jupyter-client has to be installed but “jupyter kernelspec --version” exited with code 127

Question

I have already installed Jupyter notebook in my ubuntu 16.04 machine. In jupyter notebook there is by default python installed. Now I want to use R from jupyter notebook.

Answer 1

Adding this for anyone, who googling and getting stuck on this issue may benefit from this simple learning. I got the same error, as above, followed the instructions above (while launching R in Terminal):

IRkernel::installspec() 
Error in IRkernel::installspec() : 
    jupyter-client has to be installed but “jupyter kernelspec --version” exited with code 127.

In Bash I ran:

Mac-Pro:~ $ jupyter kernelspec --version
-bash: jupyter: command not found

Duh, I had no Jupyter installed. Install Jupyter (on mac, using brew) with: brew install jupyter. Installation instructions for other OS's can be found here.

All now working fine.

Answer 2

I don't use conda - I have python virtual env and R installed. I don't know if it matters but I'm running on ubuntu for windows (WSL) (might be the cause of my error, might be unrelated). In the console, it recognized jupyter. Inside R console I installed IRkernel but IRkernel::installspec() gave me the error above (didn't recognize jupyter). I couldn't find a solution that worked, so I am writing here what solved it for me. I found the internals of installspec here. Check before hand where is your jupyter installed with which jupyter and run R from command line. Then, run the following code (adjusted from the link above):

 srcdir <- system.file('kernelspec', package = 'IRkernel')
 tmp_name <- tempfile()
 dir.create(tmp_name)
 file.copy(srcdir, tmp_name, recursive = TRUE)
 spec_path <- file.path(tmp_name, 'kernelspec', 'kernel.json')
 library(jsonlite)
 spec$argv[[1]] <- file.path(R.home('bin'), 'R')
 spec$display_name <- 'R'
 write(toJSON(spec, pretty = TRUE, auto_unbox = TRUE), file = spec_path)
 args <- c('kernelspec', 'install', '--replace', '--name', 'ir', file.path(tmp_name, 'kernelspec'))
 system2('/path/to/jupyter', args)  <--- here you copy paste the path you got earlier with pwd
 unlink(tmp_name, recursive = TRUE)

Answer 3

I am running Ubuntu 18.04.4 with R installed in /usr/bin/R and Python in /home/ubuntu/anaconda3/bin/python. I ran into a number of path related problems when trying to tell Jupyter where to find the kernel for my R installation. Regardless of how I ran R the IRkernel::installspec() would not work in R. I found success by making the connection using jupyter kernelspec from the command line. Specifically this is what worked for me.

From within R:

install.packages("devtools")
devtools::install_github("IRkernel/IRkernel")
system.file('kernelspec', package = 'IRkernel')

The last line should give you the location of Jupyter will need to find the kernel. Mine was /home/ubuntu/R/x86_64-pc-linux-gnu-library/4.0/IRkernel/kernelspec

From the command line:

1. inspect the path that you receive when you were in R. There should be a .json file in it.
2. jupyter kernelspec list (run this to be sure that jupyter is in your path, you should see information about the current available kernels.
3. jupyter kernelspec install /home/ubuntu/R/x86_64-pc-linux-gnu-library/4.0/IRkernel/kernelspec --name 'R' --user (you will use path that you received while working in R which could be different)
4. jupyter kernelspec list (this list should now include R)
5. restart jupyter

Answer 4

Assume that you have already run R in the terminal. You can first call library('IRkernel'), press Enter, then call installspec() or installspec(user=FALSE).

For me, simply calling IRkernel::installspec() will always raise the error you mentioned in your question. I don't know why what I did solved this problem on my centos server.

Answer 5

I had the same issue; I added the following in the ~/.bashrc (and source) and then tried running IRkernel::installspec() and it worked:

export PATH="~/anaconda3/bin:$PATH"

Answer 6

On some Windows systems you may not succeed even if you run R from cmd / powershell. That's the case on my machine. My workaround is to run R from Anaconda Prompt (if you installed Jupyter via Anaconda). You may need to specify the full path if R is not on your PATH.

I think this is some problem related to the PATH, however I had no luck adding Anaconda\Lib\site-packages\jupyter_client to my system PATH.