Coverage for python/lsst/daf/butler/registry/queries/_structs.py : 35%

Hot-keys on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
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__ = ["QuerySummary", "RegistryManagers"] # other classes here are local to subpackage
25from dataclasses import dataclass
26from typing import AbstractSet, Iterator, List, Optional, Union
28from sqlalchemy.sql import ColumnElement
30from lsst.sphgeom import Region
31from ...core import (
32 TimespanDatabaseRepresentation,
33 DataCoordinate,
34 DatasetType,
35 Dimension,
36 DimensionElement,
37 DimensionGraph,
38 DimensionUniverse,
39 NamedKeyDict,
40 NamedKeyMapping,
41 NamedValueAbstractSet,
42 NamedValueSet,
43 SkyPixDimension,
44)
45from ...core.utils import cached_getter, immutable
46from ..interfaces import (
47 CollectionManager,
48 DatasetRecordStorageManager,
49 DimensionRecordStorageManager,
50)
51from ..wildcards import GovernorDimensionRestriction
52# We're not trying to add typing to the lex/yacc parser code, so MyPy
53# doesn't know about some of these imports.
54from .exprParser import Node, NormalForm, NormalFormExpression, ParserYacc # type: ignore
57@immutable
58class QueryWhereExpression:
59 """A struct representing a parsed user-provided WHERE expression.
61 Parameters
62 ----------
63 expression : `str`, optional
64 The string expression to parse. If `None`, a where expression that
65 always evaluates to `True` is implied.
66 """
67 def __init__(self, expression: Optional[str] = None):
68 if expression:
69 try:
70 parser = ParserYacc()
71 self._tree = parser.parse(expression)
72 except Exception as exc:
73 raise RuntimeError(f"Failed to parse user expression `{expression}'.") from exc
74 assert self._tree is not None
75 else:
76 self._tree = None
78 def attach(
79 self,
80 graph: DimensionGraph,
81 dataId: Optional[DataCoordinate] = None,
82 region: Optional[Region] = None,
83 check: bool = True,
84 ) -> QueryWhereClause:
85 """Allow this expression to be attached to a `QuerySummary` by
86 transforming it into a `QueryWhereClause`, while checking it for both
87 internal consistency and consistency with the rest of the query.
89 Parameters
90 ----------
91 graph : `DimensionGraph`
92 The dimensions the query would include in the absence of this
93 WHERE expression.
94 dataId : `DataCoordinate`, optional
95 A fully-expanded data ID identifying dimensions known in advance.
96 If not provided, will be set to an empty data ID.
97 ``dataId.hasRecords()`` must return `True`.
98 region : `lsst.sphgeom.Region`, optional
99 A spatial region that all rows must overlap. If `None` and
100 ``dataId`` is not `None`, ``dataId.region`` will be used.
101 check : `bool`
102 If `True` (default) check the query for consistency. This may
103 reject some valid queries that resemble common mistakes (e.g.
104 queries for visits without specifying an instrument).
105 """
106 if region is None and dataId is not None:
107 region = dataId.region
108 if dataId is None:
109 dataId = DataCoordinate.makeEmpty(graph.universe)
110 restriction = GovernorDimensionRestriction(graph.universe)
111 summary: InspectionSummary
112 if self._tree is not None:
113 if check:
114 # Convert the expression to disjunctive normal form (ORs of
115 # ANDs). That's potentially super expensive in the general
116 # case (where there's a ton of nesting of ANDs and ORs). That
117 # won't be the case for the expressions we expect, and we
118 # actually use disjunctive normal instead of conjunctive (i.e.
119 # ANDs of ORs) because I think the worst-case is a long list
120 # of OR'd-together data IDs, which is already in or very close
121 # to disjunctive normal form.
122 expr = NormalFormExpression.fromTree(self._tree, NormalForm.DISJUNCTIVE)
123 from .expressions import CheckVisitor
124 # Check the expression for consistency and completeness.
