Coverage for python/lsst/daf/butler/registry/queries/_sql_query_backend.py: 16%
102 statements
« prev ^ index » next coverage.py v6.5.0, created at 2023-02-01 10:04 +0000
« prev ^ index » next coverage.py v6.5.0, created at 2023-02-01 10:04 +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__ = ("SqlQueryBackend",)
25from collections.abc import Iterable, Mapping, Sequence, Set
26from typing import TYPE_CHECKING, Any, cast
28from lsst.daf.relation import ColumnError, ColumnExpression, ColumnTag, Join, Predicate, Relation
30from ...core import (
31 ColumnCategorization,
32 DataCoordinate,
33 DatasetType,
34 DimensionGraph,
35 DimensionKeyColumnTag,
36 DimensionRecord,
37 DimensionRecordColumnTag,
38 DimensionUniverse,
39 SkyPixDimension,
40)
41from .._collectionType import CollectionType
42from .._exceptions import DataIdValueError
43from ..interfaces import CollectionRecord, Database
44from ._query_backend import QueryBackend
45from ._sql_query_context import SqlQueryContext
47if TYPE_CHECKING: 47 ↛ 48line 47 didn't jump to line 48, because the condition on line 47 was never true
48 from ..managers import RegistryManagerInstances
51class SqlQueryBackend(QueryBackend[SqlQueryContext]):
52 """An implementation of `QueryBackend` for `SqlRegistry`.
54 Parameters
55 ----------
56 db : `Database`
57 Object that abstracts the database engine.
58 managers : `RegistryManagerInstances`
59 Struct containing the manager objects that back a `SqlRegistry`.
60 """
62 def __init__(
63 self,
64 db: Database,
65 managers: RegistryManagerInstances,
66 ):
67 self._db = db
68 self._managers = managers
70 @property
71 def universe(self) -> DimensionUniverse:
72 # Docstring inherited.
73 return self._managers.dimensions.universe
75 def context(self) -> SqlQueryContext:
76 # Docstring inherited.
77 return SqlQueryContext(self._db, self._managers.column_types)
79 def get_collection_name(self, key: Any) -> str:
80 return self._managers.collections[key].name
82 def resolve_collection_wildcard(
83 self,
84 expression: Any,
85 *,
86 collection_types: Set[CollectionType] = CollectionType.all(),
87 done: set[str] | None = None,
88 flatten_chains: bool = True,
89 include_chains: bool | None = None,
90 ) -> list[CollectionRecord]:
91 # Docstring inherited.
92 return self._managers.collections.resolve_wildcard(
93 expression,
94 collection_types=collection_types,
95 done=done,
96 flatten_chains=flatten_chains,
97 include_chains=include_chains,
98 )
100 def resolve_dataset_type_wildcard(
101 self,
102 expression: Any,
103 components: bool | None = None,
104 missing: list[str] | None = None,
105 explicit_only: bool = False,
106 components_deprecated: bool = True,
107 ) -> dict[DatasetType, list[str | None]]:
108 # Docstring inherited.
109 return self._managers.datasets.resolve_wildcard(
110 expression, components, missing, explicit_only, components_deprecated
111 )
113 def filter_dataset_collections(
114 self,
115 dataset_types: Iterable[DatasetType],
116 collections: Sequence[CollectionRecord],
117 *,
118 governor_constraints: Mapping[str, Set[str]],
119 rejections: list[str] | None = None,
120 ) -> dict[DatasetType, list[CollectionRecord]]:
121 # Docstring inherited.
122 result: dict[DatasetType, list[CollectionRecord]] = {
123 dataset_type: [] for dataset_type in dataset_types
124 }
125 for dataset_type, filtered_collections in result.items():
126 for collection_record in collections:
127 if not dataset_type.isCalibration() and collection_record.type is CollectionType.CALIBRATION:
128 if rejections is not None:
129 rejections.append(
130 f"Not searching for non-calibration dataset of type {dataset_type.name!r} "
131 f"in CALIBRATION collection {collection_record.name!r}."
