Moving to the Dark Side

Leaving the Pipette for a Keyboard.

How to compile R in a specific location

Issue

I use the Institutes’s server for which I am not an administrator. Recently I noticed that the Bioconductor installed in our Debian server was two version out of date, and so I decided to install the newer version of R and upgrade the packages.

Save installed packages

Firstly I saved the list of packages previously installed so that these could be easily re-installed:

saveInstalledPkgs <- function(out_file){
    pkgs <- rownames(installed.packages())
    save(pkgs, file=out_file)
}
version <- getRversion()
saveInstalledPkgs(
    paste(
        "~/imb-kettinggr/common_bin/R/installed_packages_",
        version,
        '.Rdata',
        sep='')
        )

Compile R from source

The goal is to install it in a shared folder so that:

  1. members of our group - biologists with little knowledge of the command line -get access that to the same R installation with which they can replicate my analysis.
  2. package versions are the ones needed for our group’s analysis without polluting the system-wide installation.

Ultimately without sudo powers R needs to be installed from source and since there is no good equivalent to the excellent python virtual env it can be an involved process. This bit of could should have been all that was needed:

BIN=/fsimb/groups/imb-kettinggr/common_bin/
cd ${BIN}
wget http://cran.rstudio.com/src/base/R-3/R-3.3.2.tar.gz
tar xvf R-3.3.2.tar.gz
cd R-3.3.2
./configure --prefix=/home/adomingu/imb-kettinggr/common_bin/R/R-3.3.2 --enable-R-shlib
make && make install

and it as indeed worked in the past for an older version of R. However, for R version 3.3.2 make && install threw an error in my general direction:

checking if libcurl is version 7 and >= 7.28.0… no configure: error: libcurl >= 7.28.0 library and headers are required with support for https

Which sucks.

[everytime a tool needs to be compiled I die a little inside]

Compile curl

So I needed to compile the curl as well. Following this tutorial, #11:

BIN=/fsimb/groups/imb-kettinggr/common_bin
cd $BIN
wget --no-check-certificate https://curl.haxx.se/download/curl-7.51.0.tar.gz &&
tar xzvf curl-7.51.0.tar.gz  &&
cd curl-7.51.0  &&
./configure --prefix=$BIN/curl &&
make -j3  &&
make install 

# make R (or better gcc) aware of the new libcurl (first two for building, the last two for the ./configure step)
LIBCURL_LOCAL_DIR="$BIN/curl"
export PATH=$LIBCURL_LOCAL_DIR/bin:$PATH
export LD_LIBRARY_PATH=$LIBCURL_LOCAL_DIR/lib:$LD_LIBRARY_PATH
export LIBCURL_CFLAGS=-I$LIBCURL_LOCAL_DIR/include
export LIBCURL_LIBS=-L$LIBCURL_LOCAL_DIR/lib

cd $BIN/
tar xvf R-3.3.2.tar.gz
mv R-3.3.2/ R/
cd $BIN/R/R-3.3.2
## configure R for usage in dependency tree (adjust if needed)
./configure --prefix=/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2 --with-x --with-blas --with-lapack --with-cairo --enable-static --enable-R-static-lib --enable-R-shlib --with-libpng --with-jpeglib --with-libtiff --with-ICU --with-tcl-config=/usr/lib/tclConfig.sh --with-tk-config=/usr/lib/tkConfig.sh --with-tcltk &&
make && make install
cd ..
chmod -R ug+rw R-3.3.2

As it is easy to see there were a lot options needed for the compiling to work, but it did work in the end.

Install libraries

Install libraries in a specific folder, instructions. Remember that for each R version there will be a correspondent libraries folder:

TOOL_DEPENDENCIES="/fsimb/groups/imb-kettinggr/common_bin"
mkdir ${TOOL_DEPENDENCIES}/R/R-3.3.2_LIBS
export PATH=${TOOL_DEPENDENCIES}/R/R-3.3.2/bin:$PATH
export R_LIBS=${TOOL_DEPENDENCIES}/R-R/3.3.2_LIBS

This lines were also added to the .bashrc:

## without the folowing line R will not be able to connect to the internet - custom installation
export PATH=/fsimb/groups/imb-kettinggr/common_bin/curl/bin:$PATH
export LD_LIBRARY_PATH=/fsimb/groups/imb-kettinggr/common_bin/curl/lib:$LD_LIBRARY_PATH
export R_LIBS=/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2_LIBS

Then from ${TOOL_DEPENDENCIES}/R/R-3.3.2/bin/R to install libraries:

## old
#.libPaths()
#[1] "/fsimb/groups/imb-kettinggr/common_bin/R/R-3.2.2_LIBS"           
#[2] "/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2/lib64/R/library"

shared_libs="/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2_LIBS"
.libPaths(shared_libs)
shared_libs
# http://stackoverflow.com/a/4090208/1274242
load('~/imb-kettinggr/common_bin/R/installed_packages_3.3.2.Rdata')
new_pkgs <- pkgs[!(pkgs %in% installed.packages()[,"Package"])]
if(length(new_pkgs) > 0) { install.packages(new_pkgs)}

new_pkgs <- pkgs[!(pkgs %in% installed.packages()[,"Package"])]
source("http://bioconductor.org/biocLite.R")
biocLite(new_pkgs)

source("http://bioconductor.org/workflows.R")
workflowInstall("rnaseqGene")
devtools::install_github("genomicsclass/ERBS")

Confer access to other users with chmod -R ug+rw /fsimb/groups/imb-kettinggr/common_bin.

Bioconductor

Bioconductor packages can also be installed like the above, but through biocLite:

shared_libs="/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2_LIBS"
source("http://bioconductor.org/biocLite.R")
biocLite("Gviz", lib=shared_libs)

library(Gviz)

There were some issues installing a package dependency, ggally, solved with:

library(devtools)
install_github("ggobi/ggally")

And changing the library path is also helpful:

shared_libs="/fsimb/groups/imb-kettinggr/common_bin/R/R-3.3.2_LIBS"
.libPaths(c(.libPaths(), shared_libs))

I hope this helps someone else in my situation.