Coverage for python/lsst/daf/butler/dimensions/_group.py: 37%
179 statements
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-16 10:44 +0000
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-16 10:44 +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__ = ["DimensionGroup"]
32import itertools
33from collections.abc import Iterable, Iterator, Mapping, Set
34from types import MappingProxyType
35from typing import TYPE_CHECKING, Any
37from lsst.utils.classes import cached_getter, immutable
39from .._named import NamedValueAbstractSet, NamedValueSet
40from .._topology import TopologicalFamily, TopologicalSpace
42if TYPE_CHECKING: # Imports needed only for type annotations; may be circular.
43 from ._elements import DimensionElement
44 from ._graph import DimensionGraph
45 from ._universe import DimensionUniverse
48class SortedSequenceSet(Set[str]):
49 """A set-like interface wrapper around a tuple.
51 This delegates directly to ``tuple.__contains__``, so there is an implicit
52 assumption that `len` is small and hence O(N) lookups are not a problem, as
53 is the case for sets of dimension names.
55 Parameters
56 ----------
57 seq : `tuple` [`str`, ...]
58 Strings to see the set.
59 """
61 def __init__(self, seq: tuple[str, ...]):
62 self._seq = seq
64 __slots__ = ("_seq",)
66 def __contains__(self, x: object) -> bool:
67 return x in self._seq
69 def __iter__(self) -> Iterator[str]:
70 return iter(self._seq)
72 def __len__(self) -> int:
73 return len(self._seq)
75 def __hash__(self) -> int:
76 return hash(self._seq)
78 def __eq__(self, other: object) -> bool:
79 if seq := getattr(other, "_seq", None):
80 return seq == self._seq
81 return super().__eq__(other)
83 @classmethod
84 def _from_iterable(cls, iterable: Iterable[str]) -> set[str]:
85 # This is used by collections.abc.Set mixin methods when they need
86 # to return a new object (e.g. in `__and__`).
87 return set(iterable)
89 def __repr__(self) -> str:
90 return f"{{{', '.join(str(k) for k in self._seq)}}}"
92 def as_tuple(self) -> tuple[str, ...]:
93 """Return the underlying tuple.
95 Returns
96 -------
97 t : `tuple`
98 A tuple of all the values.
99 """
100 return self._seq
102 @property
103 def names(self) -> Set[str]:
104 """An alias to ``self``.
106 This is a backwards-compatibility API that allows `DimensionGroup`
107 to mimic the `DimensionGraph` object it is intended to replace, by
108 permitting expressions like ``x.required.names`` when ``x`` can be
109 an object of either type.
110 """
111 return self
114@immutable
115class DimensionGroup: # numpydoc ignore=PR02
116 """An immutable, dependency-complete collection of dimensions.
118 `DimensionGroup` behaves in many respects like a set of `str` dimension
119 names that maintains several special subsets and supersets of related
120 dimension elements. It does not fully implement the `collections.abc.Set`
121 interface, because it defines a few different iteration orders and does not
122 privilege any one of them by implementing ``__iter__``.
124 Parameters
125 ----------
126 universe : `DimensionUniverse`
127 Object that manages all known dimensions.
128 names : iterable of `str`, optional
129 An iterable of the names of dimensions that must be included in the
130 group. All (recursive) dependencies of these dimensions will also be
131 included. At most one of ``dimensions`` and ``names`` must be
132 provided.
133 _conform : `bool`, optional
134 If `True` (default), expand to include dependencies. `False` should
135 only be used for callers that can guarantee that other arguments are
136 already correctly expanded, and is for internal use only.
138 Notes
139 -----
140 `DimensionGroup` should be used instead of other collections in most
141 contexts where a collection of dimensions is required and a
142 `DimensionUniverse` is available. Exceptions include cases where order
143 matters (and is different from the consistent ordering defined by the
144 `DimensionUniverse`), or complete `~collection.abc.Set` semantics are
145 required.
146 """
148 def __new__(
149 cls,
150 universe: DimensionUniverse,
151 names: Iterable[str] | DimensionGroup = frozenset(),
152 _conform: bool = True,
153 ) -> DimensionGroup:
154 if isinstance(names, DimensionGroup):
155 if names.universe is universe:
156 return names
157 else:
158 names = names.names
159 if _conform:
160 # Expand dimension names to include all required and implied
161 # dependencies.
162 to_expand = set(names)
163 names = set()
164 while to_expand:
165 dimension = universe[to_expand.pop()]
166 names.add(dimension.name)
167 to_expand.update(dimension.required.names)
168 to_expand.update(dimension.implied.names)
169 to_expand.difference_update(names)
170 else:
171 names = frozenset(names)
172 # Look in the cache of existing groups, with the expanded set of names.
173 cache_key = frozenset(names)
174 self = universe._cached_groups.get(cache_key, None)
175 if self is not None:
176 return self
177 # This is apparently a new group. Create it, and add it to the cache.
178 self = super().__new__(cls)
179 universe._cached_groups[cache_key] = self
180 self.universe = universe
181 # Reorder dimensions by iterating over the universe (which is
182 # ordered already) and extracting the ones in the set.
