-
Notifications
You must be signed in to change notification settings - Fork 1.4k
Expand file tree
/
Copy pathtype_map.py
More file actions
2430 lines (2156 loc) · 87.3 KB
/
Copy pathtype_map.py
File metadata and controls
2430 lines (2156 loc) · 87.3 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
# Copyright 2019 The Feast Authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import decimal
import json
import logging
import re
import uuid as uuid_module
from collections import defaultdict
from datetime import datetime, timedelta, timezone
from typing import (
TYPE_CHECKING,
Any,
Dict,
Iterator,
List,
Optional,
Sequence,
Set,
Sized,
Tuple,
Type,
Union,
cast,
)
import numpy as np
import pandas as pd
from google.protobuf.timestamp_pb2 import Timestamp
from feast.protos.feast.types.Value_pb2 import (
BoolList,
BoolSet,
BytesList,
BytesSet,
DoubleList,
DoubleSet,
FloatList,
FloatSet,
Int32List,
Int32Set,
Int64List,
Int64Set,
Map,
MapKey,
MapList,
RepeatedValue,
ScalarMap,
ScalarMapEntry,
StringList,
StringSet,
ZonedTimestamp,
)
from feast.protos.feast.types.Value_pb2 import Value as ProtoValue
from feast.value_type import ListType, SetType, ValueType
if TYPE_CHECKING:
import pyarrow
# null timestamps get converted to -9223372036854775808
NULL_TIMESTAMP_INT_VALUE: int = np.datetime64("NaT").astype(int)
logger = logging.getLogger(__name__)
def _zone_name(tzinfo: Optional[Any]) -> str:
"""Return a storable zone string for a datetime's tzinfo.
Prefers the IANA name (e.g. ``zoneinfo.ZoneInfo`` key) so DST is preserved;
falls back to a fixed-offset string (e.g. ``-07:00``). A naive datetime
(``tzinfo is None``) yields ``""``, which decodes back as UTC.
"""
if tzinfo is None:
return ""
key = getattr(tzinfo, "key", None) # zoneinfo.ZoneInfo
if key:
return key
name = str(tzinfo)
# zoneinfo prints as the key; pytz prints the name; offsets print as "UTC-07:00"
return name
def _zone_from_name(zone: str):
"""Resolve a stored zone string back to a tzinfo. Empty → UTC."""
if not zone:
return timezone.utc
try:
from zoneinfo import ZoneInfo
return ZoneInfo(zone)
except Exception:
# Not an IANA name. It may be a fixed-offset string produced by
# ``_zone_name`` (e.g. "UTC", "UTC-07:00", "+05:30"); parse it so the
# original offset is preserved on round trips rather than silently
# shifting the wall-clock time to UTC.
offset = _fixed_offset_from_name(zone)
if offset is not None:
return offset
logger.warning("Could not resolve zone %r; decoding as UTC", zone)
return timezone.utc
def _fixed_offset_from_name(zone: str) -> Optional[timezone]:
"""Parse a fixed-offset zone string into a ``timezone``.
Accepts the forms ``_zone_name`` emits for offset-only tzinfos: a bare
``UTC``/``GMT``, or an offset like ``UTC-07:00``, ``-07:00``, ``+05:30`` or
``+0530``. Returns ``None`` if the string is not a recognizable offset.
"""
text = zone.strip()
if text in ("UTC", "GMT"):
return timezone.utc
match = re.fullmatch(
r"(?:UTC|GMT)?([+-])(\d{2}):?(\d{2})",
text,
)
if not match:
return None
sign, hours, minutes = match.groups()
delta = timedelta(hours=int(hours), minutes=int(minutes))
if sign == "-":
delta = -delta
return timezone(delta)
def feast_value_type_to_python_type(
field_value_proto: ProtoValue,
feature_type: Optional[ValueType] = None,
) -> Any:
"""
Converts field value Proto to Dict and returns each field's Feast Value Type value
in their respective Python value.
