Error in reticulate::use_python(): failed to initialize requested version of Python

[MrClearLovez]

Dear Developer:
When I run use_python function in linux,I got this error:

this is my environment:

I conda environment is this:

and my IDE is jupyter notebook.Sincere request you to help me to solve this problem!

[MrClearLovez]

and I want use_python("/home/data/t090313/miniconda3/envs/python36/bin/python")

I got this error:

[t-kalinowski]

It looks for one reason or another, Jupyter is not giving you the opportunity to see the cause for why initialization failed.

Can you try outside of Jupyter, to see the full error message, or to see if Jupyter is somehow the cause of the issue?

Also, Python 3.6 is quite old and EOL. I'd strongly recommend using a version of Python that is currently supported (i.e., v3.9+). https://devguide.python.org/versions/

[marcuslehr]

I'm also having this issue. Using R studio.

I've had this issue on other PCs as well with different versions of R/Rstudio/Windows. It is persistent.

I updated rstudio while troubleshooting and it now prints a non-descript error message like OP shows. Ie just says 'failed to initialize requested version of python'. Before upgrading Rstudio it was much more verbose. I have part of it I was using to search: "ERROR: The requested version of Python cannot be used, as another version of Python has already been initialized.". Paths were redacted to generalize.

Anyway, I found the following solution. And yes, both calls are needed.
Sys.setenv(RETICULATE_PYTHON = "path/to/python.exe")
reticulate::use_python('path/to/python.exe')
library(reticulate)

I have no .Renviron or .Rprofile- I checked. The Rstudio GUI option to activate a python was unselected and I unchecked the box just in case. I restarted the r session (advised by the original error message) and r studio several times. I think I tested hardcoding the reticulate_python variable into the .Rprofile at some point and that may have worked. Or setting the option in the Rstudio GUI. However, I frequently jump between PCs so didn't really want to resort to that solution.

But why is this process so touchy? Why not wait to init the python version until a function is actually called vs on package loading? Or just close the existing python session and start a new one. Kind of defeats the purpose of having use_python(). And directly contradicts the syntax shown in the documentation (loading the package then calling use_python).

Really would like to see the devs do something about this. Based on my searches this is a long standing issue and the behavior seems very irrational. At the very least provide an error message instructing to update .Renviron or offer to do it automatically. Kind of ridiculous tbh.

[marcuslehr]

Update:

Loading tidyverse prior to the above breaks that solution. Idk what the interaction is. Also, calling use_python() without the Sys.setenv() call sometimes works. Seems to be a coin flip. Anyways, holy cow this is absurd.

Update update:
Solution posted above is not consistent. Changed nothing other than restarting the r session. Sitting here restarting a kajillion times trying different things. Behavior is sporadic. Sometimes it works some times it doesn't. Restarting Rstudio seems to help for some reason. I may need to resort to using .Renviron

Please don't download a random vanilla version of python. Find a system version or print an error. Or maybe a warning on package load. If the user calls use_python() INITALIZE THE REQUESTED VERSION. Good lord.

[t-kalinowski]

@marcuslehr Sorry you're experiencing issues.

Can you please open a new issue with a description of what reticulate behavior you're observing and expecting?