Coverage for python/lsst/ap/association/association.py : 21%

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 ap_association.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (https://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 <https://www.gnu.org/licenses/>.
22"""A simple implementation of source association task for ap_verify.
23"""
25__all__ = ["AssociationConfig", "AssociationTask"]
27import numpy as np
28import pandas as pd
29from scipy.spatial import cKDTree
31import lsst.geom as geom
32import lsst.pex.config as pexConfig
33import lsst.pipe.base as pipeBase
35# Enforce an error for unsafe column/array value setting in pandas.
36pd.options.mode.chained_assignment = 'raise'
39class AssociationConfig(pexConfig.Config):
40 """Config class for AssociationTask.
41 """
42 maxDistArcSeconds = pexConfig.Field(
43 dtype=float,
44 doc='Maximum distance in arcseconds to test for a DIASource to be a '
45 'match to a DIAObject.',
46 default=1.0,
47 )
50class AssociationTask(pipeBase.Task):
51 """Associate DIAOSources into existing DIAObjects.
53 This task performs the association of detected DIASources in a visit
54 with the previous DIAObjects detected over time. It also creates new
55 DIAObjects out of DIASources that cannot be associated with previously
56 detected DIAObjects.
57 """
59 ConfigClass = AssociationConfig
60 _DefaultName = "association"
62 @pipeBase.timeMethod
63 def run(self,
64 diaSources,
65 diaObjects,
66 diaSourceHistory):
67 """Associate the new DiaSources with existing or new DiaObjects,
68 updating the DiaObjects.
70 Parameters
71 ----------
72 diaSources : `pandas.DataFrame`
73 New DIASources to be associated with existing DIAObjects.
74 diaObjects : `pandas.DataFrame`
75 Existing diaObjects from the Apdb.
76 diaSourceHistory : `pandas.DataFrame`
77 12 month DiaSource history of the loaded ``diaObjects``.
79 Returns
80 -------
81 result : `lsst.pipe.base.Struct`
82 Results struct with components.
84 - ``diaObjects`` : Complete set of dia_objects covering the input
85 exposure. Catalog contains newly created, updated, and untouched
86 diaObjects. (`pandas.DataFrame`)
87 - ``updatedDiaObjects`` : Subset of DiaObjects that were updated
88 or created during processing. (`pandas.DataFrame`)
89 - ``matchedDiaObjectIds`` : DiaSources detected in this ccdVisit with
90 associated diaObjectIds. (`numpy.ndarray`)
91 """
92 diaSources = self.check_dia_source_radec(diaSources)
94 matchResult = self.associate_sources(diaObjects, diaSources)
96 # Now that we know the DiaObjects our new DiaSources are associated
97 # with, we index the new DiaSources the same way as the full history
98 # and merge the tables.
99 diaSources.set_index(["diaObjectId", "filterName", "diaSourceId"],
100 drop=False,
101 inplace=True)
102 # Append the newly created DiaObjectds.
103 diaObjects = diaObjects.append(matchResult.new_dia_objects,
104 sort=True)
105 # Double check to make sure there are no duplicates in the DiaObject
106 # table after association.
107 if diaObjects.index.has_duplicates:
108 raise RuntimeError(
109 "Duplicate DiaObjects created after association. This is "
110 "likely due to re-running data with an already populated "
111 "Apdb. If this was not the case then there was an unexpected "
112 "failure in Association while matching and creating new "
113 "DiaObjects and should be reported. Exiting.")
115 return pipeBase.Struct(
116 diaObjects=diaObjects,
117 diaSources=diaSources,
118 matchedDiaObjectIds=matchResult.associated_dia_object_ids,
119 )
121 def check_dia_source_radec(self, dia_sources):
122 """Check that all DiaSources have non-NaN values for RA/DEC.
124 If one or more DiaSources are found to have NaN values, throw a
125 warning to the log with the ids of the offending sources. Drop them
126 from the table.