Args:
field_value_proto: Field value Proto
Returns:
Python native type representation/version of the given field_value_proto
"""
val_attr = field_value_proto.WhichOneof("val")
if val_attr is None:
return None
val = getattr(field_value_proto, val_attr)
# Handle JSON types — stored as strings but returned as parsed Python objects
if val_attr == "json_val":
try:
return json.loads(val)
except (json.JSONDecodeError, TypeError):
return val
elif val_attr == "json_list_val":
result = []
for v in val.val:
if isinstance(v, str):
try:
result.append(json.loads(v))
except (json.JSONDecodeError, TypeError):
result.append(v)
else:
result.append(v)
return result
# Handle nested collection types (list_val, set_val)
if val_attr in ("list_val", "set_val"):
return _handle_nested_collection_value(val)
# Handle Struct types — stored using Map proto, returned as dicts
if val_attr == "struct_val":
return _handle_map_value(val)
elif val_attr == "struct_list_val":
return _handle_map_list_value(val)
# Handle Map and MapList types FIRST (before generic list processing)
if val_attr == "map_val":
return _handle_map_value(val)
elif val_attr == "map_list_val":
return _handle_map_list_value(val)
elif val_attr == "scalar_map_val":
return _handle_scalar_map_value(val)
# Zoned timestamp: a (instant, zone) message → tz-aware datetime in its own zone.
if val_attr == "zoned_timestamp_val":
if val.unix_timestamp == NULL_TIMESTAMP_INT_VALUE:
return None
tz = _zone_from_name(val.zone)
return datetime.fromtimestamp(val.unix_timestamp, tz=tz)
# If it's a _LIST or _SET type extract the values.
if hasattr(val, "val"):
val = list(val.val)
# Convert UNIX_TIMESTAMP values to `datetime`
if val_attr == "unix_timestamp_list_val":
val = [
(
datetime.fromtimestamp(v, tz=timezone.utc)
if v != NULL_TIMESTAMP_INT_VALUE
else None
)
for v in val
]
elif val_attr == "unix_timestamp_set_val":
val = set(
[
(
datetime.fromtimestamp(v, tz=timezone.utc)
if v != NULL_TIMESTAMP_INT_VALUE
else None
)
for v in val
]
)
elif val_attr == "unix_timestamp_val":
val = (
datetime.fromtimestamp(val, tz=timezone.utc)
if val != NULL_TIMESTAMP_INT_VALUE
else None
)
# Convert _SET types to Python sets
elif val_attr.endswith("_set_val") and val_attr != "unix_timestamp_set_val":
val = set(val)
# Convert UUID values to uuid.UUID objects
if val_attr in ("uuid_val", "time_uuid_val"):
return uuid_module.UUID(val) if isinstance(val, str) else val
if val_attr in ("uuid_list_val", "time_uuid_list_val"):
return [uuid_module.UUID(v) if isinstance(v, str) else v for v in val]
if val_attr in ("uuid_set_val", "time_uuid_set_val"):
return {uuid_module.UUID(v) if isinstance(v, str) else v for v in val}
# Convert DECIMAL values to decimal.Decimal objects
if val_attr == "decimal_val":
return decimal.Decimal(val) if isinstance(val, str) else val
if val_attr == "decimal_list_val":
return [decimal.Decimal(v) if isinstance(v, str) else v for v in val]
if val_attr == "decimal_set_val":
return {decimal.Decimal(v) if isinstance(v, str) else v for v in val}
# Backward compatibility: handle UUIDs stored as string_val/string_list_val with feature_type hint
if feature_type in (ValueType.UUID, ValueType.TIME_UUID) and isinstance(val, str):
return uuid_module.UUID(val)
if feature_type in (ValueType.UUID_LIST, ValueType.TIME_UUID_LIST) and isinstance(
val, list
):
return [uuid_module.UUID(v) if isinstance(v, str) else v for v in val]
if feature_type in (ValueType.UUID_SET, ValueType.TIME_UUID_SET) and isinstance(
val, set
):
return {uuid_module.UUID(v) if isinstance(v, str) else v for v in val}
return val
def _handle_map_value(map_message) -> Dict[str, Any]:
"""Handle Map proto message containing map<string, Value> val."""
result = {}
for key, value in map_message.val.items():
# Recursively handle the Value message
result[key] = feast_value_type_to_python_type(value)
return result
def _handle_map_list_value(map_list_message) -> List[Dict[str, Any]]:
"""Handle MapList proto message containing repeated Map val."""
result = []
for map_item in map_list_message.val:
# Handle each Map in the list
processed_map = _handle_map_value(map_item)
result.append(processed_map)
return result
def _handle_nested_collection_value(repeated_value) -> List[Any]:
"""Handle nested collection proto (RepeatedValue containing Values).
