def test_transform(self, feature_set_dataframe):
test_key = TimestampFeature()
df = test_key.transform(feature_set_dataframe)
assert df.schema[TIMESTAMP_COLUMN].dataType == DataType.TIMESTAMP.spark
def test_transform_ms(self, feature_set_dataframe_ms):
test_key = TimestampFeature(from_ms=True)
df = test_key.transform(feature_set_dataframe_ms).orderBy("timestamp")
df = df.withColumn("timestamp",
df["timestamp"].cast(StringType())).collect()
assert df[0]["timestamp"] == "2020-02-12 21:18:31"
assert df[1]["timestamp"] == "2020-02-12 21:18:42"
def test_transform_mask(self, feature_set_dataframe_date):
test_key = TimestampFeature(mask="yyyy-MM-dd")
df = test_key.transform(feature_set_dataframe_date).orderBy(
"timestamp")
df = df.withColumn("timestamp",
df["timestamp"].cast(StringType())).collect()
assert df[0]["timestamp"] == "2020-02-07 00:00:00"
assert df[1]["timestamp"] == "2020-02-08 00:00:00"
def test_construct(
self, feature_set_dataframe, fixed_windows_output_feature_set_dataframe
):
# given
spark_client = SparkClient()
# arrange
feature_set = FeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=SparkFunctionTransform(
functions=[
Function(F.avg, DataType.FLOAT),
Function(F.stddev_pop, DataType.FLOAT),
]
).with_window(
partition_by="id",
order_by=TIMESTAMP_COLUMN,
mode="fixed_windows",
window_definition=["2 minutes", "15 minutes"],
),
),
Feature(
name="divided_feature",
description="unit test",
dtype=DataType.FLOAT,
transformation=CustomTransform(
transformer=divide, column1="feature1", column2="feature2",
),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(),
)
output_df = (
feature_set.construct(feature_set_dataframe, client=spark_client)
.orderBy(feature_set.timestamp_column)
.select(feature_set.columns)
)
target_df = fixed_windows_output_feature_set_dataframe.orderBy(
feature_set.timestamp_column
).select(feature_set.columns)
# assert
assert_dataframe_equality(output_df, target_df)
def test_h3_feature_set(self, h3_input_df, h3_target_df):
spark_client = SparkClient()
feature_set = AggregatedFeatureSet(
name="h3_test",
entity="h3geolocation",
description="Test",
keys=[
KeyFeature(
name="h3_id",
description="The h3 hash ID",
dtype=DataType.DOUBLE,
transformation=H3HashTransform(
h3_resolutions=[6, 7, 8, 9, 10, 11, 12],
lat_column="lat",
lng_column="lng",
).with_stack(),
)
],
timestamp=TimestampFeature(),
features=[
Feature(
name="house_id",
description="Count of house ids over a day.",
transformation=AggregatedTransform(
functions=[Function(F.count, DataType.BIGINT)]),
),
],
).with_windows(definitions=["1 day"])
output_df = feature_set.construct(h3_input_df,
client=spark_client,
end_date="2016-04-14")
assert_dataframe_equality(output_df, h3_target_df)
def test_args_without_transformation(self):
test_key = TimestampFeature(from_column="ts")
assert test_key.name == TIMESTAMP_COLUMN
assert test_key.from_column == "ts"
assert test_key.dtype == DataType.TIMESTAMP
def test_run_with_repartition(self, spark_session):
test_pipeline = FeatureSetPipeline(
spark_client=SparkClient(),
source=Mock(
spec=Source,
readers=[TableReader(id="source_a", database="db", table="table",)],
query="select * from source_a",
),
feature_set=Mock(
spec=FeatureSet,
name="feature_set",
entity="entity",
description="description",
keys=[
KeyFeature(
name="user_id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(from_column="ts"),
features=[
Feature(
name="listing_page_viewed__rent_per_month",
description="Average of something.",
transformation=SparkFunctionTransform(
functions=[
Function(functions.avg, DataType.FLOAT),
Function(functions.stddev_pop, DataType.FLOAT),
],
).with_window(
partition_by="user_id",
order_by=TIMESTAMP_COLUMN,
window_definition=["7 days", "2 weeks"],
mode="fixed_windows",
),
),
],
),
sink=Mock(
spec=Sink, writers=[HistoricalFeatureStoreWriter(db_config=None)],
),
)
# feature_set need to return a real df for streaming validation
sample_df = spark_session.createDataFrame([{"a": "x", "b": "y", "c": "3"}])
test_pipeline.feature_set.construct.return_value = sample_df
test_pipeline.run(partition_by=["id"])
test_pipeline.source.construct.assert_called_once()
test_pipeline.feature_set.construct.assert_called_once()
test_pipeline.sink.flush.assert_called_once()
test_pipeline.sink.validate.assert_called_once()
def timestamp(self, value: TimestampFeature):
if not isinstance(value, TimestampFeature):
raise ValueError(
"timestamp needs to be a TimestampFeature object.")