128 Parameters
129 ----------
130 dia_sources : `pandas.DataFrame`
131 Input DiaSources to check for NaN values.
133 Returns
134 -------
135 trimmed_sources : `pandas.DataFrame`
136 DataFrame of DiaSources trimmed of all entries with NaN values for
137 RA/DEC.
138 """
139 nan_mask = (dia_sources.loc[:, "ra"].isnull()
140 | dia_sources.loc[:, "decl"].isnull())
141 if np.any(nan_mask):
142 nan_idxs = np.argwhere(nan_mask.to_numpy()).flatten()
143 for nan_idx in nan_idxs:
144 self.log.warning(
145 "DiaSource %i has NaN value for RA/DEC, "
146 "dropping from association." %
147 dia_sources.loc[nan_idx, "diaSourceId"])
148 dia_sources = dia_sources[~nan_mask]
149 return dia_sources
151 @pipeBase.timeMethod
152 def associate_sources(self, dia_objects, dia_sources):
153 """Associate the input DIASources with the catalog of DIAObjects.
155 DiaObject DataFrame must be indexed on ``diaObjectId``.
157 Parameters
158 ----------
159 dia_objects : `pandas.DataFrame`
160 Catalog of DIAObjects to attempt to associate the input
161 DIASources into.
162 dia_sources : `pandas.DataFrame`
163 DIASources to associate into the DIAObjectCollection.
165 Returns
166 -------
167 result : `lsst.pipeBase.Struct`
168 Results struct with components:
170 - ``updated_and_new_dia_object_ids`` : ids of new and updated
171 dia_objects as the result of association. (`list` of `int`).
172 - ``new_dia_objects`` : Newly created DiaObjects from
173 unassociated diaSources. (`pandas.DataFrame`)
174 - ``n_updated_dia_objects`` : Number of previously known
175 dia_objects with newly associated DIASources. (`int`).
176 - ``n_new_dia_objects`` : Number of newly created DIAObjects from
177 unassociated DIASources (`int`).
178 - ``n_unupdated_dia_objects`` : Number of previous DIAObjects that
179 were not associated to a new DIASource (`int`).
180 """
182 scores = self.score(
183 dia_objects, dia_sources,
184 self.config.maxDistArcSeconds * geom.arcseconds)
185 match_result = self.match(dia_objects, dia_sources, scores)
187 self._add_association_meta_data(match_result)
189 return match_result
191 @pipeBase.timeMethod
192 def score(self, dia_objects, dia_sources, max_dist):
193 """Compute a quality score for each dia_source/dia_object pair
194 between this catalog of DIAObjects and the input DIASource catalog.
196 ``max_dist`` sets maximum separation in arcseconds to consider a
197 dia_source a possible match to a dia_object. If the pair is
198 beyond this distance no score is computed.
200 Parameters
201 ----------
202 dia_objects : `pandas.DataFrame`
203 A contiguous catalog of DIAObjects to score against dia_sources.
204 dia_sources : `pandas.DataFrame`
205 A contiguous catalog of dia_sources to "score" based on distance
206 and (in the future) other metrics.
207 max_dist : `lsst.geom.Angle`
208 Maximum allowed distance to compute a score for a given DIAObject
209 DIASource pair.
211 Returns
212 -------
213 result : `lsst.pipe.base.Struct`
214 Results struct with components:
216 - ``scores``: array of floats of match quality updated DIAObjects
217 (array-like of `float`).
218 - ``obj_idxs``: indexes of the matched DIAObjects in the catalog.
219 (array-like of `int`)
220 - ``obj_ids``: array of floats of match quality updated DIAObjects
221 (array-like of `int`).
223 Default values for these arrays are
224 INF, -1, and -1 respectively for unassociated sources.
225 """
226 scores = np.full(len(dia_sources), np.inf, dtype=np.float64)
227 obj_idxs = np.full(len(dia_sources), -1, dtype=np.int64)
228 obj_ids = np.full(len(dia_sources), 0, dtype=np.int64)
230 if len(dia_objects) == 0:
231 return pipeBase.Struct(
232 scores=scores,
233 obj_idxs=obj_idxs,
234 obj_ids=obj_ids)
236 spatial_tree = self._make_spatial_tree(dia_objects)
238 max_dist_rad = max_dist.asRadians()
240 vectors = self._radec_to_xyz(dia_sources)
242 scores, obj_idxs = spatial_tree.query(
243 vectors,
244 distance_upper_bound=max_dist_rad)
245 matched_src_idxs = np.argwhere(np.isfinite(scores))
246 obj_ids[matched_src_idxs] = dia_objects.index.to_numpy()[
247 obj_idxs[matched_src_idxs]]
249 return pipeBase.Struct(
250 scores=scores,
251 obj_idxs=obj_idxs,
252 obj_ids=obj_ids)
254 def _make_spatial_tree(self, dia_objects):
255 """Create a searchable kd-tree the input dia_object positions.