Each inner Value is itself a list/set proto. We recursively convert
each inner Value to a Python list/set via feast_value_type_to_python_type.
"""
result = []
for inner_value in repeated_value.val:
result.append(feast_value_type_to_python_type(inner_value))
return result
def _map_key_to_python_value(map_key: MapKey) -> Any:
"""Convert a MapKey proto to its Python equivalent."""
key_attr = map_key.WhichOneof("key")
if key_attr is None:
return None
val = getattr(map_key, key_attr)
if key_attr in ("int32_key", "int64_key"):
return int(val)
if key_attr in ("float_key", "double_key"):
return float(val)
if key_attr == "bool_key":
return bool(val)
if key_attr == "unix_timestamp_key":
return (
datetime.fromtimestamp(val, tz=timezone.utc)
if val != NULL_TIMESTAMP_INT_VALUE
else None
)
if key_attr == "bytes_key":
return bytes(val)
if key_attr in ("uuid_key", "time_uuid_key"):
return uuid_module.UUID(val)
if key_attr == "decimal_key":
return decimal.Decimal(val)
return val
def _handle_scalar_map_value(value_map_message: ScalarMap) -> Dict[Any, Any]:
"""Handle ScalarMap proto message (repeated ScalarMapEntry) → Python dict."""
result: Dict[Any, Any] = {}
for entry in value_map_message.val:
key = _map_key_to_python_value(entry.key)
value = feast_value_type_to_python_type(entry.value)
result[key] = value
return result
def feast_value_type_to_pandas_type(value_type: ValueType) -> Any:
value_type_to_pandas_type: Dict[ValueType, str] = {
ValueType.FLOAT: "float",
ValueType.INT32: "int",
ValueType.INT64: "int",
ValueType.STRING: "str",
ValueType.DOUBLE: "float",
ValueType.BYTES: "bytes",
ValueType.BOOL: "bool",
ValueType.UNIX_TIMESTAMP: "datetime64[ns]",
ValueType.UUID: "str",
ValueType.TIME_UUID: "str",
ValueType.DECIMAL: "object",
}
if (
value_type.name in ("MAP", "JSON", "STRUCT", "VALUE_LIST", "VALUE_SET")
or value_type.name.endswith("_LIST")
or value_type.name.endswith("_SET")
):
return "object"
if value_type in value_type_to_pandas_type:
return value_type_to_pandas_type[value_type]
raise TypeError(
f"Casting to pandas type for type {value_type} failed. "
f"Type {value_type} not found"
)
def python_type_to_feast_value_type(
name: str,
value: Optional[Any] = None,
recurse: bool = True,
type_name: Optional[str] = None,
) -> ValueType:
"""
Finds the equivalent Feast Value Type for a Python value. Both native
and Pandas types are supported. This function will recursively look
for nested types when arrays are detected. All types must be homogenous.
