Best practices: column meta information and descriptions?

Hi all,

I’m a long-time Alteryx user switching over to KNIME. Very impressed by the platform and the community.

I wanted to know what the best practices are for column documentation and what everyone is using. There does not seem to be a way of adding metadata column descriptions in KNIME, i.e. not the name, but a textual description of what is inside the column that persists throughout the workflow.

It would be great to have the ability to document in detail the contents of the column for easy reference if the workflow has to be edited a year or so later. Currently, the only way I have found is either a.) just putting text as an annotation somewhere on the workflow or b.) branching and transposing the dataset to a new table with an added descriptions column.

Are there any best practices for this or better solutions that everyone is using.
ChatGPT suggest using a “Column Properties” node, but I could not find it/it does not seem to exist.

Thanks in advance.


1 Like

Hi @Ink ,

The recording method you use is also the one I am currently using. However, if there are too many columns, I think Excel or specific format text can be used for editing and recording. Finally, the nodes in Knime can be used to read and manage them.



What exactly would you like to record?

@rfeigel Textual descriptions of the data in each column. This is useful to me and others when trying to determine what exactly the output data is later down the line without having to go through the entire workflow. When someone needs to use the said data, they can check the description to see how the data has been aggregated and what modifications have been applied to it, e.g. have outliers been removed via IQR, if so what method has been used, etc. (Alteryx supports a meta field called Description specifically for this purpose and retains it when data is output to its proprietary format. Obviously, this is lost when exporting to CSV or similar.)

I can create a separate table with just the descriptions and save it along with the data, but I was wondering how everyone else does this and see if there are more elegant solutions.

@tomljh Thanks for your tips.