Getting columns back

Hey all,

The forums are great, they've answered many questions for me that others have also posed. Now I've posted my first questions two days in a row ;)

Ok so the question is regarding some sentiment analysis. I'm defining columns to Author, Category etc through the Strings to Document node, then doing sentiment tagging, bag of words creator etc. Everything is going great. I then get to a point where I use the Document Data Extractor to bring out some of those earlier columns.

But many of my other columns I also need to bring back deep into the workflow but I can't becuase they dropped off back at the dictionary taggers and bag of words node.

Hope this makes sense.

How do I bring back in other columns that are important. I've tried using the Meta Info inserter and exractor too.

Just not sure what to join back on, or what nodes would reintroduce all of the columns.

Thoughts?

Cheers, Macca

Hi Macca,

nothing prevents you from "branching out" a copy of a table at any point in a workflow and use it much farther down the road to bring back some of columns that have been dropped in the mean time. This is common practice when working with KNIME.

Now on the most appropriate way to re-join some columns from the original table to the new table, the answer is, as usual, it depends.

If the new table still has the number of rows, row keys and row order of the original table, you can simply use a Column Appender node. Make sure you read its description to familiarize yourself with some of the pre-requisites.The Column Appender node can also deal with altered row keys and will basically stick the original columns side by side with the new ones.

If the new table has a different number of rows, row keys and/or row order, the Joiner node may be more appropriate, provided you have a "key" column or UID column in both tables telling you how the rows are joined together. If you don't have a key or UID column in your original data set, either you rely on KNIME's own row ID's (mind that they can altered by some nodes) or you add your own key/UID using for example the Counter Generation node. You will have to bring along the key/UID column with the new data table up to the point where you re-join it with the original data table.

I hope this makes sense to you and is useful to solve your problem. Otherwise, please post again here.

Cheers,
Marco.