timestamp_columns = value.get_output_columns()
if len(timestamp_columns) > 1:
raise ValueError(
"TimestampFeature will produce multiple output columns.")
self.__timestamp = value
def test_construct_rolling_windows_with_end_date(
self,
feature_set_dataframe,
rolling_windows_output_feature_set_dataframe_base_date,
):
# given
spark_client = SparkClient()
# arrange
feature_set = AggregatedFeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=AggregatedTransform(functions=[
Function(F.avg, DataType.DOUBLE),
Function(F.stddev_pop, DataType.DOUBLE),
], ),
),
Feature(
name="feature2",
description="test",
transformation=AggregatedTransform(functions=[
Function(F.avg, DataType.DOUBLE),
Function(F.stddev_pop, DataType.DOUBLE),
], ),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(),
).with_windows(definitions=["1 day", "1 week"])
# act
output_df = feature_set.construct(
feature_set_dataframe, client=spark_client,
end_date="2016-04-18").orderBy("timestamp")
target_df = rolling_windows_output_feature_set_dataframe_base_date.orderBy(
feature_set.timestamp_column).select(feature_set.columns)
# assert
assert_dataframe_equality(output_df, target_df)
def test_construct_without_window(
self,
feature_set_dataframe,
target_df_without_window,
):
# given
spark_client = SparkClient()
# arrange
feature_set = AggregatedFeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
dtype=DataType.DOUBLE,
transformation=AggregatedTransform(
functions=[Function(F.avg, DataType.DOUBLE)]),
),
Feature(
name="feature2",
description="test",
dtype=DataType.FLOAT,
transformation=AggregatedTransform(
functions=[Function(F.count, DataType.BIGINT)]),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(from_column="fixed_ts"),
)
# act
output_df = feature_set.construct(feature_set_dataframe,
client=spark_client)
# assert
assert_dataframe_equality(output_df, target_df_without_window)
def test_feature_set_raise(self):
with pytest.raises(
ValueError, match="feature_set must be a FeatureSet instance"
):
FeatureSetPipeline(
spark_client=SparkClient(),
source=Mock(
spec=Source,
readers=[
TableReader(id="source_a", database="db", table="table",),
],
query="select * from source_a",
),
feature_set=Mock(
name="feature_set",
entity="entity",
description="description",
keys=[
KeyFeature(
name="user_id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(from_column="ts"),
features=[
Feature(
name="listing_page_viewed__rent_per_month",
description="Average of something.",
transformation=SparkFunctionTransform(
functions=[
Function(functions.avg, DataType.FLOAT),
Function(functions.stddev_pop, DataType.FLOAT),
],
).with_window(
partition_by="user_id",
order_by=TIMESTAMP_COLUMN,
window_definition=["7 days", "2 weeks"],
mode="fixed_windows",
),
),
],
),
sink=Mock(
spec=Sink, writers=[HistoricalFeatureStoreWriter(db_config=None)],
),
)
def test_construct_with_pivot(
self,
feature_set_df_pivot,
target_df_pivot_agg,
):
# given
spark_client = SparkClient()
# arrange
feature_set = AggregatedFeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature",
description="unit test",
transformation=AggregatedTransform(functions=[
Function(F.avg, DataType.FLOAT),
Function(F.stddev_pop, DataType.DOUBLE),
], ),
from_column="feature1",
)
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(from_column="fixed_ts"),
).with_pivot("pivot_col", ["S", "N"])
# act
output_df = feature_set.construct(feature_set_df_pivot,
client=spark_client)
# assert
assert_dataframe_equality(output_df, target_df_pivot_agg)
def feature_set():
key_features = [
KeyFeature(name="id",
description="Description",
dtype=DataType.INTEGER)
]
