Coverage for python/lsst/daf/butler/core/dimensions/_coordinate.py: 27%
345 statements
« prev ^ index » next coverage.py v7.2.4, created at 2023-04-29 02:58 -0700
« prev ^ index » next coverage.py v7.2.4, created at 2023-04-29 02:58 -0700
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/>.
22#
23# Design notes for this module are in
24# doc/lsst.daf.butler/dev/dataCoordinate.py.
25#
27from __future__ import annotations
29__all__ = ("DataCoordinate", "DataId", "DataIdKey", "DataIdValue", "SerializedDataCoordinate")
31import numbers
32from abc import abstractmethod
33from typing import (
34 TYPE_CHECKING,
35 AbstractSet,
36 Any,
37 ClassVar,
38 Dict,
39 Iterator,
40 Literal,
41 Mapping,
42 Optional,
43 Tuple,
44 Union,
45 overload,
46)
48from deprecated.sphinx import deprecated
49from lsst.sphgeom import IntersectionRegion, Region
50from pydantic import BaseModel
52from ..json import from_json_pydantic, to_json_pydantic
53from ..named import NamedKeyDict, NamedKeyMapping, NamedValueAbstractSet, NameLookupMapping
54from ..timespan import Timespan
55from ._elements import Dimension, DimensionElement
56from ._graph import DimensionGraph
57from ._records import DimensionRecord, SerializedDimensionRecord
59if TYPE_CHECKING: # Imports needed only for type annotations; may be circular.
60 from ...registry import Registry
61 from ._universe import DimensionUniverse
63DataIdKey = Union[str, Dimension]
64"""Type annotation alias for the keys that can be used to index a
65DataCoordinate.
66"""
68# Pydantic will cast int to str if str is first in the Union.
69DataIdValue = Union[int, str, None]
70"""Type annotation alias for the values that can be present in a
71DataCoordinate or other data ID.
72"""
75class SerializedDataCoordinate(BaseModel):
76 """Simplified model for serializing a `DataCoordinate`."""
78 dataId: Dict[str, DataIdValue]
79 records: Optional[Dict[str, SerializedDimensionRecord]] = None
81 @classmethod
82 def direct(cls, *, dataId: Dict[str, DataIdValue], records: Dict[str, Dict]) -> SerializedDataCoordinate:
83 """Construct a `SerializedDataCoordinate` directly without validators.
85 This differs from the pydantic "construct" method in that the arguments
86 are explicitly what the model requires, and it will recurse through
87 members, constructing them from their corresponding `direct` methods.
89 This method should only be called when the inputs are trusted.
90 """
91 node = SerializedDataCoordinate.__new__(cls)
92 setter = object.__setattr__
93 setter(node, "dataId", dataId)
94 setter(
95 node,
96 "records",
97 records
98 if records is None
99 else {k: SerializedDimensionRecord.direct(**v) for k, v in records.items()},
100 )
101 setter(node, "__fields_set__", {"dataId", "records"})
102 return node
105def _intersectRegions(*args: Region) -> Optional[Region]:
106 """Return the intersection of several regions.
108 For internal use by `ExpandedDataCoordinate` only.
110 If no regions are provided, returns `None`.
111 """
112 if len(args) == 0:
113 return None
114 else:
115 result = args[0]
116 for n in range(1, len(args)):
117 result = IntersectionRegion(result, args[n])
118 return result
121class DataCoordinate(NamedKeyMapping[Dimension, DataIdValue]):
122 """Data ID dictionary.
124 An immutable data ID dictionary that guarantees that its key-value pairs
125 identify at least all required dimensions in a `DimensionGraph`.
127 `DataCoordinate` itself is an ABC, but provides `staticmethod` factory
128 functions for private concrete implementations that should be sufficient
129 for most purposes. `standardize` is the most flexible and safe of these;
130 the others (`makeEmpty`, `fromRequiredValues`, and `fromFullValues`) are
131 more specialized and perform little or no checking of inputs.
133 Notes
134 -----
135 Like any data ID class, `DataCoordinate` behaves like a dictionary, but
136 with some subtleties:
138 - Both `Dimension` instances and `str` names thereof may be used as keys
139 in lookup operations, but iteration (and `keys`) will yield `Dimension`
140 instances. The `names` property can be used to obtain the corresponding
141 `str` names.
143 - Lookups for implied dimensions (those in ``self.graph.implied``) are
144 supported if and only if `hasFull` returns `True`, and are never
145 included in iteration or `keys`. The `full` property may be used to
146 obtain a mapping whose keys do include implied dimensions.