Args:
name: Name of the value or field
value: Value that will be inspected
recurse: Whether to recursively look for nested types in arrays
Returns:
Feast Value Type
"""
type_name = (type_name or type(value).__name__).lower()
type_map = {
"int": ValueType.INT64,
"str": ValueType.STRING,
"string": ValueType.STRING, # pandas.StringDtype
"float": ValueType.DOUBLE,
"bytes": ValueType.BYTES,
"float64": ValueType.DOUBLE,
"float32": ValueType.FLOAT,
"int64": ValueType.INT64,
"uint64": ValueType.INT64,
"int32": ValueType.INT32,
"uint32": ValueType.INT32,
"int16": ValueType.INT32,
"uint16": ValueType.INT32,
"uint8": ValueType.INT32,
"int8": ValueType.INT32,
"bool_": ValueType.BOOL, # np.bool_
"bool": ValueType.BOOL,
"boolean": ValueType.BOOL,
"timedelta": ValueType.UNIX_TIMESTAMP,
"timestamp": ValueType.UNIX_TIMESTAMP,
"datetime": ValueType.UNIX_TIMESTAMP,
"datetime64[ns]": ValueType.UNIX_TIMESTAMP,
"datetime64[ns, tz]": ValueType.UNIX_TIMESTAMP, # special dtype of pandas
"datetime64[ns, utc]": ValueType.UNIX_TIMESTAMP,
"date": ValueType.UNIX_TIMESTAMP,
"category": ValueType.STRING,
"uuid": ValueType.UUID,
"decimal": ValueType.DECIMAL,
}
if type_name in type_map:
return type_map[type_name]
# Handle pandas "object" dtype by inspecting the actual value
if type_name == "object" and value is not None:
# Check the actual type of the value
actual_type = type(value).__name__.lower()
if actual_type == "str":
return ValueType.STRING
# Check if it's a dictionary (could be a Map)
elif actual_type == "dict":
return ValueType.MAP
# If it's a different type wrapped in object, try to infer from the value
elif actual_type in type_map:
return type_map[actual_type]
if isinstance(value, np.ndarray) and str(value.dtype) in type_map:
item_type = type_map[str(value.dtype)]
return ValueType[item_type.name + "_LIST"]
if isinstance(value, (list, np.ndarray)):
# Check if it's a list of maps
if value and isinstance(value[0], dict):
return ValueType.MAP_LIST
# if the value's type is "ndarray" and we couldn't infer from "value.dtype"
# this is most probably array of "object",
# so we need to iterate over objects and try to infer type of each item
if not recurse:
raise ValueError(
f"Value type for field {name} is {type(value)} but "
f"recursion is not allowed. Nested collection types cannot be "
f"inferred automatically; use an explicit Field dtype instead "
f"(e.g., dtype=Array(Array(Int32)))."
)
# This is the final type which we infer from the list
common_item_value_type = None
for item in value:
if isinstance(item, ProtoValue):
current_item_value_type: ValueType = _proto_value_to_value_type(item)
else:
# Get the type from the current item, only one level deep
current_item_value_type = python_type_to_feast_value_type(
name=name, value=item, recurse=False
)
# Validate whether the type stays consistent
if (
common_item_value_type
and not common_item_value_type == current_item_value_type
):
raise ValueError(
f"List value type for field {name} is inconsistent. "
f"{common_item_value_type} different from "
f"{current_item_value_type}."
)
common_item_value_type = current_item_value_type
if common_item_value_type is None:
return ValueType.UNKNOWN
return ValueType[common_item_value_type.name + "_LIST"]
# Check if it's a set (Set type)
if isinstance(value, set):
if not recurse:
raise ValueError(
f"Value type for field {name} is {type(value)} but "
f"recursion is not allowed. Set types can only be one level "
f"deep."
)
# Infer the type from set elements
common_set_item_type = None
for item in value:
if isinstance(item, ProtoValue):
current_set_item_type: ValueType = _proto_value_to_value_type(item)
else:
# Get the type from the current item, only one level deep
current_set_item_type = python_type_to_feast_value_type(
name=name, value=item, recurse=False
)
# Validate whether the type stays consistent
if (
common_set_item_type
and not common_set_item_type == current_set_item_type
):
raise ValueError(
f"Set value type for field {name} is inconsistent. "
f"{common_set_item_type} different from "
f"{current_set_item_type}."
)
common_set_item_type = current_set_item_type
if common_set_item_type is None:
return ValueType.UNKNOWN
return ValueType[common_set_item_type.name + "_SET"]
# Check if it's a dictionary (Map type)
if isinstance(value, dict):
# Non-string keys require ScalarMap; string keys (or empty dict) use Map
if value and not isinstance(next(iter(value)), str):
return ValueType.SCALAR_MAP
return ValueType.MAP
raise ValueError(
f"Value with native type {type_name} cannot be converted into Feast value type"
)
def python_values_to_feast_value_type(
name: str, values: Any, recurse: bool = True
) -> ValueType:
inferred_dtype = ValueType.UNKNOWN
for row in values:
current_dtype = python_type_to_feast_value_type(
name, value=row, recurse=recurse
)
if inferred_dtype is ValueType.UNKNOWN:
inferred_dtype = current_dtype
else:
if current_dtype != inferred_dtype and current_dtype not in (
ValueType.UNKNOWN,
ValueType.NULL,
):
raise TypeError(
f"Input entity {name} has mixed types, {current_dtype} and {inferred_dtype}. That is not allowed. "
)
if inferred_dtype in (ValueType.UNKNOWN, ValueType.NULL):
raise ValueError(
f"field {name} cannot have all null values for type inference."
