-
Notifications
You must be signed in to change notification settings - Fork 1.4k
Expand file tree
/
Copy pathmonitoring_service.py
More file actions
1350 lines (1190 loc) · 47.7 KB
/
Copy pathmonitoring_service.py
File metadata and controls
1350 lines (1190 loc) · 47.7 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import logging
import math
import time
from collections import defaultdict
from datetime import date, datetime, timedelta, timezone
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
from feast.feature_logging import LOG_TIMESTAMP_FIELD, FeatureServiceLoggingSource
from feast.infra.offline_stores.offline_store import OfflineStore
from feast.monitoring.dqm_job_manager import DQMJobManager
from feast.monitoring.metrics_calculator import MetricsCalculator
from feast.monitoring.monitoring_utils import build_view_aggregate
if TYPE_CHECKING:
from feast.feature_store import FeatureStore
logger = logging.getLogger(__name__)
VALID_GRANULARITIES = OfflineStore.MONITORING_VALID_GRANULARITIES
_EPOCH = datetime(1970, 1, 1, tzinfo=timezone.utc)
_FAR_FUTURE = datetime(2099, 12, 31, 23, 59, 59, tzinfo=timezone.utc)
GRANULARITY_WINDOWS = {
"daily": timedelta(days=1),
"weekly": timedelta(days=7),
"biweekly": timedelta(days=14),
"monthly": timedelta(days=30),
"quarterly": timedelta(days=90),
}
def _as_utc_datetime(val: Any) -> Optional[datetime]:
"""Parse a timestamp-like value to a timezone-aware UTC datetime."""
if val is None:
return None
if isinstance(val, datetime):
return val if val.tzinfo else val.replace(tzinfo=timezone.utc)
if isinstance(val, date):
return datetime.combine(val, datetime.min.time(), tzinfo=timezone.utc)
if isinstance(val, str):
parsed = datetime.fromisoformat(val.replace("Z", "+00:00"))
return parsed if parsed.tzinfo else parsed.replace(tzinfo=timezone.utc)
return None
def _newest_event_in_window(
max_ts: Optional[datetime],
start_dt: datetime,
end_dt: datetime,
) -> Optional[datetime]:
"""Newest source event that falls inside ``[start_dt, end_dt]``.
``max_ts`` is MAX(event_timestamp) over the whole source. If that
timestamp is after the window, the window end is used as an upper bound.
"""
aware_max = _as_utc_datetime(max_ts)
if aware_max is None or aware_max < start_dt:
return None
return min(aware_max, end_dt)
_FLOAT_FIELDS = frozenset(
{
"null_rate",
"mean",
"stddev",
"min_val",
"max_val",
"p50",
"p75",
"p90",
"p95",
"p99",
"avg_null_rate",
"max_null_rate",
}
)
def _sanitize_floats(row: Dict[str, Any]) -> Dict[str, Any]:
"""Replace NaN/Inf float values with None so JSON serialization succeeds."""
for key in _FLOAT_FIELDS:
val = row.get(key)
if isinstance(val, float) and (math.isnan(val) or math.isinf(val)):
row[key] = None
return row
class MonitoringService:
def __init__(self, store: "FeatureStore"):
self._store = store
self._job_manager: Optional[DQMJobManager] = None
self._calculator = MetricsCalculator()
self._monitoring_tables_ensured = False
self._offline_store_cache = None
def _get_offline_store(self):
if self._offline_store_cache is None:
self._offline_store_cache = self._store._get_provider().offline_store
return self._offline_store_cache
def _ensure_monitoring_tables(self):
if not self._monitoring_tables_ensured:
self._get_offline_store().ensure_monitoring_tables(self._store.config)
self._monitoring_tables_ensured = True
@property
def job_manager(self) -> DQMJobManager:
if self._job_manager is None:
self._job_manager = DQMJobManager(
self._get_offline_store(), self._store.config
)
self._job_manager.ensure_table()
return self._job_manager
# ------------------------------------------------------------------ #
# Auto-compute: detect dates, compute all granularities
# ------------------------------------------------------------------ #
def auto_compute(
self,
project: Optional[str] = None,
feature_view_name: Optional[str] = None,
) -> Dict[str, Any]:
"""Detect date ranges from source data and compute all granularities.
Also computes baseline for any features that don't have one yet
(idempotent -- features with existing baselines are skipped).