125 try:
126 summary = expr.visit(CheckVisitor(dataId, graph))
127 except RuntimeError as err:
128 exprOriginal = str(self._tree)
129 exprNormal = str(expr.toTree())
130 if exprNormal == exprOriginal:
131 msg = f'Error in query expression "{exprOriginal}": {err}'
132 else:
133 msg = (
134 f'Error in query expression "{exprOriginal}" '
135 f'(normalized to "{exprNormal}"): {err}'
136 )
137 raise RuntimeError(msg) from None
138 restriction = GovernorDimensionRestriction(
139 graph.universe,
140 **summary.governors.byName(),
141 )
142 else:
143 from .expressions import InspectionVisitor
144 summary = self._tree.visit(InspectionVisitor(graph.universe))
145 else:
146 from .expressions import InspectionSummary
147 summary = InspectionSummary()
148 return QueryWhereClause(
149 self._tree,
150 dataId,
151 dimensions=summary.dimensions,
152 columns=summary.columns,
153 restriction=restriction,
154 region=region,
155 )
158@dataclass(frozen=True)
159class QueryWhereClause:
160 """Structure holding various contributions to a query's WHERE clause.
162 Instances of this class should only be created by
163 `QueryWhereExpression.attach`, which guarantees the consistency of its
164 attributes.
165 """
167 tree: Optional[Node]
168 """A parsed string expression tree., or `None` if there was no string
169 expression.
170 """
172 dataId: DataCoordinate
173 """A data ID identifying dimensions known before query construction
174 (`DataCoordinate`).
176 ``dataId.hasRecords()`` is guaranteed to return `True`.
177 """
179 dimensions: NamedValueAbstractSet[Dimension]
180 """Dimensions whose primary keys or dependencies were referenced anywhere
181 in the string expression (`NamedValueAbstractSet` [ `Dimension` ]).
182 """
184 columns: NamedKeyMapping[DimensionElement, AbstractSet[str]]
185 """Dimension element tables whose non-key columns were referenced anywhere
186 in the string expression
187 (`NamedKeyMapping` [ `DimensionElement`, `Set` [ `str` ] ]).
188 """
190 region: Optional[Region]
191 """A spatial region that all result rows must overlap
192 (`lsst.sphgeom.Region` or `None`).
193 """
195 restriction: GovernorDimensionRestriction
196 """Restrictions on the values governor dimensions can take in this query,
197 imposed by the string expression or data ID
198 (`GovernorDimensionRestriction`).
199 """
202@immutable
203class QuerySummary:
204 """A struct that holds and categorizes the dimensions involved in a query.
206 A `QuerySummary` instance is necessary to construct a `QueryBuilder`, and
207 it needs to include all of the dimensions that will be included in the
208 query (including any needed for querying datasets).
210 Parameters
211 ----------
212 requested : `DimensionGraph`
213 The dimensions whose primary keys should be included in the result rows
214 of the query.
215 dataId : `DataCoordinate`, optional
216 A fully-expanded data ID identifying dimensions known in advance. If
217 not provided, will be set to an empty data ID. ``dataId.hasRecords()``
218 must return `True`.
219 expression : `str` or `QueryWhereExpression`, optional
220 A user-provided string WHERE expression.
221 whereRegion : `lsst.sphgeom.Region`, optional
222 A spatial region that all rows must overlap. If `None` and ``dataId``
223 is not `None`, ``dataId.region`` will be used.
224 check : `bool`
225 If `True` (default) check the query for consistency. This may reject
226 some valid queries that resemble common mistakes (e.g. queries for
227 visits without specifying an instrument).
228 """
229 def __init__(self, requested: DimensionGraph, *,
230 dataId: Optional[DataCoordinate] = None,
231 expression: Optional[Union[str, QueryWhereExpression]] = None,
232 whereRegion: Optional[Region] = None,
233 check: bool = True):
234 self.requested = requested
235 if expression is None:
236 expression = QueryWhereExpression(None)
237 elif isinstance(expression, str):
238 expression = QueryWhereExpression(expression)
239 self.where = expression.attach(self.requested, dataId=dataId, region=whereRegion, check=check)
241 requested: DimensionGraph
242 """Dimensions whose primary keys should be included in the result rows of
243 the query (`DimensionGraph`).
244 """
246 where: QueryWhereClause
247 """Structure containing objects that contribute to the WHERE clause of the
248 query (`QueryWhereClause`).
