Coverage for python/lsst/daf/butler/registry/datasets/byDimensions/_storage.py: 95%
241 statements
« prev ^ index » next coverage.py v7.2.5, created at 2023-05-06 09:31 +0000
« prev ^ index » next coverage.py v7.2.5, created at 2023-05-06 09:31 +0000
1# This file is part of daf_butler.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <http://www.gnu.org/licenses/>.
23from __future__ import annotations
25__all__ = ("ByDimensionsDatasetRecordStorage",)
27import uuid
28from collections.abc import Iterable, Iterator, Sequence, Set
29from datetime import datetime
30from typing import TYPE_CHECKING
32import astropy.time
33import sqlalchemy
34from lsst.daf.relation import Relation, sql
36from ....core import (
37 DataCoordinate,
38 DatasetColumnTag,
39 DatasetId,
40 DatasetIdFactory,
41 DatasetIdGenEnum,
42 DatasetRef,
43 DatasetType,
44 DimensionKeyColumnTag,
45 LogicalColumn,
46 Timespan,
47 ddl,
48)
49from ..._collection_summary import CollectionSummary
50from ..._collectionType import CollectionType
51from ..._exceptions import CollectionTypeError, ConflictingDefinitionError
52from ...interfaces import DatasetRecordStorage
53from ...queries import SqlQueryContext
54from .tables import makeTagTableSpec
56if TYPE_CHECKING:
57 from ...interfaces import CollectionManager, CollectionRecord, Database, RunRecord
58 from .summaries import CollectionSummaryManager
59 from .tables import StaticDatasetTablesTuple
62class ByDimensionsDatasetRecordStorage(DatasetRecordStorage):
63 """Dataset record storage implementation paired with
64 `ByDimensionsDatasetRecordStorageManagerUUID`; see that class for more
65 information.
67 Instances of this class should never be constructed directly; use
68 `DatasetRecordStorageManager.register` instead.
69 """
71 def __init__(
72 self,
73 *,
74 datasetType: DatasetType,
75 db: Database,
76 dataset_type_id: int,
77 collections: CollectionManager,
78 static: StaticDatasetTablesTuple,
79 summaries: CollectionSummaryManager,
80 tags: sqlalchemy.schema.Table,
81 use_astropy_ingest_date: bool,
82 calibs: sqlalchemy.schema.Table | None,
83 ):
84 super().__init__(datasetType=datasetType)
85 self._dataset_type_id = dataset_type_id
86 self._db = db
87 self._collections = collections
88 self._static = static
89 self._summaries = summaries
90 self._tags = tags
91 self._calibs = calibs
92 self._runKeyColumn = collections.getRunForeignKeyName()
93 self._use_astropy = use_astropy_ingest_date
95 def delete(self, datasets: Iterable[DatasetRef]) -> None:
96 # Docstring inherited from DatasetRecordStorage.
97 # Only delete from common dataset table; ON DELETE foreign key clauses
98 # will handle the rest.
99 self._db.delete(
100 self._static.dataset,
101 ["id"],
102 *[{"id": dataset.getCheckedId()} for dataset in datasets],
103 )
105 def associate(self, collection: CollectionRecord, datasets: Iterable[DatasetRef]) -> None:
106 # Docstring inherited from DatasetRecordStorage.
107 if collection.type is not CollectionType.TAGGED: 107 ↛ 108line 107 didn't jump to line 108, because the condition on line 107 was never true
108 raise TypeError(
109 f"Cannot associate into collection '{collection.name}' "
110 f"of type {collection.type.name}; must be TAGGED."
111 )
112 protoRow = {
113 self._collections.getCollectionForeignKeyName(): collection.key,
114 "dataset_type_id": self._dataset_type_id,
115 }
116 rows = []
117 summary = CollectionSummary()
118 for dataset in summary.add_datasets_generator(datasets):
119 row = dict(protoRow, dataset_id=dataset.getCheckedId())
120 for dimension, value in dataset.dataId.items():
121 row[dimension.name] = value
122 rows.append(row)
123 # Update the summary tables for this collection in case this is the
124 # first time this dataset type or these governor values will be
125 # inserted there.
126 self._summaries.update(collection, [self._dataset_type_id], summary)
127 # Update the tag table itself.