)
return inferred_dtype
def _convert_value_type_str_to_value_type(type_str: str) -> ValueType:
type_map = {
"UNKNOWN": ValueType.UNKNOWN,
"BYTES": ValueType.BYTES,
"STRING": ValueType.STRING,
"INT32": ValueType.INT32,
"INT64": ValueType.INT64,
"DOUBLE": ValueType.DOUBLE,
"FLOAT": ValueType.FLOAT,
"FLOAT32": ValueType.FLOAT,
"BOOL": ValueType.BOOL,
"NULL": ValueType.NULL,
"UNIX_TIMESTAMP": ValueType.UNIX_TIMESTAMP,
"BYTES_LIST": ValueType.BYTES_LIST,
"STRING_LIST": ValueType.STRING_LIST,
"INT32_LIST ": ValueType.INT32_LIST,
"INT64_LIST": ValueType.INT64_LIST,
"DOUBLE_LIST": ValueType.DOUBLE_LIST,
"FLOAT_LIST": ValueType.FLOAT_LIST,
"BOOL_LIST": ValueType.BOOL_LIST,
"UNIX_TIMESTAMP_LIST": ValueType.UNIX_TIMESTAMP_LIST,
"MAP": ValueType.MAP,
"MAP_LIST": ValueType.MAP_LIST,
"JSON": ValueType.JSON,
"JSON_LIST": ValueType.JSON_LIST,
"STRUCT": ValueType.STRUCT,
"STRUCT_LIST": ValueType.STRUCT_LIST,
"BYTES_SET": ValueType.BYTES_SET,
"STRING_SET": ValueType.STRING_SET,
"INT32_SET": ValueType.INT32_SET,
"INT64_SET": ValueType.INT64_SET,
"DOUBLE_SET": ValueType.DOUBLE_SET,
"FLOAT_SET": ValueType.FLOAT_SET,
"BOOL_SET": ValueType.BOOL_SET,
"UNIX_TIMESTAMP_SET": ValueType.UNIX_TIMESTAMP_SET,
"UUID": ValueType.UUID,
"TIME_UUID": ValueType.TIME_UUID,
"UUID_LIST": ValueType.UUID_LIST,
"TIME_UUID_LIST": ValueType.TIME_UUID_LIST,
"UUID_SET": ValueType.UUID_SET,
"TIME_UUID_SET": ValueType.TIME_UUID_SET,
"VALUE_LIST": ValueType.VALUE_LIST,
"VALUE_SET": ValueType.VALUE_SET,
"DECIMAL": ValueType.DECIMAL,
"DECIMAL_LIST": ValueType.DECIMAL_LIST,
"DECIMAL_SET": ValueType.DECIMAL_SET,
"SCALAR_MAP": ValueType.SCALAR_MAP,
"ZONED_TIMESTAMP": ValueType.ZONED_TIMESTAMP,
}
return type_map.get(type_str, ValueType.STRING)
def _type_err(item, dtype):
raise TypeError(f'Value "{item}" is of type {type(item)} not of type {dtype}')
PYTHON_LIST_VALUE_TYPE_TO_PROTO_VALUE: Dict[
ValueType, Tuple[ListType, str, List[Type]]
] = {
ValueType.FLOAT_LIST: (
FloatList,
"float_list_val",
[np.float32, np.float64, float],
),
ValueType.DOUBLE_LIST: (
DoubleList,
"double_list_val",
[np.float64, np.float32, float],
),
ValueType.INT32_LIST: (Int32List, "int32_list_val", [np.int64, np.int32, int]),
ValueType.INT64_LIST: (Int64List, "int64_list_val", [np.int64, np.int32, int]),
ValueType.UNIX_TIMESTAMP_LIST: (
Int64List,
"int64_list_val",
[np.datetime64, np.int64, np.int32, int, datetime, Timestamp],
),
ValueType.STRING_LIST: (StringList, "string_list_val", [np.str_, str]),
ValueType.BOOL_LIST: (BoolList, "bool_list_val", [np.bool_, bool]),
ValueType.BYTES_LIST: (BytesList, "bytes_list_val", [np.bytes_, bytes]),
