Option 4: Graphical tags. Each in their own column.
Please vote on your favourite, and tell us in the thread below… what resonates? Why do you like your preferred option? Is there any option we are missing? Anything you don’t like? Mock-ups of your own encouraged!
Graph Manager / Node Status Extension Direction
Extension Proposed Features:
Collects every node in a graph and displays various node states
Enables you to search and filter the node list based on these node states
Enables you to export the results to Excel
Will collect a “quick view” count of all node states in the lower right (Same horizontal line as the run bar) that when you click will cycle through nodes in that state.
Will also create a clickable link in the messaging next to the run button (As well as introduce an Errors message that takes hierarchical precedence). You will be able to open the extension from here, or via the Extensions menu.
Individual columns for State, Status, and Issues makes this way easier to read in my opinion. Text vs graphics doesn’t make much of an difference for me but some sort of colored identifier (text, icon, border, etc.) for at least warnings and errors seems important as those will likely be the higher priority node states.
This is great, but will there be any additional interaction with the node states directly from the extension? For example: toggle Active/Inactive, Freeze/Thaw, or hide warnings?
In my mind the Workspace Reference extension would be enough to cover package issues. I don’t think I’d need to see it in multiple places. First instance of a warning for nulls or empty lists would be nice, but if the nodes are listed in execution order that might be enough.
My thought was that by allowing a filter for just a given package you can see if you have one package causing all the nodes with a warning or info state (ie: this has Python 2 stuff) or many packages on a per graph basis. This would facilitate review and planning of package management for larger firms. It helps to know ‘what packages’, but equally if not more importantly is ‘what packages are problematic’.
Full output parsing isn’t a requirement per day, but the ‘problematic’ data types of null and empty lists are the root cause for a lot of issues. Doubly so when some nodes returning null intentionally but not surfacing a warning. Everything thereafter will fail and novice users wouldn’t know that the resulting null is the root cause.
I think the beauty in the way this is presented makes it feel like ‘another way to see my graph at a high level’, and merging with things like TuneUp may actually simplify the experience for users, and so I may be over reaching. But to me of the team is exploring a second way to list all nodes on the canvas and report back information about them… why not do double duty with one extension, simplifying the user experience, the development maintenance, and the end user deployment?
I get you now. I immediately went to missing packages but you’re talking about missing or otherwise unhandled references within a loaded package that results in broken nodes. That could be helpful.
I also agree that there needs to be some sort of way to identify “valid” node outputs that result in all nulls or empty lists and cause warnings later on. Don’t have any opinions on how that should be handled at this point though.
Yes, doable. But in that case, since there will be more columns per node, design needs to be updated and maybe we want to let the user pick which column to expose. Or on the other hand, if we want to expose all the info including runtime + node state, then we need a more consolidated UI.
Will the results be able to be retrieved by the API? I have a VE a little similar to this (without UI) that collects warnings/errors in the workspace and pushes to a DB so we can monitor the health of our deployed scripts.
@solamour, sorry if it was mentionned above, did not have time to read the whole thread, but while clicking on the icons, let’s say the warning one, will it give a window listing all warnings texts, nodes outputs and nodes name / origin ?