We encountered at many occasion with any python nodes the following events :
impossible to use Enter to add a new line or to be able to use arrows (left, right, down, up). The only solution we found was to copy our code - remove the node - create a new one and paste the code.
We work on Knime 3.5.3, windows 10 - 64 bit. The behaviour was tested on diffrent machines.
Iris wrote me it was written as solved…Well it’s not.
Sorry but this one is really driving me crazy ;o)
Sorry for the trouble. Can you give us a hint how we can reproduce the problem? Is there a typical set of actions leading to the issue or something?
Hi Christian, in fact no, I couldn’t identify what happens and when. Only while the node is open and you are coding suddently the “enter” doesn’t work anymore. The solutions I have found are :
1 close Knime and reopen.
2 or paste a new python node and copy the code of the old one inside.
Tell me if I can copy and send you some mysterious backgroud file considering the dysfunctionnal node in order to help you.
I have no solution only to say that now and then I also have this issue.
Thanks for reporting! We’re now trying to reproduce the problem locally and let you know as soon as we come up with a solution!
My colleagues and i have the same problem. Whenever we work with Python Scripts, it almost always happens after a while. It is sufficient to close and reopen the Workflow, you don’t need to close and reboot KNIME.
Hi @Fabien_Couprie, @beginner, @MH,
finally, we’ve got it somewhat reproduced! It looks like this has something to do with the jedi-auto-completion. I’ve got this Enter-and-Arrow-Freezing you described a few times when I entered a dot in a comment (like ‘#.’) and hit strg+space. Then the auto-completion kicks in and sometimes something hangs there. Can you confirm this?
Thanks! Now we can work on a fix.
I for myself can not reproduce the comment dot problem. But i think you might be on the right track. I encountered it once or twice that the auto completion did not work after a few minutes of scripting. It probably happens more often and might indeed be the reason for the other mischief.
@christian.dietz pointed out, that it might be a workaround to simply deinstall jedi (‘conda remove -n py35_knime jedi’, if py35_knime is your conda environment). I quickly checked and the node seemed to still be working, you’d have to live with a few ‘jedi not found’ errors, however.
I have not tried the Workaround yet. But the Bug is still present in KNIME 3.6 with Python 3.6.5.