in MUI the workflow overview / outline windows is missing. Even if it would get added it only helps improving workflow navigation partially.
I’d like to suggest to add a functionality which allows to create a bookmark / jump marker to immediately move to that position in thee workflow.
Thinking about Microsoft Word hierarchical representation of the heading structure, or Wordpress or an organigram software, the idea of a bookmark could be nicely paired with some sort of structural representation of the workflow consisting of components or metanodes.
We are aware that navigating large workflows can be problematic in Modern UI. This topic is already on our radar for the first half of 2025. As always, do not understand this as a promise
about the organigram. While prepping an image, a SVG export from a workflow, it occurred to me that the layer pane in Adobe Illustrator are a much better explanation compared to the Hx-structure of Wordpress:
Imagine if you’ve got a tree view starting where each node, component, shared component, meta node etc. with it’s respective icon listed in a sorted may based on it’s position in the grid / pane.
You could also display i.e. a waring or error icon which would open the Monitor pane. You could right click to configure, you could jump to the innermost node without having to find your way through the workspace. You could apply colors in the pane for better navigation.
That approach would open so much possibilities about quick-nav, workflow orientation and more.
I really like the idea of bookmarking nodes together with a list of bookmarks, which would practically enable jumping between views on the data.
However, I’m not sure how any outline, summary or meta view could improve on the actual view on the nodes, considering the wide variety of shapes that workflows can take.
I’d expect most ETL workflows to follow a reversed tree structure, in which some parallel pipelines flow from left to right into a single stump. Compare this with data mining flows, in particular EDA, which would expand from a single (or only a very few) left sources(s) into many directions with some flowing back into each other - so a directed graph rather than a tree structure. Things become even more convoluted when ETL and EDA are done in the same workflow.
My solution to the problem involves using Metanodes (the closest approximation of a layer) as organizing method and forcing my workflows to fit the view screen as much as possible. The disadvantage of this method is that at the top level you tend to “forget” what’s going on within the sub-nodes (think Inception). It also makes rearranging the node order or consolidating nodes more cumbersome after the fact.
Thinking about Metanodes, I could imagine hovering with the mouse cursor over a metanode for eg 1-2 seconds with a floating outline view popping up to show the underlying workflow - all as a switch on/off option of course.
Additionally, it could be useful to run a network analysis kind of summary over a workflow (by node, by layer/metanode, by edge, by network), with the possibility to highlight any given node upon simple click within that summary. In its most purist interpretation, imagine a Workflow to Network node following Workflow Reader or any workflow capturing node - with the option of weighting the edges by the amount of data flowing through them.
Finally, the quest for a better workflow overview is the expression of the wish to keep things organised and clean as well as to share your workflow with someone else. This requires workflow tools which have mostly disappeared in modern UI: formatting node annotations, customising workflow annotations more flexibly, aligning nodes vertically/horizontally/automatically, etc.
Imagine if you’ve got a tree view starting where each node, component, shared component, meta node etc. with it’s respective icon listed in a sorted may based on it’s position in the grid / pane.
Yes, I could see that. Another possible ordering might be breadth-first based on the workflow graph.
Though, as @Geo has pointed out, this will not suffice alone as an overview. But I do agree it might be helpful. For instance the search for nodes in a workflow could be the filtering of such a sorted list.
Indeed, we also already observed that it could be practical to have some level in between full encapsulation in metanodes and none. Quite often you see parts of the workflow being grouped by putting an annotation behind them. An early concept of ours was also e.g. to be able to expand/collapse metanodes in-place.
Yep, this is one of the few remaining (I hope) areas in which we don’t yet have parity with the classic UI.
Automatic node alignment is high on our roadmap right now.
Richer formatting for workflow annotations has been requested (NXT-2855) but unfortunately I cannot give an ETA on that right now.
Formatting in node annotations is something I am not entirely convinced we even want. Will need to discuss about this.
This is intruiging to me personally as someone interested in network analysis Do you have any specific problems/use-cases in mind?
Thank you for taking the time to read and to reflect on the suggestions made in this thread!
The main application would be summarizing one or several workflows by their number of nodes and links, most used types or categories of nodes (eg number of manipulation nodes, number of mining nodes, critical source analysis, etc), by describing their structure, by finding loops (not loop nodes) in the data treatment, etc.
Such summaries are important for understanding or assessing workflow complexity, when explaining a project to someone else or for specifying project requirements, and so forth.
Given the integrated deployment and the network analysis nodes, the thought that something like this could be done within KNIME does not seem too far fetched