148 - Equality comparison with other mappings is supported, but it always
149 considers only required dimensions (as well as requiring both operands
150 to identify the same dimensions). This is not quite consistent with the
151 way mappings usually work - normally differing keys imply unequal
152 mappings - but it makes sense in this context because data IDs with the
153 same values for required dimensions but different values for implied
154 dimensions represent a serious problem with the data that
155 `DataCoordinate` cannot generally recognize on its own, and a data ID
156 that knows implied dimension values should still be able to compare as
157 equal to one that does not. This is of course not the way comparisons
158 between simple `dict` data IDs work, and hence using a `DataCoordinate`
159 instance for at least one operand in any data ID comparison is strongly
160 recommended.
162 See also
163 --------
164 :ref:`lsst.daf.butler-dimensions_data_ids`
165 """
167 __slots__ = ()
169 _serializedType = SerializedDataCoordinate
171 @staticmethod
172 def standardize(
173 mapping: Optional[NameLookupMapping[Dimension, DataIdValue]] = None,
174 *,
175 graph: Optional[DimensionGraph] = None,
176 universe: Optional[DimensionUniverse] = None,
177 defaults: Optional[DataCoordinate] = None,
178 **kwargs: Any,
179 ) -> DataCoordinate:
180 """Standardize the supplied dataId.
182 Adapts an arbitrary mapping and/or additional arguments into a true
183 `DataCoordinate`, or augment an existing one.
185 Parameters
186 ----------
187 mapping : `~collections.abc.Mapping`, optional
188 An informal data ID that maps dimensions or dimension names to
189 their primary key values (may also be a true `DataCoordinate`).
190 graph : `DimensionGraph`
191 The dimensions to be identified by the new `DataCoordinate`.
192 If not provided, will be inferred from the keys of ``mapping`` and
193 ``**kwargs``, and ``universe`` must be provided unless ``mapping``
194 is already a `DataCoordinate`.
195 universe : `DimensionUniverse`
196 All known dimensions and their relationships; used to expand
197 and validate dependencies when ``graph`` is not provided.
198 defaults : `DataCoordinate`, optional
199 Default dimension key-value pairs to use when needed. These are
200 never used to infer ``graph``, and are ignored if a different value
201 is provided for the same key in ``mapping`` or `**kwargs``.
202 **kwargs
203 Additional keyword arguments are treated like additional key-value
204 pairs in ``mapping``.
206 Returns
207 -------
208 coordinate : `DataCoordinate`
209 A validated `DataCoordinate` instance.
211 Raises
212 ------
213 TypeError
214 Raised if the set of optional arguments provided is not supported.
215 KeyError
216 Raised if a key-value pair for a required dimension is missing.
217 """
218 d: Dict[str, DataIdValue] = {}
219 if isinstance(mapping, DataCoordinate):
220 if graph is None:
221 if not kwargs:
222 # Already standardized to exactly what we want.
223 return mapping
224 elif kwargs.keys().isdisjoint(graph.dimensions.names):
225 # User provided kwargs, but told us not to use them by
226 # passing in dimensions that are disjoint from those kwargs.
227 # This is not necessarily user error - it's a useful pattern
228 # to pass in all of the key-value pairs you have and let the
229 # code here pull out only what it needs.
230 return mapping.subset(graph)
231 assert universe is None or universe == mapping.universe
232 universe = mapping.universe
233 d.update((name, mapping[name]) for name in mapping.graph.required.names)
234 if mapping.hasFull():
235 d.update((name, mapping[name]) for name in mapping.graph.implied.names)
236 elif isinstance(mapping, NamedKeyMapping):
237 d.update(mapping.byName())
238 elif mapping is not None:
239 d.update(mapping)
240 d.update(kwargs)
241 if graph is None:
242 if defaults is not None:
243 universe = defaults.universe
244 elif universe is None:
245 raise TypeError("universe must be provided if graph is not.")
246 graph = DimensionGraph(universe, names=d.keys())
247 if not graph.dimensions:
248 return DataCoordinate.makeEmpty(graph.universe)
249 if defaults is not None:
250 if defaults.hasFull():
251 for k, v in defaults.full.items():
252 d.setdefault(k.name, v)
253 else:
254 for k, v in defaults.items():
255 d.setdefault(k.name, v)
256 if d.keys() >= graph.dimensions.names:
257 values = tuple(d[name] for name in graph._dataCoordinateIndices.keys())
258 else:
259 try:
260 values = tuple(d[name] for name in graph.required.names)
261 except KeyError as err:
262 raise KeyError(f"No value in data ID ({mapping}) for required dimension {err}.") from err
263 # Some backends cannot handle numpy.int64 type which is a subclass of
264 # numbers.Integral; convert that to int.
265 values = tuple(
266 int(val) if isinstance(val, numbers.Integral) else val for val in values # type: ignore
267 )
268 return _BasicTupleDataCoordinate(graph, values)
270 @staticmethod
271 def makeEmpty(universe: DimensionUniverse) -> DataCoordinate:
272 """Return an empty `DataCoordinate`.
