Hide keyboard shortcuts

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 

22 

23__all__ = ("QueryBuilder",) 

24 

25from typing import List, Iterable, TYPE_CHECKING 

26 

27from sqlalchemy.sql import ColumnElement, and_, literal, bindparam, select, FromClause 

28import sqlalchemy.sql 

29from sqlalchemy.engine import Connection 

30 

31from ...core import ( 

32 DimensionElement, 

33 SkyPixDimension, 

34 Dimension, 

35 DatasetType, 

36 Timespan, 

37) 

38from ...core.utils import NamedValueSet 

39 

40from ._structs import QuerySummary, QueryColumns, QueryParameters, GivenTime 

41from ._datasets import DatasetRegistryStorage, CollectionsExpression 

42from .expressions import ClauseVisitor 

43from ._query import Query 

44 

45if TYPE_CHECKING: 45 ↛ 46line 45 didn't jump to line 46, because the condition on line 45 was never true

46 from ..interfaces import DimensionRecordStorageManager 

47 

48 

49class QueryBuilder: 

50 """A builder for potentially complex queries that join tables based 

51 on dimension relationships. 

52 

53 Parameters 

54 ---------- 

55 connection : `sqlalchemy.engine.Connection` 

56 SQLAlchemy connection object. This is only used to pass through 

57 to the `Query` object returned by `finish`. 

58 summary : `QuerySummary` 

59 Struct organizing the dimensions involved in the query. 

60 dimensionStorage : `DimensionRecordStorageManager` 

61 Manager for storage backend objects that abstract access to dimension 

62 tables. 

63 datasetStorage : `DatasetRegistryStorage` 

64 Storage backend object that abstracts access to dataset tables. 

65 """ 

66 

67 def __init__(self, connection: Connection, summary: QuerySummary, 

68 dimensionStorage: DimensionRecordStorageManager, 

69 datasetStorage: DatasetRegistryStorage): 

70 self.summary = summary 

71 self._connection = connection 

72 self._dimensionStorage = dimensionStorage 

73 self._datasetStorage = datasetStorage 

74 self._sql = None 

75 self._elements: NamedValueSet[DimensionElement] = NamedValueSet() 

76 self._columns = QueryColumns() 

77 

78 def hasDimensionKey(self, dimension: Dimension) -> bool: 

79 """Return `True` if the given dimension's primary key column has 

80 been included in the query (possibly via a foreign key column on some 

81 other table). 

82 """ 

83 return dimension in self._columns.keys 

84 

85 def joinDimensionElement(self, element: DimensionElement): 

86 """Add the table for a `DimensionElement` to the query. 

87 

88 This automatically joins the element table to all other tables in the 

89 query with which it is related, via both dimension keys and spatial 

90 and temporal relationships. 

91 

92 External calls to this method should rarely be necessary; `finish` will 

93 automatically call it if the `DimensionElement` has been identified as 

94 one that must be included. 

95 

96 Parameters 

97 ---------- 

98 element : `DimensionElement` 

99 Element for which a table should be added. The element must be 

100 associated with a database table (see `DimensionElement.hasTable`). 

101 """ 

102 assert element not in self._elements, "Element already included in query." 

103 storage = self._dimensionStorage[element] 

104 storage.join( 

105 self, 

106 regions=self._columns.regions if element in self.summary.spatial else None, 

107 timespans=self._columns.timespans if element in self.summary.temporal else None, 

108 ) 

109 self._elements.add(element) 

110 

111 def joinDataset(self, datasetType: DatasetType, collections: CollectionsExpression, *, 

112 isResult: bool = True, addRank: bool = False): 

113 """Add a dataset search or constraint to the query. 

114 

115 Unlike other `QueryBuilder` join methods, this *must* be called 

116 directly to search for datasets of a particular type or constrain the 

117 query results based on the exists of datasets. However, all dimensions 

118 used to identify the dataset type must have already been included in 

119 `QuerySummary.requested` when initializing the `QueryBuilder`. 

120 

121 Parameters 

122 ---------- 

123 datasetType : `DatasetType` 

124 The type of datasets to search for. 

125 collections : sequence of `str` or `Like`, or ``...`` 

126 An expression describing the collections in which to search for 

127 the datasets. ``...`` indicates that all collections should be 

128 searched. 

129 isResult : `bool`, optional 

130 If `True` (default), include the ``dataset_id`` column in the 

131 result columns of the query, allowing complete `DatasetRef` 

132 instances to be produced from the query results for this dataset 

133 type. If `False`, the existence of datasets of this type is used 

134 only to constrain the data IDs returned by the query. 

135 addRank : `bool`, optional 

136 If `True` (`False` is default), also include a calculated column 

137 that ranks the collection in which the dataset was found (lower 

138 is better). Requires that all entries in ``collections`` be 

139 regular strings, so there is a clear search order. Ignored if 

140 ``isResult`` is `False`. 

