Thanks for the update @DanielBog!
As mentioned by notsure: The ability to pop out data tables is in my opinion essential. This is because a user spends most of their time on analyzing the output of their tables with the full focus needed, plus it enables a user to compare two data tables and hold the tables next to each other, or switch quickly between them via “Alt + Tabulator” short key
I can’t find a way to tag KNIME 5.
Missing commands in the nodes context menus
Hi, I’m currently working on a text classification problem. To explore the available data I created a text tokenizer which produces a tag cloud.
Using the new UI I’m missing some important commands in the nodes context menu to display the node output. When the output to be displayed is a table, there is no problem: the lower part of the screen shows the data for each output port; but when it comes to displaying images the user has not enough possibilities to customize the way the image is shown. So I had to revert back to the classic UI.
With the new UI I didn’t find a way to resize or even save the resulting tag cloud as image:
Am I doing something wrong or the feature is actually missing?
In general I really like the new sleek interface and can envision how it helps especially people new to KNIME to get up and running faster. As this is also the type of people I usually teach about KNIME, I can envision that onboarding will be much faster. Also some of the new nodes, will definitely especially care for e. g. “Excel people” → that Value Lookup (or maybe shortened VLookup ) Node definitely rang a bell
I am also fully aware that the current version I downloaded (5.0.0 from March 1st) is something like an Alpha version so lots of things are probably still in development. And sometimes it might be just me not finding the right way to do it, but that then is in itself a feedback about User Experience.
With that said, I tried to structure my feedback into three categories:
- What I like
… and I also tried to provide a reasoning for my comments which hopefully makes it more understandable for devs and readers.
Oh, and I liked the previous feedbacks from other KNIME users assuming that posts with a lot of hearts are considered as an upvote for those post / topics.
Let’s go …
- Double clicking a node in the repository does not “add” it to the current workflow (if a node was selected previously).
– this slows down the workflow creation quite a bit as for an experienced KNIME user this is probably one of the standard ways to add nodes
- Node labels can’t be formatted
– This makes workflows less understandable. For example I bold column names if I mention them in the label (such as Sum of Costs). Additionally I put all flow variables I create e. g. in bold and red
- Node repository search bar seems to be in “fuzzy mode” by default
– as an experienced KNIMEr I know a lot of the nodes I want to use. However, this “fuzzy mode” suggests me a lot of the nodes I don’t want to see. If I search for the “One Row to Many” Node I expect it to appear first, when I search for it exactly like that. This makes my workflow creation slower
- I can’t adjust the names of component output ports
– working more and more with components also for local workflows, descriptive output port names make my workflows much more usable, especially when sharing them with others (which I quite often do). An output port named “Revenue by Month” is much more understandable than “Output Port 1”. This also should carry over to the massively enhanced Node Monitor which I really like
- I couldn’t figure out how to add configuration / widget nodes within an already existing component
– I tried drag and drop but somehow the config nodes simply don’t appear. Only when I first create everything in the main workflow and then wrap it in a component can I use config nodes. This however is a massive step back in workflow and component creation for me.
- Keyboard shortcuts (the usual ones) don’t seem to work right now
– I use the keyboard quite a bit and one of my main way of moving nodes (I just need to align them, yes, I know, this is kind of OCD ) is using CTRL+SHIFT+Arrows. Slows down workflow creation for me and is less user friendly
- Leaving a component seems to close it (so I have to re-open it when going back in)
– in 4.7 and prior I had the option to have both parts (the main workflow and the sub workflow in the component) open at the same time. This is more convenient for me, because sometimes you just need to jump into the other part. Not having that option slows down workflow creation for me.
- Due to the fact that a node result is now ONLY visible in the enhanced Node Monitor, I also don’t have the result table spec anymore.
– This was a quick go to place to look for some info about my result data (e. g. upper, lower bound) without having to use e. g. the Data Explorer Node (which basically does this and something more). Definitely a slow down in how I work with data (adding an additional node is simply more complicated than having a standard way to view the table spec)
- New workflow coach, that appears, when I drag a connection to an empty space, recommends nodes with a table input also if I started dragging from a flow variable outport.
