Repository navigation
Expand file tree
/
Copy pathinference.py
More file actions
352 lines (322 loc) · 14.4 KB
/
Copy pathinference.py
File metadata and controls
352 lines (322 loc) · 14.4 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
import re
from typing import List, Optional, Set, Union
from feast.data_source import DataSource, PushSource, RequestSource
from feast.entity import Entity
from feast.errors import RegistryInferenceFailure
from feast.feature_view import DUMMY_ENTITY_ID, DUMMY_ENTITY_NAME, FeatureView
from feast.field import Field, from_value_type
from feast.infra.offline_stores.bigquery_source import BigQuerySource
from feast.infra.offline_stores.contrib.mssql_offline_store.mssqlserver_source import (
MsSqlServerSource,
)
from feast.infra.offline_stores.file_source import FileSource
from feast.infra.offline_stores.redshift_source import RedshiftSource
from feast.infra.offline_stores.snowflake_source import SnowflakeSource
from feast.infra.provider import Provider
from feast.on_demand_feature_view import OnDemandFeatureView
from feast.repo_config import RepoConfig
from feast.stream_feature_view import StreamFeatureView
from feast.types import String
from feast.value_type import ValueType
def update_data_sources_with_inferred_event_timestamp_col(
data_sources: List[DataSource], config: RepoConfig
) -> None:
ERROR_MSG_PREFIX = "Unable to infer DataSource timestamp_field"
for data_source in data_sources:
if data_source is None:
continue
if isinstance(data_source, RequestSource):
continue
if isinstance(data_source, PushSource):
if not isinstance(data_source.batch_source, DataSource):
continue
else:
data_source = data_source.batch_source
if data_source.timestamp_field is None or data_source.timestamp_field == "":
# prepare right match pattern for data source
ts_column_type_regex_pattern: str
# TODO(adchia): Move Spark source inference out of this logic
if (
isinstance(data_source, FileSource)
or "SparkSource" == data_source.__class__.__name__
):
ts_column_type_regex_pattern = r"^timestamp"
elif isinstance(data_source, BigQuerySource):
ts_column_type_regex_pattern = "TIMESTAMP|DATETIME"
elif isinstance(data_source, RedshiftSource):
ts_column_type_regex_pattern = "TIMESTAMP[A-Z]*"
elif isinstance(data_source, SnowflakeSource):
ts_column_type_regex_pattern = "TIMESTAMP_[A-Z]*"
elif isinstance(data_source, MsSqlServerSource):
ts_column_type_regex_pattern = "TIMESTAMP|DATETIME"
else:
raise RegistryInferenceFailure(
"DataSource",
f"""
DataSource inferencing of timestamp_field is currently only supported
for FileSource, SparkSource, BigQuerySource, RedshiftSource, SnowflakeSource, MsSqlSource.
Attempting to infer from {data_source}.
""",
)
# for informing the type checker
assert (
isinstance(data_source, FileSource)
or isinstance(data_source, BigQuerySource)
or isinstance(data_source, RedshiftSource)
or isinstance(data_source, SnowflakeSource)
or isinstance(data_source, MsSqlServerSource)
or "SparkSource" == data_source.__class__.__name__
)
# loop through table columns to find singular match
timestamp_fields = []
for (
col_name,
col_datatype,
) in data_source.get_table_column_names_and_types(config):
if re.match(ts_column_type_regex_pattern, col_datatype):
timestamp_fields.append(col_name)
if len(timestamp_fields) > 1:
raise RegistryInferenceFailure(
"DataSource",
f"""{ERROR_MSG_PREFIX}; found multiple possible columns of timestamp type.
Data source type: {data_source.__class__.__name__},
Timestamp regex: `{ts_column_type_regex_pattern}`, columns: {timestamp_fields}""",
)
elif len(timestamp_fields) == 1:
data_source.timestamp_field = timestamp_fields[0]
else:
raise RegistryInferenceFailure(
"DataSource",
f"""
{ERROR_MSG_PREFIX}; Found no columns of timestamp type.
Data source type: {data_source.__class__.__name__},
Timestamp regex: `{ts_column_type_regex_pattern}`.
""",
)
def update_feature_views_with_inferred_features_and_entities(
provider: Provider,
fvs: Union[List[FeatureView], List[StreamFeatureView], List[OnDemandFeatureView]],
entities: List[Entity],
config: RepoConfig,
) -> None:
"""
Infers the features and entities associated with each feature view and updates it in place.
Columns whose names match a join key of an entity are considered to be entity columns; all
other columns except designated timestamp columns are considered to be feature columns. If
the feature view already has features, feature inference is skipped.
Note that this inference logic currently does not take any transformations (either a UDF or
aggregations) into account. For example, even if a stream feature view has a transformation,
this method assumes that the batch source contains transformed data with the correct final schema.
Args:
fvs: The feature views to be updated.
entities: A list containing entities associated with the feature views.
config: The config for the current feature store.