ValueType.UUID_LIST: (
StringList,
"uuid_list_val",
[np.str_, str, uuid_module.UUID],
),
ValueType.TIME_UUID_LIST: (
StringList,
"time_uuid_list_val",
[np.str_, str, uuid_module.UUID],
),
ValueType.DECIMAL_LIST: (
StringList,
"decimal_list_val",
[np.str_, str, decimal.Decimal],
),
}
PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE: Dict[
ValueType, Tuple[SetType, str, List[Type]]
] = {
ValueType.FLOAT_SET: (
FloatSet,
"float_set_val",
[np.float32, np.float64, float],
),
ValueType.DOUBLE_SET: (
DoubleSet,
"double_set_val",
[np.float64, np.float32, float],
),
ValueType.INT32_SET: (Int32Set, "int32_set_val", [np.int64, np.int32, int]),
ValueType.INT64_SET: (Int64Set, "int64_set_val", [np.int64, np.int32, int]),
ValueType.UNIX_TIMESTAMP_SET: (
Int64Set,
"unix_timestamp_set_val",
[np.datetime64, np.int64, np.int32, int, datetime, Timestamp],
),
ValueType.STRING_SET: (StringSet, "string_set_val", [np.str_, str]),
ValueType.BOOL_SET: (BoolSet, "bool_set_val", [np.bool_, bool]),
ValueType.BYTES_SET: (BytesSet, "bytes_set_val", [np.bytes_, bytes]),
ValueType.UUID_SET: (StringSet, "uuid_set_val", [np.str_, str, uuid_module.UUID]),
ValueType.TIME_UUID_SET: (
StringSet,
"time_uuid_set_val",
[np.str_, str, uuid_module.UUID],
),
ValueType.DECIMAL_SET: (
StringSet,
"decimal_set_val",
[np.str_, str, decimal.Decimal],
),
}
PYTHON_SCALAR_VALUE_TYPE_TO_PROTO_VALUE: Dict[
ValueType, Tuple[str, Any, Optional[Set[Type]]]
] = {
ValueType.INT32: ("int32_val", lambda x: int(x), None),
ValueType.INT64: (
"int64_val",
lambda x: (
int(x.timestamp())
if isinstance(x, pd._libs.tslibs.timestamps.Timestamp)
else int(x)
),
None,
),
ValueType.FLOAT: ("float_val", lambda x: float(x), None),
ValueType.DOUBLE: (
"double_val",
lambda x: x,
{float, np.float64, int, np.int_, decimal.Decimal},
),
ValueType.STRING: ("string_val", lambda x: str(x), None),
ValueType.BYTES: ("bytes_val", lambda x: x, {bytes}),
ValueType.IMAGE_BYTES: ("bytes_val", lambda x: x, {bytes}),
ValueType.BOOL: ("bool_val", lambda x: x, {bool, np.bool_, int, np.int_}),
ValueType.UUID: ("uuid_val", lambda x: str(x), {str, uuid_module.UUID}),
ValueType.TIME_UUID: ("time_uuid_val", lambda x: str(x), {str, uuid_module.UUID}),
ValueType.DECIMAL: ("decimal_val", lambda x: str(x), {decimal.Decimal, str}),
}
def _python_datetime_to_int_timestamp(
values: Sequence[Any],
) -> Sequence[Union[int, np.int_]]:
# Fast path for Numpy array.
if isinstance(values, np.ndarray) and isinstance(values.dtype, np.datetime64):
if values.ndim != 1:
raise ValueError("Only 1 dimensional arrays are supported.")
