Coverage for python/lsst/daf/butler/queries/overlaps.py: 19%
114 statements
« prev ^ index » next coverage.py v7.4.4, created at 2024-04-19 10:53 +0000
« prev ^ index » next coverage.py v7.4.4, created at 2024-04-19 10:53 +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 software is dual licensed under the GNU General Public License and also
10# under a 3-clause BSD license. Recipients may choose which of these licenses
11# to use; please see the files gpl-3.0.txt and/or bsd_license.txt,
12# respectively. If you choose the GPL option then the following text applies
13# (but note that there is still no warranty even if you opt for BSD instead):
14#
15# This program is free software: you can redistribute it and/or modify
16# it under the terms of the GNU General Public License as published by
17# the Free Software Foundation, either version 3 of the License, or
18# (at your option) any later version.
19#
20# This program is distributed in the hope that it will be useful,
21# but WITHOUT ANY WARRANTY; without even the implied warranty of
22# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23# GNU General Public License for more details.
24#
25# You should have received a copy of the GNU General Public License
26# along with this program. If not, see <http://www.gnu.org/licenses/>.
28from __future__ import annotations
30__all__ = ("OverlapsVisitor",)
32import itertools
33from collections.abc import Hashable, Iterable, Sequence, Set
34from typing import Generic, Literal, TypeVar, cast
36from lsst.sphgeom import Region
38from .._topology import TopologicalFamily
39from ..dimensions import DimensionElement, DimensionGroup
40from . import tree
41from .visitors import PredicateVisitFlags, SimplePredicateVisitor
43_T = TypeVar("_T", bound=Hashable)
46class _NaiveDisjointSet(Generic[_T]):
47 """A very naive (but simple) implementation of a "disjoint set" data
48 structure for strings, with mostly O(N) performance.
50 This class should not be used in any context where the number of elements
51 in the data structure is large. It intentionally implements a subset of
52 the interface of `scipy.cluster.DisJointSet` so that non-naive
53 implementation could be swapped in if desired.
55 Parameters
56 ----------
57 superset : `~collections.abc.Iterable` [ `str` ]
58 Elements to initialize the disjoint set, with each in its own
59 single-element subset.
60 """
62 def __init__(self, superset: Iterable[_T]):
63 self._subsets = [{k} for k in superset]
64 self._subsets.sort(key=len, reverse=True)
66 def merge(self, a: _T, b: _T) -> bool: # numpydoc ignore=PR04
67 """Merge the subsets containing the given elements.
69 Parameters
70 ----------
71 a :
72 Element whose subset should be merged.
73 b :
74 Element whose subset should be merged.
76 Returns
77 -------
78 merged : `bool`
79 `True` if a merge occurred, `False` if the elements were already in
80 the same subset.
81 """
82 for i, subset in enumerate(self._subsets):
83 if a in subset:
84 break
85 else:
86 raise KeyError(f"Merge argument {a!r} not in disjoin set {self._subsets}.")
87 for j, subset in enumerate(self._subsets):
88 if b in subset:
89 break
90 else:
91 raise KeyError(f"Merge argument {b!r} not in disjoin set {self._subsets}.")
92 if i == j:
93 return False
94 i, j = sorted((i, j))
95 self._subsets[i].update(self._subsets[j])
96 del self._subsets[j]
97 self._subsets.sort(key=len, reverse=True)
98 return True
100 def subsets(self) -> Sequence[Set[_T]]:
101 """Return the current subsets, ordered from largest to smallest."""
102 return self._subsets
104 @property
105 def n_subsets(self) -> int:
106 """The number of subsets."""
107 return len(self._subsets)
110class OverlapsVisitor(SimplePredicateVisitor):
111 """A helper class for dealing with spatial and temporal overlaps in a
112 query.
114 Parameters
115 ----------
116 dimensions : `DimensionGroup`
117 Dimensions of the query.
119 Notes
120 -----
121 This class includes logic for extracting explicit spatial and temporal
122 joins from a WHERE-clause predicate and computing automatic joins given the
123 dimensions of the query. It is designed to be subclassed by query driver
124 implementations that want to rewrite the predicate at the same time.