257 Parameters
258 ----------
259 dia_objects : `pandas.DataFrame`
260 A catalog of DIAObjects to create the tree from.
262 Returns
263 -------
264 kd_tree : `scipy.spatical.cKDTree`
265 Searchable kd-tree created from the positions of the DIAObjects.
266 """
267 vectors = self._radec_to_xyz(dia_objects)
268 return cKDTree(vectors)
270 def _radec_to_xyz(self, catalog):
271 """Convert input ra/dec coordinates to spherical unit-vectors.
273 Parameters
274 ----------
275 catalog : `pandas.DataFrame`
276 Catalog to produce spherical unit-vector from.
278 Returns
279 -------
280 vectors : `numpy.ndarray`, (N, 3)
281 Output unit-vectors
282 """
283 ras = np.radians(catalog["ra"])
284 decs = np.radians(catalog["decl"])
285 vectors = np.empty((len(ras), 3))
287 sin_dec = np.sin(np.pi / 2 - decs)
288 vectors[:, 0] = sin_dec * np.cos(ras)
289 vectors[:, 1] = sin_dec * np.sin(ras)
290 vectors[:, 2] = np.cos(np.pi / 2 - decs)
292 return vectors
294 @pipeBase.timeMethod
295 def match(self, dia_objects, dia_sources, score_struct):
296 """Match DIAsources to DIAObjects given a score and create new
297 DIAObject Ids for new unassociated DIASources.
299 Parameters
300 ----------
301 dia_objects : `pandas.DataFrame`
302 A SourceCatalog of DIAObjects to associate to DIASources.
303 dia_sources : `pandas.DataFrame`
304 A contiguous catalog of dia_sources for which the set of scores
305 has been computed on with DIAObjectCollection.score.
306 score_struct : `lsst.pipe.base.Struct`
307 Results struct with components:
309 - ``scores``: array of floats of match quality
310 updated DIAObjects (array-like of `float`).
311 - ``obj_ids``: array of floats of match quality
312 updated DIAObjects (array-like of `int`).
313 - ``obj_idxs``: indexes of the matched DIAObjects in the catalog.
314 (array-like of `int`)
316 Default values for these arrays are
317 INF, -1 and -1 respectively for unassociated sources.
319 Returns
320 -------
321 result : `lsst.pipeBase.Struct`
322 Results struct with components:
324 - ``updated_and_new_dia_object_ids`` : ids of new and updated
325 dia_objects as the result of association. (`list` of `int`).
326 - ``new_dia_objects`` : Newly created DiaObjects from unassociated
327 diaSources. (`pandas.DataFrame`)
328 - ``n_updated_dia_objects`` : Number of previously know dia_objects
329 with newly associated DIASources. (`int`).
330 - ``n_new_dia_objects`` : Number of newly created DIAObjects from
331 unassociated DIASources (`int`).
332 - ``n_unupdated_dia_objects`` : Number of previous DIAObjects that
333 were not associated to a new DIASource (`int`).
334 """
336 n_previous_dia_objects = len(dia_objects)
337 used_dia_object = np.zeros(n_previous_dia_objects, dtype=bool)
338 used_dia_source = np.zeros(len(dia_sources), dtype=bool)
339 associated_dia_object_ids = np.zeros(len(dia_sources),
340 dtype=np.uint64)
341 new_dia_objects = []
343 n_updated_dia_objects = 0
344 n_new_dia_objects = 0
346 # We sort from best match to worst to effectively perform a
347 # "handshake" match where both the DIASources and DIAObjects agree
348 # their the best match. By sorting this way, scores with NaN (those
349 # sources that have no match and will create new DIAObjects) will be
350 # placed at the end of the array.
351 score_args = score_struct.scores.argsort(axis=None)
352 for score_idx in score_args:
353 if not np.isfinite(score_struct.scores[score_idx]):
354 # Thanks to the sorting the rest of the sources will be
355 # NaN for their score. We therefore exit the loop to append
356 # sources to a existing DIAObject, leaving these for
357 # the loop creating new objects.