141 """ 

142 assert datasetType.dimensions.issubset(self.summary.requested) 

143 table = self._datasetStorage.getDatasetSubquery(datasetType, collections=collections, 

144 dataId=self.summary.dataId, 

145 isResult=isResult, addRank=addRank) 

146 self.joinTable(table, datasetType.dimensions) 

147 if isResult: 

148 self._columns.datasets[datasetType] = (table.columns["dataset_id"], 

149 table.columns["rank"] if addRank else None) 

150 

151 def joinTable(self, table: FromClause, dimensions: Iterable[Dimension]): 

152 """Join an arbitrary table to the query via dimension relationships. 

153 

154 External calls to this method should only be necessary for tables whose 

155 records represent neither dataset nor dimension elements (i.e. 

156 extensions to the standard `Registry` schema). 

157 

158 Parameters 

159 ---------- 

160 table : `sqlalchemy.sql.FromClause` 

161 SQLAlchemy object representing the logical table (which may be a 

162 join or subquery expression) to be joined. 

163 dimensions : iterable of `Dimension` 

164 The dimensions that relate this table to others that may be in the 

165 query. The table must have columns with the names of the 

166 dimensions. 

167 """ 

168 joinOn = self.startJoin(table, dimensions, dimensions.names) 

169 self.finishJoin(table, joinOn) 

170 

171 def startJoin(self, table: FromClause, dimensions: Iterable[Dimension], columnNames: Iterable[str] 

172 ) -> List[ColumnElement]: 

173 """Begin a join on dimensions. 

174 

175 Must be followed by call to `finishJoin`. 

176 

177 Parameters 

178 ---------- 

179 table : `sqlalchemy.sql.FromClause` 

180 SQLAlchemy object representing the logical table (which may be a 

181 join or subquery expression) to be joined. 

182 dimensions : iterable of `Dimension` 

183 The dimensions that relate this table to others that may be in the 

184 query. The table must have columns with the names of the 

185 dimensions. 

186 columnNames : iterable of `str` 

187 Names of the columns that correspond to dimension key values; must 

188 be `zip` iterable with ``dimensions``. 

189 

190 Returns 

191 ------- 

192 joinOn : `list` of `sqlalchemy.sql.ColumnElement` 

193 Sequence of boolean expressions that should be combined with AND 

194 to form (part of) the ON expression for this JOIN. 

195 """ 

196 joinOn = [] 

197 for dimension, columnName in zip(dimensions, columnNames): 

198 columnInTable = table.columns[columnName] 

199 columnsInQuery = self._columns.keys.setdefault(dimension, []) 

200 for columnInQuery in columnsInQuery: 

201 joinOn.append(columnInQuery == columnInTable) 

202 columnsInQuery.append(columnInTable) 

203 return joinOn 

204 

205 def finishJoin(self, table, joinOn): 

206 """Complete a join on dimensions. 

207 

208 Must be preceded by call to `startJoin`. 

209 

210 Parameters 

211 ---------- 

212 table : `sqlalchemy.sql.FromClause` 

213 SQLAlchemy object representing the logical table (which may be a 

214 join or subquery expression) to be joined. Must be the same object 

215 passed to `startJoin`. 

216 joinOn : `list` of `sqlalchemy.sql.ColumnElement` 

217 Sequence of boolean expressions that should be combined with AND 

218 to form (part of) the ON expression for this JOIN. Should include 

219 at least the elements of the list returned by `startJoin`. 

220 """ 

221 if joinOn: 

222 self._sql = self._sql.join(table, and_(*joinOn)) 

223 elif self._sql is None: 

224 self._sql = table 

225 else: 

226 # New table is completely unrelated to all already-included 

227 # tables. We need a cross join here but SQLAlchemy does not 

228 # have a specific method for that. Using join() without 

229 # `onclause` will try to join on FK and will raise an exception 

230 # for unrelated tables, so we have to use `onclause` which is 

231 # always true. 

232 self._sql = self._sql.join(table, literal(True) == literal(True)) 

233 

234 def _joinMissingDimensionElements(self): 

235 """Join all dimension element tables that were identified as necessary 

236 by `QuerySummary` and have not yet been joined. 

237 

238 For internal use by `QueryBuilder` only; will be called (and should 

239 only by called) by `finish`. 

240 """ 

241 # Join all DimensionElement tables that we need for spatial/temporal 

242 # joins/filters or a nontrivial WHERE expression. 

243 # We iterate over these in *reverse* topological order to minimize the 

244 # number of tables joined. For example, the "visit" table provides 

245 # the primary key value for the "instrument" table it depends on, so we 

246 # don't need to join "instrument" as well unless we had a nontrivial 

247 # expression on it (and hence included it already above). 

248 for element in self.summary.universe.sorted(self.summary.mustHaveTableJoined, reverse=True): 

249 self.joinDimensionElement(element) 

250 # Join in any requested Dimension tables that don't already have their 

251 # primary keys identified by the query. 