125 """
127 def __init__(self, dimensions: DimensionGroup):
128 self.dimensions = dimensions
129 self._spatial_connections = _NaiveDisjointSet(self.dimensions.spatial)
130 self._temporal_connections = _NaiveDisjointSet(self.dimensions.temporal)
132 def run(self, predicate: tree.Predicate, join_operands: Iterable[DimensionGroup]) -> tree.Predicate:
133 """Process the given predicate to extract spatial and temporal
134 overlaps.
136 Parameters
137 ----------
138 predicate : `tree.Predicate`
139 Predicate to process.
140 join_operands : `~collections.abc.Iterable` [ `DimensionGroup` ]
141 The dimensions of logical tables being joined into this query;
142 these can included embedded spatial and temporal joins that can
143 make it unnecessary to add new ones.
145 Returns
146 -------
147 predicate : `tree.Predicate`
148 A possibly-modified predicate that should replace the original.
149 """
150 result = predicate.visit(self)
151 if result is None:
152 result = predicate
153 for join_operand_dimensions in join_operands:
154 self.add_join_operand_connections(join_operand_dimensions)
155 for a, b in self.compute_automatic_spatial_joins():
156 join_predicate = self.visit_spatial_join(a, b, PredicateVisitFlags.HAS_AND_SIBLINGS)
157 if join_predicate is None:
158 join_predicate = tree.Predicate.compare(
159 tree.DimensionFieldReference.model_construct(element=a, field="region"),
160 "overlaps",
161 tree.DimensionFieldReference.model_construct(element=b, field="region"),
162 )
163 result = result.logical_and(join_predicate)
164 for a, b in self.compute_automatic_temporal_joins():
165 join_predicate = self.visit_temporal_dimension_join(a, b, PredicateVisitFlags.HAS_AND_SIBLINGS)
166 if join_predicate is None:
167 join_predicate = tree.Predicate.compare(
168 tree.DimensionFieldReference.model_construct(element=a, field="timespan"),
169 "overlaps",
170 tree.DimensionFieldReference.model_construct(element=b, field="timespan"),
171 )
172 result = result.logical_and(join_predicate)
173 return result
175 def visit_comparison(
176 self,
177 a: tree.ColumnExpression,
178 operator: tree.ComparisonOperator,
179 b: tree.ColumnExpression,
180 flags: PredicateVisitFlags,
181 ) -> tree.Predicate | None:
182 # Docstring inherited.
183 if operator == "overlaps":
184 if a.column_type == "region":
185 return self.visit_spatial_overlap(a, b, flags)
186 elif b.column_type == "timespan":
187 return self.visit_temporal_overlap(a, b, flags)
188 else:
189 raise AssertionError(f"Unexpected column type {a.column_type} for overlap.")
190 return None
192 def add_join_operand_connections(self, operand_dimensions: DimensionGroup) -> None:
193 """Add overlap connections implied by a table or subquery.
195 Parameters
196 ----------
197 operand_dimensions : `DimensionGroup`
198 Dimensions of of the table or subquery.
200 Notes
201 -----
202 We assume each join operand to a `tree.Select` has its own
203 complete set of spatial and temporal joins that went into generating
204 its rows. That will naturally be true for relations originating from
205 the butler database, like dataset searches and materializations, and if
206 it isn't true for a data ID upload, that would represent an intentional
207 association between non-overlapping things that we'd want to respect by
208 *not* adding a more restrictive automatic join.
209 """
210 for a_family, b_family in itertools.pairwise(operand_dimensions.spatial):
211 self._spatial_connections.merge(a_family, b_family)
212 for a_family, b_family in itertools.pairwise(operand_dimensions.temporal):
213 self._temporal_connections.merge(a_family, b_family)
215 def compute_automatic_spatial_joins(self) -> list[tuple[DimensionElement, DimensionElement]]:
216 """Return pairs of dimension elements that should be spatially joined.
