Coverage for python/lsst/daf/butler/registry/bridge/monolithic.py: 23%
99 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-14 09:11 +0000
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-14 09:11 +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/>.
21from __future__ import annotations
23__all__ = ("MonolithicDatastoreRegistryBridgeManager", "MonolithicDatastoreRegistryBridge")
25import copy
26from collections import namedtuple
27from collections.abc import Iterable, Iterator
28from contextlib import contextmanager
29from typing import TYPE_CHECKING, cast
31import sqlalchemy
33from ...core import NamedValueSet, StoredDatastoreItemInfo, ddl
34from ..interfaces import (
35 DatasetIdRef,
36 DatastoreRegistryBridge,
37 DatastoreRegistryBridgeManager,
38 FakeDatasetRef,
39 OpaqueTableStorage,
40 VersionTuple,
41)
42from ..opaque import ByNameOpaqueTableStorage
43from .ephemeral import EphemeralDatastoreRegistryBridge
45if TYPE_CHECKING:
46 from ...core import DimensionUniverse
47 from ...core.datastore import DatastoreTransaction
48 from ..interfaces import (
49 Database,
50 DatasetRecordStorageManager,
51 OpaqueTableStorageManager,
52 StaticTablesContext,
53 )
55_TablesTuple = namedtuple(
56 "_TablesTuple",
57 [
58 "dataset_location",
59 "dataset_location_trash",
60 ],
61)
63# This has to be updated on every schema change
64_VERSION = VersionTuple(0, 2, 0)
67def _makeTableSpecs(datasets: type[DatasetRecordStorageManager]) -> _TablesTuple:
68 """Construct specifications for tables used by the monolithic datastore
69 bridge classes.
71 Parameters
72 ----------
73 universe : `DimensionUniverse`
74 All dimensions known to the `Registry`.
75 datasets : subclass of `DatasetRecordStorageManager`
76 Manager class for datasets; used only to create foreign key fields.
78 Returns
79 -------
80 specs : `_TablesTuple`
81 A named tuple containing `ddl.TableSpec` instances.
82 """
83 # We want the dataset_location and dataset_location_trash tables
84 # to have the same definition, aside from the behavior of their link
85 # to the dataset table: the trash table has no foreign key constraint.
86 dataset_location_spec = ddl.TableSpec(
87 doc=(
88 "A table that provides information on whether a dataset is stored in "
89 "one or more Datastores. The presence or absence of a record in this "
90 "table itself indicates whether the dataset is present in that "
91 "Datastore. "
92 ),
93 fields=NamedValueSet(
94 [
95 ddl.FieldSpec(
96 name="datastore_name",
97 dtype=sqlalchemy.String,
98 length=256,
99 primaryKey=True,
100 nullable=False,
101 doc="Name of the Datastore this entry corresponds to.",
102 ),
103 ]
104 ),
105 )
106 dataset_location = copy.deepcopy(dataset_location_spec)
107 datasets.addDatasetForeignKey(dataset_location, primaryKey=True)
108 dataset_location_trash = copy.deepcopy(dataset_location_spec)
109 datasets.addDatasetForeignKey(dataset_location_trash, primaryKey=True, constraint=False)
110 return _TablesTuple(
111 dataset_location=dataset_location,
112 dataset_location_trash=dataset_location_trash,
113 )
116class MonolithicDatastoreRegistryBridge(DatastoreRegistryBridge):
117 """An implementation of `DatastoreRegistryBridge` that uses the same two
118 tables for all non-ephemeral datastores.
120 Parameters
121 ----------
122 datastoreName : `str`
123 Name of the `Datastore` as it should appear in `Registry` tables
124 referencing it.
125 db : `Database`
126 Object providing a database connection and generic distractions.
127 tables : `_TablesTuple`
128 Named tuple containing `sqlalchemy.schema.Table` instances.
129 """
131 def __init__(self, datastoreName: str, *, db: Database, tables: _TablesTuple):
132 super().__init__(datastoreName)
133 self._db = db
134 self._tables = tables
136 def _refsToRows(self, refs: Iterable[DatasetIdRef]) -> list[dict]:
137 """Transform an iterable of `DatasetRef` or `FakeDatasetRef` objects to
138 a list of dictionaries that match the schema of the tables used by this
139 class.
