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/>. 

21 

22from __future__ import annotations 

23 

24__all__ = ["DimensionGraph"] 

25 

26import itertools 

27from typing import ( 

28 Any, 

29 Dict, 

30 Iterable, 

31 Iterator, 

32 KeysView, 

33 Optional, 

34 Set, 

35 Tuple, 

36 TYPE_CHECKING, 

37 Union, 

38) 

39 

40from ..named import NamedValueSet 

41from ..utils import immutable 

42 

43if TYPE_CHECKING: # Imports needed only for type annotations; may be circular. 43 ↛ 44line 43 didn't jump to line 44, because the condition on line 43 was never true

44 from .universe import DimensionUniverse 

45 from .elements import DimensionElement, Dimension 

46 

47 

48@immutable 

49class DimensionGraph: 

50 """An immutable, dependency-complete collection of dimensions. 

51 

52 `DimensionGraph` behaves in many respects like a set of `Dimension` 

53 instances that maintains several special subsets and supersets of 

54 related `DimensionElement` instances. It does not fully implement the 

55 `collections.abc.Set` interface, as its automatic expansion of dependencies 

56 would make set difference and XOR operations behave surprisingly. 

57 

58 It also provides dict-like lookup of `DimensionElement` instances from 

59 their names. 

60 

61 Parameters 

62 ---------- 

63 universe : `DimensionUniverse` 

64 The special graph of all known dimensions of which this graph will be 

65 a subset. 

66 dimensions : iterable of `Dimension`, optional 

67 An iterable of `Dimension` instances that must be included in the 

68 graph. All (recursive) dependencies of these dimensions will also 

69 be included. At most one of ``dimensions`` and ``names`` must be 

70 provided. 

71 names : iterable of `str`, optional 

72 An iterable of the names of dimensiosn that must be included in the 

73 graph. All (recursive) dependencies of these dimensions will also 

74 be included. At most one of ``dimensions`` and ``names`` must be 

75 provided. 

76 conform : `bool`, optional 

77 If `True` (default), expand to include dependencies. `False` should 

78 only be used for callers that can guarantee that other arguments are 

79 already correctly expanded, and is primarily for internal use. 

80 

81 Notes 

82 ----- 

83 `DimensionGraph` should be used instead of other collections in any context 

84 where a collection of dimensions is required and a `DimensionUniverse` is 

85 available. 

86 

87 While `DimensionUniverse` inherits from `DimensionGraph`, it should 

88 otherwise not be used as a base class. 

89 """ 

90 

91 def __new__(cls, universe: DimensionUniverse, 

92 dimensions: Optional[Iterable[Dimension]] = None, 

93 names: Optional[Iterable[str]] = None, 

94 conform: bool = True) -> DimensionGraph: 

95 conformedNames: Set[str] 

96 if names is None: 

97 if dimensions is None: 

98 conformedNames = set() 

99 else: 

100 try: 

101 # Optimize for NamedValueSet/NamedKeyDict, though that's 

102 # not required. 

103 conformedNames = set(dimensions.names) # type: ignore 

104 except AttributeError: 

105 conformedNames = set(d.name for d in dimensions) 

106 else: 

107 if dimensions is not None: 

108 raise TypeError("Only one of 'dimensions' and 'names' may be provided.") 

109 conformedNames = set(names) 

110 if conform: 

111 # Expand given dimensions to include all dependencies. 

112 for name in tuple(conformedNames): # iterate over a temporary copy so we can modify the original 

113 conformedNames.update(universe[name]._related.dependencies) 

114 # Look in the cache of existing graphs, with the expanded set of names. 

115 cacheKey = frozenset(conformedNames) 

116 self = universe._cache.get(cacheKey, None) 

117 if self is not None: 

118 return self 

119 # This is apparently a new graph. Create it, and add it to the cache. 

120 self = super().__new__(cls) 

121 universe._cache[cacheKey] = self 

122 self.universe = universe 

123 # Reorder dimensions by iterating over the universe (which is 

124 # ordered already) and extracting the ones in the set. 

125 self.dimensions = NamedValueSet(universe.sorted(conformedNames)) 

126 # Make a set that includes both the dimensions and any 

127 # DimensionElements whose dependencies are in self.dimensions. 

