import contextlib
import uuid
from datetime import datetime
from pathlib import Path
from typing import (
Callable,
ContextManager,
Dict,
Iterator,
List,
Optional,
Tuple,
Union,
)
import numpy as np
import pandas as pd
import pyarrow
import pyarrow as pa
from pydantic import StrictStr
from pydantic.typing import Literal
from pytz import utc
from feast import OnDemandFeatureView
from feast.data_source import DataSource
from feast.errors import InvalidEntityType
from feast.feature_logging import LoggingConfig, LoggingSource
from feast.feature_view import DUMMY_ENTITY_ID, DUMMY_ENTITY_VAL, FeatureView
from feast.infra.offline_stores import offline_utils
from feast.infra.offline_stores.contrib.athena_offline_store.athena_source import (
AthenaLoggingDestination,
AthenaSource,
SavedDatasetAthenaStorage,
)
from feast.infra.offline_stores.offline_store import (
OfflineStore,
RetrievalJob,
RetrievalMetadata,
)
from feast.infra.registry.base_registry import BaseRegistry
from feast.infra.registry.registry import Registry
from feast.infra.utils import aws_utils
from feast.repo_config import FeastConfigBaseModel, RepoConfig
from feast.saved_dataset import SavedDatasetStorage
from feast.usage import log_exceptions_and_usage
[docs]class AthenaOfflineStoreConfig(FeastConfigBaseModel):
"""Offline store config for AWS Athena"""
type: Literal["athena"] = "athena"
""" Offline store type selector"""
data_source: StrictStr
""" athena data source ex) AwsDataCatalog """
region: StrictStr
""" Athena's AWS region """
database: StrictStr
""" Athena database name """
workgroup: StrictStr
""" Athena workgroup name """
s3_staging_location: StrictStr
""" S3 path for importing & exporting data to Athena """
[docs]class AthenaOfflineStore(OfflineStore):
[docs] @staticmethod
@log_exceptions_and_usage(offline_store="athena")
def pull_latest_from_table_or_query(
config: RepoConfig,
data_source: DataSource,
join_key_columns: List[str],
feature_name_columns: List[str],
timestamp_field: str,
created_timestamp_column: Optional[str],
start_date: datetime,
end_date: datetime,
) -> RetrievalJob:
assert isinstance(config.offline_store, AthenaOfflineStoreConfig)
assert isinstance(data_source, AthenaSource)
from_expression = data_source.get_table_query_string(config)
partition_by_join_key_string = ", ".join(join_key_columns)
if partition_by_join_key_string != "":
partition_by_join_key_string = (
"PARTITION BY " + partition_by_join_key_string
)
timestamp_columns = [timestamp_field]
if created_timestamp_column:
timestamp_columns.append(created_timestamp_column)
timestamp_desc_string = " DESC, ".join(timestamp_columns) + " DESC"
field_string = ", ".join(
join_key_columns + feature_name_columns + timestamp_columns
)
date_partition_column = data_source.date_partition_column
athena_client = aws_utils.get_athena_data_client(config.offline_store.region)
s3_resource = aws_utils.get_s3_resource(config.offline_store.region)
start_date = start_date.astimezone(tz=utc)
end_date = end_date.astimezone(tz=utc)
query = f"""
SELECT
{field_string}
{f", {repr(DUMMY_ENTITY_VAL)} AS {DUMMY_ENTITY_ID}" if not join_key_columns else ""}
FROM (
SELECT {field_string},
