How to loop over one table?

[mpuckhaber]

Hi,

I am struggling with the following task.
I have read a table from a PDF that contains a data line by line like this.

Legend number1 number2 percent

lalala 200 0 10,00 %
blabla 40 60 24,50 %
dummy
dumdum 30 300 22,22 %
dada
da
something 70 5 99,00 %

My target is a table that contains all the legend info in the first column and then the remaining data in separate columns. Like

Legend | number1 | number2 | percent

lalala | 200 | 0 | 10,00 %
blabla | 40 | 60 | 24,50 %
dummy dumdum | 30 | 300 | 22,22 %
dada da something | 70 | 5 | 99,00 %

I manage to flag the lines with data with a regular expression and can also split the columns with a “Regex Split” node. With some SQL-experience, I am probably blind to an elegant way of giving the rows that belong together a unique key to then group the rows.

So my I idea was to loop over the table, join it with the next row of itself (skipping data-rows joined with the following row) and combine always one “legend only”-row with the following data-row. I found no way of feeding the processed table into the loop again until no “legend only”-row is left. (And I agree the approach is not elegant either.)

Of course, my goal is to find a way to give the rows that belong together a unique key. But there must also be a way to loop over a table.

Thanks in advance for hints for both ways of solving this problem!

[mlauber71]

@mpuckhaber you might want to take a look at this concept of identifying different blocks in one sheet.

Move head rows to a column

@willy_oracle typically you would mark all lines with a Counter. Then identify the ‘headers’ with a rule engine [IF Col1 LIKE “Head*” => Counter] and keep that (starting) row in a separate column. Then you fill the missings with the previous value so a block would emerge marking the data table: [image] So now you have the blocks marked an can use a Table Row to Variable Loop Start – KNIME Community Hub or Group Loop Start – KNIME Community Hub to do something with the blocks (export them delet…

[mpuckhaber]

Hi mlauber71,

thanks a lot for the solution. Missing value replacement!
By just changing to fill missing values with the next value (rather than last value) it works even without sorting back and forth. That is actually the elegant solution I was looking for! So thanks again!

Just out of curiosity: Is the other approach impossible? The idea was to loop over the same table, i.e. change the table in each loop step. In every step I would join the table with itself using the row number and the following row number, combine the content of the legend of the neighbouring rows into the legend of the data-row (except the first is a data-row) and delete the integrated row.