128 self.elements = NamedValueSet(e for e in universe.getStaticElements() 

129 if e._shouldBeInGraph(self.dimensions.names)) 

130 self._finish() 

131 return self 

132 

133 def _finish(self) -> None: 

134 """Complete construction of the graph. 

135 

136 This is intended for internal use by `DimensionGraph` and 

137 `DimensionUniverse` only. 

138 """ 

139 # Freeze the sets the constructor is responsible for populating. 

140 self.dimensions.freeze() 

141 self.elements.freeze() 

142 

143 # Split dependencies up into "required" and "implied" subsets. 

144 # Note that a dimension may be required in one graph and implied in 

145 # another. 

146 self.required = NamedValueSet() 

147 self.implied = NamedValueSet() 

148 for i1, dim1 in enumerate(self.dimensions): 

149 for i2, dim2 in enumerate(self.dimensions): 

150 if dim1.name in dim2._related.implied: 

151 self.implied.add(dim1) 

152 break 

153 else: 

154 # If no other dimension implies dim1, it's required. 

155 self.required.add(dim1) 

156 self.required.freeze() 

157 self.implied.freeze() 

158 

159 # Compute sets of spatial and temporal elements. 

160 # This contain the values of the `.spatial` and `.temporal` attributes 

161 # of all elements, unless those attributes are not in the graph. 

162 # In that case, the element whose attribute is not in the graph is 

163 # added instead. This ensures that these sets contain the 

164 # most-specific spatial and temporal elements, not the summary elements 

165 # that aggregate them, unless the summaries are all that we have. 

166 self.spatial = NamedValueSet() 

167 self.temporal = NamedValueSet() 

168 for element in self.elements: 

169 if element.spatial is not None: 

170 if element.spatial in self.elements: 

171 self.spatial.add(element.spatial) 

172 else: 

173 self.spatial.add(element) 

174 if element.temporal is not None: 

175 if element.temporal in self.elements: 

176 self.temporal.add(element.temporal) 

177 else: 

178 self.temporal.add(element) 

179 self.spatial.freeze() 

180 self.temporal.freeze() 

181 

182 # Build mappings from dimension to index; this is really for 

183 # DataCoordinate, but we put it in DimensionGraph because many 

184 # (many!) DataCoordinates will share the same DimensionGraph, and 

185 # we want them to be lightweight. The order here is what's convenient 

186 # for DataCoordinate: all required dimensions before all implied 

187 # dimensions. 

188 self._dataCoordinateIndices: Dict[str, int] = { 

189 name: i for i, name in enumerate(itertools.chain(self.required.names, self.implied.names)) 

190 } 

191 

192 def __getnewargs__(self) -> tuple: 

193 return (self.universe, None, tuple(self.dimensions.names), False) 

194 

195 @property 

196 def names(self) -> KeysView[str]: 

197 """A set of the names of all dimensions in the graph (`KeysView`). 

198 """ 

199 return self.dimensions.names 

200 

201 def __iter__(self) -> Iterator[Dimension]: 

202 """Iterate over all dimensions in the graph (and true `Dimension` 

203 instances only). 

204 """ 

205 return iter(self.dimensions) 

206 

207 def __len__(self) -> int: 

208 """Return the number of dimensions in the graph (and true `Dimension` 

209 instances only). 

210 """ 

211 return len(self.dimensions) 

212 

213 def __contains__(self, element: Union[str, DimensionElement]) -> bool: 

214 """Return `True` if the given element or element name is in the graph. 

215 

216 This test covers all `DimensionElement` instances in ``self.elements``, 

217 not just true `Dimension` instances). 

218 """ 

219 return element in self.elements 

220 

221 def __getitem__(self, name: str) -> DimensionElement: 

222 """Return the element with the given name. 

223 

224 This lookup covers all `DimensionElement` instances in 

225 ``self.elements``, not just true `Dimension` instances). 

226 """ 

227 return self.elements[name] 

228 

229 def get(self, name: str, default: Any = None) -> DimensionElement: 

230 """Return the element with the given name. 

231 

232 This lookup covers all `DimensionElement` instances in 

233 ``self.elements``, not just true `Dimension` instances). 

