The PMML transformer currently operates in "appending" mode - the Transformer#transform(Dataset<Row>) method appends prediction columns to the input dataset.
The "appending mode" appears to be Apache Spark convention. However, in some situations (eg. demonstration purposes) it might be desirable to operate in "simple" (aka "standalone") mode - the transform method creates a new dataset that only contains prediction columns.
Could be easily achieved by introducing a TransformerBuilder#setTransformationMode(Mode.APPENDING|Mode.STANDALONE) configuration method.
The PMML transformer currently operates in "appending" mode - the
Transformer#transform(Dataset<Row>)
method appends prediction columns to the input dataset.The "appending mode" appears to be Apache Spark convention. However, in some situations (eg. demonstration purposes) it might be desirable to operate in "simple" (aka "standalone") mode - the transform method creates a new dataset that only contains prediction columns.
Could be easily achieved by introducing a
TransformerBuilder#setTransformationMode(Mode.APPENDING|Mode.STANDALONE)
configuration method.