I like the new interface but I think it would be better if it allowed:
1-Copy & paste nodes (or groups of nodes) as this can save a lot of time in some situations
2-adding labels or descriptions for nodes
3-dropping nodes onto the connectors to place the node between two existing nodes
4-an easy way to copy all values from the output of a given node (i used to just open the table, hit crtl+A and then crtl+C, but this no longer works and selecting the whole table manually doesn’t work as well for pasting into excel or google sheets
I love the first steps of this major evolution although it is not yet completely functional.
At first sight, here are some proposals for a future release:
Allow renaming columns from the preview panel by double clicking on the column name (a small visual symbol could indicate the column has been renamed - this would have the same effect as the column rename node
Allow sorting columns from the preview. Either one by one using small handles in col headers or by multiselecting columns. The multiselect and options to move to start/to end are missing also in the recent table manipulation node and would be handy in this context. Also missing is the auto-fill of the original name when renaming, as usually, the name is similar to original.
Allow column filtering from the new preview pane
Allow row sorting from the new preview pane
Allow row filtering from the new preview pane
I understand this could be a challange to keep the UI sleek with all this but it may be possible with some activation switches (to activate modifiers when needed) or by making it accessible when making the preview frame larger (higher).
It could also be handful to be able to:
activate some basic data profiling on demand (distribution, % of missing, nb of distinct…)
being able to isolate a particular table preview in a separate and persistant modal window so that it can be viewed in parallel with editing on a separate screen for instance
Another cool feature could be to have inline viewers.
Those would allow to view some simple aggregations results directly in the workflow editor without having to open or click on something. Could look like some KPI cards with just and (editable) name and a value. So many time I add some group by nodes that I need to execute and open to control the number of distinct values or some totals or average value…
We could even make it (optionally) auto-executable when input changes. Some auto-formating rules and automatic historisation of the n last values would also allow to highlight exceptions and trends.
I agree with you on selecting node. In the meantime, you can use the great actions added by the Vernalis Extension : which are doing exactly what you are looking for (except the trick with the mouse that i’ve seen in another software …)
Hi,
great preview. I am missing “Available Renders” option for double values. Right now, all values are shown in full precision, which is great for scientific purposed, but of limited use in the business context.
It looks amazing. But you need to add the possibility to open other workflows. I guess that is just because it is a beta version. Also, please that one can modify the connecting lines. I am obsessed to put the lines that look as organized as I want.
Hi everyone,
I was going through the post of other users and agree with most of the points mentioned. Understand that the UI preview is a start and personally, enjoy the look and feel of the interface.
I am not sure if the point was covered in the discussed commentaries specific to open views. I was not clear on the point around execute and open view discussed previously.
Currently, I prefer the option to run the workflow and have the option to view say running decision tree or scatter plots at my convenience. I did not find the open view option in the new UI and am unsure if the same will be have similar functionality going forward. At times, there is quite a bit of lag between the node execution and view being opened which can be difficult to work with, in case the execute and open view option becomes a standard feature as i seemed to understand from the prior posts.
Thanks to KNIME team and enjoying the KNIME experience.
Thanking knime for the new interface, I give the following points as feedback
That the interface was multi-language, with the option to choose the preferred language.
I think that eliminating the functionality called “outline” is not good, since for massive flows it becomes difficult to search or identify traceability
The new interface needs a more intuitive approach to placing nodes from the repository by drag and drop and by keyboard entry.
Inserting a node into an existing connection by drag and drop is not working on my end. This is crucial and there should be clear visual feedback on the connection line when a node will be inserted into a flow during a drag and drop.
Inserting a node by keyboard or double-click (when a node is selected that has no outgoing connection in place) connects the two nodes on insert as expected, however when a node is selected in the workflow that has an existing outgoing connection (or when selecting the connection line itself) placing a new node from the repository via keyboard or double-click results in the node being dropped separately onto the canvas. It would be much more useful if the selection of a connected node or connection line inserted the node directly into the connection behind the selected node or on the selected connection line.
I also don’t see how to access files and components in this new UI. I have said before that I think it would make more sense to allow users to show components in the “Repository”. This would enable some much needed tools for components such as fuzzy search, keyboard placement and give users a single point of access to all of our solving tools.
I use external controllers to insert new nodes from the repository, so this simple UI is particularly appealing to me as I can operate with just a clean view of the Workflow / Node Monitor windows without the need to access the other support windows.
Control + = is not working for zoom in on my end either, but Control + - is working for zoom out…
Ability to choose/switch between “personas” where the UI and examples will be tailored to the persona (obviously with the ability to add more).
–Three basic categories are probably sufficient, with some sort of “more/custom” button:
— Level 1: Reporting/Data Blending/Basic Analytics. Simple I/O nodes, basic transforms
— Level 2: Advanced Analytics/Data Mining/Data Apps: More advanced nodes, workflow abstraction, workflow management, and some info on variable flow.
— Level 3: Kninjas/Data Scientists/Engineers: More advanced things like looping, error handling.
— More: Server admins, Biosciences, academics/statisticians.
– Perhaps the ability to click a persona, then click some check-boxes on specific areas of interest that will help users down a learning path.