183 self.names = SortedSequenceSet(tuple(d.name for d in universe.sorted(names)))
184 # Make a set that includes both the dimensions and any
185 # DimensionElements whose dependencies are in self.dimensions.
186 self.elements = SortedSequenceSet(
187 tuple(e.name for e in universe.elements if e.required.names <= self.names)
188 )
189 self.governors = SortedSequenceSet(
190 tuple(d for d in self.names if d in universe.governor_dimensions.names)
191 )
192 self.skypix = SortedSequenceSet(tuple(d for d in self.names if d in universe.skypix_dimensions.names))
193 # Split dependencies up into "required" and "implied" subsets.
194 # Note that a dimension may be required in one group and implied in
195 # another.
196 required: list[str] = []
197 implied: list[str] = []
198 for dim1 in self.names:
199 for dim2 in self.names:
200 if dim1 in universe[dim2].implied.names:
201 implied.append(dim1)
202 break
203 else:
204 # If no other dimension implies dim1, it's required.
205 required.append(dim1)
206 self.required = SortedSequenceSet(tuple(required))
207 self.implied = SortedSequenceSet(tuple(implied))
209 self._space_families = MappingProxyType(
210 {
211 space: NamedValueSet(
212 universe[e].topology[space] for e in self.elements if space in universe[e].topology
213 ).freeze()
214 for space in TopologicalSpace.__members__.values()
215 }
216 )
218 # Build mappings from dimension to index; this is really for
219 # DataCoordinate, but we put it in DimensionGroup because many (many!)
220 # DataCoordinates will share the same DimensionGroup, and we want them
221 # to be lightweight. The order here is what's convenient for
222 # DataCoordinate: all required dimensions before all implied
223 # dimensions.
224 self._data_coordinate_indices = {
225 name: i for i, name in enumerate(itertools.chain(self.required, self.implied))
226 }
227 return self
229 def __getnewargs__(self) -> tuple:
230 return (self.universe, self.names._seq, False)
232 def __deepcopy__(self, memo: dict) -> DimensionGroup:
233 # DimensionGroup is recursively immutable; see note in @immutable
234 # decorator.
235 return self
237 def __len__(self) -> int:
238 return len(self.names)
240 def __contains__(self, element: str) -> bool:
241 if element in self.elements:
242 return True
243 else:
244 from ._elements import DimensionElement
246 if isinstance(element, DimensionElement): # type: ignore[unreachable]
247 raise TypeError(
248 "DimensionGroup does not support membership tests using DimensionElement "
249 "instances; use their names instead."
250 )
251 return False
253 def __str__(self) -> str:
254 return str(self.names)
256 def __repr__(self) -> str:
257 return f"DimensionGroup({self.names})"
259 def as_group(self) -> DimensionGroup:
260 """Return ``self``.
262 Returns
263 -------
264 group : `DimensionGroup`
265 Returns itself.
267 Notes
268 -----
269 This is a backwards-compatibility API that allows both `DimensionGraph`
270 and `DimensionGroup` to be coerced to the latter.
271 """
272 return self
274 @cached_getter
275 def _as_graph(self) -> DimensionGraph:
276 """Return a view of ``self`` as a `DimensionGraph`.
278 Returns
279 -------
280 graph : `DimensionGraph`
281 The deprecated form of `DimensionGroup`.
283 Notes
284 -----
285 This is provided as a convenience for methods and properties that must
286 return a `DimensionGraph` for backwards compatibility (until v27). It
287 is the only way of making a `DimensionGraph` that does not produce
288 a warning.
289 """
290 from ._graph import DimensionGraph
292 result = object.__new__(DimensionGraph)
293 result._group = self
294 return result
296 def isdisjoint(self, other: DimensionGroup) -> bool:
297 """Test whether the intersection of two groups is empty.
299 Parameters
300 ----------
301 other : `DimensionGroup`
302 Other group to compare with.
304 Returns
305 -------
306 is_disjoin : `bool`
307 Returns `True` if either operand is the empty.
308 """
309 return self.names.isdisjoint(other.names)
311 def issubset(self, other: DimensionGroup) -> bool:
312 """Test whether all dimensions in ``self`` are also in ``other``.
314 Parameters
315 ----------
316 other : `DimensionGroup`
317 Other group to compare with.
319 Returns
320 -------
321 is_subset : `bool`
322 Returns `True` if ``self`` is empty.
323 """
324 return self.names <= other.names
326 def issuperset(self, other: DimensionGroup) -> bool:
327 """Test whether all dimensions in ``other`` are also in ``self``.
329 Parameters
330 ----------
331 other : `DimensionGroup`
332 Other group to compare with.
334 Returns
335 -------
336 is_superset : `bool`
337 Returns `True` if ``other`` is empty.
338 """
339 return self.names >= other.names
341 def __eq__(self, other: Any) -> bool:
342 from ._graph import DimensionGraph
344 # TODO: Drop DimensionGraph support here on DM-41326.
