Rules and Guidelines for Transformations
Certain transformations are valid with restrictions with the Window transformation. The following table describes the rules and guidelines for transformations:
Transformation | Rules and Guidelines |
|---|
Aggregator | The Aggregator transformation is a multiple-group active transformation. You must use a Window transformation before an Aggregator transformation in a Streaming mapping. If you do not specify the group by ports to define groups for aggregations, the transformation sends a value of 0 when it receives no data. |
Joiner | The Joiner transformation is a multiple-group active transformation. The following rules apply to Joiner transformations: - - You must use a Window transformation between the streaming source and any Joiner transformation in a Streaming mapping.
- - The upstream Window transformations in pipelines to a Joiner transformation must have the same slide intervals.
|
Lookup | The following rules apply to Lookup transformations: - - A Lookup transformation does not require a Window transformation between a streaming source and itself.
- - You can include a Lookup transformation only if the mapping has flat file, Hive, HBase, or JDBC sources.
|
Rank | You must use a Window transformation before a Rank transformation in a Streaming mapping. |
Sorter | The Sorter transformation is a multiple-group active transformation. You must use a Window transformation between the streaming source and the Sorter transformation in a Streaming mapping. |
Union | The following rules apply to Union transformations: - - A Union transformation does not require a Window transformation between a streaming source and itself.
- - If one pipeline to a Union transformation has a Window transformation, all streaming pipelines must have a Window transformation. All upstream Window transformations in the pipelines to the Union transformations must have the same slide interval.
- - A Union transformation cannot be used to merge data from streaming and non-streaming pipelines.
|
Window | The following rules apply to Window transformations: - - You cannot add a Window transformation to an Logical Data Object mapping, REST mapping, or mapplet.
- - A Window transformation must have at least one upstream streaming source.
- - All Window transformations must have a slide interval that is a multiple of the mapping batch interval.
- - A Window transformation that is downstream from another Window transformation must have a slide interval that is a multiple of the slide interval of the upstream Window transformation.
- - The slide interval of a sliding Window transformation must be less than window size.
- - The format of the parameter of the window size must have the TimeDuration parameter type.
- - The window size and slide interval of a Window transformation must be greater than 0.
|