274 It identifies the null set of dimensions.
276 Parameters
277 ----------
278 universe : `DimensionUniverse`
279 Universe to which this null dimension set belongs.
281 Returns
282 -------
283 dataId : `DataCoordinate`
284 A data ID object that identifies no dimensions. `hasFull` and
285 `hasRecords` are guaranteed to return `True`, because both `full`
286 and `records` are just empty mappings.
287 """
288 return _ExpandedTupleDataCoordinate(universe.empty, (), {})
290 @staticmethod
291 def fromRequiredValues(graph: DimensionGraph, values: Tuple[DataIdValue, ...]) -> DataCoordinate:
292 """Construct a `DataCoordinate` from required dimension values.
294 This is a low-level interface with at most assertion-level checking of
295 inputs. Most callers should use `standardize` instead.
297 Parameters
298 ----------
299 graph : `DimensionGraph`
300 Dimensions this data ID will identify.
301 values : `tuple` [ `int` or `str` ]
302 Tuple of primary key values corresponding to ``graph.required``,
303 in that order.
305 Returns
306 -------
307 dataId : `DataCoordinate`
308 A data ID object that identifies the given dimensions.
309 ``dataId.hasFull()`` will return `True` if and only if
310 ``graph.implied`` is empty, and ``dataId.hasRecords()`` will never
311 return `True`.
312 """
313 assert len(graph.required) == len(
314 values
315 ), f"Inconsistency between dimensions {graph.required} and required values {values}."
316 return _BasicTupleDataCoordinate(graph, values)
318 @staticmethod
319 def fromFullValues(graph: DimensionGraph, values: Tuple[DataIdValue, ...]) -> DataCoordinate:
320 """Construct a `DataCoordinate` from all dimension values.
322 This is a low-level interface with at most assertion-level checking of
323 inputs. Most callers should use `standardize` instead.
325 Parameters
326 ----------
327 graph : `DimensionGraph`
328 Dimensions this data ID will identify.
329 values : `tuple` [ `int` or `str` ]
330 Tuple of primary key values corresponding to
331 ``itertools.chain(graph.required, graph.implied)``, in that order.
332 Note that this is _not_ the same order as ``graph.dimensions``,
333 though these contain the same elements.
335 Returns
336 -------
337 dataId : `DataCoordinate`
338 A data ID object that identifies the given dimensions.
339 ``dataId.hasFull()`` will return `True` if and only if
340 ``graph.implied`` is empty, and ``dataId.hasRecords()`` will never
341 return `True`.
342 """
343 assert len(graph.dimensions) == len(
344 values
345 ), f"Inconsistency between dimensions {graph.dimensions} and full values {values}."
346 return _BasicTupleDataCoordinate(graph, values)
348 def __hash__(self) -> int:
349 return hash((self.graph,) + tuple(self[d.name] for d in self.graph.required))
351 def __eq__(self, other: Any) -> bool:
352 if not isinstance(other, DataCoordinate):
353 other = DataCoordinate.standardize(other, universe=self.universe)
354 return self.graph == other.graph and all(self[d.name] == other[d.name] for d in self.graph.required)
356 def __repr__(self) -> str:
357 # We can't make repr yield something that could be exec'd here without
358 # printing out the whole DimensionUniverse the graph is derived from.
359 # So we print something that mostly looks like a dict, but doesn't
360 # quote its keys: that's both more compact and something that can't
361 # be mistaken for an actual dict or something that could be exec'd.
362 terms = [f"{d}: {self[d]!r}" for d in self.graph.required.names]
363 if self.hasFull() and self.graph.required != self.graph.dimensions:
364 terms.append("...")
365 return "{{{}}}".format(", ".join(terms))
367 def __lt__(self, other: Any) -> bool:
368 # Allow DataCoordinate to be sorted
369 if not isinstance(other, type(self)):
370 return NotImplemented
371 # Form tuple of tuples for each DataCoordinate:
372 # Unlike repr() we only use required keys here to ensure that
373 # __eq__ can not be true simultaneously with __lt__ being true.
374 self_kv = tuple(self.items())
375 other_kv = tuple(other.items())
377 return self_kv < other_kv
379 def __iter__(self) -> Iterator[Dimension]:
380 return iter(self.keys())
382 def __len__(self) -> int:
383 return len(self.keys())
385 def keys(self) -> NamedValueAbstractSet[Dimension]: # type: ignore
386 return self.graph.required
388 @property
389 def names(self) -> AbstractSet[str]:
390 """Names of the required dimensions identified by this data ID.
392 They are returned in the same order as `keys`
393 (`collections.abc.Set` [ `str` ]).