ROW_NUMBER() OVER({partition_by_join_key_string} ORDER BY {timestamp_desc_string}) AS _feast_row
FROM {from_expression}
WHERE {timestamp_field} BETWEEN TIMESTAMP '{start_date.strftime('%Y-%m-%d %H:%M:%S')}' AND TIMESTAMP '{end_date.strftime('%Y-%m-%d %H:%M:%S')}'
{"AND "+date_partition_column+" >= '"+start_date.strftime('%Y-%m-%d')+"' AND "+date_partition_column+" <= '"+end_date.strftime('%Y-%m-%d')+"' " if date_partition_column != "" and date_partition_column is not None else ''}
)
WHERE _feast_row = 1
"""
# When materializing a single feature view, we don't need full feature names. On demand transforms aren't materialized
return AthenaRetrievalJob(
query=query,
athena_client=athena_client,
s3_resource=s3_resource,
config=config,
full_feature_names=False,
)
[docs] @staticmethod
@log_exceptions_and_usage(offline_store="athena")
def pull_all_from_table_or_query(
config: RepoConfig,
data_source: DataSource,
join_key_columns: List[str],
feature_name_columns: List[str],
timestamp_field: str,
start_date: datetime,
end_date: datetime,
) -> RetrievalJob:
assert isinstance(config.offline_store, AthenaOfflineStoreConfig)
assert isinstance(data_source, AthenaSource)
from_expression = data_source.get_table_query_string(config)
field_string = ", ".join(
join_key_columns + feature_name_columns + [timestamp_field]
)
athena_client = aws_utils.get_athena_data_client(config.offline_store.region)
s3_resource = aws_utils.get_s3_resource(config.offline_store.region)
date_partition_column = data_source.date_partition_column
query = f"""
SELECT {field_string}
FROM {from_expression}
WHERE {timestamp_field} BETWEEN TIMESTAMP '{start_date.astimezone(tz=utc).strftime("%Y-%m-%d %H:%M:%S.%f")[:-3]}' AND TIMESTAMP '{end_date.astimezone(tz=utc).strftime("%Y-%m-%d %H:%M:%S.%f")[:-3]}'
{"AND "+date_partition_column+" >= '"+start_date.strftime('%Y-%m-%d')+"' AND "+date_partition_column+" <= '"+end_date.strftime('%Y-%m-%d')+"' " if date_partition_column != "" and date_partition_column is not None else ''}
"""
return AthenaRetrievalJob(
query=query,
athena_client=athena_client,
s3_resource=s3_resource,
config=config,
full_feature_names=False,
)
[docs] @staticmethod
@log_exceptions_and_usage(offline_store="athena")
def get_historical_features(
config: RepoConfig,
feature_views: List[FeatureView],
feature_refs: List[str],
entity_df: Union[pd.DataFrame, str],
registry: Registry,
project: str,
full_feature_names: bool = False,
) -> RetrievalJob:
assert isinstance(config.offline_store, AthenaOfflineStoreConfig)
for fv in feature_views:
assert isinstance(fv.batch_source, AthenaSource)
athena_client = aws_utils.get_athena_data_client(config.offline_store.region)
s3_resource = aws_utils.get_s3_resource(config.offline_store.region)
# get pandas dataframe consisting of 1 row (LIMIT 1) and generate the schema out of it
entity_schema = _get_entity_schema(
entity_df, athena_client, config, s3_resource
)
# find timestamp column of entity df.(default = "event_timestamp"). Exception occurs if there are more than two timestamp columns.
entity_df_event_timestamp_col = (
offline_utils.infer_event_timestamp_from_entity_df(entity_schema)
)