132 )
133 else:
134 collection_summary = self._managers.datasets.getCollectionSummary(collection_record)
135 if collection_summary.is_compatible_with(
136 dataset_type,
137 governor_constraints,
138 rejections=rejections,
139 name=collection_record.name,
140 ):
141 filtered_collections.append(collection_record)
142 return result
144 def make_dataset_query_relation(
145 self,
146 dataset_type: DatasetType,
147 collections: Sequence[CollectionRecord],
148 columns: Set[str],
149 context: SqlQueryContext,
150 ) -> Relation:
151 # Docstring inherited.
152 assert len(collections) > 0, (
153 "Caller is responsible for handling the case of all collections being rejected (we can't "
154 "write a good error message without knowing why collections were rejected)."
155 )
156 dataset_storage = self._managers.datasets.find(dataset_type.name)
157 if dataset_storage is None:
158 # Unrecognized dataset type means no results.
159 return self.make_doomed_dataset_relation(
160 dataset_type,
161 columns,
162 messages=[
163 f"Dataset type {dataset_type.name!r} is not registered, "
164 "so no instances of it can exist in any collection."
165 ],
166 context=context,
167 )
168 else:
169 return dataset_storage.make_relation(
170 *collections,
171 columns=columns,
172 context=context,
173 )
175 def make_dimension_relation(
176 self,
177 dimensions: DimensionGraph,
178 columns: Set[ColumnTag],
179 context: SqlQueryContext,
180 *,
181 initial_relation: Relation | None = None,
182 initial_join_max_columns: frozenset[ColumnTag] | None = None,
183 initial_dimension_relationships: Set[frozenset[str]] | None = None,
184 spatial_joins: Iterable[tuple[str, str]] = (),
185 governor_constraints: Mapping[str, Set[str]],
186 ) -> Relation:
187 # Docstring inherited.
189 default_join = Join(max_columns=initial_join_max_columns)
191 # Set up the relation variable we'll update as we join more relations
192 # in, and ensure it is in the SQL engine.
193 relation = context.make_initial_relation(initial_relation)
195 if initial_dimension_relationships is None:
196 initial_dimension_relationships = self.extract_dimension_relationships(relation)
198 # Make a mutable copy of the columns argument.
199 columns_required = set(columns)
201 # Next we'll handle spatial joins, since those can require refinement
202 # predicates that will need region columns to be included in the
203 # relations we'll join.
204 predicate: Predicate = Predicate.literal(True)
205 for element1, element2 in spatial_joins:
206 overlaps, needs_refinement = self._managers.dimensions.make_spatial_join_relation(
207 element1, element2, context=context, governor_constraints=governor_constraints
208 )
209 if needs_refinement:
210 predicate = predicate.logical_and(
211 context.make_spatial_region_overlap_predicate(
212 ColumnExpression.reference(DimensionRecordColumnTag(element1, "region")),
213 ColumnExpression.reference(DimensionRecordColumnTag(element2, "region")),
214 )
215 )
216 columns_required.add(DimensionRecordColumnTag(element1, "region"))
217 columns_required.add(DimensionRecordColumnTag(element2, "region"))
218 relation = relation.join(overlaps)
220 # All skypix columns need to come from either the initial_relation or a
221 # spatial join, since we need all dimension key columns present in the
222 # SQL engine and skypix regions are added by postprocessing in the
223 # native iteration engine.
224 for dimension in dimensions:
225 if DimensionKeyColumnTag(dimension.name) not in relation.columns and isinstance(
226 dimension, SkyPixDimension
227 ):
228 raise NotImplementedError(
229 f"Cannot construct query involving skypix dimension {dimension.name} unless "
230 "it is part of a dataset subquery, spatial join, or other initial relation."
231 )
233 # Categorize columns not yet included in the relation to associate them
234 # with dimension elements and detect bad inputs.
235 missing_columns = ColumnCategorization.from_iterable(columns_required - relation.columns)
236 if not (missing_columns.dimension_keys <= dimensions.names):
237 raise ColumnError(
238 "Cannot add dimension key column(s) "
239 f"{{{', '.join(name for name in missing_columns.dimension_keys)}}} "
240 f"that were not included in the given dimensions {dimensions}."
