Troubleshooting r server oddities in KNIME

I have an R snippet that I imported from someone else’s R code into KNIME (3.7.1 windows 64bit). When the author of this code wrote it they used to generate several figures in R studio from this one piece of code.

I have since gone through to carefully comment out parts of the code in the hopes of splitting it into several snippets (or R plot modules) in KNIME so that the plots can be generated individually. One problem I keep running in to though is that when KNIME calls up R Server (especially to draw a figure through R Snippet) the plot is drawn and then R Server is shown as un-responsive for some time. If I put this same code into an R plot module, no plot is drawn at all. I have other pieces of R code that do successfully draw plots in an R plot module.

My R version in windows is
3.5.2 (2018-12-20) – “Eggshell Igloo”

It seems you’re using the community extension developed at MPI-CBG (Community Nodes > R Scripting ).

Did you also try the newer KNIME integration (Scripting > R )? It works quite well in my experience.

Any chance that you can share (part of) this code? This would definitely make it easier for others to give advice.

Could you check these examples how to capture R graphic output:

Then also it might be an option to determine the local path and write the PNG file to disk without using the KNIME graphic port - I do not fully grasp how the various R graphic nodes capture their plots and graphics.

Thank you for the reply! I am using the node from Scripting rather than community (hence the one that takes a table in as “knime.in”, rather than the one that takes a table in as “kIn”). Should this one not be calling up R Server? It is certainly possible that I have KNIME configured incorrectly for R on this system.

As for sharing the code, I’ll have to reach out to the original author for clearance on this. I understand that it would of course make troubleshooting vastly more straightforward. This code does however also take in data from upstream sources to the point where running it in vacuo would be impossible; I’ll have to get permission to release an example input table then as well.