"""
start_time = time.time()
self._ensure_monitoring_tables()
if project is None:
project = self._store.config.project
feature_views = self._resolve_feature_views(project, feature_view_name)
total_features = 0
total_views = 0
baseline_features = 0
granularities_computed = set()
for fv in feature_views:
try:
feature_fields = self._classify_fields(fv)
if not feature_fields:
continue
max_ts = self._get_max_timestamp(fv)
if max_ts is None:
logger.warning(
"No data found for feature view '%s', skipping", fv.name
)
continue
now = datetime.now(timezone.utc)
# Compute baseline for features that don't have one yet
fields_needing_baseline = self._get_features_without_baseline(
project, fv
)
if fields_needing_baseline:
bl_fields = self._classify_fields(
fv, fields=fields_needing_baseline
)
if bl_fields:
bl_metrics = self._compute_feature_metrics(
fv,
bl_fields,
_EPOCH,
_FAR_FUTURE,
)
self._save_computed_metrics(
project=project,
feature_view=fv,
metrics_list=bl_metrics,
metric_date=date.today(),
granularity="baseline",
set_baseline=True,
now=now,
max_event_timestamp=max_ts,
)
baseline_features += len(bl_metrics)
for granularity, window in GRANULARITY_WINDOWS.items():
window_start = max_ts - window
metrics_list = self._compute_feature_metrics(
fv,
feature_fields,
window_start,
max_ts,
)
self._save_computed_metrics(
project=project,
feature_view=fv,
metrics_list=metrics_list,
metric_date=window_start.date(),
granularity=granularity,
set_baseline=False,
now=now,
max_event_timestamp=max_ts,
)
self._compute_feature_service_metrics(
project=project,
granularity=granularity,
metric_dates=[window_start.date()],
set_baseline=False,
)
total_features += len(metrics_list)
granularities_computed.add(granularity)
total_views += 1
except Exception:
logger.exception(
"Failed to auto-compute metrics for feature view '%s'", fv.name
)
duration_ms = int((time.time() - start_time) * 1000)
return {
"status": "completed",
"computed_feature_views": total_views,
"computed_features": total_features,
"baseline_features": baseline_features,
"granularities": sorted(granularities_computed),
"duration_ms": duration_ms,
}
# ------------------------------------------------------------------ #
# Log source: compute metrics from feature serving logs
# ------------------------------------------------------------------ #
def compute_log_metrics(
self,
project: str,
feature_service_name: str,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
granularity: str = "daily",
set_baseline: bool = False,
) -> Dict[str, Any]:
"""Compute monitoring metrics from feature serving logs.
Requires the feature service to have a logging_config with a
LoggingDestination that can be converted to a DataSource.
"""
self._ensure_monitoring_tables()
if granularity not in VALID_GRANULARITIES:
raise ValueError(
f"Invalid granularity '{granularity}'. "
f"Must be one of {VALID_GRANULARITIES}"
)
start_time = time.time()
start_dt, end_dt = self._to_date_range(start_date, end_date)
fs = self._store.registry.get_feature_service(
name=feature_service_name, project=project
)
log_source = self._resolve_log_source(fs)
if log_source is None:
return {
"status": "skipped",
"reason": f"Feature service '{feature_service_name}' has no logging configured",
"duration_ms": int((time.time() - start_time) * 1000),
}
data_source, ts_field, feature_fields, log_col_map = log_source
metrics_list = self._compute_from_source(
data_source,
ts_field,
feature_fields,
start_dt,
end_dt,
)
now = datetime.now(timezone.utc)
metric_date = start_dt.date()
self._save_log_metrics(
project=project,
feature_service_name=feature_service_name,
log_col_map=log_col_map,
metrics_list=metrics_list,
metric_date=metric_date,
granularity=granularity,
set_baseline=set_baseline,
now=now,
max_event_timestamp=_newest_event_in_window(
self._get_max_timestamp_for_source(data_source, ts_field),
start_dt,
end_dt,
),
)
duration_ms = int((time.time() - start_time) * 1000)
return {
"status": "completed",
"data_source_type": "log",
"feature_service_name": feature_service_name,
"granularity": granularity,
"computed_features": len(metrics_list),
"metric_date": metric_date.isoformat(),
"duration_ms": duration_ms,
}
def auto_compute_log_metrics(
self,
project: Optional[str] = None,
feature_service_name: Optional[str] = None,
) -> Dict[str, Any]:
"""Auto-detect date ranges from log data and compute all granularities."""
