Big Data Management User Guide

Back Next

Windowing Configuration

When you include a window function in an Expression transformation, you configure the windowing properties associated with the function. Windowing properties define the partitioning, ordering, and frame boundaries associated with a particular input row.

Configure a transformation for windowing on the Windowing tab.

The Windowing tab shows three configuration sections. The frame section has a start offset and end offset selection box, with "All Rows Preceding" and "All Rows Following" checkboxes. You can specify the order keys and partition keys by value or parameter, and choose ports from the transformation. You can choose Ascending or Descending for the order key.

You configure the following groups of properties on the Windowing tab:

Frame: Defines the rows that are included in the frame for the current input row, based on physical offsets from the position of the current input row.
You configure a frame if you use an aggregate function as a window function. The window functions LEAD and LAG reference individual rows and ignore the frame specification.
Partition Keys: Separate the input rows into different partitions. If you do not define partition keys, all rows belong to a single partition.
Order Keys: Define how rows in a partition are ordered. The ports you choose determine the position of a row within a partition. The order key can be ascending or descending. If you do not define order keys, the rows have no particular order.

Stateful Computing on the Spark Engine

Frame

Partition and Order Keys

Rules and Guidelines for Windowing Configuration

Download Guide

Watch

Comments

Communities

Knowledge Base

Success Portal