KNIME Analytics Platform 4.5 released

Hi everyone,

New releases of KNIME Analytics Platform 4.5 and KNIME Server 4.14 are now online and available for download.

This release features major updates for Python coders, more flexibility for workflow deployment, improvements to data apps, and more.

See the What’s New page for a highlight summary and the changelog for the full list of new features.

We’d really like to hear what you think about the new functionality and encourage you to post your thoughts and questions here.

The KNIME Team.

9 Likes

I will start! Because I am super excited about this new feature: Dynamic Loop Ports!!! I would update only to get this. This will make a few of my workflows so much lighter :slight_smile:

11 Likes

This is great! While I do some work in Python, a lot of my work is in R – Are there any plans to enable similar speed ups with R integration? The current way of RServe can be quite slow.

5 Likes

@marc-bux thank you for the release. I just tried the new Python integration and it does work, sort of.

I have one problem. I do not seem to be able to open the variable window (for the flow variables) on the left side of the code editor on my Mac.

image

3 Likes

Unfortunately under Windows I have the same problem. The sidebar in the Python Script node is not there and I cannot see the flow variables, although they so work if you execute the script.

The processing time is much faster :slight_smile:

3 Likes

Hi @mlauber71,

This is (partly) intentional. We wanted to use the opportunity a new node in Labs status brings to experiment with the dialog a little. In particular, we wanted to make things more “code-centric”.
Besides removing the node’s Options tab (you can apply all settings that were previously located there directly in the script now), we also removed all variables-related panels except for the workspace on the right side of the editor. The general idea was to get rid of those panels displaying information that users can also access via their script. For the flow variables, that could for example be done by adding a line print(knio.flow_variables) to the script, selecting it, and pressing Execute selected lines.
The separator bar that is still visible in the dialog is a leftover of these changes and should have been removed, sorry for the confusion!

That being said, we are not 100% sure if we like that particular change ourselves. So any feedback is highly appreciated! :slight_smile: We are already working on a nicer node dialog that should improve the development experience in the editor and make it feel more like working in a “real” editor. So this is the perfect moment for suggestions, etc.

Marcel

7 Likes

@MarcelW OK I understand. From my perspective it would make sense to have the presentation of the flow variables consitant with the other tools, because otherwise it would be more complicated to see the variables being present.

If you are looking for improvements a better code editor would go a long way to make coding easier (cf. https://pandastutor.com/). Maybe integrate auto-completion (I saw some mentioning of auto completion not working when I clicked).

The speed is better than before. What I see is that the setup of Python environment still can be a challenge; and I think not alle people are aware of the legal (commercial) implications of using Anaconda (or Miniconda for that matter). If KNIME could offer more guidance there I am sure this would help.

Maybe in a future web-based version of KNIME one could integrate a notebook concept like Jupyter.

8 Likes

I am kind of confused. I upgraded and I still see the window with flow variables? Why is that?
Speed seems to be better at first glance (only tested one file so far)
I totally agree with @mlauber71 regarding the visibility. And also the editor. Right now I only paste my code into the KNIME editor but if a future version would be a jupyter experience that would be awesome.
br

2 Likes

You need to switch to the Python Scripting (Labs) node to see performance speed-ups (especially in conjunction with the Columnar Table Backend). See KNIME Python Integration Guide

7 Likes

I toyed around furtehr with the new Python integration and I think I covered several examples of table, object/pickle and image in- and -out-ports:

One thing I noticed is that the conda.environment variable would not run thru across several nodes. One would have to run a ‘pure’ flow variable in order to make it work. Maybe that could be adressed in the future (or would I have to allow a flow variable to be passed thru like in components?).

5 Likes

@Snowy have you tried to use Parquet files (or SQLite) to transfer data between KNIME and R? That way you could keep the column types:

3 Likes

Yes, this is a bug. Thanks for reporting! (And great to see you trying out the node so thoroughly, much appreciated! :slight_smile:)

5 Likes

The v4.5 feels faster indeed. While the dynamic ports on the Loops is a very great addition and super useful, that means that some of the Loop nodes have been updated, and it seems including also the Loop End node, meaning that these nodes in existing workflows are now marked as Deprecated and need to be manually changed/updated in these workflows.

2 Likes

The 4.5 upgrade broke the SharePoint connector. When KNIME writes files “successfully” to SharePoint, they are not visible in SharePoint/OneDrive. If I try to write the file again it says the file already exists. When I use a delete file node to remove the file, it says that the file exists, but cannot be deleted.

1 Like

@MarcelW at least under MacOS the Python Object Reader seems to have the same problem and also there seems to be a problem with the dynamic ports (the three little dots) - which are not working.

Also with the Flow Variable problem: The “conda.environment” has to be on the top port when using a “Merge Variable” node, otherwise it would also fail.

I have not tested the Python Object Writer but it might also be affected. So it might be best to check the whole concept “conda.environment” propagation on all platforms and with all Python nodes.

3 Likes

Hey @usabledata,

thanks for providing feedback, unfortunately I could not reproduce the issue. Is this still relevant as I saw that you deleted your thread (4.5 SharePoint Connector). If yes could you please use the thread and provide some more information about the issue? :slight_smile:

Best regards
Lars

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.