import contextlib
from dataclasses import asdict
from datetime import datetime
from typing import (
Any,
Callable,
ContextManager,
Dict,
Iterator,
KeysView,
List,
Optional,
Tuple,
Union,
)
import numpy as np
import pandas as pd
import pyarrow as pa
from jinja2 import BaseLoader, Environment
from psycopg2 import sql
from pydantic.typing import Literal
from pytz import utc
from feast.data_source import DataSource
from feast.errors import InvalidEntityType
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.postgres_offline_store.postgres_source import (
SavedDatasetPostgreSQLStorage,
)
from feast.infra.offline_stores.offline_store import (
OfflineStore,
RetrievalJob,
RetrievalMetadata,
)
from feast.infra.registry.registry import Registry
from feast.infra.utils.postgres.connection_utils import (
_get_conn,
df_to_postgres_table,
get_query_schema,
)
from feast.infra.utils.postgres.postgres_config import PostgreSQLConfig
from feast.on_demand_feature_view import OnDemandFeatureView
from feast.repo_config import RepoConfig
from feast.saved_dataset import SavedDatasetStorage
from feast.type_map import pg_type_code_to_arrow
from feast.usage import log_exceptions_and_usage
from .postgres_source import PostgreSQLSource
[docs]class PostgreSQLOfflineStoreConfig(PostgreSQLConfig):
type: Literal["postgres"] = "postgres"
[docs]class PostgreSQLOfflineStore(OfflineStore):
[docs] @staticmethod
@log_exceptions_and_usage(offline_store="postgres")
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, PostgreSQLOfflineStoreConfig)
assert isinstance(data_source, PostgreSQLSource)
from_expression = data_source.get_table_query_string()
partition_by_join_key_string = ", ".join(_append_alias(join_key_columns, "a"))
if partition_by_join_key_string != "":
partition_by_join_key_string = (
"PARTITION BY " + partition_by_join_key_string
)
timestamps = [timestamp_field]
if created_timestamp_column:
timestamps.append(created_timestamp_column)
timestamp_desc_string = " DESC, ".join(_append_alias(timestamps, "a")) + " DESC"
a_field_string = ", ".join(
_append_alias(join_key_columns + feature_name_columns + timestamps, "a")
)
b_field_string = ", ".join(
_append_alias(join_key_columns + feature_name_columns + timestamps, "b")
)
query = f"""
SELECT
{b_field_string}
{f", {repr(DUMMY_ENTITY_VAL)} AS {DUMMY_ENTITY_ID}" if not join_key_columns else ""}
FROM (
SELECT {a_field_string},
ROW_NUMBER() OVER({partition_by_join_key_string} ORDER BY {timestamp_desc_string}) AS _feast_row
FROM ({from_expression}) a
WHERE a."{timestamp_field}" BETWEEN '{start_date}'::timestamptz AND '{end_date}'::timestamptz
) b
WHERE _feast_row = 1
"""
return PostgreSQLRetrievalJob(
query=query,
config=config,
full_feature_names=False,
on_demand_feature_views=None,
)
[docs] @staticmethod
@log_exceptions_and_usage(offline_store="postgres")
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, PostgreSQLOfflineStoreConfig)
for fv in feature_views:
assert isinstance(fv.batch_source, PostgreSQLSource)
entity_schema = _get_entity_schema(entity_df, config)
entity_df_event_timestamp_col = (
offline_utils.infer_event_timestamp_from_entity_df(entity_schema)
)
entity_df_event_timestamp_range = _get_entity_df_event_timestamp_range(
entity_df,
entity_df_event_timestamp_col,
config,
)
@contextlib.contextmanager
def query_generator() -> Iterator[str]:
table_name = offline_utils.get_temp_entity_table_name()
_upload_entity_df(config, entity_df, 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
)
query_context = offline_utils.get_feature_view_query_context(
feature_refs,
feature_views,
registry,
project,
entity_df_event_timestamp_range,
)
query_context_dict = [asdict(context) for context in query_context]
# Hack for query_context.entity_selections to support uppercase in columns
for context in query_context_dict:
context["entity_selections"] = [
f'''"{entity_selection.replace(' AS ', '" AS "')}\"'''
for entity_selection in context["entity_selections"]
]
try:
yield build_point_in_time_query(
query_context_dict,
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,
)
finally:
if table_name:
with _get_conn(config.offline_store) as conn, conn.cursor() as cur:
cur.execute(
sql.SQL(
"""
DROP TABLE IF EXISTS {};
"""
).format(sql.Identifier(table_name)),
)
return PostgreSQLRetrievalJob(
query=query_generator,
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
@log_exceptions_and_usage(offline_store="postgres")
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, PostgreSQLOfflineStoreConfig)
assert isinstance(data_source, PostgreSQLSource)
from_expression = data_source.get_table_query_string()
field_string = ", ".join(
join_key_columns + feature_name_columns + [timestamp_field]
)
start_date = start_date.astimezone(tz=utc)
end_date = end_date.astimezone(tz=utc)
query = f"""
SELECT {field_string}
FROM {from_expression} AS paftoq_alias
WHERE "{timestamp_field}" BETWEEN '{start_date}'::timestamptz AND '{end_date}'::timestamptz
"""
return PostgreSQLRetrievalJob(
query=query,
config=config,
full_feature_names=False,
on_demand_feature_views=None,
)
[docs]class PostgreSQLRetrievalJob(RetrievalJob):
def __init__(
self,
query: Union[str, Callable[[], ContextManager[str]]],
config: RepoConfig,
full_feature_names: bool,
on_demand_feature_views: Optional[List[OnDemandFeatureView]],
metadata: Optional[RetrievalMetadata] = None,
):
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.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
def _to_df_internal(self) -> pd.DataFrame:
# We use arrow format because it gives better control of the table schema
return self._to_arrow_internal().to_pandas()
[docs] def to_sql(self) -> str:
with self._query_generator() as query:
return query
def _to_arrow_internal(self) -> pa.Table:
with self._query_generator() as query:
with _get_conn(self.config.offline_store) as conn, conn.cursor() as cur:
conn.set_session(readonly=True)
cur.execute(query)
fields = [
(c.name, pg_type_code_to_arrow(c.type_code))
for c in cur.description
]
data = cur.fetchall()
schema = pa.schema(fields)
# TODO: Fix...
data_transposed: List[List[Any]] = []
for col in range(len(fields)):
data_transposed.append([])
for row in range(len(data)):
data_transposed[col].append(data[row][col])
table = pa.Table.from_arrays(
[pa.array(row) for row in data_transposed], schema=schema
)
return table
@property
def metadata(self) -> Optional[RetrievalMetadata]:
return self._metadata
[docs] def persist(self, storage: SavedDatasetStorage, allow_overwrite: bool = False):
assert isinstance(storage, SavedDatasetPostgreSQLStorage)
df_to_postgres_table(
config=self.config.offline_store,
df=self.to_df(),
table_name=storage.postgres_options._table,
)
def _get_entity_df_event_timestamp_range(
entity_df: Union[pd.DataFrame, str],
entity_df_event_timestamp_col: str,
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
with _get_conn(config.offline_store) as conn, conn.cursor() as cur:
cur.execute(
f"SELECT MIN({entity_df_event_timestamp_col}) AS min, MAX({entity_df_event_timestamp_col}) AS max FROM ({entity_df}) as tmp_alias"
),
res = cur.fetchone()
entity_df_event_timestamp_range = (res[0], res[1])
else:
raise InvalidEntityType(type(entity_df))
return entity_df_event_timestamp_range
def _append_alias(field_names: List[str], alias: str) -> List[str]:
return [f'{alias}."{field_name}"' for field_name in field_names]
[docs]def build_point_in_time_query(
feature_view_query_contexts: List[dict],
left_table_query_string: str,
entity_df_event_timestamp_col: str,
entity_df_columns: KeysView[str],