234 """ 

235 return self.elements.get(name, default) 

236 

237 def __str__(self) -> str: 

238 return str(self.dimensions) 

239 

240 def __repr__(self) -> str: 

241 return f"DimensionGraph({str(self)})" 

242 

243 @classmethod 

244 def decode(cls, encoded: bytes, *, universe: DimensionUniverse) -> DimensionGraph: 

245 """Construct a `DimensionGraph` from its encoded representation. 

246 

247 Parameters 

248 ---------- 

249 encoded : `bytes` 

250 Byte string produced by `DimensionGraph.encode`. 

251 universe : `DimensionUniverse` 

252 Universe the new graph is a part of. Must have the same dimensions 

253 as the original universe. 

254 

255 Returns 

256 ------- 

257 graph : `DimensionGraph` 

258 A new (or possibly cached) `DimensionGraph` instance matching the 

259 given encoding. 

260 """ 

261 dimensions = [] 

262 mask = int.from_bytes(encoded, "big") 

263 for dimension in universe.getStaticDimensions(): 

264 index = universe.getDimensionIndex(dimension.name) 

265 if mask & (1 << index): 

266 dimensions.append(dimension) 

267 return cls(universe, dimensions=dimensions, conform=False) 

268 

269 def encode(self) -> bytes: 

270 """Encode a `DimensionGraph` into a byte string. 

271 

272 Returns 

273 ------- 

274 encoded : `bytes` 

275 Encoded representation of the graph. Length is guaranteed to be 

276 equal to `DimensionUniverse.getEncodeLength`. 

277 """ 

278 mask = 0 

279 for dimension in self.dimensions: 

280 index = self.universe.getDimensionIndex(dimension.name) 

281 mask |= (1 << index) 

282 return mask.to_bytes(self.universe.getEncodeLength(), byteorder="big") 

283 

284 def isdisjoint(self, other: DimensionGraph) -> bool: 

285 """Test whether the intersection of two graphs is empty. 

286 

287 Returns `True` if either operand is the empty. 

288 """ 

289 return self.dimensions.isdisjoint(other.dimensions) 

290 

291 def issubset(self, other: DimensionGraph) -> bool: 

292 """Test whether all dimensions in ``self`` are also in ``other``. 

293 

294 Returns `True` if ``self`` is empty. 

295 """ 

296 return self.dimensions.issubset(other.dimensions) 

297 

298 def issuperset(self, other: DimensionGraph) -> bool: 

299 """Test whether all dimensions in ``other`` are also in ``self``. 

300 

301 Returns `True` if ``other`` is empty. 

302 """ 

303 return self.dimensions.issuperset(other.dimensions) 

304 

305 def __eq__(self, other: Any) -> bool: 

306 """Test whether ``self`` and ``other`` have exactly the same dimensions 

307 and elements. 

308 """ 

309 if isinstance(other, DimensionGraph): 

310 return self.dimensions == other.dimensions 

311 else: 

312 return False 

313 

314 def __hash__(self) -> int: 

315 return hash(tuple(self.dimensions.names)) 

316 

317 def __le__(self, other: DimensionGraph) -> bool: 

318 """Test whether ``self`` is a subset of ``other``. 

319 """ 

320 return self.dimensions <= other.dimensions 

321 

322 def __ge__(self, other: DimensionGraph) -> bool: 

323 """Test whether ``self`` is a superset of ``other``. 

324 """ 

325 return self.dimensions >= other.dimensions 

326 

327 def __lt__(self, other: DimensionGraph) -> bool: 

328 """Test whether ``self`` is a strict subset of ``other``. 

329 """ 

330 return self.dimensions < other.dimensions 

331 

332 def __gt__(self, other: DimensionGraph) -> bool: 

333 """Test whether ``self`` is a strict superset of ``other``. 

334 """ 

335 return self.dimensions > other.dimensions 

336 

337 def union(self, *others: DimensionGraph) -> DimensionGraph: 

338 """Construct a new graph containing all dimensions in any of the 

339 operands. 

