I’m testing the nightly build, and there is a bug within the Python configuration. A small bug (or new feature) but it still work with it.
Indeed, if you have to python installed (conda and miniconda for example), you have to specify a path (See Figure below). In the available conda environment, you see all the environment from all the conda instances. You were able to choose.
In the nightly build, this option is not supported anymore. If the conda path is defined and you’re selecting an environment from the other instance, you’ll have the error below.
Thanks for reporting this issue!
Would you mind sharing a list of the names of all folders that reside inside your C:\ProgramData\Miniconda3\envs directory?
In our current nightly build, we assume that the folders inside the envs directory have the exact same name like their corresponding environments (so for example, py3_knime_4.2.0 in your case). Maybe this is not the case because Conda for example replaces dots by underscores or the like.
Actually, because of the 2 instances I have some environment in the folders C:\ProgramData\Miniconda3\envs and C:\Users\jerome.conda\envs.
That’s why the executor can’t find the correct path.
It’s not a big deal, but with the previous version, it work fine even with different paths.
I change the way I configured my conda environment but I can reproduce it if you want to ?
Ah, now I understand - sorry!
We are going to take a look at this in more detail. On the one hand, if this used to work in previous versions of KNIME, it should continue to work in the upcoming version, too. On the other hand, maybe we should consider the old behavior a bug since it was never intended to allow choosing environments from other Conda instances apart from the one configured in KNIME.
Anyway - again, thanks for reporting (and the explanation)! No need to reproduce the problem again.
Just a quick follow-up: we have fixed that behavior in the meantime. In the recently released version 4.3 of KNIME Analytics Platform, you will be able to select Conda environments located at arbitrary locations (including environments of other Conda instances) again.
This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.