ts_feature = TimestampFeature(from_column="timestamp")
features = [
Feature(name="feature",
description="Description",
dtype=DataType.FLOAT),
]
return FeatureSet(
"test_sink_feature_set",
"test_sink_entity",
"description",
keys=key_features,
timestamp=ts_feature,
features=features,
)
def test_construct_rolling_windows_without_end_date(
self, feature_set_dataframe,
rolling_windows_output_feature_set_dataframe):
# given
spark_client = SparkClient()
# arrange
feature_set = AggregatedFeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=AggregatedTransform(functions=[
Function(F.avg, DataType.DOUBLE),
Function(F.stddev_pop, DataType.DOUBLE),
], ),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(),
).with_windows(definitions=["1 day", "1 week"], )
# act & assert
with pytest.raises(ValueError):
_ = feature_set.construct(feature_set_dataframe,
client=spark_client)
def test_get_schema(self):
expected_schema = [
{
"column_name": "id",
"type": LongType(),
"primary_key": True
},
{
"column_name": "timestamp",
"type": TimestampType(),
"primary_key": False
},
{
"column_name": "feature1__avg_over_2_minutes_fixed_windows",
"type": FloatType(),
"primary_key": False,
},
{
"column_name": "feature1__avg_over_15_minutes_fixed_windows",
"type": FloatType(),
"primary_key": False,
},
{
"column_name":
"feature1__stddev_pop_over_2_minutes_fixed_windows",
"type": FloatType(),
"primary_key": False,
},
{
"column_name":
"feature1__stddev_pop_over_15_minutes_fixed_windows",
"type": FloatType(),
"primary_key": False,
},
]
feature_set = FeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=SparkFunctionTransform(functions=[
Function(F.avg, DataType.FLOAT),
Function(F.stddev_pop, DataType.FLOAT),
]).with_window(
partition_by="id",
order_by=TIMESTAMP_COLUMN,
mode="fixed_windows",
window_definition=["2 minutes", "15 minutes"],
),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.BIGINT,
)
],
timestamp=TimestampFeature(),
)
schema = feature_set.get_schema()
assert schema == expected_schema
def test_get_schema(self):
expected_schema = [
{
"column_name": "id",
"type": LongType(),
"primary_key": True
},
{
"column_name": "timestamp",
"type": TimestampType(),
"primary_key": False
},
{
"column_name": "feature1__avg_over_1_week_rolling_windows",
"type": DoubleType(),
"primary_key": False,
},
{
"column_name": "feature1__avg_over_2_days_rolling_windows",
"type": DoubleType(),
"primary_key": False,
},
{
"column_name":
"feature1__stddev_pop_over_1_week_rolling_windows",
"type": DoubleType(),
"primary_key": False,
},
{
"column_name":
"feature1__stddev_pop_over_2_days_rolling_windows",
"type": DoubleType(),
"primary_key": False,
},
{
"column_name": "feature2__count_over_1_week_rolling_windows",
"type": ArrayType(StringType(), True),
"primary_key": False,
},
{
"column_name": "feature2__count_over_2_days_rolling_windows",
"type": ArrayType(StringType(), True),
"primary_key": False,
},
]
feature_set = AggregatedFeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=AggregatedTransform(functions=[
Function(functions.avg, DataType.DOUBLE),
Function(functions.stddev_pop, DataType.DOUBLE),
], ),
),
Feature(
name="feature2",
description="test",
transformation=AggregatedTransform(functions=[
Function(functions.count, DataType.ARRAY_STRING)
]),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.BIGINT,
)
],
timestamp=TimestampFeature(),
).with_windows(definitions=["1 week", "2 days"])
schema = feature_set.get_schema()
assert schema == expected_schema
def test_feature_transform_with_data_type_array(self, spark_context,
spark_session):
# arrange
input_data = [
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 10
},