394 """
395 return self.keys().names
397 @abstractmethod
398 def subset(self, graph: DimensionGraph) -> DataCoordinate:
399 """Return a `DataCoordinate` whose graph is a subset of ``self.graph``.
401 Parameters
402 ----------
403 graph : `DimensionGraph`
404 The dimensions identified by the returned `DataCoordinate`.
406 Returns
407 -------
408 coordinate : `DataCoordinate`
409 A `DataCoordinate` instance that identifies only the given
410 dimensions. May be ``self`` if ``graph == self.graph``.
412 Raises
413 ------
414 KeyError
415 Raised if the primary key value for one or more required dimensions
416 is unknown. This may happen if ``graph.issubset(self.graph)`` is
417 `False`, or even if ``graph.issubset(self.graph)`` is `True`, if
418 ``self.hasFull()`` is `False` and
419 ``graph.required.issubset(self.graph.required)`` is `False`. As
420 an example of the latter case, consider trying to go from a data ID
421 with dimensions {instrument, physical_filter, band} to
422 just {instrument, band}; band is implied by
423 physical_filter and hence would have no value in the original data
424 ID if ``self.hasFull()`` is `False`.
426 Notes
427 -----
428 If `hasFull` and `hasRecords` return `True` on ``self``, they will
429 return `True` (respectively) on the returned `DataCoordinate` as well.
430 The converse does not hold.
431 """
432 raise NotImplementedError()
434 @abstractmethod
435 def union(self, other: DataCoordinate) -> DataCoordinate:
436 """Combine two data IDs.
438 Yields a new one that identifies all dimensions that either of them
439 identify.
441 Parameters
442 ----------
443 other : `DataCoordinate`
444 Data ID to combine with ``self``.
446 Returns
447 -------
448 unioned : `DataCoordinate`
449 A `DataCoordinate` instance that satisfies
450 ``unioned.graph == self.graph.union(other.graph)``. Will preserve
451 ``hasFull`` and ``hasRecords`` whenever possible.
453 Notes
454 -----
455 No checking for consistency is performed on values for keys that
456 ``self`` and ``other`` have in common, and which value is included in
457 the returned data ID is not specified.
458 """
459 raise NotImplementedError()
461 @abstractmethod
462 def expanded(
463 self, records: NameLookupMapping[DimensionElement, Optional[DimensionRecord]]
464 ) -> DataCoordinate:
465 """Return a `DataCoordinate` that holds the given records.
467 Guarantees that `hasRecords` returns `True`.
469 This is a low-level interface with at most assertion-level checking of
470 inputs. Most callers should use `Registry.expandDataId` instead.
472 Parameters
473 ----------
474 records : `Mapping` [ `str`, `DimensionRecord` or `None` ]
475 A `NamedKeyMapping` with `DimensionElement` keys or a regular
476 `Mapping` with `str` (`DimensionElement` name) keys and
477 `DimensionRecord` values. Keys must cover all elements in
478 ``self.graph.elements``. Values may be `None`, but only to reflect
479 actual NULL values in the database, not just records that have not
480 been fetched.
481 """
482 raise NotImplementedError()
484 @property
485 def universe(self) -> DimensionUniverse:
486 """Universe that defines all known compatible dimensions.
488 The univers will be compatible with this coordinate
489 (`DimensionUniverse`).
490 """
491 return self.graph.universe
493 @property
494 @abstractmethod
495 def graph(self) -> DimensionGraph:
496 """Dimensions identified by this data ID (`DimensionGraph`).
498 Note that values are only required to be present for dimensions in
499 ``self.graph.required``; all others may be retrieved (from a
500 `Registry`) given these.
501 """
502 raise NotImplementedError()
504 @abstractmethod
505 def hasFull(self) -> bool:
506 """Whether this data ID contains implied and required values.
508 Returns
509 -------
510 state : `bool`
511 If `True`, `__getitem__`, `get`, and `__contains__` (but not
512 `keys`!) will act as though the mapping includes key-value pairs
513 for implied dimensions, and the `full` property may be used. If
514 `False`, these operations only include key-value pairs for required
515 dimensions, and accessing `full` is an error. Always `True` if
516 there are no implied dimensions.
517 """
518 raise NotImplementedError()
520 @property
521 def full(self) -> NamedKeyMapping[Dimension, DataIdValue]:
522 """Return mapping for all dimensions in ``self.graph``.
524 The mapping includes key-value pairs for all dimensions in
525 ``self.graph``, including implied (`NamedKeyMapping`).
527 Accessing this attribute if `hasFull` returns `False` is a logic error
528 that may raise an exception of unspecified type either immediately or
529 when implied keys are accessed via the returned mapping, depending on
530 the implementation and whether assertions are enabled.
531 """
532 assert self.hasFull(), "full may only be accessed if hasFull() returns True."