– this does not make sense to me, as I simply cannot connect any of these suggestions to e. g. a component or node that only has a flow variable output port. Confuses me.
- No console or some other “place” to see why a node threw an error
– this is maybe my BIGGEST issue so far. I must have a spot where I can read why something went wrong (e. g. wrong data type or column missing etc). That was a huge feature over other tools like Excel, PowerBI etc that simply say stupid stuff like “division by zero” or “reference error”. It does not necessarily have to be the console, could be any other window or pop or place in KNIME where I can see why a selected (!) node didn’t execute
- copy & paste of nodes acts strangely → the pasted objects seem to appear directly under the copied node which basically makes them invisible.
– I only discovered them when I wrapped everything in a component and that space become “unoccupied” (so to say). That was very confusing to me, especially because I even can’t move them with CTRL+SHIFT+Arrow (see above) even if I know of this behavior. So I would expect the past action to put the pasted node slightly off of the original node (like a little bit lower and a little bit to the right, which seems to be standard behavior in a lot of applications)
- Not having an Outline View
– especially when your workflows grow, this is a biggie (and I would dare to say that workflows grow pretty fast for 99% of the users as it is the nature of KNIME to have Nodes take 1 task and 1 task only). Yes, I can now zoom to the whole workflow with CTRL+2 but that is a slightly slower thing than for example simply pointing and clicking my mouse to a spot in the Outline view. This slows down the workflow creation for me.
- Node repository seems to have only 4 categories: IO, Manipulation, Views, Other)
– as an experienced user I sometimes had a rough feeling where I probably would look for the matching node and then simply went through a few trial & errors. This not only massively improves “learning the tool” but is also a good way to approach data handling in a more explorative way. Now I have to either know the nodes or go to the Hub which also is not the most convenient way. As others have said this can be eventually done through some individual (like “advanced mode”) settings
I have only one (so far): the same search functionality as in the Node Repository would be great in the new “popup workflow coach” - that would massively speed up workflow creation.
What I like
- drawing backward connections from input port to output port
– this makes workflow connection faster
- config and play buttons direclty above the node
– it feels like this is where they belong. Increase the speed of the workflow creation and makes the whole process feel much smoother
- The execute button in the node monitor
– this is a really nice feature. As I probably look more and more in the new node monitor, having a button there to execute definitely improves my user experience (and saves mouse travel time)
- Data Types in Node Monitor
– This is something I have wished for quite some time as it enables me to see immediately if I need a type conversion node or if my data is already in the right data type
- search function in the Node Monitor
– great, massive improvement over some other rather improvised ways of searching for data. Love it!
- snapping of connections
– I find this especially useful for the now “appearing” flow variables as previously it sometimes was a bit of a guessing game, if I “hit” the right corner spot. Good time improvement
Hope this helps
Nice @kowisoft ! When someone asks Phillip for detailed feedback at Summit, they get detailed feedback
Hi all, some notes on the first impression about v5.
What I definitely like the most is the new output window. Every time when I inspected the results of an executed node - and this is for sure the most wanted action on an executed node, to inspect the results - the popup window was in the way. Now without any effort, I will always get the output presented.
Some drawbacks I discovered:
- I have a node with two output port, one special (custom) project port, and one regular table port. But in the output panel there is only the message: “The data at the output port is not supported by any viewer.”
- I don’t really understand the grouping with the node repository. I have loaded our own extensions, but I can reach the nodes only by searching the specific node names. When I type the extension name, it will be shown as a filter option, but when I click on this filter, no nodes will be displayed.
- Like kowisoft, I also miss the console or some error message window.
I would think that these issues are not principle design problems but some detail problems to get old functionality into the new design.
Question: Is there an update of the API for node developers?
I’m excited about the final 5.0 to come!
The new interface is a step backward from a professional user’s point of view.