# get min,max of event_timestamp.
entity_df_event_timestamp_range = _get_entity_df_event_timestamp_range(
entity_df,
entity_df_event_timestamp_col,
athena_client,
config,
)
@contextlib.contextmanager
def query_generator() -> Iterator[str]:
table_name = offline_utils.get_temp_entity_table_name()
_upload_entity_df(entity_df, athena_client, config, s3_resource, table_name)
expected_join_keys = offline_utils.get_expected_join_keys(
project, feature_views, registry
)
offline_utils.assert_expected_columns_in_entity_df(
entity_schema, expected_join_keys, entity_df_event_timestamp_col
)
# Build a query context containing all information required to template the Athena SQL query
query_context = offline_utils.get_feature_view_query_context(
feature_refs,
feature_views,
registry,
project,
entity_df_event_timestamp_range,
)
# Generate the Athena SQL query from the query context
query = offline_utils.build_point_in_time_query(
query_context,
left_table_query_string=table_name,
entity_df_event_timestamp_col=entity_df_event_timestamp_col,
entity_df_columns=entity_schema.keys(),
query_template=MULTIPLE_FEATURE_VIEW_POINT_IN_TIME_JOIN,
full_feature_names=full_feature_names,
)
try:
yield query
finally:
# Always clean up the temp Athena table
aws_utils.execute_athena_query(
athena_client,
config.offline_store.data_source,
config.offline_store.database,
config.offline_store.workgroup,
f"DROP TABLE IF EXISTS {config.offline_store.database}.{table_name}",
)
bucket = config.offline_store.s3_staging_location.replace(
"s3://", ""
).split("/", 1)[0]
aws_utils.delete_s3_directory(
s3_resource, bucket, "entity_df/" + table_name + "/"
)
return AthenaRetrievalJob(
query=query_generator,
athena_client=athena_client,
s3_resource=s3_resource,
config=config,
full_feature_names=full_feature_names,
on_demand_feature_views=OnDemandFeatureView.get_requested_odfvs(
feature_refs, project, registry
),
metadata=RetrievalMetadata(
features=feature_refs,
keys=list(entity_schema.keys() - {entity_df_event_timestamp_col}),
min_event_timestamp=entity_df_event_timestamp_range[0],
max_event_timestamp=entity_df_event_timestamp_range[1],
),
)
[docs] @staticmethod
def write_logged_features(
config: RepoConfig,
data: Union[pyarrow.Table, Path],
source: LoggingSource,
logging_config: LoggingConfig,
registry: BaseRegistry,
):
destination = logging_config.destination
assert isinstance(destination, AthenaLoggingDestination)
athena_client = aws_utils.get_athena_data_client(config.offline_store.region)
s3_resource = aws_utils.get_s3_resource(config.offline_store.region)
if isinstance(data, Path):
s3_path = f"{config.offline_store.s3_staging_location}/logged_features/{uuid.uuid4()}"
else:
s3_path = f"{config.offline_store.s3_staging_location}/logged_features/{uuid.uuid4()}.parquet"
aws_utils.upload_arrow_table_to_athena(
table=data,
athena_client=athena_client,
data_source=config.offline_store.data_source,
database=config.offline_store.database,
workgroup=config.offline_store.workgroup,
s3_resource=s3_resource,
s3_path=s3_path,
table_name=destination.table_name,
schema=source.get_schema(registry),
fail_if_exists=False,
)
[docs]class AthenaRetrievalJob(RetrievalJob):
def __init__(
self,
query: Union[str, Callable[[], ContextManager[str]]],
athena_client,
s3_resource,
config: RepoConfig,
full_feature_names: bool,
on_demand_feature_views: Optional[List[OnDemandFeatureView]] = None,
metadata: Optional[RetrievalMetadata] = None,
):
"""Initialize AthenaRetrievalJob object.
Args:
query: Athena SQL query to execute. Either a string, or a generator function that handles the artifact cleanup.
athena_client: boto3 athena client
s3_resource: boto3 s3 resource object
config: Feast repo config
full_feature_names: Whether to add the feature view prefixes to the feature names
on_demand_feature_views (optional): A list of on demand transforms to apply at retrieval time
"""
if not isinstance(query, str):
self._query_generator = query
else:
@contextlib.contextmanager
def query_generator() -> Iterator[str]:
assert isinstance(query, str)
yield query
self._query_generator = query_generator
self._athena_client = athena_client
self._s3_resource = s3_resource
self._config = config
self._full_feature_names = full_feature_names
self._on_demand_feature_views = on_demand_feature_views or []
self._metadata = metadata
@property
def full_feature_names(self) -> bool:
return self._full_feature_names
@property
def on_demand_feature_views(self) -> List[OnDemandFeatureView]:
return self._on_demand_feature_views