533 return _DataCoordinateFullView(self)
535 @abstractmethod
536 def hasRecords(self) -> bool:
537 """Whether this data ID contains records.
539 These are the records for all of the dimension elements it identifies.
541 Returns
542 -------
543 state : `bool`
544 If `True`, the following attributes may be accessed:
546 - `records`
547 - `region`
548 - `timespan`
549 - `pack`
551 If `False`, accessing any of these is considered a logic error.
552 """
553 raise NotImplementedError()
555 @property
556 def records(self) -> NamedKeyMapping[DimensionElement, Optional[DimensionRecord]]:
557 """Return the records.
559 Returns a mapping that contains `DimensionRecord` objects for all
560 elements identified by this data ID (`NamedKeyMapping`).
562 The values of this mapping may be `None` if and only if there is no
563 record for that element with these dimensions in the database (which
564 means some foreign key field must have a NULL value).
566 Accessing this attribute if `hasRecords` returns `False` is a logic
567 error that may raise an exception of unspecified type either
568 immediately or when the returned mapping is used, depending on the
569 implementation and whether assertions are enabled.
570 """
571 assert self.hasRecords(), "records may only be accessed if hasRecords() returns True."
572 return _DataCoordinateRecordsView(self)
574 @abstractmethod
575 def _record(self, name: str) -> Optional[DimensionRecord]:
576 """Protected implementation hook that backs the ``records`` attribute.
578 Parameters
579 ----------
580 name : `str`
581 The name of a `DimensionElement`, guaranteed to be in
582 ``self.graph.elements.names``.
584 Returns
585 -------
586 record : `DimensionRecord` or `None`
587 The dimension record for the given element identified by this
588 data ID, or `None` if there is no such record.
589 """
590 raise NotImplementedError()
592 @property
593 def region(self) -> Optional[Region]:
594 """Spatial region associated with this data ID.
596 (`lsst.sphgeom.Region` or `None`).
598 This is `None` if and only if ``self.graph.spatial`` is empty.
600 Accessing this attribute if `hasRecords` returns `False` is a logic
601 error that may or may not raise an exception, depending on the
602 implementation and whether assertions are enabled.
603 """
604 assert self.hasRecords(), "region may only be accessed if hasRecords() returns True."
605 regions = []
606 for family in self.graph.spatial:
607 element = family.choose(self.graph.elements)
608 record = self._record(element.name)
609 if record is None or record.region is None:
610 return None
611 else:
612 regions.append(record.region)
613 return _intersectRegions(*regions)
615 @property
616 def timespan(self) -> Optional[Timespan]:
617 """Temporal interval associated with this data ID.
619 (`Timespan` or `None`).
621 This is `None` if and only if ``self.graph.timespan`` is empty.
623 Accessing this attribute if `hasRecords` returns `False` is a logic
624 error that may or may not raise an exception, depending on the
625 implementation and whether assertions are enabled.
626 """
627 assert self.hasRecords(), "timespan may only be accessed if hasRecords() returns True."
628 timespans = []
629 for family in self.graph.temporal:
630 element = family.choose(self.graph.elements)
631 record = self._record(element.name)
632 # DimensionRecord subclasses for temporal elements always have
633 # .timespan, but they're dynamic so this can't be type-checked.
634 if record is None or record.timespan is None:
635 return None
636 else:
637 timespans.append(record.timespan)
638 if not timespans:
639 return None
640 elif len(timespans) == 1:
641 return timespans[0]
642 else:
643 return Timespan.intersection(*timespans)
645 @overload
646 def pack(self, name: str, *, returnMaxBits: Literal[True]) -> Tuple[int, int]:
647 ...
649 @overload
650 def pack(self, name: str, *, returnMaxBits: Literal[False]) -> int:
651 ...
653 # TODO: Remove this method and its overloads above on DM-38687.
654 @deprecated(
655 "Deprecated in favor of configurable dimension packers. Will be removed after v27.",
656 version="v26",
657 category=FutureWarning,
658 )
659 def pack(self, name: str, *, returnMaxBits: bool = False) -> Union[Tuple[int, int], int]:
660 """Pack this data ID into an integer.
662 Parameters
663 ----------
664 name : `str`
665 Name of the `DimensionPacker` algorithm (as defined in the
666 dimension configuration).
667 returnMaxBits : `bool`, optional
668 If `True` (`False` is default), return the maximum number of
669 nonzero bits in the returned integer across all data IDs.
671 Returns
672 -------
673 packed : `int`
674 Integer ID. This ID is unique only across data IDs that have
675 the same values for the packer's "fixed" dimensions.
676 maxBits : `int`, optional
677 Maximum number of nonzero bits in ``packed``. Not returned unless
678 ``returnMaxBits`` is `True`.