start_time = time.time()
self._ensure_monitoring_tables()
if project is None:
project = self._store.config.project
if feature_service_name:
services = [
self._store.registry.get_feature_service(
name=feature_service_name, project=project
)
]
else:
services = self._store.registry.list_feature_services(project=project)
total_features = 0
total_services = 0
granularities_computed: set = set()
for fs in services:
try:
log_source = self._resolve_log_source(fs)
if log_source is None:
continue
data_source, ts_field, feature_fields, log_col_map = log_source
max_ts = self._get_max_timestamp_for_source(data_source, ts_field)
if max_ts is None:
logger.warning(
"No log data found for feature service '%s', skipping",
fs.name,
)
continue
now = datetime.now(timezone.utc)
for gran, window in GRANULARITY_WINDOWS.items():
window_start = max_ts - window
metrics_list = self._compute_from_source(
data_source,
ts_field,
feature_fields,
window_start,
max_ts,
)
self._save_log_metrics(
project=project,
feature_service_name=fs.name,
log_col_map=log_col_map,
metrics_list=metrics_list,
metric_date=window_start.date(),
granularity=gran,
set_baseline=False,
now=now,
max_event_timestamp=max_ts,
)
total_features += len(metrics_list)
granularities_computed.add(gran)
total_services += 1
except Exception:
logger.exception(
"Failed to auto-compute log metrics for feature service '%s'",
fs.name,
)
duration_ms = int((time.time() - start_time) * 1000)
return {
"status": "completed",
"data_source_type": "log",
"computed_feature_services": total_services,
"computed_features": total_features,
"granularities": sorted(granularities_computed),
"duration_ms": duration_ms,
}
# ------------------------------------------------------------------ #
# Baseline: compute from all available source data
# ------------------------------------------------------------------ #
def compute_baseline(
self,
project: Optional[str] = None,
feature_view_name: Optional[str] = None,
feature_names: Optional[List[str]] = None,
) -> Dict[str, Any]:
"""Compute baseline metrics from all available source data.
Idempotent: only features without existing baselines are computed.
"""
start_time = time.time()
self._ensure_monitoring_tables()
if project is None:
project = self._store.config.project
feature_views = self._resolve_feature_views(project, feature_view_name)
total_features = 0
total_views = 0
for fv in feature_views:
try:
fields_needing_baseline = self._get_features_without_baseline(
project, fv, feature_names
)
if not fields_needing_baseline:
logger.info(
"All features in '%s' already have baselines, skipping",
fv.name,
)
continue
feature_fields = self._classify_fields(
fv, fields=fields_needing_baseline
)
if not feature_fields:
continue
metrics_list = self._compute_feature_metrics(
fv,
feature_fields,
_EPOCH,
_FAR_FUTURE,
)
now = datetime.now(timezone.utc)
self._save_computed_metrics(
project=project,
feature_view=fv,
metrics_list=metrics_list,
metric_date=date.today(),
granularity="baseline",
set_baseline=True,
now=now,
max_event_timestamp=self._get_max_timestamp(fv),
)
total_features += len(metrics_list)
total_views += 1
except Exception:
logger.exception(
"Failed to compute baseline for feature view '%s'", fv.name
)
duration_ms = int((time.time() - start_time) * 1000)
return {
"status": "completed",
"computed_features": total_features,
"computed_feature_views": total_views,
"is_baseline": True,
"duration_ms": duration_ms,
}
# ------------------------------------------------------------------ #
# Compute: explicit dates + granularity (stored)
# ------------------------------------------------------------------ #
def compute_metrics(
self,
project: str,
feature_view_name: Optional[str] = None,
feature_names: Optional[List[str]] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
granularity: str = "daily",
set_baseline: bool = False,
) -> Dict[str, Any]:
self._ensure_monitoring_tables()
if granularity not in VALID_GRANULARITIES:
raise ValueError(
f"Invalid granularity '{granularity}'. "
f"Must be one of {VALID_GRANULARITIES}"
)
start_time = time.time()
start_dt, end_dt = self._to_date_range(start_date, end_date)
feature_views = self._resolve_feature_views(project, feature_view_name)
total_features = 0
total_views = 0
computed_dates: set = set()
for fv in feature_views:
try:
max_ts = self._get_max_timestamp(fv)
fv_metrics = self._compute_for_feature_view(
project=project,
feature_view=fv,
feature_names=feature_names,
start_dt=start_dt,
end_dt=end_dt,
granularity=granularity,
