Controlling RStudio Python Child Processes

June 23, 2014
By

(This article was first published on Data Driven Security, and kindly contributed to R-bloggers)

I’ve been using RStudio’s new ability to run Python scripts since I often need to analyze/process data in R but then run web services with said data in Python (usually via Flask). I’d rather live with the foibles of the RStudio editor than use a separate one and run code on the command line.

Everyting below is for OS X, but I suspect holds true for Linux and can be adapted pretty readily for Windows via PowerShell

Running the “Right” Python

The first “gotcha” was running the version of Python I wanted to. By default, it will use the system Python and I rarely use that since I’d rather have Homebrew manage my config. My Homebrew Python binary is in /usr/local/bin and all that needs to be done in R to make it look there for the system() call it uses to run Python scripts is to add:

Sys.setenv(PATH = paste("/usr/local/bin", Sys.getenv("PATH"), sep=":"))

to your .Rprofile. There are other ways to do this, but I prefer this method. Feel free to share yours in the comments.

Killing Python

Running a long Python job from RStudio, or running something like a Flask app from within RStudio means a having blocked rsession until the child process exits. That’s fine if the process is going to exit, but tasks like a Flask app will just keep running and the RStudio “stop” icon won’t just kill the child process it will try to terminate the entire R session (#notcool). You can kill off the Python process by adding the following bash function to your shell startup script:

rpspy() { pgrep rsession | while read ppid ; do pgrep -lP $ppid ; done }

run it:

rpspy
24451 Python

then do a normal kill 24451 to stop the job. I was going to combine the search & destroy operations it into one function, but sometimes I just want to have the PID at hand vs kill it (and there’s no need for two shell functions when it’s so little typing to kill the program). You can add such a feature as a function option if you desire.

While you could just do a pkill python, I have other jobs running that I don’t want killed, so the rpspy function will find the Python binary with a parent process of RStudio’s rsession and just kill it.

The prospect of having RStudio as a completely unified data science developent environemnt is pretty exciting and I’m looking for further tool integration and features from the RStudio team.

To leave a comment for the author, please follow the link and comment on his blog: Data Driven Security.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.