345 if isinstance(other, (DimensionGroup, DimensionGraph)):
346 return self.names == other.names
347 else:
348 return False
350 def __hash__(self) -> int:
351 return hash(self.required._seq)
353 def __le__(self, other: DimensionGroup) -> bool:
354 return self.names <= other.names
356 def __ge__(self, other: DimensionGroup) -> bool:
357 return self.names >= other.names
359 def __lt__(self, other: DimensionGroup) -> bool:
360 return self.names < other.names
362 def __gt__(self, other: DimensionGroup) -> bool:
363 return self.names > other.names
365 def union(self, *others: DimensionGroup) -> DimensionGroup:
366 """Construct a new group with all dimensions in any of the operands.
368 Parameters
369 ----------
370 *others : `DimensionGroup`
371 Other groups to join with.
373 Returns
374 -------
375 union : `DimensionGroup`
376 Union of all the groups.
378 Notes
379 -----
380 The elements of the returned group may exceed the naive union of their
381 elements, as some dimension elements are included in groups whenever
382 multiple dimensions are present, and those dependency dimensions could
383 have been provided by different operands.
384 """
385 names = set(self.names).union(*[other.names for other in others])
386 return DimensionGroup(self.universe, names)
388 def intersection(self, *others: DimensionGroup) -> DimensionGroup:
389 """Construct a new group with only dimensions in all of the operands.
391 Parameters
392 ----------
393 *others : `DimensionGroup`
394 Other groups to compare with.
396 Returns
397 -------
398 inter : `DimensionGroup`
399 Intersection of all the groups.
401 Notes
402 -----
403 See also `union`.
404 """
405 names = set(self.names).intersection(*[other.names for other in others])
406 return DimensionGroup(self.universe, names=names)
408 def __or__(self, other: DimensionGroup) -> DimensionGroup:
409 return self.union(other)
411 def __and__(self, other: DimensionGroup) -> DimensionGroup:
412 return self.intersection(other)
414 @property
415 def data_coordinate_keys(self) -> Set[str]:
416 """A set of dimensions ordered like `DataCoordinate.mapping`.
418 This order is defined as all required dimensions followed by all
419 implied dimensions.
420 """
421 return self._data_coordinate_indices.keys()
423 @property
424 @cached_getter
425 def lookup_order(self) -> tuple[str, ...]:
426 """A tuple of all elements in the order needed to find their records.
428 Unlike the table definition/topological order (which is what
429 `DimensionUniverse.sorted` gives you), when dimension A implies
430 dimension B, dimension A appears first.
431 """
432 done: set[str] = set()
433 order: list[str] = []
435 def add_to_order(element: DimensionElement) -> None:
436 if element.name in done:
437 return
438 predecessors = set(element.required.names)
439 predecessors.discard(element.name)
440 if not done.issuperset(predecessors):
441 return
442 order.append(element.name)
443 done.add(element.name)
444 for other in element.implied:
445 add_to_order(other)
447 while not done.issuperset(self.required):
448 for dimension in self.required:
449 add_to_order(self.universe[dimension])
451 order.extend(element for element in self.elements if element not in done)
452 return tuple(order)
454 @property
455 def spatial(self) -> NamedValueAbstractSet[TopologicalFamily]:
456 """Families represented by the spatial elements in this graph."""
457 return self._space_families[TopologicalSpace.SPATIAL]
459 @property
460 def temporal(self) -> NamedValueAbstractSet[TopologicalFamily]:
461 """Families represented by the temporal elements in this graph."""
462 return self._space_families[TopologicalSpace.TEMPORAL]
464 # Class attributes below are shadowed by instance attributes, and are
465 # present just to hold the docstrings for those instance attributes.
467 universe: DimensionUniverse
468 """The set of all known dimensions, of which this group is a subset
469 (`DimensionUniverse`).
470 """
472 names: SortedSequenceSet
473 """A true `~collections.abc.Set` of the dimension names.
475 Iteration order is consist with `DimensionUniverse.sorted`: each dimension
476 is preceded by its required and implied dependencies.
477 """
479 elements: SortedSequenceSet
480 """A true `~collections.abc.Set` of all dimension element names in the
481 group; a superset of `dimensions`.
482 """
484 governors: SortedSequenceSet
485 """A true `~collections.abc.Set` of all governor dimension names in the
486 group.
487 """
489 skypix: SortedSequenceSet
490 """A true `~collections.abc.Set` of all skypix dimension names in the "
491 group.
492 """
494 required: SortedSequenceSet
495 """The dimensions that must be directly identified via their primary keys
496 in a data ID in order to identify the rest of the elements in the group.
497 """
499 implied: SortedSequenceSet
500 """The dimensions that need not be directly identified via their primary
501 keys in a data ID.
502 """
504 _space_families: Mapping[TopologicalSpace, NamedValueAbstractSet[TopologicalFamily]]
505 """Families of elements in this graph that exist in topological spaces
506 relationships (`~collections.abc.Mapping` from `TopologicalSpace` to
507 `NamedValueAbstractSet` of `TopologicalFamily`).
508 """
510 _data_coordinate_indices: dict[str, int]