241 )
242 if missing_columns.datasets:
243 raise ColumnError(
244 f"Unexpected dataset columns {missing_columns.datasets} in call to make_dimension_relation; "
245 "use make_dataset_query_relation or make_dataset_search relation instead, or filter them "
246 "out if they have already been added or will be added later."
247 )
248 for element_name in missing_columns.dimension_records.keys():
249 if element_name not in dimensions.elements.names:
250 raise ColumnError(
251 f"Cannot join dimension element {element_name} whose dimensions are not a "
252 f"subset of {dimensions}."
253 )
255 # Iterate over all dimension elements whose relations definitely have
256 # to be joined in. The order doesn't matter as long as we can assume
257 # the database query optimizer is going to try to reorder them anyway.
258 for element in dimensions.elements:
259 columns_still_needed = missing_columns.dimension_records[element.name]
260 # Two separate conditions in play here:
261 # - if we need a record column (not just key columns) from this
262 # element, we have to join in its relation;
263 # - if the element establishes a relationship between key columns
264 # that wasn't already established by the initial relation, we
265 # always join that element's relation. Any element with
266 # implied dependencies or the alwaysJoin flag establishes such a
267 # relationship.
268 if columns_still_needed or (
269 (element.alwaysJoin or element.implied)
270 and frozenset(element.dimensions.names) not in initial_dimension_relationships
271 ):
272 storage = self._managers.dimensions[element]
273 relation = storage.join(relation, default_join, context)
274 # At this point we've joined in all of the element relations that
275 # definitely need to be included, but we may not have all of the
276 # dimension key columns in the query that we want. To fill out that
277 # set, we iterate over just the given DimensionGraph's dimensions (not
278 # all dimension *elements*) in reverse topological order. That order
279 # should reduce the total number of tables we bring in, since each
280 # dimension will bring in keys for its required dependencies before we
281 # get to those required dependencies.
282 for dimension in self.universe.sorted(dimensions, reverse=True):
283 if DimensionKeyColumnTag(dimension.name) not in relation.columns:
284 storage = self._managers.dimensions[dimension]
285 relation = storage.join(relation, default_join, context)
287 # Add the predicates we constructed earlier, with a transfer to native
288 # iteration first if necessary.
289 if not predicate.as_trivial():
290 relation = relation.with_rows_satisfying(
291 predicate, preferred_engine=context.iteration_engine, transfer=True
292 )
294 # Finally project the new relation down to just the columns in the
295 # initial relation, the dimension key columns, and the new columns
296 # requested.
297 columns_kept = set(columns)
298 if initial_relation is not None:
299 columns_kept.update(initial_relation.columns)
300 columns_kept.update(DimensionKeyColumnTag.generate(dimensions.names))
301 relation = relation.with_only_columns(columns_kept, preferred_engine=context.preferred_engine)
303 return relation
305 def resolve_governor_constraints(
306 self, dimensions: DimensionGraph, constraints: Mapping[str, Set[str]], context: SqlQueryContext
307 ) -> Mapping[str, Set[str]]:
308 # Docstring inherited.
309 result: dict[str, Set[str]] = {}
310 for dimension in dimensions.governors:
311 storage = self._managers.dimensions[dimension]
312 records = storage.get_record_cache(context)
313 assert records is not None, "Governor dimensions are always cached."
314 all_values = {cast(str, data_id[dimension.name]) for data_id in records.keys()}
315 if (constraint_values := constraints.get(dimension.name)) is not None:
316 if not (constraint_values <= all_values):
317 raise DataIdValueError(
318 f"Unknown values specified for governor dimension {dimension.name}: "
319 f"{constraint_values - all_values}."
320 )
321 result[dimension.name] = constraint_values
322 else:
323 result[dimension.name] = all_values
324 return result
326 def get_dimension_record_cache(
327 self,
328 element_name: str,
329 context: SqlQueryContext,
330 ) -> Mapping[DataCoordinate, DimensionRecord] | None:
331 return self._managers.dimensions[element_name].get_record_cache(context)