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 20
},
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 30
},
{
"id": 2,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 10
},
]
target_data = [
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature__collect_set": [30.0, 20.0, 10.0],
},
{
"id": 2,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature__collect_set": [10.0],
},
]
input_df = create_df_from_collection(
input_data, spark_context, spark_session).withColumn(
"timestamp",
functions.to_timestamp(functions.col("timestamp")))
target_df = create_df_from_collection(
target_data, spark_context, spark_session).withColumn(
"timestamp",
functions.to_timestamp(functions.col("timestamp")))
fs = AggregatedFeatureSet(
name="name",
entity="entity",
description="description",
keys=[
KeyFeature(name="id",
description="test",
dtype=DataType.INTEGER)
],
timestamp=TimestampFeature(),
features=[
Feature(
name="feature",
description="aggregations with ",
dtype=DataType.BIGINT,
transformation=AggregatedTransform(functions=[
Function(functions.collect_set, DataType.ARRAY_FLOAT),
], ),
from_column="feature",
),
],
)
# act
output_df = fs.construct(input_df, SparkClient())
# assert
assert_dataframe_equality(target_df, output_df)
def test_feature_transform_with_filter_expression(self, spark_context,
spark_session):
# arrange
input_data = [
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 10,
"type": "a",
},
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 20,
"type": "a",
},
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 30,
"type": "b",
},
{
"id": 2,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature": 10,
"type": "a",
},
]
target_data = [
{
"id": 1,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature_only_type_a__avg": 15.0,
"feature_only_type_a__min": 10,
"feature_only_type_a__max": 20,
},
{
"id": 2,
"timestamp": "2020-04-22T00:00:00+00:00",
"feature_only_type_a__avg": 10.0,
"feature_only_type_a__min": 10,
"feature_only_type_a__max": 10,
},
]
input_df = create_df_from_collection(
input_data, spark_context, spark_session).withColumn(
"timestamp",
functions.to_timestamp(functions.col("timestamp")))
target_df = create_df_from_collection(
target_data, spark_context, spark_session).withColumn(
"timestamp",
functions.to_timestamp(functions.col("timestamp")))
fs = AggregatedFeatureSet(
name="name",
entity="entity",
description="description",
keys=[
KeyFeature(name="id",
description="test",
dtype=DataType.INTEGER)
],
timestamp=TimestampFeature(),
features=[
Feature(
name="feature_only_type_a",
description="aggregations only when type = a",
dtype=DataType.BIGINT,
transformation=AggregatedTransform(
functions=[
Function(functions.avg, DataType.FLOAT),
Function(functions.min, DataType.FLOAT),
Function(functions.max, DataType.FLOAT),
],
filter_expression="type = 'a'",
),
from_column="feature",
),
],
)
# act
output_df = fs.construct(input_df, SparkClient())
# assert
assert_dataframe_equality(target_df, output_df)
def test_feature_set_args(self):
# arrange and act
out_columns = [
"user_id",
"timestamp",
"listing_page_viewed__rent_per_month__avg_over_7_days_fixed_windows",
"listing_page_viewed__rent_per_month__avg_over_2_weeks_fixed_windows",
"listing_page_viewed__rent_per_month__stddev_pop_over_7_days_fixed_windows",
"listing_page_viewed__rent_per_month__"
"stddev_pop_over_2_weeks_fixed_windows",
# noqa
]
pipeline = FeatureSetPipeline(
source=Source(
readers=[
TableReader(id="source_a", database="db", table="table",),
FileReader(id="source_b", path="path", format="parquet",),
],
query="select a.*, b.specific_feature "
"from source_a left join source_b on a.id=b.id",