set_baseline=set_baseline,
max_event_timestamp=max_ts,
)
total_features += fv_metrics["feature_count"]
total_views += 1
computed_dates.update(fv_metrics["dates"])
except Exception:
logger.exception(
"Failed to compute metrics for feature view '%s'", fv.name
)
total_services = self._compute_feature_service_metrics(
project=project,
granularity=granularity,
metric_dates=list(computed_dates),
set_baseline=set_baseline,
)
duration_ms = int((time.time() - start_time) * 1000)
return {
"status": "completed",
"granularity": granularity,
"computed_features": total_features,
"computed_feature_views": total_views,
"computed_feature_services": total_services,
"metric_dates": sorted(d.isoformat() for d in computed_dates),
"duration_ms": duration_ms,
}
# ------------------------------------------------------------------ #
# Transient compute (not stored)
# ------------------------------------------------------------------ #
def compute_transient(
self,
project: str,
feature_view_name: str,
feature_names: Optional[List[str]] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
) -> Dict[str, Any]:
"""Compute metrics on-the-fly for an arbitrary date range without persisting."""
start_time = time.time()
start_dt, end_dt = self._to_date_range(start_date, end_date)
effective_start = start_date or (date.today() - timedelta(days=1))
effective_end = end_date or date.today()
fv = self._store.registry.get_feature_view(
name=feature_view_name, project=project
)
feature_fields = self._classify_fields(fv, feature_names=feature_names)
if not feature_fields:
return {
"status": "completed",
"feature_view_name": feature_view_name,
"start_date": effective_start.isoformat(),
"end_date": effective_end.isoformat(),
"metrics": [],
"duration_ms": int((time.time() - start_time) * 1000),
}
metrics_list = self._compute_feature_metrics(
fv,
feature_fields,
start_dt,
end_dt,
)
for m in metrics_list:
m["feature_view_name"] = feature_view_name
m["start_date"] = effective_start.isoformat()
m["end_date"] = effective_end.isoformat()
return {
"status": "completed",
"feature_view_name": feature_view_name,
"start_date": effective_start.isoformat(),
"end_date": effective_end.isoformat(),
"metrics": metrics_list,
"duration_ms": int((time.time() - start_time) * 1000),
}
# ------------------------------------------------------------------ #
# DQM Job helpers
# ------------------------------------------------------------------ #
def submit_job(
self,
project: str,
job_type: str,
feature_view_name: Optional[str] = None,
parameters: Optional[Dict[str, Any]] = None,
) -> str:
return self.job_manager.submit(
project=project,
job_type=job_type,
feature_view_name=feature_view_name,
parameters=parameters,
)
def get_job(self, job_id: str) -> Optional[Dict[str, Any]]:
return self.job_manager.get_job(job_id)
def execute_job(self, job_id: str) -> Dict[str, Any]:
return self.job_manager.execute_job(job_id, self)
# ------------------------------------------------------------------ #
# Read helpers (delegate to offline store)
# ------------------------------------------------------------------ #
def _query(
self,
metric_type: str,
project: str,
filters=None,
start_date=None,
end_date=None,
):
self._ensure_monitoring_tables()
rows = self._get_offline_store().query_monitoring_metrics(
config=self._store.config,
project=project,
metric_type=metric_type,
filters=filters,
start_date=start_date,
end_date=end_date,
)
return [_sanitize_floats(r) for r in rows]
def get_feature_metrics(
self,
project: str,
feature_service_name: Optional[str] = None,
feature_view_name: Optional[str] = None,
feature_name: Optional[str] = None,
granularity: Optional[str] = None,
data_source_type: Optional[str] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
) -> List[Dict[str, Any]]:
filters = {
"feature_view_name": feature_view_name,
"feature_name": feature_name,
"granularity": granularity,
"data_source_type": data_source_type,
}
if feature_service_name:
return self._get_metrics_by_service(
project,
feature_service_name,
lambda fv_name: self._query(
"feature",
project,
{**filters, "feature_view_name": fv_name},
start_date,
end_date,
),
)
return self._query("feature", project, filters, start_date, end_date)
def get_feature_view_metrics(
self,
project: str,
feature_service_name: Optional[str] = None,
feature_view_name: Optional[str] = None,
granularity: Optional[str] = None,
data_source_type: Optional[str] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
) -> List[Dict[str, Any]]:
filters = {
"feature_view_name": feature_view_name,