141 Parameters
142 ----------
143 refs : `~collections.abc.Iterable` [ `DatasetRef` or `FakeDatasetRef` ]
144 Datasets to transform.
146 Returns
147 -------
148 rows : `list` [ `dict` ]
149 List of dictionaries, with "datastoreName" and "dataset_id" keys.
150 """
151 return [{"datastore_name": self.datastoreName, "dataset_id": ref.id} for ref in refs]
153 def insert(self, refs: Iterable[DatasetIdRef]) -> None:
154 # Docstring inherited from DatastoreRegistryBridge
155 self._db.insert(self._tables.dataset_location, *self._refsToRows(refs))
157 def forget(self, refs: Iterable[DatasetIdRef]) -> None:
158 # Docstring inherited from DatastoreRegistryBridge
159 rows = self._refsToRows(self.check(refs))
160 self._db.delete(self._tables.dataset_location, ["datastore_name", "dataset_id"], *rows)
162 def moveToTrash(self, refs: Iterable[DatasetIdRef], transaction: DatastoreTransaction | None) -> None:
163 # Docstring inherited from DatastoreRegistryBridge
164 # TODO: avoid self.check() call via queries like
165 # INSERT INTO dataset_location_trash
166 # SELECT datastore_name, dataset_id FROM dataset_location
167 # WHERE datastore_name=? AND dataset_id IN (?);
168 # DELETE FROM dataset_location
169 # WHERE datastore_name=? AND dataset_id IN (?);
170 # ...but the Database interface doesn't support those kinds of queries
171 # right now.
172 rows = self._refsToRows(self.check(refs))
173 with self._db.transaction():
174 self._db.delete(self._tables.dataset_location, ["datastore_name", "dataset_id"], *rows)
175 self._db.insert(self._tables.dataset_location_trash, *rows)
177 def check(self, refs: Iterable[DatasetIdRef]) -> Iterable[DatasetIdRef]:
178 # Docstring inherited from DatastoreRegistryBridge
179 byId = {ref.id: ref for ref in refs}
180 sql = (
181 sqlalchemy.sql.select(self._tables.dataset_location.columns.dataset_id)
182 .select_from(self._tables.dataset_location)
183 .where(
184 sqlalchemy.sql.and_(
185 self._tables.dataset_location.columns.datastore_name == self.datastoreName,
186 self._tables.dataset_location.columns.dataset_id.in_(byId.keys()),
187 )
188 )
189 )
190 with self._db.query(sql) as sql_result:
191 sql_rows = sql_result.fetchall()
192 for row in sql_rows:
193 yield byId[row.dataset_id]
195 @contextmanager
196 def emptyTrash(
197 self,
198 records_table: OpaqueTableStorage | None = None,
199 record_class: type[StoredDatastoreItemInfo] | None = None,
200 record_column: str | None = None,
201 ) -> Iterator[tuple[Iterable[tuple[DatasetIdRef, StoredDatastoreItemInfo | None]], set[str] | None]]:
202 # Docstring inherited from DatastoreRegistryBridge
204 if records_table is None:
205 raise ValueError("This implementation requires a records table.")
207 assert isinstance(
208 records_table, ByNameOpaqueTableStorage
209 ), f"Records table must support hidden attributes. Got {type(records_table)}."
211 if record_class is None:
212 raise ValueError("Record class must be provided if records table is given.")
214 # Helper closure to generate the common join+where clause.
215 def join_records(
216 select: sqlalchemy.sql.Select, location_table: sqlalchemy.schema.Table
217 ) -> sqlalchemy.sql.Select:
218 # mypy needs to be sure
219 assert isinstance(records_table, ByNameOpaqueTableStorage)
220 return select.select_from(
221 records_table._table.join(
222 location_table,
223 onclause=records_table._table.columns.dataset_id == location_table.columns.dataset_id,
224 )
225 ).where(location_table.columns.datastore_name == self.datastoreName)
227 # SELECT records.dataset_id, records.path FROM records
228 # JOIN records on dataset_location.dataset_id == records.dataset_id
229 # WHERE dataset_location.datastore_name = datastoreName
231 # It's possible that we may end up with a ref listed in the trash
232 # table that is not listed in the records table. Such an
233 # inconsistency would be missed by this query.
234 info_in_trash = join_records(records_table._table.select(), self._tables.dataset_location_trash)
236 # Run query, transform results into a list of dicts that we can later
237 # use to delete.
238 with self._db.query(info_in_trash) as sql_result:
239 rows = [dict(row, datastore_name=self.datastoreName) for row in sql_result.mappings()]
241 # It is possible for trashed refs to be linked to artifacts that
242 # are still associated with refs that are not to be trashed. We
243 # need to be careful to consider those and indicate to the caller
244 # that those artifacts should be retained. Can only do this check
245 # if the caller provides a column name that can map to multiple
246 # refs.
247 preserved: set[str] | None = None
248 if record_column is not None:
249 # Some helper subqueries
250 items_not_in_trash = join_records(
251 sqlalchemy.sql.select(records_table._table.columns[record_column]),
252 self._tables.dataset_location,
253 ).alias("items_not_in_trash")
254 items_in_trash = join_records(
255 sqlalchemy.sql.select(records_table._table.columns[record_column]),
256 self._tables.dataset_location_trash,
257 ).alias("items_in_trash")
259 # A query for paths that are referenced by datasets in the trash
260 # and datasets not in the trash.