218 Returns
219 -------
220 joins : `list` [ `tuple` [ `DimensionElement`, `DimensionElement` ] ]
221 Automatic joins.
223 Notes
224 -----
225 This takes into account explicit joins extracted by `run` and implicit
226 joins added by `add_join_operand_connections`, and only returns
227 additional joins if there is an unambiguous way to spatially connect
228 any dimensions that are not already spatially connected. Automatic
229 joins are always the most fine-grained join between sets of dimensions
230 (i.e. ``visit_detector_region`` and ``patch`` instead of ``visit`` and
231 ``tract``), but explicitly adding a coarser join between sets of
232 elements will prevent the fine-grained join from being added.
233 """
234 return self._compute_automatic_joins("spatial", self._spatial_connections)
236 def compute_automatic_temporal_joins(self) -> list[tuple[DimensionElement, DimensionElement]]:
237 """Return pairs of dimension elements that should be spatially joined.
239 Returns
240 -------
241 joins : `list` [ `tuple` [ `DimensionElement`, `DimensionElement` ] ]
242 Automatic joins.
244 Notes
245 -----
246 See `compute_automatic_spatial_joins` for information on how automatic
247 joins are determined. Joins to dataset validity ranges are never
248 automatic.
249 """
250 return self._compute_automatic_joins("temporal", self._temporal_connections)
252 def _compute_automatic_joins(
253 self, kind: Literal["spatial", "temporal"], connections: _NaiveDisjointSet[TopologicalFamily]
254 ) -> list[tuple[DimensionElement, DimensionElement]]:
255 if connections.n_subsets <= 1:
256 # All of the joins we need are already present.
257 return []
258 if connections.n_subsets > 2:
259 raise tree.InvalidQueryError(
260 f"Too many disconnected sets of {kind} families for an automatic "
261 f"join: {connections.subsets()}. Add explicit {kind} joins to avoid this error."
262 )
263 a_subset, b_subset = connections.subsets()
264 if len(a_subset) > 1 or len(b_subset) > 1:
265 raise tree.InvalidQueryError(
266 f"A {kind} join is needed between {a_subset} and {b_subset}, but which join to "
267 "add is ambiguous. Add an explicit spatial join to avoid this error."
268 )
269 # We have a pair of families that are not explicitly or implicitly
270 # connected to any other families; add an automatic join between their
271 # most fine-grained members.
272 (a_family,) = a_subset
273 (b_family,) = b_subset
274 return [
275 (
276 cast(DimensionElement, a_family.choose(self.dimensions.elements, self.dimensions.universe)),
277 cast(DimensionElement, b_family.choose(self.dimensions.elements, self.dimensions.universe)),
278 )
279 ]
281 def visit_spatial_overlap(
282 self, a: tree.ColumnExpression, b: tree.ColumnExpression, flags: PredicateVisitFlags
283 ) -> tree.Predicate | None:
284 """Dispatch a spatial overlap comparison predicate to handlers.
286 This method should rarely (if ever) need to be overridden.
288 Parameters
289 ----------
290 a : `tree.ColumnExpression`
291 First operand.
292 b : `tree.ColumnExpression`
293 Second operand.
294 flags : `tree.PredicateLeafFlags`
295 Information about where this overlap comparison appears in the
296 larger predicate tree.
298 Returns
299 -------
300 replaced : `tree.Predicate` or `None`
301 The predicate to be inserted instead in the processed tree, or
302 `None` if no substitution is needed.
303 """
304 match a, b:
305 case tree.DimensionFieldReference(element=a_element), tree.DimensionFieldReference(
306 element=b_element
307 ):
308 return self.visit_spatial_join(a_element, b_element, flags)
309 case tree.DimensionFieldReference(element=element), region_expression:
310 pass
311 case region_expression, tree.DimensionFieldReference(element=element):
312 pass
313 case _:
314 raise AssertionError(f"Unexpected arguments for spatial overlap: {a}, {b}.")
315 if region := region_expression.get_literal_value():
316 return self.visit_spatial_constraint(element, region, flags)
317 raise AssertionError(f"Unexpected argument for spatial overlap: {region_expression}.")
319 def visit_temporal_overlap(
320 self, a: tree.ColumnExpression, b: tree.ColumnExpression, flags: PredicateVisitFlags
321 ) -> tree.Predicate | None:
322 """Dispatch a temporal overlap comparison predicate to handlers.
