The way the current "Hilite Collector" node works is a little confusing. Please can it be improved such that the node remains in execution mode until the user enters their desired content in the Hilite Collector window and clicks on a Finish button (so working in the same way as the "NG Visual Clusterer (beta)" node). The trouble with the current way the node works is that running the workflow just runs straight through the "Hilite Collector" node, also making any annotations when the "Hilite Collector" node has already executed results in the annotations not being applied which can be annoying if you spent several minutes doing this without realising the status of the node.
As a second improvement request, can we have the above applied to a new node which is a modified version of the "Interactive Table Viewer" node, where the node is the same as the "Interactive Table Viewer" but you can make modifications to the cells, add/remove columns and rows etc. Again, this working in the same was as the "NG VIsual Clusterer" node where the node remains in execution node until the user clicks FINISH button in the window.
I second both the requests. Would be great to see them in future releases. They should be quite useful, especially the "HiLite Collector". Several of my collaborators spent hours going through their data NOT realizing the state of this node.
Although I can certainly see the possible advantages of a “Table Editor”-node, I think it is not available in Knime (or other workflow tools) for good reasons.
Currently, a workflow will translate a given input in a clearly defined way into an output. If none of the nodes introduces some kind of randomness, a workflow will always produce the same result if it is applied to the same input. Therefore, the workflow itself is already a good documentation of what has been done to a dataset. If you allow a user to modify the data within a workflow, you introduce nondeterminism and loose that property.
Having said this, I think the main problem is that it is currently difficult to edit tables in their native format. If you have to export and re-import a table using non-native formats (such as CSV, XLS), you run the risk of introducing conversion errors. One possible solution could be to add a “Table Manipulation”-perspective/view to the Knime Workbench that allows one to load, manipulate and save Knime tables. It might even be possible to launch such an editor in a simple way from the Table Reader/Writer-nodes. Unfortunately, this would certainly require a lot of work.