128 self._db.replace(self._tags, *rows)
130 def disassociate(self, collection: CollectionRecord, datasets: Iterable[DatasetRef]) -> None:
131 # Docstring inherited from DatasetRecordStorage.
132 if collection.type is not CollectionType.TAGGED: 132 ↛ 133line 132 didn't jump to line 133, because the condition on line 132 was never true
133 raise TypeError(
134 f"Cannot disassociate from collection '{collection.name}' "
135 f"of type {collection.type.name}; must be TAGGED."
136 )
137 rows = [
138 {
139 "dataset_id": dataset.getCheckedId(),
140 self._collections.getCollectionForeignKeyName(): collection.key,
141 }
142 for dataset in datasets
143 ]
144 self._db.delete(self._tags, ["dataset_id", self._collections.getCollectionForeignKeyName()], *rows)
146 def _buildCalibOverlapQuery(
147 self,
148 collection: CollectionRecord,
149 data_ids: set[DataCoordinate] | None,
150 timespan: Timespan,
151 context: SqlQueryContext,
152 ) -> Relation:
153 relation = self.make_relation(
154 collection, columns={"timespan", "dataset_id", "calib_pkey"}, context=context
155 ).with_rows_satisfying(
156 context.make_timespan_overlap_predicate(
157 DatasetColumnTag(self.datasetType.name, "timespan"), timespan
158 ),
159 )
160 if data_ids is not None:
161 relation = relation.join(
162 context.make_data_id_relation(
163 data_ids, self.datasetType.dimensions.required.names
164 ).transferred_to(context.sql_engine),
165 )
166 return relation
168 def certify(
169 self,
170 collection: CollectionRecord,
171 datasets: Iterable[DatasetRef],
172 timespan: Timespan,
173 context: SqlQueryContext,
174 ) -> None:
175 # Docstring inherited from DatasetRecordStorage.
176 if self._calibs is None: 176 ↛ 177line 176 didn't jump to line 177, because the condition on line 176 was never true
177 raise CollectionTypeError(
178 f"Cannot certify datasets of type {self.datasetType.name}, for which "
179 "DatasetType.isCalibration() is False."
180 )
181 if collection.type is not CollectionType.CALIBRATION: 181 ↛ 182line 181 didn't jump to line 182, because the condition on line 181 was never true
182 raise CollectionTypeError(
183 f"Cannot certify into collection '{collection.name}' "
184 f"of type {collection.type.name}; must be CALIBRATION."
185 )
186 TimespanReprClass = self._db.getTimespanRepresentation()
187 protoRow = {
188 self._collections.getCollectionForeignKeyName(): collection.key,
189 "dataset_type_id": self._dataset_type_id,
190 }
191 rows = []
192 dataIds: set[DataCoordinate] | None = (
193 set() if not TimespanReprClass.hasExclusionConstraint() else None
194 )
195 summary = CollectionSummary()
196 for dataset in summary.add_datasets_generator(datasets):
197 row = dict(protoRow, dataset_id=dataset.getCheckedId())
198 for dimension, value in dataset.dataId.items():
199 row[dimension.name] = value
200 TimespanReprClass.update(timespan, result=row)
201 rows.append(row)
202 if dataIds is not None: 202 ↛ 196line 202 didn't jump to line 196, because the condition on line 202 was never false
203 dataIds.add(dataset.dataId)
204 # Update the summary tables for this collection in case this is the
205 # first time this dataset type or these governor values will be
206 # inserted there.
207 self._summaries.update(collection, [self._dataset_type_id], summary)
208 # Update the association table itself.
209 if TimespanReprClass.hasExclusionConstraint(): 209 ↛ 212line 209 didn't jump to line 212, because the condition on line 209 was never true
210 # Rely on database constraint to enforce invariants; we just
211 # reraise the exception for consistency across DB engines.
212 try:
213 self._db.insert(self._calibs, *rows)
214 except sqlalchemy.exc.IntegrityError as err:
215 raise ConflictingDefinitionError(
216 f"Validity range conflict certifying datasets of type {self.datasetType.name} "
217 f"into {collection.name} for range [{timespan.begin}, {timespan.end})."
218 ) from err
219 else:
220 # Have to implement exclusion constraint ourselves.
