No interruptions. That was our team’s goal for RStudio Package Manager 1.1.0 - we set out to make R package installation fast enough that it wouldn’t interrupt your work. More and more data scientists use Linux environments, whether to access extra horsepower during development or to run production code in containers. Unfortunately, the rise in Linux environments has seen a corresponding increase in package installation pain. For Windows and Mac OS, CRAN provides pre-compiled binary packages that install almost instantly. But the same binaries are not available on Linux. As a result, data scientists can lose their train of thought, or put off trying out a new method, all because they have to wait for new packages to compile and install. New users often face a tedious hour-long setup process before they can try out environments. IT/DevOps engineers are forced to wait anytime they want to build a new image, deploy to production, or restore an environment.
I attempted to try the example in the blog but with 'bionic' instead, since I'm running Ubuntu 18.04. However, it installed from source instead of using the binary when following the instructions.
Is the example meant to simply demo the capability with just 'xenial'? I would be interested in pursuing the package manager to deliver binaries down the line with 'bionic' support, and understand if the demo an example for just one distro to prove out how fast it works. If so, I can just demo the 'bionic' binaries at a later time.
@dustinkucko sorry for the trouble! The demo server has bionic binaries available as well. The most likely culprit for why you're seeing installation from source is if system isn't specifying a user agent that tells Package Manager what version of R to use. This is a problem we've fixed in the newest versions of RStudio, and we are working with the R core team for a fix in R 3.7 as well. In the meantime, these steps should help you test things out:
Setting the HTTPUserAgent option did the trick. Thank you for help Sean! Also, using the Package Manager binary saved 5min of compiling on my machine (at least for dplyr). Great tool!
This is awesome! I currently use cran2deb4ubuntu for fast binary installations on Ubuntu, but it only includes packages listed in the CRAN Task Views. What's the long-term plan for https://demo.rstudiopm.com? Could I switch to using it for my Ubuntu R setup, or could it disappear at any time?
