chaning a columnrearrange node from replace to replaceAppend


(happy New Year!!!, 2.3 has some VERY nice improvements!!! Great work!!!!)

I have a question...

I am about to rewrite one of my nodes that previously used the columnrearranger to replace a cell from one column using SingleCellFactory... Now I want to also append a column without calling two SingleCellFactories as this would imply duplicating the work.

Now my question: I guess it is not possible under these circumstance to still use the columnRearranger concept but I would rather replace the whole row using the BufferedDataTable. Is this true?

The benefit of using the columnRearranger is 1. that only one column is being affected and the other columns are not passed through to the next node thereby saving disk space. 2. the concept of using the SingleCellFactory enables one to parallelize things at some point. Is this correct as well?

Thanks for the clarification...


If you extend from AbstractCellFactory, you can produce multiple columns in one go. You can also use a ColumnRearranger and apply two SingleCellFactories (just call the replace or append method on the ColumnRearranger twice). I favor the AbstractCellFactory method and then shift the appended columns using the move method.

In any case, do not create a new BufferedDataTable using a container if you can avoid it (and here you can) as this would duplicate the data.

As for the benefits: Both points are correct. You save disk space and we can add the ability to parallelize row computation at some point (already working on a prototype).