[docs] def get_temp_s3_path(self) -> str:
return (
self._config.offline_store.s3_staging_location
+ "/unload/"
+ str(uuid.uuid4())
)
@log_exceptions_and_usage
def _to_df_internal(self, timeout: Optional[int] = None) -> pd.DataFrame:
with self._query_generator() as query:
temp_table_name = "_" + str(uuid.uuid4()).replace("-", "")
temp_external_location = self.get_temp_s3_path()
return aws_utils.unload_athena_query_to_df(
self._athena_client,
self._config.offline_store.data_source,
self._config.offline_store.database,
self._config.offline_store.workgroup,
self._s3_resource,
temp_external_location,
self.get_temp_table_dml_header(temp_table_name, temp_external_location)
+ query,
temp_table_name,
)
@log_exceptions_and_usage
def _to_arrow_internal(self, timeout: Optional[int] = None) -> pa.Table:
with self._query_generator() as query:
temp_table_name = "_" + str(uuid.uuid4()).replace("-", "")
temp_external_location = self.get_temp_s3_path()
return aws_utils.unload_athena_query_to_pa(
self._athena_client,
self._config.offline_store.data_source,
self._config.offline_store.database,
self._config.offline_store.workgroup,
self._s3_resource,
temp_external_location,
self.get_temp_table_dml_header(temp_table_name, temp_external_location)
+ query,
temp_table_name,
)
@property
def metadata(self) -> Optional[RetrievalMetadata]:
return self._metadata
[docs] def persist(
self,
storage: SavedDatasetStorage,
allow_overwrite: Optional[bool] = False,
timeout: Optional[int] = None,
):
assert isinstance(storage, SavedDatasetAthenaStorage)
self.to_athena(table_name=storage.athena_options.table)
[docs] @log_exceptions_and_usage
def to_athena(self, table_name: str) -> None:
if self.on_demand_feature_views:
transformed_df = self.to_df()
_upload_entity_df(
transformed_df,
self._athena_client,
self._config,
self._s3_resource,
table_name,
)
return
with self._query_generator() as query:
query = f'CREATE TABLE "{table_name}" AS ({query});\n'
aws_utils.execute_athena_query(
self._athena_client,
self._config.offline_store.data_source,
self._config.offline_store.database,
self._config.offline_store.workgroup,
query,
)
def _upload_entity_df(
entity_df: Union[pd.DataFrame, str],
athena_client,
config: RepoConfig,
s3_resource,
table_name: str,
):
if isinstance(entity_df, pd.DataFrame):
# If the entity_df is a pandas dataframe, upload it to Athena
aws_utils.upload_df_to_athena(
athena_client,
config.offline_store.data_source,
config.offline_store.database,
config.offline_store.workgroup,
s3_resource,
f"{config.offline_store.s3_staging_location}/entity_df/{table_name}/{table_name}.parquet",
table_name,
entity_df,
)
elif isinstance(entity_df, str):
# If the entity_df is a string (SQL query), create a Athena table out of it
aws_utils.execute_athena_query(
athena_client,
config.offline_store.data_source,
config.offline_store.database,
config.offline_store.workgroup,
f"CREATE TABLE {table_name} AS ({entity_df})",
)
else:
raise InvalidEntityType(type(entity_df))
def _get_entity_schema(
entity_df: Union[pd.DataFrame, str],
athena_client,
config: RepoConfig,
s3_resource,
) -> Dict[str, np.dtype]:
if isinstance(entity_df, pd.DataFrame):
return dict(zip(entity_df.columns, entity_df.dtypes))
elif isinstance(entity_df, str):
# get pandas dataframe consisting of 1 row (LIMIT 1) and generate the schema out of it
entity_df_sample = AthenaRetrievalJob(
f"SELECT * FROM ({entity_df}) LIMIT 1",
athena_client,
s3_resource,
config,
full_feature_names=False,
).to_df()
return dict(zip(entity_df_sample.columns, entity_df_sample.dtypes))
else:
raise InvalidEntityType(type(entity_df))
def _get_entity_df_event_timestamp_range(
entity_df: Union[pd.DataFrame, str],
entity_df_event_timestamp_col: str,
athena_client,
config: RepoConfig,
) -> Tuple[datetime, datetime]:
if isinstance(entity_df, pd.DataFrame):
entity_df_event_timestamp = entity_df.loc[
:, entity_df_event_timestamp_col
].infer_objects()
if pd.api.types.is_string_dtype(entity_df_event_timestamp):
entity_df_event_timestamp = pd.to_datetime(
entity_df_event_timestamp, utc=True
)
entity_df_event_timestamp_range = (
entity_df_event_timestamp.min().to_pydatetime(),
entity_df_event_timestamp.max().to_pydatetime(),
)
elif isinstance(entity_df, str):
# If the entity_df is a string (SQL query), determine range
# from table
statement_id = aws_utils.execute_athena_query(
athena_client,
config.offline_store.data_source,
config.offline_store.database,
config.offline_store.workgroup,