return cast(Sequence[np.int_], values.astype("datetime64[s]").astype(np.int_))
int_timestamps = []
for value in values:
if isinstance(value, datetime):
int_timestamps.append(int(value.timestamp()))
elif isinstance(value, Timestamp):
int_timestamps.append(int(value.ToSeconds()))
elif isinstance(value, np.datetime64):
int_timestamps.append(value.astype("datetime64[s]").astype(np.int_)) # type: ignore[attr-defined]
elif isinstance(value, type(np.nan)):
int_timestamps.append(NULL_TIMESTAMP_INT_VALUE)
else:
int_timestamps.append(int(value))
return int_timestamps
def _convert_timestamp_collection_to_proto(
values: List[Any],
proto_field: str,
proto_type: type,
) -> List[ProtoValue]:
"""Convert timestamp collection values (list or set) to proto.
Args:
values: List of timestamp collections to convert.
proto_field: The proto field name (e.g., 'unix_timestamp_list_val').
proto_type: The proto type class (e.g., Int64List).
Returns:
List of ProtoValue with converted timestamps.
"""
result = []
for value in values:
if value is not None:
result.append(
ProtoValue(
**{
proto_field: proto_type(
val=_python_datetime_to_int_timestamp(value)
)
} # type: ignore
)
)
else:
result.append(ProtoValue())
return result
def _convert_bool_collection_to_proto(
values: List[Any],
proto_field: str,
proto_type: type,
) -> List[ProtoValue]:
"""Convert boolean collection values (list or set) to proto.
ProtoValue does not support direct conversion of np.bool_, so we need to
explicitly convert each element to Python bool.
Args:
values: List of boolean collections to convert.
proto_field: The proto field name (e.g., 'bool_list_val').
proto_type: The proto type class (e.g., BoolList).
Returns:
List of ProtoValue with converted booleans.
"""
result = []
for value in values:
if value is not None:
result.append(
ProtoValue(**{proto_field: proto_type(val=[bool(e) for e in value])}) # type: ignore
)
else:
result.append(ProtoValue())
return result
def _validate_collection_item_types(
sample: Any,
valid_types: List[Type],
feast_value_type: ValueType,
) -> None:
"""Validate that collection items match expected types.
Args:
sample: A sample collection value to check.
valid_types: List of valid Python types for items.
feast_value_type: The Feast value type for error messages.
Raises:
TypeError: If any item in sample is not a valid type.
"""
if sample is None:
return
if all(type(item) in valid_types for item in sample if item is not None):
return
# to_numpy() upcasts INT32/INT64 with NULL to Float64 automatically
int_collection_types = [
ValueType.INT32_LIST,
ValueType.INT64_LIST,
ValueType.INT32_SET,
ValueType.INT64_SET,
]
for item in sample:
if item is None:
continue # None elements in STRING_LIST are replaced with ""; for other types they are dropped
if type(item) not in valid_types:
if feast_value_type in int_collection_types:
# Check if the float values are due to NULL upcast
if not any(np.isnan(i) for i in sample if isinstance(i, float)):
logger.error(
f"{feast_value_type.name} has NULL values. to_numpy() upcasts to Float64 automatically."
)
raise _type_err(item, valid_types[0])
def _python_set_to_proto_values(
feast_value_type: ValueType, values: List[Any]
) -> List[ProtoValue]:
"""
Converts Python set values to Feast Proto Values.