252 for dimension in self.summary.universe.sorted(self.summary.mustHaveKeysJoined, reverse=True): 

253 if dimension not in self._columns.keys: 

254 self.joinDimensionElement(dimension) 

255 

256 def _addWhereClause(self): 

257 """Add a WHERE clause to the query under construction, connecting all 

258 joined dimensions to the expression and given dimensions from 

259 `QuerySummary`. 

260 

261 For internal use by `QueryBuilder` only; will be called (and should 

262 only by called) by `finish`. 

263 """ 

264 parameters = QueryParameters() 

265 whereTerms = [] 

266 if self.summary.expression.tree is not None: 

267 visitor = ClauseVisitor(self.summary.universe, self._columns, self._elements) 

268 whereTerms.append(self.summary.expression.tree.visit(visitor)) 

269 for dimension, columnsInQuery in self._columns.keys.items(): 

270 if dimension in self.summary.given: 

271 if self.summary.whenIsDimensionGiven(dimension) == GivenTime.AT_EXECUTION: 

272 givenKey = bindparam(f"_given_later_{dimension.name}") 

273 parameters.keys[dimension] = givenKey 

274 else: 

275 givenKey = self.summary.dataId[dimension] 

276 # Add a WHERE term for each column that corresponds to each 

277 # key. This is redundant with the JOIN ON clauses that make 

278 # them equal to each other, but more constraints have a chance 

279 # of making things easier on the DB's query optimizer. 

280 for columnInQuery in columnsInQuery: 

281 whereTerms.append(columnInQuery == givenKey) 

282 else: 

283 # Dimension is not fully identified, but it might be a skypix 

284 # dimension that's constrained by a given region. 

285 if self.summary.given.spatial and isinstance(dimension, SkyPixDimension): 

286 if self.summary.whenIsRegionGiven() == GivenTime.AT_CONSTRUCTION: 

287 # We know the region now. 

288 givenSkyPixIds = [] 

289 for begin, end in dimension.pixelization.envelope(self.summary.dataId.region): 

290 givenSkyPixIds.extend(range(begin, end)) 

291 else: 

292 # We'll know the region later (there might be a region 

293 # now, too, but we'll know a more precise one later, 

294 # and hence we'll ignore the one we know now). 

295 givenSkyPixIds = bindparam(f"_given_later_{dimension.name}") 

296 parameters.skypix[dimension] = givenSkyPixIds 

297 for columnInQuery in columnsInQuery: 

298 whereTerms.append(columnInQuery.in_(givenSkyPixIds)) 

299 # If we are [to be] given an dataId with a timespan, and there are 

300 # one or more timespans in the query that aren't given, add a WHERE 

301 # expression for each of them. 

302 if self.summary.given.temporal and self.summary.temporal: 

303 if self.summary.whenIsTimespanGiven() == GivenTime.AT_CONSTRUCTION: 

304 # Timespan is known now. 

305 givenInterval = self.summary.dataId.timespan 

306 else: 

307 # We'll know the timespan later (there might be a timespan now, 

308 # too, but we'll know a more precise one later, and hence we'll 

309 # ignore the one we know now). 

310 givenInterval = Timespan( 

311 begin=bindparam(f"_given_later_timespan_begin"), 

312 end=bindparam(f"_given_later_timespan_end"), 

313 ) 

314 for element, intervalInQuery in self._columns.timespans.items(): 

315 assert element not in self.summary.given.elements 

316 whereTerms.append(intervalInQuery.overlaps(givenInterval, ops=sqlalchemy.sql)) 

317 # AND-together the full WHERE clause, and combine it with the FROM 

318 # clause. 

319 self._sql = self._sql.where(and_(*whereTerms)) 

320 return parameters 

321 

322 def _addSelectClause(self): 

323 """Add a SELECT clause to the query under construction containing all 

324 output columns identified by the `QuerySummary` and requested in calls 

325 to `joinDataset` with ``isResult=True``. 

326 

327 For internal use by `QueryBuilder` only; will be called (and should 

328 only by called) by `finish`. 

329 """ 

330 columns = [] 

331 for dimension in self.summary.requested: 

332 columns.append(self._columns.getKeyColumn(dimension)) 

333 for datasetType, columnPair in self._columns.datasets.items(): 

334 columns.extend(columnPair) 

335 for element, column in self._columns.regions.items(): 

336 columns.append(column) 

337 self._sql = select(columns).select_from(self._sql) 

338 

339 def finish(self) -> Query: 

340 """Finish query constructing, returning a new `Query` instance. 

341 

342 This automatically joins any missing dimension element tables 

343 (according to the categorization of the `QuerySummary` the builder was 

344 constructed with). 

345 

346 This consumes the `QueryBuilder`; no other methods should be called 

347 after this one. 

348 

349 Returns 

350 ------- 

351 query : `Query` 

352 A `Query` object that can be executed (possibly multiple times 

353 with different bind parameter values) and used to interpret result 

354 rows. 

355 """ 

356 self._joinMissingDimensionElements() 

357 self._addSelectClause() 

358 parameters = self._addWhereClause() 

359 return Query(summary=self.summary, connection=self._connection, 

360 sql=self._sql, columns=self._columns, parameters=parameters)