tecton.FilteredSource

Summary

FilteredSource is a convenience utility that can be used to pre-filter sources in Feature View definitions.

Description

Filters source to the time range [start_time + start_time_offset, end_time) before executing the feature view transformation. When materializing a range, the output of Feature Views is automatically filtered by Tecton to the range [start_time, end_time). In most cases, using a FilteredSource is a performance optimization for when the query engine fails to “push down” timestamp filtering.

If source has configured tecton.declarative.DatetimePartitionColumn, Tecton will automatically apply partition filters, which can significantly improve performance.

To filter only on the end time, use FilteredSource(credit_scores, start_time_offset=datetime.timedelta.min).

Example

from tecton import batch_feature_view, BatchSource, HiveConfig, materialization_context, DatetimePartitionColumn
from tecton import FilteredSource

# Declare a BatchSource. `timestamp_field` is normally optional, but it is required when the source is used
# with a FilteredSource. `datetime_partition_columns` is also recommended and will help improve time
# filtering efficiency.
partition_columns = [
    DatetimePartitionColumn(column_name="partition_0", datepart="year", zero_padded=True),
    DatetimePartitionColumn(column_name="partition_1", datepart="month", zero_padded=True),
    DatetimePartitionColumn(column_name="partition_2", datepart="day", zero_padded=True),
]
credit_scores = BatchSource(
    name="credit_scores_batch",
    batch_config=HiveConfig(
        database="demo_fraud",
        table="credit_scores",
        timestamp_field="timestamp",
        datetime_partition_columns=partition_columns,
    ),
)

# Declare a batch feature view with `credit_scores` as a source.
@batch_feature_view(sources=[FilteredSource(credit_scores)], ...)
def user_credit_score(credit_scores):
    return f"""
        SELECT user_id, timestamp, credit_score
        FROM {credit_scores}
        """


# The above feature view query is equivalent to:
@batch_feature_view(sources=[credit_scores], ...)
def user_credit_score(credit_scores, context=materialization_context()):
    return f"""
        WITH FILTERED_CREDIT_SCORES AS (
            SELECT *
            FROM {credit_scores}
            WHERE
                timestamp >= to_timestamp("{context.start_time}")
                AND timestamp < to_timestamp("{context.end_time}")
                -- AND additional clauses if the data source has datetime_partition_columns configured.
        )
        SELECT user_id, timestamp, credit_score
        FROM FILTERED_CREDIT_SCORES
        """

Attributes

The attributes are the same as the __init__ method parameters. See below.

Methods

__init__(...)

Parameters

• source (DataSource) – Data Source that this FilteredSource class wraps.

• start_time_offset (Optional[timedelta]) – FilteredSource will pre-filter the data source to the time range [start_time + start_time_offset, end_time). Must be zero or negative. Defaults to zero.

Returns

A FilteredSource to pass into a Feature View.