CBC Reviews: Revolution R (in which this doesn’t go well)

Update!: The latest version of Revolution R, which added support for RHEL 6, appears to work (it appears to at least install, run, and perform basic tasks). See this post for more details.

I’ve come to enjoy using R. I had dabbled with it in the past, but found it painfully opaque, and the Effort:Reward ratio when I already used SAS just enough to keep me interested. But then a couple things happened – I went off and learned Python, and all of a sudden about half the things I found “quirky” about R made sense, and I found myself needing packages in R to do things SAS is pretty bad at – meta-analysis, and one particular researcher’s code that comes in an R package, complete with a handy tutorial.

So now I use R maybe…35% of the time?

But this is something of a side note. This review is on Revolution R, a commercial version of R that has some promising stuff in it. Or this review would have been on it, but for some…problems. More after the jump.

Revolution R seems, essentially, to be a distribution of R being billed as “Faster, Better, Stronger!” by Revolution analytics. Some of the underlying math libraries have been tweaked to make it faster in benchmarks, it’s got its own ritzy IDE, and its supposed to handle “Big Data” better – R sometimes struggles with very large data sets.

All of this was interesting to me, and lo, there’s a free academic version.

First problem: No Mac version. This is part of what sold me on the whole R thing in the first place, not having a partition on my machine that might as well be named “SAS and Mass Effect 2″. When asked over Twitter about Mac development, I was told this:

@epigrad We’d love to, but our roadmap is focused on Windows and RedHat for now. But it’s on our list for the future.

Fair enough says I. And I’ve got Parallels, I can get around this! Not being a huge fan of Windows, and the whole point of this being to get away from something that needs a Windows license as well, I decide to check out the Linux version. I admittedly do not have Red Hat Enterprise Linux 5. What I do have is its extremely closely intertwined cousin, Fedora. Sponsored by the same company. Both use Yum as a software package manager. Fedora, according to Wikipedia, “Fedora serves as upstream for future versions of RHEL. RHEL trees are forked off the Fedora repository, and released after a substantial stabilization and quality assurance effort.”

Outstanding. And things designed for “Red Hat Linux” have worked on my Fedora install before – namely SAS and the Enthought Python Distribution.

Not Revolution R my friends. Three attempts to install, a few walls of failed dependency installations, and I gave up and emailed for help, detailing that I was using Fedora 15, and what was happening. What I got back was this:

The Linux version of Revolution R Enterprise is officially only supported on Redhat and CentOS. Our installer requires ‘yum’ to install all of the components. It also relies on ome third party libraries and will attempt to use ‘yum’ to download and install those dependencies.

We list all of these dependencies in our Installation Guide on page 7, in the section ‘Package Dependencies’.  The error message you received is indicating that the installation could not complete successfully, because some of these dependencies were not found. All of the dependent packages need to be installed first, for the main part of the install to proceed.

I would suggest that you attempt to install the software on a Redhat system.

Otherwise, you will need to see if you can find versions of the dependent packages for Fedora 15 in their official repository.

To be blunt…no. I tried, I really did, but R installed like a charm, and I’m not buying another commercial operating system to get your thing to work, or going hunting for my own packages. The whole point of commercial software on Linux OS’s is that you shouldn’t have to do that. For a product that charges $1,000 for a single user workstation license to not work on a nearly identical OS? If your product really is only compatible with Red Hat Enterprise Linux, that’s…a pretty narrow slice of the market. I’d also note that CentOS isn’t listed on the site as one of the supported operating systems – apparently one similar-to-RHEL Os works, but the other doesn’t.

I give up. Back to regular R it is. I’m sure Revolution R is an excellent product, and if you’re using Windows or RHEL, you’re set. But honestly, at that point, I’m back to the things I don’t like about SAS, and away from the things that pulled me to R.