324 This method should rarely (if ever) need to be overridden.
326 Parameters
327 ----------
328 a : `tree.ColumnExpression`-
329 First operand.
330 b : `tree.ColumnExpression`
331 Second operand.
332 flags : `tree.PredicateLeafFlags`
333 Information about where this overlap comparison appears in the
334 larger predicate tree.
336 Returns
337 -------
338 replaced : `tree.Predicate` or `None`
339 The predicate to be inserted instead in the processed tree, or
340 `None` if no substitution is needed.
341 """
342 match a, b:
343 case tree.DimensionFieldReference(element=a_element), tree.DimensionFieldReference(
344 element=b_element
345 ):
346 return self.visit_temporal_dimension_join(a_element, b_element, flags)
347 case _:
348 # We don't bother differentiating any other kind of temporal
349 # comparison, because in all foreseeable database schemas we
350 # wouldn't have to do anything special with them, since they
351 # don't participate in automatic join calculations and they
352 # should be straightforwardly convertible to SQL.
353 return None
355 def visit_spatial_join(
356 self, a: DimensionElement, b: DimensionElement, flags: PredicateVisitFlags
357 ) -> tree.Predicate | None:
358 """Handle a spatial overlap comparison between two dimension elements.
360 The default implementation updates the set of known spatial connections
361 (for use by `compute_automatic_spatial_joins`) and returns `None`.
363 Parameters
364 ----------
365 a : `DimensionElement`
366 One element in the join.
367 b : `DimensionElement`
368 The other element in the join.
369 flags : `tree.PredicateLeafFlags`
370 Information about where this overlap comparison appears in the
371 larger predicate tree.
373 Returns
374 -------
375 replaced : `tree.Predicate` or `None`
376 The predicate to be inserted instead in the processed tree, or
377 `None` if no substitution is needed.
378 """
379 if a.spatial == b.spatial:
380 raise tree.InvalidQueryError(f"Spatial join between {a} and {b} is not necessary.")
381 self._spatial_connections.merge(
382 cast(TopologicalFamily, a.spatial), cast(TopologicalFamily, b.spatial)
383 )
384 return None
386 def visit_spatial_constraint(
387 self,
388 element: DimensionElement,
389 region: Region,
390 flags: PredicateVisitFlags,
391 ) -> tree.Predicate | None:
392 """Handle a spatial overlap comparison between a dimension element and
393 a literal region.
395 The default implementation just returns `None`.
397 Parameters
398 ----------
399 element : `DimensionElement`
400 The dimension element in the comparison.
401 region : `lsst.sphgeom.Region`
402 The literal region in the comparison.
403 flags : `tree.PredicateLeafFlags`
404 Information about where this overlap comparison appears in the
405 larger predicate tree.
407 Returns
408 -------
409 replaced : `tree.Predicate` or `None`
410 The predicate to be inserted instead in the processed tree, or
411 `None` if no substitution is needed.
412 """
413 return None
415 def visit_temporal_dimension_join(
416 self, a: DimensionElement, b: DimensionElement, flags: PredicateVisitFlags
417 ) -> tree.Predicate | None:
418 """Handle a temporal overlap comparison between two dimension elements.
420 The default implementation updates the set of known temporal
421 connections (for use by `compute_automatic_temporal_joins`) and returns
422 `None`.
424 Parameters
425 ----------
426 a : `DimensionElement`
427 One element in the join.
428 b : `DimensionElement`
429 The other element in the join.
430 flags : `tree.PredicateLeafFlags`
431 Information about where this overlap comparison appears in the
432 larger predicate tree.
434 Returns
435 -------
436 replaced : `tree.Predicate` or `None`
437 The predicate to be inserted instead in the processed tree, or
438 `None` if no substitution is needed.
439 """
440 if a.temporal == b.temporal:
441 raise tree.InvalidQueryError(f"Temporal join between {a} and {b} is not necessary.")
442 self._temporal_connections.merge(
443 cast(TopologicalFamily, a.temporal), cast(TopologicalFamily, b.temporal)
444 )
445 return None