358 break
359 dia_obj_idx = score_struct.obj_idxs[score_idx]
360 if used_dia_object[dia_obj_idx]:
361 continue
362 used_dia_object[dia_obj_idx] = True
363 used_dia_source[score_idx] = True
364 obj_id = score_struct.obj_ids[score_idx]
365 associated_dia_object_ids[score_idx] = obj_id
366 n_updated_dia_objects += 1
367 dia_sources.loc[score_idx, "diaObjectId"] = obj_id
369 # Argwhere returns a array shape (N, 1) so we access the index
370 # thusly to retrieve the value rather than the tuple
371 for (src_idx,) in np.argwhere(np.logical_not(used_dia_source)):
372 src_id = dia_sources.loc[src_idx, "diaSourceId"]
373 new_dia_objects.append(self._initialize_dia_object(src_id))
374 associated_dia_object_ids[src_idx] = src_id
375 dia_sources.loc[src_idx, "diaObjectId"] = src_id
376 n_new_dia_objects += 1
378 if len(new_dia_objects) > 0:
379 new_dia_objects = pd.DataFrame(data=new_dia_objects)
380 else:
381 # Create a junk DiaObject to get the columns.
382 tmpObj = self._initialize_dia_object(0)
383 new_dia_objects = pd.DataFrame(data=new_dia_objects,
384 columns=tmpObj.keys())
385 new_dia_objects.set_index("diaObjectId", inplace=True, drop=False)
387 # Return the ids of the DIAObjects in this DIAObjectCollection that
388 # were updated or newly created.
389 n_unassociated_dia_objects = \
390 n_previous_dia_objects - n_updated_dia_objects
391 return pipeBase.Struct(
392 associated_dia_object_ids=associated_dia_object_ids,
393 new_dia_objects=new_dia_objects,
394 n_updated_dia_objects=n_updated_dia_objects,
395 n_new_dia_objects=n_new_dia_objects,
396 n_unassociated_dia_objects=n_unassociated_dia_objects,)
398 def _initialize_dia_object(self, objId):
399 """Create a new DiaObject with values required to be initialized by the
400 Ppdb.
402 Parameters
403 ----------
404 objid : `int`
405 ``diaObjectId`` value for the of the new DiaObject.
407 Returns
408 -------
409 diaObject : `dict`
410 Newly created DiaObject with keys:
412 ``diaObjectId``
413 Unique DiaObjectId (`int`).
414 ``pmParallaxNdata``
415 Number of data points used for parallax calculation (`int`).
416 ``nearbyObj1``
417 Id of the a nearbyObject in the Object table (`int`).
418 ``nearbyObj2``
419 Id of the a nearbyObject in the Object table (`int`).
420 ``nearbyObj3``
421 Id of the a nearbyObject in the Object table (`int`).
422 ``?PSFluxData``
423 Number of data points used to calculate point source flux
424 summary statistics in each bandpass (`int`).
425 """
426 new_dia_object = {"diaObjectId": objId,
427 "pmParallaxNdata": 0,
428 "nearbyObj1": 0,
429 "nearbyObj2": 0,
430 "nearbyObj3": 0,
431 "flags": 0}
432 for f in ["u", "g", "r", "i", "z", "y"]:
433 new_dia_object["%sPSFluxNdata" % f] = 0
434 return new_dia_object
436 def _add_association_meta_data(self, match_result):
437 """Store summaries of the association step in the task metadata.
439 Parameters
440 ----------
441 match_result : `lsst.pipeBase.Struct`
442 Results struct with components:
444 - ``updated_and_new_dia_object_ids`` : ids new and updated
445 dia_objects in the collection (`list` of `int`).
446 - ``n_updated_dia_objects`` : Number of previously know dia_objects
447 with newly associated DIASources. (`int`).
448 - ``n_new_dia_objects`` : Number of newly created DIAObjects from
449 unassociated DIASources (`int`).
450 - ``n_unupdated_dia_objects`` : Number of previous DIAObjects that
451 were not associated to a new DIASource (`int`).
452 """
453 self.metadata.add('numUpdatedDiaObjects',
454 match_result.n_updated_dia_objects)
455 self.metadata.add('numNewDiaObjects',
456 match_result.n_new_dia_objects)
457 self.metadata.add('numUnassociatedDiaObjects',
458 match_result.n_unassociated_dia_objects)