249 """
251 @property
252 def universe(self) -> DimensionUniverse:
253 """All known dimensions (`DimensionUniverse`).
254 """
255 return self.requested.universe
257 @property # type: ignore
258 @cached_getter
259 def spatial(self) -> NamedValueSet[DimensionElement]:
260 """Dimension elements whose regions and skypix IDs should be included
261 in the query (`NamedValueSet` of `DimensionElement`).
262 """
263 # An element may participate spatially in the query if:
264 # - it's the most precise spatial element for its system in the
265 # requested dimensions (i.e. in `self.requested.spatial`);
266 # - it isn't also given at query construction time.
267 result: NamedValueSet[DimensionElement] = NamedValueSet()
268 for family in self.mustHaveKeysJoined.spatial:
269 element = family.choose(self.mustHaveKeysJoined.elements)
270 assert isinstance(element, DimensionElement)
271 if element not in self.where.dataId.graph.elements:
272 result.add(element)
273 if len(result) == 1:
274 # There's no spatial join, but there might be a WHERE filter based
275 # on a given region.
276 if self.where.dataId.graph.spatial:
277 # We can only perform those filters against SkyPix dimensions,
278 # so if what we have isn't one, add the common SkyPix dimension
279 # to the query; the element we have will be joined to that.
280 element, = result
281 if not isinstance(element, SkyPixDimension):
282 result.add(self.universe.commonSkyPix)
283 else:
284 # There is no spatial join or filter in this query. Even
285 # if this element might be associated with spatial
286 # information, we don't need it for this query.
287 return NamedValueSet()
288 elif len(result) > 1:
289 # There's a spatial join. Those require the common SkyPix
290 # system to be included in the query in order to connect them.
291 result.add(self.universe.commonSkyPix)
292 return result
294 @property # type: ignore
295 @cached_getter
296 def temporal(self) -> NamedValueSet[DimensionElement]:
297 """Dimension elements whose timespans should be included in the
298 query (`NamedValueSet` of `DimensionElement`).
299 """
300 # An element may participate temporally in the query if:
301 # - it's the most precise temporal element for its system in the
302 # requested dimensions (i.e. in `self.requested.temporal`);
303 # - it isn't also given at query construction time.
304 result: NamedValueSet[DimensionElement] = NamedValueSet()
305 for family in self.mustHaveKeysJoined.temporal:
306 element = family.choose(self.mustHaveKeysJoined.elements)
307 assert isinstance(element, DimensionElement)
308 if element not in self.where.dataId.graph.elements:
309 result.add(element)
310 if len(result) == 1 and not self.where.dataId.graph.temporal:
311 # No temporal join or filter. Even if this element might be
312 # associated with temporal information, we don't need it for this
313 # query.
314 return NamedValueSet()
315 return result
317 @property # type: ignore
318 @cached_getter
319 def mustHaveKeysJoined(self) -> DimensionGraph:
320 """Dimensions whose primary keys must be used in the JOIN ON clauses
321 of the query, even if their tables do not appear (`DimensionGraph`).
323 A `Dimension` primary key can appear in a join clause without its table
324 via a foreign key column in table of a dependent dimension element or
325 dataset.
326 """
327 names = set(self.requested.names | self.where.dimensions.names)
328 return DimensionGraph(self.universe, names=names)
330 @property # type: ignore
331 @cached_getter
332 def mustHaveTableJoined(self) -> NamedValueSet[DimensionElement]:
333 """Dimension elements whose associated tables must appear in the
334 query's FROM clause (`NamedValueSet` of `DimensionElement`).
335 """
336 result = NamedValueSet(self.spatial | self.temporal | self.where.columns.keys())
337 for dimension in self.mustHaveKeysJoined:
338 if dimension.implied:
339 result.add(dimension)
340 for element in self.mustHaveKeysJoined.union(self.where.dataId.graph).elements:
341 if element.alwaysJoin:
342 result.add(element)
343 return result
346@dataclass
347class DatasetQueryColumns:
348 """A struct containing the columns used to reconstruct `DatasetRef`
349 instances from query results.