),
feature_set=FeatureSet(
name="feature_set",
entity="entity",
description="description",
keys=[
KeyFeature(
name="user_id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(from_column="ts"),
features=[
Feature(
name="listing_page_viewed__rent_per_month",
description="Average of something.",
transformation=SparkFunctionTransform(
functions=[
Function(functions.avg, DataType.FLOAT),
Function(functions.stddev_pop, DataType.FLOAT),
],
).with_window(
partition_by="user_id",
order_by=TIMESTAMP_COLUMN,
window_definition=["7 days", "2 weeks"],
mode="fixed_windows",
),
),
],
),
sink=Sink(
writers=[
HistoricalFeatureStoreWriter(db_config=None),
OnlineFeatureStoreWriter(db_config=None),
],
),
)
assert isinstance(pipeline.spark_client, SparkClient)
assert len(pipeline.source.readers) == 2
assert all(isinstance(reader, Reader) for reader in pipeline.source.readers)
assert isinstance(pipeline.source.query, str)
assert pipeline.feature_set.name == "feature_set"
assert pipeline.feature_set.entity == "entity"
assert pipeline.feature_set.description == "description"
assert isinstance(pipeline.feature_set.timestamp, TimestampFeature)
assert len(pipeline.feature_set.keys) == 1
assert all(isinstance(k, KeyFeature) for k in pipeline.feature_set.keys)
assert len(pipeline.feature_set.features) == 1
assert all(
isinstance(feature, Feature) for feature in pipeline.feature_set.features
)
assert pipeline.feature_set.columns == out_columns
assert len(pipeline.sink.writers) == 2
assert all(isinstance(writer, Writer) for writer in pipeline.sink.writers)
def timestamp_c():
return TimestampFeature()
def test_feature_set_pipeline(self, mocked_df, spark_session,
fixed_windows_output_feature_set_dataframe):
# arrange
table_reader_id = "a_source"
table_reader_table = "table"
table_reader_db = environment.get_variable(
"FEATURE_STORE_HISTORICAL_DATABASE")
create_temp_view(dataframe=mocked_df, name=table_reader_id)
create_db_and_table(
spark=spark_session,
table_reader_id=table_reader_id,
table_reader_db=table_reader_db,
table_reader_table=table_reader_table,
)
dbconfig = Mock()
dbconfig.get_options = Mock(
return_value={
"mode": "overwrite",
"format_": "parquet",
"path": "test_folder/historical/entity/feature_set",
})
# act
test_pipeline = FeatureSetPipeline(
source=Source(
readers=[
TableReader(
id=table_reader_id,
database=table_reader_db,
table=table_reader_table,
),
],
query=f"select * from {table_reader_id} ", # noqa
),
feature_set=FeatureSet(
name="feature_set",
entity="entity",
description="description",
features=[
Feature(
name="feature1",
description="test",
transformation=SparkFunctionTransform(functions=[
Function(F.avg, DataType.FLOAT),
Function(F.stddev_pop, DataType.FLOAT),
], ).with_window(
partition_by="id",
order_by=TIMESTAMP_COLUMN,
mode="fixed_windows",
window_definition=["2 minutes", "15 minutes"],
),
),
Feature(
name="divided_feature",
description="unit test",
dtype=DataType.FLOAT,
transformation=CustomTransform(
transformer=divide,
column1="feature1",
column2="feature2",
),
),
],
keys=[
KeyFeature(
name="id",
description="The user's Main ID or device ID",
dtype=DataType.INTEGER,
)
],
timestamp=TimestampFeature(),
),
sink=Sink(
writers=[HistoricalFeatureStoreWriter(db_config=dbconfig)], ),
)
test_pipeline.run()
# assert
path = dbconfig.get_options("historical/entity/feature_set").get(
"path")
df = spark_session.read.parquet(path).orderBy(TIMESTAMP_COLUMN)
target_df = fixed_windows_output_feature_set_dataframe.orderBy(
test_pipeline.feature_set.timestamp_column)
# assert
assert_dataframe_equality(df, target_df)
# tear down
shutil.rmtree("test_folder")