221 # Start by building a SELECT query for any rows that would overlap
222 # this one.
223 relation = self._buildCalibOverlapQuery(collection, dataIds, timespan, context)
224 # Acquire a table lock to ensure there are no concurrent writes
225 # could invalidate our checking before we finish the inserts. We
226 # use a SAVEPOINT in case there is an outer transaction that a
227 # failure here should not roll back.
228 with self._db.transaction(lock=[self._calibs], savepoint=True):
229 # Enter SqlQueryContext in case we need to use a temporary
230 # table to include the give data IDs in the query. Note that
231 # by doing this inside the transaction, we make sure it doesn't
232 # attempt to close the session when its done, since it just
233 # sees an already-open session that it knows it shouldn't
234 # manage.
235 with context:
236 # Run the check SELECT query.
237 conflicting = context.count(context.process(relation))
238 if conflicting > 0:
239 raise ConflictingDefinitionError(
240 f"{conflicting} validity range conflicts certifying datasets of type "
241 f"{self.datasetType.name} into {collection.name} for range "
242 f"[{timespan.begin}, {timespan.end})."
243 )
244 # Proceed with the insert.
245 self._db.insert(self._calibs, *rows)
247 def decertify(
248 self,
249 collection: CollectionRecord,
250 timespan: Timespan,
251 *,
252 dataIds: Iterable[DataCoordinate] | None = None,
253 context: SqlQueryContext,
254 ) -> None:
255 # Docstring inherited from DatasetRecordStorage.
256 if self._calibs is None: 256 ↛ 257line 256 didn't jump to line 257, because the condition on line 256 was never true
257 raise CollectionTypeError(
258 f"Cannot decertify datasets of type {self.datasetType.name}, for which "
259 "DatasetType.isCalibration() is False."
260 )
261 if collection.type is not CollectionType.CALIBRATION: 261 ↛ 262line 261 didn't jump to line 262, because the condition on line 261 was never true
262 raise CollectionTypeError(
263 f"Cannot decertify from collection '{collection.name}' "
264 f"of type {collection.type.name}; must be CALIBRATION."
265 )
266 TimespanReprClass = self._db.getTimespanRepresentation()
267 # Construct a SELECT query to find all rows that overlap our inputs.
268 dataIdSet: set[DataCoordinate] | None
269 if dataIds is not None:
270 dataIdSet = set(dataIds)
271 else:
272 dataIdSet = None
273 relation = self._buildCalibOverlapQuery(collection, dataIdSet, timespan, context)
274 calib_pkey_tag = DatasetColumnTag(self.datasetType.name, "calib_pkey")
275 dataset_id_tag = DatasetColumnTag(self.datasetType.name, "dataset_id")
276 timespan_tag = DatasetColumnTag(self.datasetType.name, "timespan")
277 data_id_tags = [
278 (name, DimensionKeyColumnTag(name)) for name in self.datasetType.dimensions.required.names
279 ]
280 # Set up collections to populate with the rows we'll want to modify.
281 # The insert rows will have the same values for collection and
282 # dataset type.
283 protoInsertRow = {
284 self._collections.getCollectionForeignKeyName(): collection.key,
285 "dataset_type_id": self._dataset_type_id,
286 }
287 rowsToDelete = []
288 rowsToInsert = []
289 # Acquire a table lock to ensure there are no concurrent writes
290 # between the SELECT and the DELETE and INSERT queries based on it.
291 with self._db.transaction(lock=[self._calibs], savepoint=True):
292 # Enter SqlQueryContext in case we need to use a temporary table to
293 # include the give data IDs in the query (see similar block in
294 # certify for details).
295 with context:
296 for row in context.fetch_iterable(relation):
297 rowsToDelete.append({"id": row[calib_pkey_tag]})
298 # Construct the insert row(s) by copying the prototype row,
299 # then adding the dimension column values, then adding
300 # what's left of the timespan from that row after we
301 # subtract the given timespan.
302 newInsertRow = protoInsertRow.copy()
303 newInsertRow["dataset_id"] = row[dataset_id_tag]
304 for name, tag in data_id_tags:
305 newInsertRow[name] = row[tag]
306 rowTimespan = row[timespan_tag]
307 assert rowTimespan is not None, "Field should have a NOT NULL constraint."