f"SELECT MIN({entity_df_event_timestamp_col}) AS min, MAX({entity_df_event_timestamp_col}) AS max "
f"FROM ({entity_df})",
)
res = aws_utils.get_athena_query_result(athena_client, statement_id)
entity_df_event_timestamp_range = (
datetime.strptime(
res["Rows"][1]["Data"][0]["VarCharValue"], "%Y-%m-%d %H:%M:%S.%f"
),
datetime.strptime(
res["Rows"][1]["Data"][1]["VarCharValue"], "%Y-%m-%d %H:%M:%S.%f"
),
)
else:
raise InvalidEntityType(type(entity_df))
return entity_df_event_timestamp_range
MULTIPLE_FEATURE_VIEW_POINT_IN_TIME_JOIN = """
/*
Compute a deterministic hash for the `left_table_query_string` that will be used throughout
all the logic as the field to GROUP BY the data
*/
WITH entity_dataframe AS (
SELECT *,
{{entity_df_event_timestamp_col}} AS entity_timestamp
{% for featureview in featureviews %}
{% if featureview.entities %}
,(
{% for entity in featureview.entities %}
CAST({{entity}} as VARCHAR) ||
{% endfor %}
CAST({{entity_df_event_timestamp_col}} AS VARCHAR)
) AS {{featureview.name}}__entity_row_unique_id
{% else %}
,CAST({{entity_df_event_timestamp_col}} AS VARCHAR) AS {{featureview.name}}__entity_row_unique_id
{% endif %}
{% endfor %}
FROM {{ left_table_query_string }}
),
{% for featureview in featureviews %}
{{ featureview.name }}__entity_dataframe AS (
SELECT
{{ featureview.entities | join(', ')}}{% if featureview.entities %},{% else %}{% endif %}
entity_timestamp,
{{featureview.name}}__entity_row_unique_id
FROM entity_dataframe
GROUP BY
{{ featureview.entities | join(', ')}}{% if featureview.entities %},{% else %}{% endif %}
entity_timestamp,
{{featureview.name}}__entity_row_unique_id
),
/*
This query template performs the point-in-time correctness join for a single feature set table
to the provided entity table.
1. We first join the current feature_view to the entity dataframe that has been passed.
This JOIN has the following logic:
- For each row of the entity dataframe, only keep the rows where the `timestamp_field`
is less than the one provided in the entity dataframe
- If there a TTL for the current feature_view, also keep the rows where the `timestamp_field`
is higher the the one provided minus the TTL
- For each row, Join on the entity key and retrieve the `entity_row_unique_id` that has been
computed previously
The output of this CTE will contain all the necessary information and already filtered out most
of the data that is not relevant.
*/
{{ featureview.name }}__subquery AS (
SELECT
{{ featureview.timestamp_field }} as event_timestamp,
{{ featureview.created_timestamp_column ~ ' as created_timestamp,' if featureview.created_timestamp_column else '' }}
{{ featureview.entity_selections | join(', ')}}{% if featureview.entity_selections %},{% else %}{% endif %}
{% for feature in featureview.features %}
{{ feature }} as {% if full_feature_names %}{{ featureview.name }}__{{featureview.field_mapping.get(feature, feature)}}{% else %}{{ featureview.field_mapping.get(feature, feature) }}{% endif %}{% if loop.last %}{% else %}, {% endif %}
{% endfor %}
FROM {{ featureview.table_subquery }}
WHERE {{ featureview.timestamp_field }} <= from_iso8601_timestamp('{{ featureview.max_event_timestamp }}')
{% if featureview.date_partition_column != "" and featureview.date_partition_column is not none %}
AND {{ featureview.date_partition_column }} <= '{{ featureview.max_event_timestamp[:10] }}'
{% endif %}
{% if featureview.ttl == 0 %}{% else %}
AND {{ featureview.timestamp_field }} >= from_iso8601_timestamp('{{ featureview.min_event_timestamp }}')
{% if featureview.date_partition_column != "" and featureview.date_partition_column is not none %}
AND {{ featureview.date_partition_column }} >= '{{ featureview.min_event_timestamp[:10] }}'
{% endif %}
{% endif %}
),
{{ featureview.name }}__base AS (
SELECT
subquery.*,
entity_dataframe.entity_timestamp,
entity_dataframe.{{featureview.name}}__entity_row_unique_id
FROM {{ featureview.name }}__subquery AS subquery
INNER JOIN {{ featureview.name }}__entity_dataframe AS entity_dataframe
ON TRUE
AND subquery.event_timestamp <= entity_dataframe.entity_timestamp
{% if featureview.ttl == 0 %}{% else %}
AND subquery.event_timestamp >= entity_dataframe.entity_timestamp - {{ featureview.ttl }} * interval '1' second
{% endif %}
{% for entity in featureview.entities %}
AND subquery.{{ entity }} = entity_dataframe.{{ entity }}
{% endfor %}
),
/*
2. If the `created_timestamp_column` has been set, we need to
deduplicate the data first. This is done by calculating the
`MAX(created_at_timestamp)` for each event_timestamp.