"granularity": granularity,
"data_source_type": data_source_type,
}
if feature_service_name:
return self._get_metrics_by_service(
project,
feature_service_name,
lambda fv_name: self._query(
"feature_view",
project,
{**filters, "feature_view_name": fv_name},
start_date,
end_date,
),
)
return self._query("feature_view", project, filters, start_date, end_date)
def get_feature_service_metrics(
self,
project: str,
feature_service_name: Optional[str] = None,
granularity: Optional[str] = None,
data_source_type: Optional[str] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
) -> List[Dict[str, Any]]:
filters = {
"feature_service_name": feature_service_name,
"granularity": granularity,
"data_source_type": data_source_type,
}
return self._query("feature_service", project, filters, start_date, end_date)
def get_baseline(
self,
project: str,
feature_view_name: Optional[str] = None,
feature_name: Optional[str] = None,
data_source_type: Optional[str] = None,
) -> List[Dict[str, Any]]:
filters = {
"feature_view_name": feature_view_name,
"feature_name": feature_name,
"data_source_type": data_source_type,
"is_baseline": True,
}
return self._query("feature", project, filters)
def get_timeseries(
self,
project: str,
feature_view_name: Optional[str] = None,
feature_name: Optional[str] = None,
feature_service_name: Optional[str] = None,
granularity: Optional[str] = None,
data_source_type: Optional[str] = None,
start_date: Optional[date] = None,
end_date: Optional[date] = None,
) -> List[Dict[str, Any]]:
return self.get_feature_metrics(
project=project,
feature_service_name=feature_service_name,
feature_view_name=feature_view_name,
feature_name=feature_name,
granularity=granularity,
data_source_type=data_source_type,
start_date=start_date,
end_date=end_date,
)
# ------------------------------------------------------------------ #
# Auto-baseline trigger for feast apply
# ------------------------------------------------------------------ #
def submit_baseline_for_new_features(
self,
project: str,
feature_views: Optional[List] = None,
) -> List[str]:
"""Submit baseline DQM jobs for feature views with new features.
Called from feast apply. Returns list of submitted job IDs.
Idempotent — only features without existing baselines are included.
"""
if project is None:
project = self._store.config.project
if feature_views is None:
feature_views = self._store.registry.list_feature_views(project=project)
job_ids = []
for fv in feature_views:
new_features = self._get_features_without_baseline(project, fv)
if not new_features:
continue
feature_names = [f.name for f in new_features]
job_id = self.job_manager.submit(
project=project,
job_type="baseline",
feature_view_name=fv.name,
parameters={"feature_names": feature_names},
)
job_ids.append(job_id)
logger.info(
"Queued baseline computation for '%s' features %s (job: %s)",
fv.name,
feature_names,
job_id,
)
return job_ids
# ------------------------------------------------------------------ #
# Private: compute engine dispatch (SQL push-down → Python fallback)
# ------------------------------------------------------------------ #
def _compute_feature_metrics(
self,
feature_view,
feature_fields: List[Tuple[str, str]],
start_dt: datetime,
end_dt: datetime,
) -> List[Dict[str, Any]]:
"""Compute metrics from a feature view's batch source."""
return self._compute_from_source(
feature_view.batch_source,
feature_view.batch_source.timestamp_field,
feature_fields,
start_dt,
end_dt,
)
def _get_max_timestamp(self, feature_view) -> Optional[datetime]:
"""Query the batch source for MAX(event_timestamp)."""
return self._get_max_timestamp_for_source(
feature_view.batch_source,
feature_view.batch_source.timestamp_field,
)
# ------------------------------------------------------------------ #
# Private: shared helpers (DRY)
# ------------------------------------------------------------------ #
@staticmethod
def _to_date_range(
start_date: Optional[date], end_date: Optional[date]
) -> Tuple[datetime, datetime]:
today = date.today()
if end_date is None:
end_date = today
if start_date is None:
start_date = end_date - timedelta(days=1)
start_dt = datetime(
start_date.year, start_date.month, start_date.day, tzinfo=timezone.utc
)
end_dt = datetime(
end_date.year, end_date.month, end_date.day, 23, 59, 59, tzinfo=timezone.utc
)
return start_dt, end_dt
@staticmethod
def _classify_fields(
feature_view,
feature_names=None,
fields=None,
) -> List[Tuple[str, str]]:
"""Extract and classify features as numeric/categorical.