308 for diffTimespan in rowTimespan.difference(timespan):
309 rowsToInsert.append(
310 TimespanReprClass.update(diffTimespan, result=newInsertRow.copy())
311 )
312 # Run the DELETE and INSERT queries.
313 self._db.delete(self._calibs, ["id"], *rowsToDelete)
314 self._db.insert(self._calibs, *rowsToInsert)
316 def make_relation(
317 self,
318 *collections: CollectionRecord,
319 columns: Set[str],
320 context: SqlQueryContext,
321 ) -> Relation:
322 # Docstring inherited from DatasetRecordStorage.
323 collection_types = {collection.type for collection in collections}
324 assert CollectionType.CHAINED not in collection_types, "CHAINED collections must be flattened."
325 TimespanReprClass = self._db.getTimespanRepresentation()
326 #
327 # There are two kinds of table in play here:
328 #
329 # - the static dataset table (with the dataset ID, dataset type ID,
330 # run ID/name, and ingest date);
331 #
332 # - the dynamic tags/calibs table (with the dataset ID, dataset type
333 # type ID, collection ID/name, data ID, and possibly validity
334 # range).
335 #
336 # That means that we might want to return a query against either table
337 # or a JOIN of both, depending on which quantities the caller wants.
338 # But the data ID is always included, which means we'll always include
339 # the tags/calibs table and join in the static dataset table only if we
340 # need things from it that we can't get from the tags/calibs table.
341 #
342 # Note that it's important that we include a WHERE constraint on both
343 # tables for any column (e.g. dataset_type_id) that is in both when
344 # it's given explicitly; not doing can prevent the query planner from
345 # using very important indexes. At present, we don't include those
346 # redundant columns in the JOIN ON expression, however, because the
347 # FOREIGN KEY (and its index) are defined only on dataset_id.
348 tag_relation: Relation | None = None
349 calib_relation: Relation | None = None
350 if collection_types != {CollectionType.CALIBRATION}:
351 # We'll need a subquery for the tags table if any of the given
352 # collections are not a CALIBRATION collection. This intentionally
353 # also fires when the list of collections is empty as a way to
354 # create a dummy subquery that we know will fail.
355 # We give the table an alias because it might appear multiple times
356 # in the same query, for different dataset types.
357 tags_parts = sql.Payload[LogicalColumn](self._tags.alias(f"{self.datasetType.name}_tags"))
358 if "timespan" in columns:
359 tags_parts.columns_available[
360 DatasetColumnTag(self.datasetType.name, "timespan")
361 ] = TimespanReprClass.fromLiteral(Timespan(None, None))
362 tag_relation = self._finish_single_relation(
363 tags_parts,
364 columns,
365 [
366 (record, rank)
367 for rank, record in enumerate(collections)
368 if record.type is not CollectionType.CALIBRATION
369 ],
370 context,
371 )
372 assert "calib_pkey" not in columns, "For internal use only, and only for pure-calib queries."
373 if CollectionType.CALIBRATION in collection_types:
374 # If at least one collection is a CALIBRATION collection, we'll
375 # need a subquery for the calibs table, and could include the
376 # timespan as a result or constraint.
377 assert (
378 self._calibs is not None
379 ), "DatasetTypes with isCalibration() == False can never be found in a CALIBRATION collection."
380 calibs_parts = sql.Payload[LogicalColumn](self._calibs.alias(f"{self.datasetType.name}_calibs"))
381 if "timespan" in columns:
382 calibs_parts.columns_available[
383 DatasetColumnTag(self.datasetType.name, "timespan")
384 ] = TimespanReprClass.from_columns(calibs_parts.from_clause.columns)
385 if "calib_pkey" in columns:
386 # This is a private extension not included in the base class
387 # interface, for internal use only in _buildCalibOverlapQuery,
388 # which needs access to the autoincrement primary key for the
389 # calib association table.
390 calibs_parts.columns_available[
391 DatasetColumnTag(self.datasetType.name, "calib_pkey")
392 ] = calibs_parts.from_clause.columns.id
393 calib_relation = self._finish_single_relation(
394 calibs_parts,
395 columns,
396 [
397 (record, rank)
398 for rank, record in enumerate(collections)
399 if record.type is CollectionType.CALIBRATION
400 ],
401 context,
402 )
403 if tag_relation is not None:
404 if calib_relation is not None:
405 # daf_relation's chain operation does not automatically
406 # deduplicate; it's more like SQL's UNION ALL. To get UNION
407 # in SQL here, we add an explicit deduplication.
