_callable, 

1592 arg, 

1593 kw, 

1594 ) 

1595 

1596 

1597def _load_scalar_attributes(mapper, state, attribute_names, passive): 

1598 """initiate a column-based attribute refresh operation.""" 

1599 

1600 # assert mapper is _state_mapper(state) 

1601 session = state.session 

1602 if not session: 

1603 raise orm_exc.DetachedInstanceError( 

1604 "Instance %s is not bound to a Session; " 

1605 "attribute refresh operation cannot proceed" % (state_str(state)) 

1606 ) 

1607 

1608 no_autoflush = bool(passive & attributes.NO_AUTOFLUSH) 

1609 

1610 # in the case of inheritance, particularly concrete and abstract 

1611 # concrete inheritance, the class manager might have some keys 

1612 # of attributes on the superclass that we didn't actually map. 

1613 # These could be mapped as "concrete, don't load" or could be completely 

1614 # excluded from the mapping and we know nothing about them. Filter them 

1615 # here to prevent them from coming through. 

1616 if attribute_names: 

1617 attribute_names = attribute_names.intersection(mapper.attrs.keys()) 

1618 

1619 if mapper.inherits and not mapper.concrete: 

1620 # load based on committed attributes in the object, formed into 

1621 # a truncated SELECT that only includes relevant tables. does not 

1622 # currently use state.key 

1623 statement = mapper._optimized_get_statement(state, attribute_names) 

1624 if statement is not None: 

1625 # undefer() isn't needed here because statement has the 

1626 # columns needed already, this implicitly undefers that column 

1627 stmt = FromStatement(mapper, statement) 

1628 

1629 return _load_on_ident( 

1630 session, 

1631 stmt, 

1632 None, 

1633 only_load_props=attribute_names, 

1634 refresh_state=state, 

1635 no_autoflush=no_autoflush, 

1636 ) 

1637 

1638 # normal load, use state.key as the identity to SELECT 

1639 has_key = bool(state.key) 

1640 

1641 if has_key: 

1642 identity_key = state.key 

1643 else: 

1644 # this codepath is rare - only valid when inside a flush, and the 

1645 # object is becoming persistent but hasn't yet been assigned 

1646 # an identity_key. 

1647 # check here to ensure we have the attrs we need. 

1648 pk_attrs = [ 

1649 mapper._columntoproperty[col].key for col in mapper.primary_key 

1650 ] 

1651 if state.expired_attributes.intersection(pk_attrs): 

1652 raise sa_exc.InvalidRequestError( 

1653 "Instance %s cannot be refreshed - it's not " 

1654 " persistent and does not " 

1655 "contain a full primary key." % state_str(state) 

1656 ) 

1657 identity_key = mapper._identity_key_from_state(state) 

1658 

1659 if ( 

1660 _none_set.issubset(identity_key) and not mapper.allow_partial_pks 

1661 ) or _none_set.issuperset(identity_key): 

1662 util.warn_limited( 

1663 "Instance %s to be refreshed doesn't " 

1664 "contain a full primary key - can't be refreshed " 

1665 "(and shouldn't be expired, either).", 

1666 state_str(state), 

1667 ) 

1668 return 

1669 

1670 result = _load_on_ident( 

1671 session, 

1672 select(mapper).set_label_style(LABEL_STYLE_TABLENAME_PLUS_COL), 

1673 identity_key, 

1674 refresh_state=state, 

1675 only_load_props=attribute_names, 

1676 no_autoflush=no_autoflush, 

1677 ) 

1678 

1679 # if instance is pending, a refresh operation 

1680 # may not complete (even if PK attributes are assigned) 

1681 if has_key and result is None: 

1682 raise orm_exc.ObjectDeletedError(state)