The ability to focus on the workflow and calculations in the nodes, rather than on a constantly updating data table as in Excel, is something that made KNIME a tool for serious computation, modeling, and data science.
When working with large workflows, the large data window over the whole display is annoying; on the contrary, the ability to view the data tables at different nodes is missing.
As a user of KNIME for more than ten years, I humbly recommend focusing on solving some chronic problems of the platform, rather than adding new half-baked features and breaking what works.
As an example, I would mention solving charting as a chronic problem. KNIME has about four or five different tools to plot graphs, but none that are of good quality and usable for creating professional and customizable graphs that can be exported and used in reports or publications. The new interactive graphs look cool, but they’re not useful for serious work. Instead of adding such a new tool, it would be better to fine-tune the usability of some of the existing ones. Or better yet, give up developing a new tool and create a user-friendly interface for one of the classic libraries, e.g. Matplotlib, since Python is already embedded in KNIME.
This would save a lot of time for the user because now quality graphs have to be created with external tools - I do it with Python, for example. Yes, it can be done, but it lacks the elegance that a tool like KNIME could bring to users so their energy can be spent on data analysis rather than on needed knowledge of code syntax when they need more than a basic plot.
Serious general request - please, for version 5 and beyond, keep the option to switch to the classic view - just as there is now an option to switch to the new UI.
As you have started the tabbed window in knime, you can implement tableau or power bi visuals in knime tab, keeping the tableau or power bi in the backend. All tabbed windows should support some of interactive reporting option and exporting of the reports for dash and plotly.
There should be other features to be incorposated like PMML etc more learning transfer even for deep learning transfer, so tha ML engineer can work with the transfer. So that we can contribute in the wide range ML engineering section directly.
Data I/O should be more automated, it now recquire more manipulation steps then other softwares. Sometime it recquires advanced programming within knime. Even excel can handle those I/O automatically, but Knime can not detect without appropriate coding.
A little bit off topic. One option could be to pre-configure Phyton based graphics so a user would have some flexibility but also a basic framework without too much code.
A list of the (too?) many graphic options and packages is at the end of the article
I just reinstalled 4.7.3. It came with an embedded V. 5. I previously installed V.5 on its own and created a workplace for it. When I open the new version, there’s no way I can see to open my V.5 workplace.
A post was split to a new topic: Version 45 Python Trouble on Mac
There is a dedicated forum thread only for the new views and I am more than happy to discuss missing functionality or bugs.
I tried to use this new UI but I had to switch back because I can’t do my work with this.
I like that when you click on a node, you see the table at the bottom of the screen – but this pane is less usable compared to earlier versions of Knime. In Knime <5 I could right-click and see the table, and do this for any number of nodes at the same time. How do you open multiple views like this to see multiple node results ? I need to be able to compare.
Also this view is in the bottom pane and so limited to that tiny window. Maybe I can see 10 rows but then have to scroll in all directions.
Maybe I have bad settings here ? What am I doing wrong here ?
I agree with you. This new UI is a huge step backwards. Maybe its done to be oversimplified for new users – so maybe that can be like a basic perspective or view. But the old UI was better, despite some of the shortcomings, because those shortcomings are all here in this new UI too.
This new UI to me doesn’t make proper use of space on the screen. Huge fonts, too much padding and spacing between things. For example, the repository is a huge pane now, but mostly with white space.
New interface looks nice, but without ability to “View” node: HiLite Row Splitter is useless because i can’t hilite.
can you elaborate a bit more? It should be possible to select points in the new view nodes and get the selected rows with the HiLite Row Splitter.
I would like to restate a concern that I raised before about the new interface for node selection. I have attached an image to illustrate my point. The old interface has a clear and organized list view with categories, which makes it easy to read and select the nodes. The new interface, on the other hand, has a cluttered and confusing layout, with no apparent sorting logic or spacing. The nodes are too far apart to each other and hard to distinguish. I find this very inconvenient and frustrating unless i am missing something?