680 Notes
681 -----
682 Accessing this attribute if `hasRecords` returns `False` is a logic
683 error that may or may not raise an exception, depending on the
684 implementation and whether assertions are enabled.
685 """
686 assert self.hasRecords(), "pack() may only be called if hasRecords() returns True."
687 return self.universe.makePacker(name, self).pack(self, returnMaxBits=returnMaxBits)
689 def to_simple(self, minimal: bool = False) -> SerializedDataCoordinate:
690 """Convert this class to a simple python type.
692 This is suitable for serialization.
694 Parameters
695 ----------
696 minimal : `bool`, optional
697 Use minimal serialization. If set the records will not be attached.
699 Returns
700 -------
701 simple : `SerializedDataCoordinate`
702 The object converted to simple form.
703 """
704 # Convert to a dict form
705 if self.hasFull():
706 dataId = self.full.byName()
707 else:
708 dataId = self.byName()
709 records: Optional[Dict[str, SerializedDimensionRecord]]
710 if not minimal and self.hasRecords():
711 records = {k: v.to_simple() for k, v in self.records.byName().items() if v is not None}
712 else:
713 records = None
715 return SerializedDataCoordinate(dataId=dataId, records=records)
717 @classmethod
718 def from_simple(
719 cls,
720 simple: SerializedDataCoordinate,
721 universe: Optional[DimensionUniverse] = None,
722 registry: Optional[Registry] = None,
723 ) -> DataCoordinate:
724 """Construct a new object from the simplified form.
726 The data is assumed to be of the form returned from the `to_simple`
727 method.
729 Parameters
730 ----------
731 simple : `dict` of [`str`, `Any`]
732 The `dict` returned by `to_simple()`.
733 universe : `DimensionUniverse`
734 The special graph of all known dimensions.
735 registry : `lsst.daf.butler.Registry`, optional
736 Registry from which a universe can be extracted. Can be `None`
737 if universe is provided explicitly.
739 Returns
740 -------
741 dataId : `DataCoordinate`
742 Newly-constructed object.
743 """
744 if universe is None and registry is None:
745 raise ValueError("One of universe or registry is required to convert a dict to a DataCoordinate")
746 if universe is None and registry is not None:
747 universe = registry.dimensions
748 if universe is None:
749 # this is for mypy
750 raise ValueError("Unable to determine a usable universe")
752 dataId = cls.standardize(simple.dataId, universe=universe)
753 if simple.records:
754 dataId = dataId.expanded(
755 {k: DimensionRecord.from_simple(v, universe=universe) for k, v in simple.records.items()}
756 )
757 return dataId
759 to_json = to_json_pydantic
760 from_json: ClassVar = classmethod(from_json_pydantic)
763DataId = Union[DataCoordinate, Mapping[str, Any]]
764"""A type-annotation alias for signatures that accept both informal data ID
765dictionaries and validated `DataCoordinate` instances.
766"""
769class _DataCoordinateFullView(NamedKeyMapping[Dimension, DataIdValue]):
770 """View class for `DataCoordinate.full`.
772 Provides the default implementation for
773 `DataCoordinate.full`.
775 Parameters
776 ----------
777 target : `DataCoordinate`
778 The `DataCoordinate` instance this object provides a view of.
779 """
781 def __init__(self, target: DataCoordinate):
782 self._target = target
784 __slots__ = ("_target",)
786 def __repr__(self) -> str:
787 terms = [f"{d}: {self[d]!r}" for d in self._target.graph.dimensions.names]
788 return "{{{}}}".format(", ".join(terms))
790 def __getitem__(self, key: DataIdKey) -> DataIdValue:
791 return self._target[key]
793 def __iter__(self) -> Iterator[Dimension]:
794 return iter(self.keys())
796 def __len__(self) -> int:
797 return len(self.keys())
799 def keys(self) -> NamedValueAbstractSet[Dimension]: # type: ignore
800 return self._target.graph.dimensions
802 @property
803 def names(self) -> AbstractSet[str]:
804 # Docstring inherited from `NamedKeyMapping`.
805 return self.keys().names
808class _DataCoordinateRecordsView(NamedKeyMapping[DimensionElement, Optional[DimensionRecord]]):
809 """View class for `DataCoordinate.records`.
811 Provides the default implementation for
812 `DataCoordinate.records`.