408 return tag_relation.chain(calib_relation).without_duplicates()
409 else:
410 return tag_relation
411 elif calib_relation is not None:
412 return calib_relation
413 else:
414 raise AssertionError("Branch should be unreachable.")
416 def _finish_single_relation(
417 self,
418 payload: sql.Payload[LogicalColumn],
419 requested_columns: Set[str],
420 collections: Sequence[tuple[CollectionRecord, int]],
421 context: SqlQueryContext,
422 ) -> Relation:
423 """Helper method for `make_relation`.
425 This handles adding columns and WHERE terms that are not specific to
426 either the tags or calibs tables.
428 Parameters
429 ----------
430 payload : `lsst.daf.relation.sql.Payload`
431 SQL query parts under construction, to be modified in-place and
432 used to construct the new relation.
433 requested_columns : `~collections.abc.Set` [ `str` ]
434 Columns the relation should include.
435 collections : `Sequence` [ `tuple` [ `CollectionRecord`, `int` ] ]
436 Collections to search for the dataset and their ranks.
437 context : `SqlQueryContext`
438 Context that manages engines and state for the query.
440 Returns
441 -------
442 relation : `lsst.daf.relation.Relation`
443 New dataset query relation.
444 """
445 payload.where.append(payload.from_clause.columns.dataset_type_id == self._dataset_type_id)
446 dataset_id_col = payload.from_clause.columns.dataset_id
447 collection_col = payload.from_clause.columns[self._collections.getCollectionForeignKeyName()]
448 # We always constrain and optionally retrieve the collection(s) via the
449 # tags/calibs table.
450 if len(collections) == 1:
451 payload.where.append(collection_col == collections[0][0].key)
452 if "collection" in requested_columns:
453 payload.columns_available[
454 DatasetColumnTag(self.datasetType.name, "collection")
455 ] = sqlalchemy.sql.literal(collections[0][0].key)
456 else:
457 assert collections, "The no-collections case should be in calling code for better diagnostics."
458 payload.where.append(collection_col.in_([collection.key for collection, _ in collections]))
459 if "collection" in requested_columns:
460 payload.columns_available[
461 DatasetColumnTag(self.datasetType.name, "collection")
462 ] = collection_col
463 # Add rank if requested as a CASE-based calculation the collection
464 # column.
465 if "rank" in requested_columns:
466 payload.columns_available[DatasetColumnTag(self.datasetType.name, "rank")] = sqlalchemy.sql.case(
467 {record.key: rank for record, rank in collections},
468 value=collection_col,
469 )
470 # Add more column definitions, starting with the data ID.
471 for dimension_name in self.datasetType.dimensions.required.names:
472 payload.columns_available[DimensionKeyColumnTag(dimension_name)] = payload.from_clause.columns[
473 dimension_name
474 ]
475 # We can always get the dataset_id from the tags/calibs table.
476 if "dataset_id" in requested_columns:
477 payload.columns_available[DatasetColumnTag(self.datasetType.name, "dataset_id")] = dataset_id_col
478 # It's possible we now have everything we need, from just the
479 # tags/calibs table. The things we might need to get from the static
480 # dataset table are the run key and the ingest date.
481 need_static_table = False
482 if "run" in requested_columns:
483 if len(collections) == 1 and collections[0][0].type is CollectionType.RUN:
484 # If we are searching exactly one RUN collection, we
485 # know that if we find the dataset in that collection,
486 # then that's the datasets's run; we don't need to
487 # query for it.
488 payload.columns_available[
489 DatasetColumnTag(self.datasetType.name, "run")
490 ] = sqlalchemy.sql.literal(collections[0][0].key)
491 else:
492 payload.columns_available[
493 DatasetColumnTag(self.datasetType.name, "run")
494 ] = self._static.dataset.columns[self._runKeyColumn]
495 need_static_table = True
496 # Ingest date can only come from the static table.