UI-Specific:
– Add tags to nodes (for example, #vlookup to Joiner, #sumif to GroupBy, #union to Concatenate, #concatenate to Column Aggregator)
— Loosen the fuzzy match on search. I have typed “Group By” 8000 times only have to go back and remove the space.
– Ability to “bundle” outgoing connections or change the curve so that it simply “rounds” what were the straight lines.
– Ability to organize connections so that they don’t easily overlap
– Annotations with a more modern feel (rounded edges, HTML, adding arrows, themes?)
– Dark Theme… my eyes…
– Toggle for Grid
— Toggle for Snap to Grid.
— Button for Snap to Grid when pressed lines up all the nodes to the nearest grid location (NOT Auto-Arrange… I kinda hate auto arrange… sorry).
– Metadata available when you click on a node (basically run the Statistics node when someone clicks on the tab, but not automatically so that resources are not used unnecessarily.
– Make Metanodes gray again.
– Ability to pick an Icon when creating Components (I know you can somehow, but it’s not apparent how).
Other general suggestions/oddities:
– Ability to grab .py files from a GIT repository and pipe into the Python node.
– Combine Widgets and Configurations back into the same category again. I always use the wrong one.
– Remove the NEED to put an empty table into a Catch node. Make it optional to add a pre-defined table with ‘…’
– Add search boxes to all drop-downs in configurations, especially in Joiner, plus anything dealing with DB metadata.
– Node ports in the search dont match the nodes in the workflow.
AAAAANNNNNDDDDDDDDD Make the font of all dialouge boxes match the new UI font.
I know that is a lot, but I want to see KNIME succeed in gaining market share. Great to see you guys again. Looking forward to chatting with you soon!
Double-click a node handle/port to grab it so you dont have to click-and-hold. This would be super useful for things like DB connectors that attach to multiple nodes that are far apart.
Make the Forum link at the top PERMANENT, but make it more general once the new interface goes live.
When searching nodes, group search results by category rather than string similarity.
Another “cool” thing would be to be able to see nodes that provide or accept ports by clicking those ports on a node in the workflow (with necessary requirements, like if a node needs a URI, only nodes that generate URIs would be listed in “Inputs”).
With the latest release we have added the following features to the Modern UI Preview
Support for dynamic ports at native nodes – add ports directly at natives nodes, the same way like for components and metanodes, by clicking the “+” button.
Add ports and connect the nodes automatically at metanodes or components or dynamic ports at native nodes by dropping ports on the “+” button.
Quick nodes adding by dragging a port out of a node. After that receive a panel with the 12 most common used nodes based on the workflow coach data. Add and connect the desired node by simply clicking on it.
Table view based on the new “Table View (Labs)” node, with the possibility to use basic filtering, basic search or resize the columns.
Application tab bar, to work with multiple opened workflows in multiple application tabs – e.g. copy and paste nodes from one to another workflow.
So far, I really enjoy using the new user interface. It looks very good!
A couple of things:
When I click on an output node (the Geospatial View node, specifically), the results pane shows the Flow Variables tab. When I then click on any other node, such as a manipulator node, the results pane remains on the Flow Variables tab. Would it be reasonable to default to the data tab for these nodes?
When I drag a node onto the canvas, can we drag the node on top of an existing connector and have it added automatically? This was possible with the older interface.
First, some great usability enhancements within the Modern UI, especially in this latest round of updates in 4.7 – exciting things are happening here so keep up the good work!
That said, my company has one major concern…
Internally, we created a visualizer (i.e., a “blue” KNIME node) built upon the Java 2D API that also leverages the “highlighting” feature within KNIME – we use this frequently to display models and trace back visual anomalies to our input tables.
In trying out the Modern UI, we noticed that our custom visualizer did not have a “magnifying glass” option to view the plot whatsoever – this seems to be the case for all non-JavaScript visualizers (e.g., the “Local (Swing)” category of nodes). It also appears that the “highlighting” feature is not accessible within the tables (i.e., you cannot see what you highlighted in the table views). We noted that you can still use the “HiLite Row Splitter” node to figure out what was highlighted so the information is still stored, but there seems to be no way to make and track selections within the tables anymore.
This raises a few questions:
Will non-JavaScript visualizers ever be viewable within the Modern UI?(If not, this would be a significantly blow to our ability to use KNIME going forward.)
Will highlighting ever be accessible within the tables views of the Modern UI?(This was one of the biggest draws of KNIME as it allowed you to trace the evolution of your analysis on specific points of data.)
At what point will the Modern UI become the “only” option?(I noticed a team member stated, “someday in the future the current user interface will not be maintained anymore” so it would be nice to have general timeline for this transition.)
Aside from this, I did note one other technical issue:
If the user has a 4K monitor and is also using the Linux version of KNIME, then the Modern UI does not “extend” across the entire screen. That is to say, only a quarter of the window displays the UI while the rest of the screen is “greyed out”. Looking back in forum posts, it seems a similar issue of 4K + Linux has come up before yet with no resolution (see Node-icons and dialogs with HiDPI (linux) - KNIME Analytics Platform - KNIME Community Forum).