814 Parameters
815 ----------
816 target : `DataCoordinate`
817 The `DataCoordinate` instance this object provides a view of.
818 """
820 def __init__(self, target: DataCoordinate):
821 self._target = target
823 __slots__ = ("_target",)
825 def __repr__(self) -> str:
826 terms = [f"{d}: {self[d]!r}" for d in self._target.graph.elements.names]
827 return "{{{}}}".format(", ".join(terms))
829 def __str__(self) -> str:
830 return "\n".join(str(v) for v in self.values())
832 def __getitem__(self, key: Union[DimensionElement, str]) -> Optional[DimensionRecord]:
833 if isinstance(key, DimensionElement):
834 key = key.name
835 return self._target._record(key)
837 def __iter__(self) -> Iterator[DimensionElement]:
838 return iter(self.keys())
840 def __len__(self) -> int:
841 return len(self.keys())
843 def keys(self) -> NamedValueAbstractSet[DimensionElement]: # type: ignore
844 return self._target.graph.elements
846 @property
847 def names(self) -> AbstractSet[str]:
848 # Docstring inherited from `NamedKeyMapping`.
849 return self.keys().names
852class _BasicTupleDataCoordinate(DataCoordinate):
853 """Standard implementation of `DataCoordinate`.
855 Backed by a tuple of values.
857 This class should only be accessed outside this module via the
858 `DataCoordinate` interface, and should only be constructed via the static
859 methods there.
861 Parameters
862 ----------
863 graph : `DimensionGraph`
864 The dimensions to be identified.
865 values : `tuple` [ `int` or `str` ]
866 Data ID values, ordered to match ``graph._dataCoordinateIndices``. May
867 include values for just required dimensions (which always come first)
868 or all dimensions.
869 """
871 def __init__(self, graph: DimensionGraph, values: Tuple[DataIdValue, ...]):
872 self._graph = graph
873 self._values = values
875 __slots__ = ("_graph", "_values")
877 @property
878 def graph(self) -> DimensionGraph:
879 # Docstring inherited from DataCoordinate.
880 return self._graph
882 def __getitem__(self, key: DataIdKey) -> DataIdValue:
883 # Docstring inherited from DataCoordinate.
884 if isinstance(key, Dimension):
885 key = key.name
886 index = self._graph._dataCoordinateIndices[key]
887 try:
888 return self._values[index]
889 except IndexError:
890 # Caller asked for an implied dimension, but this object only has
891 # values for the required ones.
892 raise KeyError(key) from None
894 def subset(self, graph: DimensionGraph) -> DataCoordinate:
895 # Docstring inherited from DataCoordinate.
896 if self._graph == graph:
897 return self
898 elif self.hasFull() or self._graph.required >= graph.dimensions:
899 return _BasicTupleDataCoordinate(
900 graph,
901 tuple(self[k] for k in graph._dataCoordinateIndices.keys()),
902 )
903 else:
904 return _BasicTupleDataCoordinate(graph, tuple(self[k] for k in graph.required.names))
906 def union(self, other: DataCoordinate) -> DataCoordinate:
907 # Docstring inherited from DataCoordinate.
908 graph = self.graph.union(other.graph)
909 # See if one or both input data IDs is already what we want to return;
910 # if so, return the most complete one we have.
911 if other.graph == graph:
912 if self.graph == graph:
913 # Input data IDs have the same graph (which is also the result
914 # graph), but may not have the same content.
915 # other might have records; self does not, so try other first.
916 # If it at least has full values, it's no worse than self.
917 if other.hasFull():
918 return other
919 else:
920 return self
921 elif other.hasFull():
922 return other
923 # There's some chance that neither self nor other has full values,
924 # but together provide enough to the union to. Let the general
925 # case below handle that.
926 elif self.graph == graph:
927 # No chance at returning records. If self has full values, it's
928 # the best we can do.
929 if self.hasFull():
930 return self
931 # General case with actual merging of dictionaries.
932 values = self.full.byName() if self.hasFull() else self.byName()
933 values.update(other.full.byName() if other.hasFull() else other.byName())
934 return DataCoordinate.standardize(values, graph=graph)
936 def expanded(
937 self, records: NameLookupMapping[DimensionElement, Optional[DimensionRecord]]
938 ) -> DataCoordinate:
939 # Docstring inherited from DataCoordinate
940 values = self._values
941 if not self.hasFull():
942 # Extract a complete values tuple from the attributes of the given
943 # records. It's possible for these to be inconsistent with
944 # self._values (which is a serious problem, of course), but we've
945 # documented this as a no-checking API.
946 values += tuple(getattr(records[d.name], d.primaryKey.name) for d in self._graph.implied)
947 return _ExpandedTupleDataCoordinate(self._graph, values, records)
949 def hasFull(self) -> bool:
950 # Docstring inherited from DataCoordinate.
951 return len(self._values) == len(self._graph._dataCoordinateIndices)
953 def hasRecords(self) -> bool:
954 # Docstring inherited from DataCoordinate.
955 return False
957 def _record(self, name: str) -> Optional[DimensionRecord]:
958 # Docstring inherited from DataCoordinate.