497 if "ingest_date" in requested_columns:
498 need_static_table = True
499 payload.columns_available[
500 DatasetColumnTag(self.datasetType.name, "ingest_date")
501 ] = self._static.dataset.columns.ingest_date
502 # If we need the static table, join it in via dataset_id and
503 # dataset_type_id
504 if need_static_table:
505 payload.from_clause = payload.from_clause.join(
506 self._static.dataset, onclause=(dataset_id_col == self._static.dataset.columns.id)
507 )
508 # Also constrain dataset_type_id in static table in case that helps
509 # generate a better plan.
510 # We could also include this in the JOIN ON clause, but my guess is
511 # that that's a good idea IFF it's in the foreign key, and right
512 # now it isn't.
513 payload.where.append(self._static.dataset.columns.dataset_type_id == self._dataset_type_id)
514 leaf = context.sql_engine.make_leaf(
515 payload.columns_available.keys(),
516 payload=payload,
517 name=self.datasetType.name,
518 parameters={record.name: rank for record, rank in collections},
519 )
520 return leaf
522 def getDataId(self, id: DatasetId) -> DataCoordinate:
523 """Return DataId for a dataset.
525 Parameters
526 ----------
527 id : `DatasetId`
528 Unique dataset identifier.
530 Returns
531 -------
532 dataId : `DataCoordinate`
533 DataId for the dataset.
534 """
535 # This query could return multiple rows (one for each tagged collection
536 # the dataset is in, plus one for its run collection), and we don't
537 # care which of those we get.
538 sql = (
539 self._tags.select()
540 .where(
541 sqlalchemy.sql.and_(
542 self._tags.columns.dataset_id == id,
543 self._tags.columns.dataset_type_id == self._dataset_type_id,
544 )
545 )
546 .limit(1)
547 )
548 with self._db.query(sql) as sql_result:
549 row = sql_result.mappings().fetchone()
550 assert row is not None, "Should be guaranteed by caller and foreign key constraints."
551 return DataCoordinate.standardize(
552 {dimension.name: row[dimension.name] for dimension in self.datasetType.dimensions.required},
553 graph=self.datasetType.dimensions,
554 )
557class ByDimensionsDatasetRecordStorageUUID(ByDimensionsDatasetRecordStorage):
558 """Implementation of ByDimensionsDatasetRecordStorage which uses UUID for
559 dataset IDs.
560 """
562 idMaker = DatasetIdFactory()
563 """Factory for dataset IDs. In the future this factory may be shared with
564 other classes (e.g. Registry)."""
566 def insert(
567 self,
568 run: RunRecord,
569 dataIds: Iterable[DataCoordinate],
570 idMode: DatasetIdGenEnum = DatasetIdGenEnum.UNIQUE,
571 ) -> Iterator[DatasetRef]:
572 # Docstring inherited from DatasetRecordStorage.
574 # Current timestamp, type depends on schema version. Use microsecond
575 # precision for astropy time to keep things consistent with
576 # TIMESTAMP(6) SQL type.
577 timestamp: datetime | astropy.time.Time
578 if self._use_astropy:
579 # Astropy `now()` precision should be the same as `utcnow()` which
580 # should mean microsecond.
581 timestamp = astropy.time.Time.now()
582 else:
583 timestamp = datetime.utcnow()
585 # Iterate over data IDs, transforming a possibly-single-pass iterable
586 # into a list.
587 dataIdList = []
588 rows = []
589 summary = CollectionSummary()
590 for dataId in summary.add_data_ids_generator(self.datasetType, dataIds):
591 dataIdList.append(dataId)
592 rows.append(
593 {
594 "id": self.idMaker.makeDatasetId(run.name, self.datasetType, dataId, idMode),
595 "dataset_type_id": self._dataset_type_id,
596 self._runKeyColumn: run.key,
597 "ingest_date": timestamp,
598 }
599 )
601 with self._db.transaction():
602 # Insert into the static dataset table.
603 self._db.insert(self._static.dataset, *rows)
604 # Update the summary tables for this collection in case this is the
605 # first time this dataset type or these governor values will be
606 # inserted there.
607 self._summaries.update(run, [self._dataset_type_id], summary)
608 # Combine the generated dataset_id values and data ID fields to
609 # form rows to be inserted into the tags table.