Args:
feature_view: FeatureView to extract fields from (used if fields is None).
feature_names: Optional filter list of feature names.
fields: Optional pre-selected Field objects (e.g., from idempotency check).
"""
if fields is None:
fields = feature_view.features
if feature_names:
fields = [f for f in fields if f.name in feature_names]
result = []
for field in fields:
ftype = MetricsCalculator.classify_feature(field.dtype)
if ftype is None:
logger.warning(
"Unsupported dtype '%s' for feature '%s', skipping",
field.dtype,
field.name,
)
continue
result.append((field.name, ftype))
return result
def _save_computed_metrics(
self,
project: str,
feature_view,
metrics_list: List[Dict[str, Any]],
metric_date: date,
granularity: str,
set_baseline: bool,
now: datetime,
max_event_timestamp: Optional[datetime] = None,
) -> None:
if not metrics_list:
return
offline_store = self._get_offline_store()
config = self._store.config
if set_baseline:
offline_store.clear_monitoring_baseline(
config=config,
project=project,
feature_view_name=feature_view.name,
)
for m in metrics_list:
m["project_id"] = project
m["feature_view_name"] = feature_view.name
m["metric_date"] = metric_date
m["granularity"] = granularity
m["data_source_type"] = "batch"
m["computed_at"] = now
m["max_event_timestamp"] = max_event_timestamp
m["is_baseline"] = set_baseline
offline_store.save_monitoring_metrics(config, "feature", metrics_list)
view_metric = {
"project_id": project,
"feature_view_name": feature_view.name,
"metric_date": metric_date,
"granularity": granularity,
"data_source_type": "batch",
"computed_at": now,
"max_event_timestamp": max_event_timestamp,
"is_baseline": set_baseline,
**build_view_aggregate(metrics_list),
}
offline_store.save_monitoring_metrics(config, "feature_view", [view_metric])
def _resolve_join_key_columns(self, feature_view) -> List[str]:
config = self._store.config
return (
[
entity.name
for entity in self._store.registry.list_entities(project=config.project)
if entity.name in (feature_view.entities or [])
]
or feature_view.entities
or []
)
def _get_metrics_by_service(
self, project: str, feature_service_name: str, query_fn
):
fs = self._store.registry.get_feature_service(
name=feature_service_name, project=project
)
fv_names = [proj.name for proj in fs.feature_view_projections]
results = []
for fv_name in fv_names:
results.extend(query_fn(fv_name))
return results
def _resolve_feature_views(self, project: str, feature_view_name: Optional[str]):
if feature_view_name:
fv = self._store.registry.get_feature_view(
name=feature_view_name, project=project
)
return [fv]
return self._store.registry.list_feature_views(project=project)
def _get_features_without_baseline(self, project, feature_view, feature_names=None):
existing = self.get_baseline(
project=project,
feature_view_name=feature_view.name,
)
existing_names = {m["feature_name"] for m in existing}
fields = feature_view.features
if feature_names:
fields = [f for f in fields if f.name in feature_names]
return [f for f in fields if f.name not in existing_names]
def _compute_for_feature_view(
self,
project: str,
feature_view,
feature_names: Optional[List[str]],
start_dt: datetime,
end_dt: datetime,
granularity: str,
set_baseline: bool,
max_event_timestamp: Optional[datetime] = None,
) -> Dict[str, Any]:
feature_fields = self._classify_fields(
feature_view, feature_names=feature_names
)
if not feature_fields:
return {"feature_count": 0, "dates": set()}
metrics_list = self._compute_feature_metrics(
feature_view,
feature_fields,
start_dt,
end_dt,
)
now = datetime.now(timezone.utc)
metric_date = start_dt.date()
self._save_computed_metrics(
project=project,
feature_view=feature_view,
metrics_list=metrics_list,
metric_date=metric_date,
granularity=granularity,
set_baseline=set_baseline,
now=now,
max_event_timestamp=_newest_event_in_window(
max_event_timestamp,