959 assert False
961 def __reduce__(self) -> tuple[Any, ...]:
962 return (_BasicTupleDataCoordinate, (self._graph, self._values))
964 def __getattr__(self, name: str) -> Any:
965 if name in self.graph.elements.names:
966 raise AttributeError(
967 f"Dimension record attribute {name!r} is only available on expanded DataCoordinates."
968 )
969 raise AttributeError(name)
972class _ExpandedTupleDataCoordinate(_BasicTupleDataCoordinate):
973 """A `DataCoordinate` implementation that can hold `DimensionRecord`.
975 This class should only be accessed outside this module via the
976 `DataCoordinate` interface, and should only be constructed via calls to
977 `DataCoordinate.expanded`.
979 Parameters
980 ----------
981 graph : `DimensionGraph`
982 The dimensions to be identified.
983 values : `tuple` [ `int` or `str` ]
984 Data ID values, ordered to match ``graph._dataCoordinateIndices``.
985 May include values for just required dimensions (which always come
986 first) or all dimensions.
987 records : `Mapping` [ `str`, `DimensionRecord` or `None` ]
988 A `NamedKeyMapping` with `DimensionElement` keys or a regular
989 `Mapping` with `str` (`DimensionElement` name) keys and
990 `DimensionRecord` values. Keys must cover all elements in
991 ``self.graph.elements``. Values may be `None`, but only to reflect
992 actual NULL values in the database, not just records that have not
993 been fetched.
994 """
996 def __init__(
997 self,
998 graph: DimensionGraph,
999 values: Tuple[DataIdValue, ...],
1000 records: NameLookupMapping[DimensionElement, Optional[DimensionRecord]],
1001 ):
1002 super().__init__(graph, values)
1003 assert super().hasFull(), "This implementation requires full dimension records."
1004 self._records = records
1006 __slots__ = ("_records",)
1008 def subset(self, graph: DimensionGraph) -> DataCoordinate:
1009 # Docstring inherited from DataCoordinate.
1010 if self._graph == graph:
1011 return self
1012 return _ExpandedTupleDataCoordinate(
1013 graph, tuple(self[k] for k in graph._dataCoordinateIndices.keys()), records=self._records
1014 )
1016 def expanded(
1017 self, records: NameLookupMapping[DimensionElement, Optional[DimensionRecord]]
1018 ) -> DataCoordinate:
1019 # Docstring inherited from DataCoordinate.
1020 return self
1022 def union(self, other: DataCoordinate) -> DataCoordinate:
1023 # Docstring inherited from DataCoordinate.
1024 graph = self.graph.union(other.graph)
1025 # See if one or both input data IDs is already what we want to return;
1026 # if so, return the most complete one we have.
1027 if self.graph == graph:
1028 # self has records, so even if other is also a valid result, it's
1029 # no better.
1030 return self
1031 if other.graph == graph:
1032 # If other has full values, and self does not identify some of
1033 # those, it's the base we can do. It may have records, too.
1034 if other.hasFull():
1035 return other
1036 # If other does not have full values, there's a chance self may
1037 # provide the values needed to complete it. For example, self
1038 # could be {band} while other could be
1039 # {instrument, physical_filter, band}, with band unknown.
1040 # General case with actual merging of dictionaries.
1041 values = self.full.byName()
1042 values.update(other.full.byName() if other.hasFull() else other.byName())
1043 basic = DataCoordinate.standardize(values, graph=graph)
1044 # See if we can add records.
1045 if self.hasRecords() and other.hasRecords():
1046 # Sometimes the elements of a union of graphs can contain elements
1047 # that weren't in either input graph (because graph unions are only
1048 # on dimensions). e.g. {visit} | {detector} brings along
1049 # visit_detector_region.
1050 elements = set(graph.elements.names)
1051 elements -= self.graph.elements.names
1052 elements -= other.graph.elements.names
1053 if not elements:
1054 records = NamedKeyDict[DimensionElement, Optional[DimensionRecord]](self.records)
1055 records.update(other.records)
1056 return basic.expanded(records.freeze())
1057 return basic
1059 def hasFull(self) -> bool:
1060 # Docstring inherited from DataCoordinate.
1061 return True
1063 def hasRecords(self) -> bool:
1064 # Docstring inherited from DataCoordinate.
1065 return True
1067 def _record(self, name: str) -> Optional[DimensionRecord]:
1068 # Docstring inherited from DataCoordinate.
1069 return self._records[name]
1071 def __reduce__(self) -> tuple[Any, ...]:
1072 return (_ExpandedTupleDataCoordinate, (self._graph, self._values, self._records))
1074 def __getattr__(self, name: str) -> Any:
1075 try:
1076 return self._record(name)
1077 except KeyError:
1078 raise AttributeError(name) from None
1080 def __dir__(self) -> list[str]:
1081 result = list(super().__dir__())
1082 result.extend(self.graph.elements.names)
1083 return result