610 protoTagsRow = {
611 "dataset_type_id": self._dataset_type_id,
612 self._collections.getCollectionForeignKeyName(): run.key,
613 }
614 tagsRows = [
615 dict(protoTagsRow, dataset_id=row["id"], **dataId.byName())
616 for dataId, row in zip(dataIdList, rows)
617 ]
618 # Insert those rows into the tags table.
619 self._db.insert(self._tags, *tagsRows)
621 for dataId, row in zip(dataIdList, rows):
622 yield DatasetRef(
623 datasetType=self.datasetType,
624 dataId=dataId,
625 id=row["id"],
626 run=run.name,
627 )
629 def import_(
630 self,
631 run: RunRecord,
632 datasets: Iterable[DatasetRef],
633 idGenerationMode: DatasetIdGenEnum = DatasetIdGenEnum.UNIQUE,
634 reuseIds: bool = False,
635 ) -> Iterator[DatasetRef]:
636 # Docstring inherited from DatasetRecordStorage.
638 # Current timestamp, type depends on schema version.
639 if self._use_astropy:
640 # Astropy `now()` precision should be the same as `utcnow()` which
641 # should mean microsecond.
642 timestamp = sqlalchemy.sql.literal(astropy.time.Time.now(), type_=ddl.AstropyTimeNsecTai)
643 else:
644 timestamp = sqlalchemy.sql.literal(datetime.utcnow())
646 # Iterate over data IDs, transforming a possibly-single-pass iterable
647 # into a list.
648 dataIds = {}
649 summary = CollectionSummary()
650 for dataset in summary.add_datasets_generator(datasets):
651 # Ignore unknown ID types, normally all IDs have the same type but
652 # this code supports mixed types or missing IDs.
653 datasetId = dataset.id if isinstance(dataset.id, uuid.UUID) else None
654 if datasetId is None:
655 datasetId = self.idMaker.makeDatasetId(
656 run.name, self.datasetType, dataset.dataId, idGenerationMode
657 )
658 dataIds[datasetId] = dataset.dataId
660 # We'll insert all new rows into a temporary table
661 tableSpec = makeTagTableSpec(self.datasetType, type(self._collections), ddl.GUID, constraints=False)
662 collFkName = self._collections.getCollectionForeignKeyName()
663 protoTagsRow = {
664 "dataset_type_id": self._dataset_type_id,
665 collFkName: run.key,
666 }
667 tmpRows = [
668 dict(protoTagsRow, dataset_id=dataset_id, **dataId.byName())
669 for dataset_id, dataId in dataIds.items()
670 ]
671 with self._db.transaction(for_temp_tables=True):
672 with self._db.temporary_table(tableSpec) as tmp_tags:
673 # store all incoming data in a temporary table
674 self._db.insert(tmp_tags, *tmpRows)
676 # There are some checks that we want to make for consistency
677 # of the new datasets with existing ones.
678 self._validateImport(tmp_tags, run)
680 # Before we merge temporary table into dataset/tags we need to
681 # drop datasets which are already there (and do not conflict).
682 self._db.deleteWhere(
683 tmp_tags,
684 tmp_tags.columns.dataset_id.in_(sqlalchemy.sql.select(self._static.dataset.columns.id)),
685 )
687 # Copy it into dataset table, need to re-label some columns.
688 self._db.insert(
689 self._static.dataset,
690 select=sqlalchemy.sql.select(
691 tmp_tags.columns.dataset_id.label("id"),
692 tmp_tags.columns.dataset_type_id,
693 tmp_tags.columns[collFkName].label(self._runKeyColumn),
694 timestamp.label("ingest_date"),
695 ),
696 )
698 # Update the summary tables for this collection in case this
699 # is the first time this dataset type or these governor values
700 # will be inserted there.
701 self._summaries.update(run, [self._dataset_type_id], summary)
703 # Copy it into tags table.
704 self._db.insert(self._tags, select=tmp_tags.select())
706 # Return refs in the same order as in the input list.
707 for dataset_id, dataId in dataIds.items():
708 yield DatasetRef(
709 datasetType=self.datasetType,
710 id=dataset_id,
711 dataId=dataId,
712 run=run.name,
713 )
715 def _validateImport(self, tmp_tags: sqlalchemy.schema.Table, run: RunRecord) -> None:
716 """Validate imported refs against existing datasets.
