1

I'm getting a little bit crazy with this issue. I'm trying to install an R package using conda in my environment (python 2.7) in my home on a cluster (i.e. without root permissions). I firstly installed R in my env using: 

conda install -c r r=3.4

Then: 

conda install -c conda-forge python-igraph

(because igraph is required by my library of interest)

and finally: 

conda install -c conda-forge r-diffusionmap

Unfortunately when I launch R the following message appears:

Error: package or namespace load failed for 'RevoUtilsMath': .onLoad failed in loadNamespace() for 'RevoUtilsMath', details: call: NULL
error: Remove Microsoft R and then re-install. Be sure to select MKL libraries as an install option.

During startup - Warning message: package 'RevoUtils' was built under R version 3.4.3

What does it mean? How can I solve this?

Thank you in advance

NewUsr_stat
  • 2,351
  • 5
  • 28
  • 38
  • You don't say what R version you're trying to run, but it appears you have a package in your library that was built for an incompatible version. You can do what the message says (reinstall Microsoft R), or delete the package and everything that depends on it. – user2554330 Jul 08 '18 at 10:50
  • I just edited with the r-version I specified – NewUsr_stat Jul 08 '18 at 11:16
  • There is no version 3.4 of R (versions have 3 parts, like 3.4.3). I don't know `conda` or what it would choose to install given that spec. – user2554330 Jul 08 '18 at 15:02
  • conda takes default 3.4.3 when nothing is specified. Takes the last version. – NewUsr_stat Jul 08 '18 at 19:48
  • 1
    The package you are using was built under 3.4.3, but the R you are running is a different version. So `conda` might not be doing what you think. (Or maybe Microsoft has screwed things up without changing the version number.) – user2554330 Jul 08 '18 at 20:03

3 Answers3

1

I had this same issue after I installed some libraries (Rcpp included) in my root R, but not my conda environment (which screwed up conda). This would cause kernel death anytime a jupyter notebook running R was even opened.

The fix for me was:

  1. Uninstall Anaconda3
  2. Reinstall Anaconda3
  3. Reinstall all the libraries I needed (mostly just Bioconductor in R)

A few other issues popped up, like package inconsistencies, but I dealt with those as described here.

Thomas Matthew
  • 2,826
  • 4
  • 34
  • 58
1

All R packages on conda-forge (or Bioconda) are compiled against one single version or R for each new release branch (usually starting from patch 1, so 3.x.1, except for 3.4.3). This is due to ABI incompatibility problems.

Also note that defaults and conda-forge channels are (where) not binary compatible (although now they should be). And that since 2018 the default anaconda channel is distributing Microsoft R Open as default R, whether all packages from conda-forge should be preferably used with R from conda-forge.

You should be able to solve this issue by installing R using conda install -c conda-forge r-base.

FabienP
  • 3,018
  • 1
  • 20
  • 25
0

the same error information for me when I open R for run code in ubuntu platform(18.4), and there is no other useful methods to solve it.My R version is 3.4.3.enter image description here

Hua Zou
  • 19
  • 1