340 

341 The elements of the returned graph may exceed the naive union of 

342 their elements, as some `DimensionElement` instances are included 

343 in graphs whenever multiple dimensions are present, and those 

344 dependency dimensions could have been provided by different operands. 

345 """ 

346 names = set(self.names).union(*[other.names for other in others]) 

347 return DimensionGraph(self.universe, names=names) 

348 

349 def intersection(self, *others: DimensionGraph) -> DimensionGraph: 

350 """Construct a new graph containing only dimensions in all of the 

351 operands. 

352 """ 

353 names = set(self.names).intersection(*[other.names for other in others]) 

354 return DimensionGraph(self.universe, names=names) 

355 

356 def __or__(self, other: DimensionGraph) -> DimensionGraph: 

357 """Construct a new graph containing all dimensions in any of the 

358 operands. 

359 

360 See `union`. 

361 """ 

362 return self.union(other) 

363 

364 def __and__(self, other: DimensionGraph) -> DimensionGraph: 

365 """Construct a new graph containing only dimensions in all of the 

366 operands. 

367 """ 

368 return self.intersection(other) 

369 

370 @property 

371 def primaryKeyTraversalOrder(self) -> Tuple[DimensionElement, ...]: 

372 """Return a tuple of all elements in an order allows records to be 

373 found given their primary keys, starting from only the primary keys of 

374 required dimensions (`tuple` [ `DimensionRecord` ]). 

375 

376 Unlike the table definition/topological order (which is what 

377 DimensionUniverse.sorted gives you), when dimension A implies 

378 dimension B, dimension A appears first. 

379 """ 

380 order = getattr(self, "_primaryKeyTraversalOrder", None) 

381 if order is None: 

382 done: Set[str] = set() 

383 order = [] 

384 

385 def addToOrder(element: DimensionElement) -> None: 

386 if element.name in done: 

387 return 

388 predecessors = set(element.required.names) 

389 predecessors.discard(element.name) 

390 if not done.issuperset(predecessors): 

391 return 

392 order.append(element) 

393 done.add(element.name) 

394 for other in element.implied: 

395 addToOrder(other) 

396 

397 while not done.issuperset(self.required): 

398 for dimension in self.required: 

399 addToOrder(dimension) 

400 

401 order.extend(element for element in self.elements if element.name not in done) 

402 order = tuple(order) 

403 self._primaryKeyTraversalOrder = order 

404 return order 

405 

406 # Class attributes below are shadowed by instance attributes, and are 

407 # present just to hold the docstrings for those instance attributes. 

408 

409 universe: DimensionUniverse 

410 """The set of all known dimensions, of which this graph is a subset 

411 (`DimensionUniverse`). 

412 """ 

413 

414 dimensions: NamedValueSet[Dimension] 

415 """A true `~collections.abc.Set` of all true `Dimension` instances in the 

416 graph (`NamedValueSet` of `Dimension`). 

417 

418 This is the set used for iteration, ``len()``, and most set-like operations 

419 on `DimensionGraph` itself. 

420 """ 

421 

422 elements: NamedValueSet[DimensionElement] 

423 """A true `~collections.abc.Set` of all `DimensionElement` instances in the 

424 graph; a superset of `dimensions` (`NamedValueSet` of `DimensionElement`). 

425 

426 This is the set used for dict-like lookups, including the ``in`` operator, 

427 on `DimensionGraph` itself. 

428 """ 

429 

430 required: NamedValueSet[Dimension] 

431 """The subset of `dimensions` whose elments must be directly identified via 

432 their primary keys in a data ID in order to identify the rest of the 

433 elements in the graph (`NamedValueSet` of `Dimension`). 

434 """ 

435 

436 implied: NamedValueSet[Dimension] 

437 """The subset of `dimensions` whose elements need not be directly 

438 identified via their primary keys in a data ID (`NamedValueSet` of 

439 `Dimension`). 

440 """ 

441 

442 spatial: NamedValueSet[DimensionElement] 

443 """Elements that are associated with independent spatial regions 

444 (`NamedValueSet` of `DimensionElement`). 

445 """ 

446 

447 temporal: NamedValueSet[DimensionElement] 

448 """Elements that are associated with independent spatial regions 

449 (`NamedValueSet` of `DimensionElement`). 

450 """