718 Parameters
719 ----------
720 tmp_tags : `sqlalchemy.schema.Table`
721 Temporary table with new datasets and the same schema as tags
722 table.
723 run : `RunRecord`
724 The record object describing the `~CollectionType.RUN` collection.
726 Raises
727 ------
728 ConflictingDefinitionError
729 Raise if new datasets conflict with existing ones.
730 """
731 dataset = self._static.dataset
732 tags = self._tags
733 collFkName = self._collections.getCollectionForeignKeyName()
735 # Check that existing datasets have the same dataset type and
736 # run.
737 query = (
738 sqlalchemy.sql.select(
739 dataset.columns.id.label("dataset_id"),
740 dataset.columns.dataset_type_id.label("dataset_type_id"),
741 tmp_tags.columns.dataset_type_id.label("new dataset_type_id"),
742 dataset.columns[self._runKeyColumn].label("run"),
743 tmp_tags.columns[collFkName].label("new run"),
744 )
745 .select_from(dataset.join(tmp_tags, dataset.columns.id == tmp_tags.columns.dataset_id))
746 .where(
747 sqlalchemy.sql.or_(
748 dataset.columns.dataset_type_id != tmp_tags.columns.dataset_type_id,
749 dataset.columns[self._runKeyColumn] != tmp_tags.columns[collFkName],
750 )
751 )
752 .limit(1)
753 )
754 with self._db.query(query) as result:
755 if (row := result.first()) is not None:
756 # Only include the first one in the exception message
757 raise ConflictingDefinitionError(
758 f"Existing dataset type or run do not match new dataset: {row._asdict()}"
759 )
761 # Check that matching dataset in tags table has the same DataId.
762 query = (
763 sqlalchemy.sql.select(
764 tags.columns.dataset_id,
765 tags.columns.dataset_type_id.label("type_id"),
766 tmp_tags.columns.dataset_type_id.label("new type_id"),
767 *[tags.columns[dim] for dim in self.datasetType.dimensions.required.names],
768 *[
769 tmp_tags.columns[dim].label(f"new {dim}")
770 for dim in self.datasetType.dimensions.required.names
771 ],
772 )
773 .select_from(tags.join(tmp_tags, tags.columns.dataset_id == tmp_tags.columns.dataset_id))
774 .where(
775 sqlalchemy.sql.or_(
776 tags.columns.dataset_type_id != tmp_tags.columns.dataset_type_id,
777 *[
778 tags.columns[dim] != tmp_tags.columns[dim]
779 for dim in self.datasetType.dimensions.required.names
780 ],
781 )
782 )
783 .limit(1)
784 )
786 with self._db.query(query) as result:
787 if (row := result.first()) is not None:
788 # Only include the first one in the exception message
789 raise ConflictingDefinitionError(
790 f"Existing dataset type or dataId do not match new dataset: {row._asdict()}"
791 )
793 # Check that matching run+dataId have the same dataset ID.
794 query = (
795 sqlalchemy.sql.select(
796 tags.columns.dataset_type_id.label("dataset_type_id"),
797 *[tags.columns[dim] for dim in self.datasetType.dimensions.required.names],
798 tags.columns.dataset_id,
799 tmp_tags.columns.dataset_id.label("new dataset_id"),
800 tags.columns[collFkName],
801 tmp_tags.columns[collFkName].label(f"new {collFkName}"),
802 )
803 .select_from(
804 tags.join(
805 tmp_tags,
806 sqlalchemy.sql.and_(
807 tags.columns.dataset_type_id == tmp_tags.columns.dataset_type_id,
808 tags.columns[collFkName] == tmp_tags.columns[collFkName],
809 *[
810 tags.columns[dim] == tmp_tags.columns[dim]
811 for dim in self.datasetType.dimensions.required.names
812 ],
813 ),
814 )
815 )
816 .where(tags.columns.dataset_id != tmp_tags.columns.dataset_id)
817 .limit(1)
818 )
819 with self._db.query(query) as result:
820 if (row := result.first()) is not None:
821 # only include the first one in the exception message
822 raise ConflictingDefinitionError(
823 f"Existing dataset type and dataId does not match new dataset: {row._asdict()}"
824 )