350 """
352 datasetType: DatasetType
353 """The dataset type being queried (`DatasetType`).
354 """
356 id: ColumnElement
357 """Column containing the unique integer ID for this dataset.
358 """
360 runKey: ColumnElement
361 """Foreign key column to the `~CollectionType.RUN` collection that holds
362 this dataset.
363 """
365 ingestDate: Optional[ColumnElement]
366 """Column containing the ingest timestamp, this is not a part of
367 `DatasetRef` but it comes from the same table.
368 """
370 def __iter__(self) -> Iterator[ColumnElement]:
371 yield self.id
372 yield self.runKey
375@dataclass
376class QueryColumns:
377 """A struct organizing the columns in an under-construction or currently-
378 executing query.
380 Takes no parameters at construction, as expected usage is to add elements
381 to its container attributes incrementally.
382 """
383 def __init__(self) -> None:
384 self.keys = NamedKeyDict()
385 self.timespans = NamedKeyDict()
386 self.regions = NamedKeyDict()
387 self.datasets = None
389 keys: NamedKeyDict[Dimension, List[ColumnElement]]
390 """Columns that correspond to the primary key values of dimensions
391 (`NamedKeyDict` mapping `Dimension` to a `list` of `ColumnElement`).
393 Each value list contains columns from multiple tables corresponding to the
394 same dimension, and the query should constrain the values of those columns
395 to be the same.
397 In a `Query`, the keys of this dictionary must include at least the
398 dimensions in `QuerySummary.requested` and `QuerySummary.dataId.graph`.
399 """
401 timespans: NamedKeyDict[DimensionElement, TimespanDatabaseRepresentation]
402 """Columns that correspond to timespans for elements that participate in a
403 temporal join or filter in the query (`NamedKeyDict` mapping
404 `DimensionElement` to `TimespanDatabaseRepresentation`).
406 In a `Query`, the keys of this dictionary must be exactly the elements
407 in `QuerySummary.temporal`.
408 """
410 regions: NamedKeyDict[DimensionElement, ColumnElement]
411 """Columns that correspond to regions for elements that participate in a
412 spatial join or filter in the query (`NamedKeyDict` mapping
413 `DimensionElement` to `ColumnElement`).
415 In a `Query`, the keys of this dictionary must be exactly the elements
416 in `QuerySummary.spatial`.
417 """
419 datasets: Optional[DatasetQueryColumns]
420 """Columns that can be used to construct `DatasetRef` instances from query
421 results.
422 (`DatasetQueryColumns` or `None`).
423 """
425 def isEmpty(self) -> bool:
426 """Return `True` if this query has no columns at all.
427 """
428 return not (self.keys or self.timespans or self.regions or self.datasets is not None)
430 def getKeyColumn(self, dimension: Union[Dimension, str]) -> ColumnElement:
431 """ Return one of the columns in self.keys for the given dimension.
433 The column selected is an implentation detail but is guaranteed to
434 be deterministic and consistent across multiple calls.
436 Parameters
437 ----------
438 dimension : `Dimension` or `str`
439 Dimension for which to obtain a key column.
441 Returns
442 -------
443 column : `sqlalchemy.sql.ColumnElement`
444 SQLAlchemy column object.
445 """
446 # Choosing the last element here is entirely for human readers of the
447 # query (e.g. developers debugging things); it makes it more likely a
448 # dimension key will be provided by the dimension's own table, or
449 # failing that, some closely related dimension, which might be less
450 # surprising to see than e.g. some dataset subquery. From the
451 # database's perspective this is entirely arbitrary, because the query
452 # guarantees they all have equal values.
453 return self.keys[dimension][-1]
456@dataclass
457class RegistryManagers:
458 """Struct used to pass around the manager objects that back a `Registry`
459 and are used internally by the query system.
460 """
462 collections: CollectionManager
463 """Manager for collections (`CollectionManager`).
464 """
466 datasets: DatasetRecordStorageManager
467 """Manager for datasets and dataset types (`DatasetRecordStorageManager`).
468 """
470 dimensions: DimensionRecordStorageManager
471 """Manager for dimensions (`DimensionRecordStorageManager`).
472 """