Parallelization: allowing different sized tables at output

Hello all,

I'm currently trying to speed up a workflow and use the full capacity of the computer at hand by using parallelization. I'm running into the problem that eachtime the section is ran, the number of column varies at the end. KNIME is not happy with that. 

I get the error:

ERROR Parallel Chunk End   0:522      Execute failed: Cell count in row "c1_r0" is not equal to length of column names array: 4 vs. 21

I need the data afterwards, does anyone know a work around this? (similar to the Loop end node with the "allow changing table specifications" parameter)



Hi Matthieu,

if you are not able to guess the number of columns before the execution of the the individual iterations of the loop you can do the following:

Use a unpivoting node, to rotate the columns into rows. Therefore select all columns as value columns.

After the loop end you can generate the columns with the pivoting node. Groups are the RowIds, Privot the Columnames and use the aggregation method first on the columnvalues.

Cheers, Iris