query_template: str,
full_feature_names: bool = False,
) -> str:
"""Build point-in-time query between each feature view table and the entity dataframe for PostgreSQL"""
template = Environment(loader=BaseLoader()).from_string(source=query_template)
final_output_feature_names = list(entity_df_columns)
final_output_feature_names.extend(
[
(
f'{fv["name"]}__{fv["field_mapping"].get(feature, feature)}'
if full_feature_names
else fv["field_mapping"].get(feature, feature)
)
for fv in feature_view_query_contexts
for feature in fv["features"]
]
)
# Add additional fields to dict
template_context = {
"left_table_query_string": left_table_query_string,
"entity_df_event_timestamp_col": entity_df_event_timestamp_col,
"unique_entity_keys": set(
[entity for fv in feature_view_query_contexts for entity in fv["entities"]]
),
"featureviews": feature_view_query_contexts,
"full_feature_names": full_feature_names,
"final_output_feature_names": final_output_feature_names,
}
query = template.render(template_context)
return query
def _upload_entity_df(
config: RepoConfig, entity_df: Union[pd.DataFrame, str], table_name: str
):
if isinstance(entity_df, pd.DataFrame):
# If the entity_df is a pandas dataframe, upload it to Postgres
df_to_postgres_table(config.offline_store, entity_df, table_name)
elif isinstance(entity_df, str):
# If the entity_df is a string (SQL query), create a Postgres table out of it
with _get_conn(config.offline_store) as conn, conn.cursor() as cur:
cur.execute(f"CREATE TABLE {table_name} AS ({entity_df})")
else:
raise InvalidEntityType(type(entity_df))
def _get_entity_schema(
entity_df: Union[pd.DataFrame, str],
config: RepoConfig,
) -> Dict[str, np.dtype]:
if isinstance(entity_df, pd.DataFrame):
return dict(zip(entity_df.columns, entity_df.dtypes))
elif isinstance(entity_df, str):
df_query = f"({entity_df}) AS sub"
return get_query_schema(config.offline_store, df_query)
else:
raise InvalidEntityType(type(entity_df))
# Copied from the Feast Redshift offline store implementation
# Note: Keep this in sync with sdk/python/feast/infra/offline_stores/redshift.py:
# MULTIPLE_FEATURE_VIEW_POINT_IN_TIME_JOIN
# https://github.com/feast-dev/feast/blob/master/sdk/python/feast/infra/offline_stores/redshift.py
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
{% if featureview.entities %}"{{ featureview.entities | join('", "') }}",{% endif %}
entity_timestamp,
"{{featureview.name}}__entity_row_unique_id"
FROM entity_dataframe
GROUP BY
{% if featureview.entities %}"{{ featureview.entities | join('", "')}}",{% 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 }} AS sub
WHERE "{{ featureview.timestamp_field }}" <= (SELECT MAX(entity_timestamp) FROM entity_dataframe)
{% if featureview.ttl == 0 %}{% else %}
AND "{{ featureview.timestamp_field }}" >= (SELECT MIN(entity_timestamp) FROM entity_dataframe) - {{ featureview.ttl }} * interval '1' second
{% 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 *,
ROW_NUMBER() OVER(
PARTITION BY "{{featureview.name}}__entity_row_unique_id"
ORDER BY event_timestamp DESC{% if featureview.created_timestamp_column %},created_timestamp DESC{% endif %}
) AS row_number
FROM "{{ featureview.name }}__base"
{% if featureview.created_timestamp_column %}
INNER JOIN "{{ featureview.name }}__dedup"
USING ("{{featureview.name}}__entity_row_unique_id", event_timestamp, created_timestamp)
{% endif %}
) AS sub
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"
USING(
"{{featureview.name}}__entity_row_unique_id",
event_timestamp
{% if featureview.created_timestamp_column %}
,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
{% 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 "{{featureview.name}}" USING ("{{featureview.name}}__entity_row_unique_id")
{% endfor %}
"""