Args:
feast_value_type: The target set value type
values: List of set values that will be converted
Returns:
List of Feast Value Proto
"""
# Feature can be set but None is still valid
if feast_value_type not in PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE:
return []
set_proto_type, set_field_name, set_valid_types = (
PYTHON_SET_VALUE_TYPE_TO_PROTO_VALUE[feast_value_type]
)
# Convert set to list for proto (proto doesn't have native set type)
def convert_set_to_list(value: Any) -> Any:
if value is None:
return None
if isinstance(value, set):
return list(value)
if isinstance(value, (list, tuple, np.ndarray)):
return list(set(value))
return value
converted_values = [convert_set_to_list(v) for v in values]
sample = next(filter(_non_empty_value, converted_values), None)
# Bytes to array type conversion
if isinstance(sample, (bytes, bytearray)):
if feast_value_type == ValueType.BYTES_SET:
raise _type_err(sample, ValueType.BYTES_SET)
json_sample = json.loads(sample)
if isinstance(json_sample, list):
json_values = [
json.loads(value) if value is not None else None
for value in converted_values
]
if feast_value_type == ValueType.BOOL_SET:
json_values = [
[bool(item) for item in list_item]
if list_item is not None
else None
for list_item in json_values
]
return [
ProtoValue(**{set_field_name: set_proto_type(val=v)}) # type: ignore[arg-type]
if v is not None
else ProtoValue()
for v in json_values
]
raise _type_err(sample, set_valid_types[0])
# Validate item types using shared helper
_validate_collection_item_types(sample, set_valid_types, feast_value_type)
# Handle special types using shared helpers
if feast_value_type == ValueType.UNIX_TIMESTAMP_SET:
return _convert_timestamp_collection_to_proto(
converted_values, "unix_timestamp_set_val", Int64Set
)
if feast_value_type == ValueType.BOOL_SET:
return _convert_bool_collection_to_proto(
converted_values, set_field_name, set_proto_type
)
if feast_value_type in (ValueType.UUID_SET, ValueType.TIME_UUID_SET):
# uuid.UUID objects must be converted to str for StringSet proto.
return [
(
ProtoValue(
**{set_field_name: set_proto_type(val=[str(e) for e in value])} # type: ignore[arg-type, misc]
)
if value is not None
else ProtoValue()
)
for value in converted_values
]
if feast_value_type == ValueType.DECIMAL_SET:
# decimal.Decimal objects must be converted to str for StringSet proto.
return [
(
ProtoValue(
**{set_field_name: set_proto_type(val=[str(e) for e in value])} # type: ignore[arg-type, misc]
)
if value is not None
else ProtoValue()
)
for value in converted_values
]
# Generic set conversion
return [
ProtoValue(**{set_field_name: set_proto_type(val=value)}) # type: ignore[arg-type]
if value is not None
else ProtoValue()
for value in converted_values
]
# Per-type default values substituted for None elements inside list columns.
# Protobuf repeated fields do not accept None, so we replace with a
# type-appropriate zero/empty value.
_LIST_NONE_DEFAULTS: Dict[ValueType, Any] = {
ValueType.STRING_LIST: "",
ValueType.BYTES_LIST: b"",
ValueType.INT32_LIST: 0,
ValueType.INT64_LIST: 0,
ValueType.FLOAT_LIST: 0.0,
ValueType.DOUBLE_LIST: 0.0,
ValueType.BOOL_LIST: False,
ValueType.UNIX_TIMESTAMP_LIST: NULL_TIMESTAMP_INT_VALUE,
ValueType.UUID_LIST: "",
ValueType.TIME_UUID_LIST: "",
ValueType.DECIMAL_LIST: "",
}
def _sanitize_list_value(value: Any, feast_value_type: ValueType) -> Any:
"""Convert ndarray to list and replace None elements with a type-appropriate default.
Arrow/Athena may deserialize array columns as numpy.ndarray with object dtype
instead of plain Python lists. Protobuf repeated fields do not accept ndarrays
or None elements, so we normalise here before building proto messages.
"""
if isinstance(value, np.ndarray):
value = value.tolist()
none_default = _LIST_NONE_DEFAULTS.get(feast_value_type)
if none_default is not None and isinstance(value, list):
value = [none_default if v is None else v for v in value]
return value
def _convert_list_values_to_proto(
feast_value_type: ValueType,
values: List[Any],
sample: Any,
) -> List[ProtoValue]:
"""Convert list-type values to proto.
Args:
feast_value_type: The target list value type.
values: List of list values to convert.
sample: First non-empty value for type checking.
Returns:
List of ProtoValue.
"""
if feast_value_type not in PYTHON_LIST_VALUE_TYPE_TO_PROTO_VALUE:
raise Exception(f"Unsupported list type: {feast_value_type}")
proto_type, field_name, valid_types = PYTHON_LIST_VALUE_TYPE_TO_PROTO_VALUE[
feast_value_type
]
values = [
_sanitize_list_value(v, feast_value_type) if v is not None else v