We then join the data on the next CTE
*/
{% if featureview.created_timestamp_column %}
{{ featureview.name }}__dedup AS (
SELECT
{{featureview.name}}__entity_row_unique_id,
event_timestamp,
MAX(created_timestamp) as created_timestamp
FROM {{ featureview.name }}__base
GROUP BY {{featureview.name}}__entity_row_unique_id, event_timestamp
),
{% endif %}
/*
3. The data has been filtered during the first CTE "*__base"
Thus we only need to compute the latest timestamp of each feature.
*/
{{ featureview.name }}__latest AS (
SELECT
event_timestamp,
{% if featureview.created_timestamp_column %}created_timestamp,{% endif %}
{{featureview.name}}__entity_row_unique_id
FROM
(
SELECT base.*,
ROW_NUMBER() OVER(
PARTITION BY base.{{featureview.name}}__entity_row_unique_id
ORDER BY base.event_timestamp DESC{% if featureview.created_timestamp_column %},base.created_timestamp DESC{% endif %}
) AS row_number
FROM {{ featureview.name }}__base as base
{% if featureview.created_timestamp_column %}
INNER JOIN {{ featureview.name }}__dedup as dedup
ON TRUE
AND base.{{featureview.name}}__entity_row_unique_id = dedup.{{featureview.name}}__entity_row_unique_id
AND base.event_timestamp = dedup.event_timestamp
AND base.created_timestamp = dedup.created_timestamp
{% endif %}
)
WHERE row_number = 1
),
/*
4. Once we know the latest value of each feature for a given timestamp,
we can join again the data back to the original "base" dataset
*/
{{ featureview.name }}__cleaned AS (
SELECT base.*
FROM {{ featureview.name }}__base as base
INNER JOIN {{ featureview.name }}__latest as latest
ON TRUE
AND base.{{featureview.name}}__entity_row_unique_id = latest.{{featureview.name}}__entity_row_unique_id
AND base.event_timestamp = latest.event_timestamp
{% if featureview.created_timestamp_column %}
AND base.created_timestamp = latest.created_timestamp
{% endif %}
){% if loop.last %}{% else %}, {% endif %}
{% endfor %}
/*
Joins the outputs of multiple time travel joins to a single table.
The entity_dataframe dataset being our source of truth here.
*/
SELECT {{ final_output_feature_names | join(', ')}}
FROM entity_dataframe as entity_df
{% for featureview in featureviews %}
LEFT JOIN (
SELECT
{{featureview.name}}__entity_row_unique_id
{% for feature in featureview.features %}
,{% if full_feature_names %}{{ featureview.name }}__{{featureview.field_mapping.get(feature, feature)}}{% else %}{{ featureview.field_mapping.get(feature, feature) }}{% endif %}
{% endfor %}
FROM {{ featureview.name }}__cleaned
) as cleaned
ON TRUE
AND entity_df.{{featureview.name}}__entity_row_unique_id = cleaned.{{featureview.name}}__entity_row_unique_id
{% endfor %}
"""