261 items_to_preserve = sqlalchemy.sql.select(items_in_trash.columns[record_column]).select_from(
262 items_not_in_trash.join(
263 items_in_trash,
264 onclause=items_in_trash.columns[record_column]
265 == items_not_in_trash.columns[record_column],
266 )
267 )
268 with self._db.query(items_to_preserve) as sql_result:
269 preserved = {row[record_column] for row in sql_result.mappings()}
271 # Convert results to a tuple of id+info and a record of the artifacts
272 # that should not be deleted from datastore. The id+info tuple is
273 # solely to allow logging to report the relevant ID.
274 id_info = ((FakeDatasetRef(row["dataset_id"]), record_class.from_record(row)) for row in rows)
276 # Start contextmanager, return results
277 yield ((id_info, preserved))
279 # No exception raised in context manager block.
280 if not rows:
281 return
283 # Delete the rows from the records table
284 records_table.delete(["dataset_id"], *[{"dataset_id": row["dataset_id"]} for row in rows])
286 # Delete those rows from the trash table.
287 self._db.delete(
288 self._tables.dataset_location_trash,
289 ["dataset_id", "datastore_name"],
290 *[{"dataset_id": row["dataset_id"], "datastore_name": row["datastore_name"]} for row in rows],
291 )
294class MonolithicDatastoreRegistryBridgeManager(DatastoreRegistryBridgeManager):
295 """An implementation of `DatastoreRegistryBridgeManager` that uses the same
296 two tables for all non-ephemeral datastores.
298 Parameters
299 ----------
300 db : `Database`
301 Object providing a database connection and generic distractions.
302 tables : `_TablesTuple`
303 Named tuple containing `sqlalchemy.schema.Table` instances.
304 opaque : `OpaqueTableStorageManager`
305 Manager object for opaque table storage in the `Registry`.
306 universe : `DimensionUniverse`
307 All dimensions know to the `Registry`.
308 datasetIdColumnType : `type`
309 Type for dataset ID column.
310 """
312 def __init__(
313 self,
314 *,
315 db: Database,
316 tables: _TablesTuple,
317 opaque: OpaqueTableStorageManager,
318 universe: DimensionUniverse,
319 datasetIdColumnType: type,
320 registry_schema_version: VersionTuple | None = None,
321 ):
322 super().__init__(
323 opaque=opaque,
324 universe=universe,
325 datasetIdColumnType=datasetIdColumnType,
326 registry_schema_version=registry_schema_version,
327 )
328 self._db = db
329 self._tables = tables
330 self._ephemeral: dict[str, EphemeralDatastoreRegistryBridge] = {}
332 @classmethod
333 def initialize(
334 cls,
335 db: Database,
336 context: StaticTablesContext,
337 *,
338 opaque: OpaqueTableStorageManager,
339 datasets: type[DatasetRecordStorageManager],
340 universe: DimensionUniverse,
341 registry_schema_version: VersionTuple | None = None,
342 ) -> DatastoreRegistryBridgeManager:
343 # Docstring inherited from DatastoreRegistryBridge
344 tables = context.addTableTuple(_makeTableSpecs(datasets))
345 return cls(
346 db=db,
347 tables=cast(_TablesTuple, tables),
348 opaque=opaque,
349 universe=universe,
350 datasetIdColumnType=datasets.getIdColumnType(),
351 registry_schema_version=registry_schema_version,
352 )
354 def refresh(self) -> None:
355 # Docstring inherited from DatastoreRegistryBridge
356 # This implementation has no in-Python state that depends on which
357 # datastores exist, so there's nothing to do.
358 pass
360 def register(self, name: str, *, ephemeral: bool = False) -> DatastoreRegistryBridge:
361 # Docstring inherited from DatastoreRegistryBridge
362 if ephemeral:
363 return self._ephemeral.setdefault(name, EphemeralDatastoreRegistryBridge(name))
364 return MonolithicDatastoreRegistryBridge(name, db=self._db, tables=self._tables)
366 def findDatastores(self, ref: DatasetIdRef) -> Iterable[str]:
367 # Docstring inherited from DatastoreRegistryBridge
368 sql = (
369 sqlalchemy.sql.select(self._tables.dataset_location.columns.datastore_name)
370 .select_from(self._tables.dataset_location)
371 .where(self._tables.dataset_location.columns.dataset_id == ref.id)
372 )
373 with self._db.query(sql) as sql_result:
374 sql_rows = sql_result.mappings().fetchall()
375 for row in sql_rows:
376 yield row[self._tables.dataset_location.columns.datastore_name]
377 for name, bridge in self._ephemeral.items():
378 if ref in bridge:
379 yield name
381 @classmethod
382 def currentVersions(cls) -> list[VersionTuple]:
383 # Docstring inherited from VersionedExtension.
384 return [_VERSION]