"""
entity_name_to_entity_map = {e.name: e for e in entities}
entity_name_to_join_key_map = {e.name: e.join_key for e in entities}
for fv in fvs:
join_keys = set(
[
entity_name_to_join_key_map.get(entity_name)
for entity_name in getattr(fv, "entities", [])
]
)
# Fields whose names match a join key are considered to be entity columns; all
# other fields are considered to be feature columns.
entity_columns = fv.entity_columns if fv.entity_columns else []
for field in fv.schema:
if field.name in join_keys:
# Do not override a preexisting field with the same name.
if field.name not in [
entity_column.name for entity_column in entity_columns
]:
entity_columns.append(field)
else:
if field.name not in [feature.name for feature in fv.features]:
fv.features.append(field)
# Respect the `value_type` attribute of the entity, if it is specified.
fv_entities = getattr(fv, "entities", [])
for entity_name in fv_entities:
entity = entity_name_to_entity_map.get(entity_name)
# pass when entity does not exist. Entityless feature view case
if entity is None:
continue
if (
entity.join_key
not in [entity_column.name for entity_column in entity_columns]
and entity.value_type != ValueType.UNKNOWN
):
entity_columns.append(
Field(
name=entity.join_key,
dtype=from_value_type(entity.value_type),
)
)
# Infer a dummy entity column for entityless feature views.
if (
len(fv_entities) == 1
and fv_entities[0] == DUMMY_ENTITY_NAME
and not entity_columns
):
entity_columns.append(Field(name=DUMMY_ENTITY_ID, dtype=String))
fv.entity_columns = entity_columns
# Run inference for entity columns if there are fewer entity fields than expected.
run_inference_for_entities = len(fv.entity_columns) < len(join_keys)
# Run inference for feature columns if there are no feature fields.
run_inference_for_features = len(fv.features) == 0
if run_inference_for_entities or run_inference_for_features:
_infer_features_and_entities(
provider,
fv,
join_keys,
run_inference_for_features,
config,
)
if not fv.features:
if isinstance(fv, OnDemandFeatureView):
return None
else:
raise RegistryInferenceFailure(
"FeatureView",
f"Could not infer Features for the FeatureView named {fv.name}.",
)
def _infer_features_and_entities(
provider: Provider,
fv: Union[FeatureView, OnDemandFeatureView],
join_keys: Set[Optional[str]],
run_inference_for_features,
config,
) -> None:
"""
Updates the specific feature in place with inferred features and entities.
Args:
fv: The feature view on which to run inference.
join_keys: The set of join keys for the feature view's entities.
run_inference_for_features: Whether to run inference for features.
config: The config for the current feature store.
"""
if isinstance(fv, OnDemandFeatureView):
return _infer_on_demand_features_and_entities(
fv, join_keys, run_inference_for_features, config
)
entity_columns: List[Field] = fv.entity_columns if fv.entity_columns else []
columns_to_exclude = {
fv.batch_source.timestamp_field,
fv.batch_source.created_timestamp_column,
}
for original_col, mapped_col in fv.batch_source.field_mapping.items():
if mapped_col in columns_to_exclude:
columns_to_exclude.remove(mapped_col)
columns_to_exclude.add(original_col)
table_column_names_and_types = (
provider.get_table_column_names_and_types_from_data_source(
config, fv.batch_source
)
)
for col_name, col_datatype in table_column_names_and_types:
if col_name in columns_to_exclude:
continue
elif col_name in join_keys:
field = Field(
name=col_name,
dtype=from_value_type(
fv.batch_source.source_datatype_to_feast_value_type()(col_datatype)
),
)
if field.name not in [
entity_column.name for entity_column in fv.entity_columns
]:
entity_columns.append(field)
elif not re.match(
"^__|__$", col_name
): # double underscores often signal an internal-use column
if run_inference_for_features:
feature_name = (
fv.batch_source.field_mapping[col_name]
if col_name in fv.batch_source.field_mapping
else col_name
)
field = Field(
name=feature_name,
dtype=from_value_type(
fv.batch_source.source_datatype_to_feast_value_type()(
col_datatype
)
),
)
if field.name not in [feature.name for feature in fv.features]:
fv.features.append(field)
fv.entity_columns = entity_columns
def _infer_on_demand_features_and_entities(
fv: OnDemandFeatureView,
join_keys: Set[Optional[str]],
run_inference_for_features,
config,
) -> None:
"""
Updates the specific feature in place with inferred features and entities.
Args:
fv: The feature view on which to run inference.
join_keys: The set of join keys for the feature view's entities.
run_inference_for_features: Whether to run inference for features.
config: The config for the current feature store.
"""
entity_columns: list[Field] = []
columns_to_exclude = set()
for (
source_feature_view_name,
source_feature_view,
) in fv.source_feature_view_projections.items():
columns_to_exclude.add(source_feature_view.timestamp_field)
columns_to_exclude.add(source_feature_view.created_timestamp_column)
batch_source = getattr(source_feature_view, "batch_source")
batch_field_mapping = getattr(batch_source or None, "field_mapping")
for (
original_col,
mapped_col,
) in batch_field_mapping.items():
if mapped_col in columns_to_exclude:
columns_to_exclude.remove(mapped_col)
columns_to_exclude.add(original_col)
table_column_names_and_types = batch_source.get_table_column_names_and_types(
config
)
batch_field_mapping = getattr(batch_source, "field_mapping", {})
for col_name, col_datatype in table_column_names_and_types:
if col_name in columns_to_exclude:
continue
elif col_name in join_keys:
field = Field(
name=col_name,
dtype=from_value_type(
batch_source.source_datatype_to_feast_value_type()(col_datatype)
),
)
if field.name not in [
entity_column.name
for entity_column in entity_columns
if hasattr(entity_column, "name")
]:
entity_columns.append(field)
elif not re.match(
"^__|__$", col_name
): # double underscores often signal an internal-use column
if run_inference_for_features:
feature_name = (
batch_field_mapping[col_name]
if col_name in batch_field_mapping
else col_name
)
field = Field(
name=feature_name,
dtype=from_value_type(
batch_source.source_datatype_to_feast_value_type()(
col_datatype
)
),
)
if field.name not in [
feature.name for feature in source_feature_view.features
]:
source_feature_view.features.append(field)
fv.entity_columns = entity_columns