INSTALL.md

In sawers-rellan-labs/grassGEA: Sorghum and Corn Genome Environment analysis

Table of Contents

• HPC installation
• Conda environment basic setup
• Install rJava
• Install rTASSEL
• Install our package grassGEA
• Add environment variables
• yq insallaion

HPC installation

Conda environment basic setup

# install R with compatible versions of BLAS/LAPACK
# for matrix algebra
conda create --prefix /usr/local/usrapps/maize/sorghum/conda/envs/r_env\
      -c conda-forge r-base blas lapack

Add packages

# Activate r_env
conda activate /usr/local/usrapps/maize/conda/envs/r_env

conda config --add channels conda-forge   
conda config --set channel_priority strict
# install mamba?
# conda install mamba
conda install r-essentials
conda install r-devtools
conda install r-xml
conda install r-raster
conda install r-rgdal
conda install openjdk

Install rJava

As discussed here

I need to find the directory where libjvm.so is:

# I know, wrong use of find but it works:
find /usr/local/usrapps/maize/sorghum/conda/envs/r_env | grep libjvm.so

It seems that you need to reset R_JAVA_LD_LIBRARY_PATH in the file R/etc/ldpaths from the R installation, according to this post

# I know, wrong use of find but it works:
find /usr/local/usrapps/maize/sorghum/conda/envs/r_env | grep ldpaths
# /usr/local/usrapps/maize/sorghum/conda/envs/r_env/lib/R/etc/ldpaths #

You could edit it by hand but the intended way is giving the path to R CMD javareconf. As stated in the help page, R CMD javareconf --help, you can give it a path for JAVA_LD_LIBRARY_PATH.

#from tcsh
R CMD javareconf \
   JAVA_HOME=$CONDA_PREFIX/jre
   JAVA_LD_LIBRARY_PATH=$CONDA_PREFIX/jre/lib/amd64/server

This was not enough and I finally had to add the $LD_LIBRARY_PATH to the environment via conda I'll also add some othe useful environment variables for the installation.

Add environment variables

# in tcsh
conda env config vars set env R_ENV=$CONDA_PREFIX
conda env config vars set env OLD_LD_LIBRARY_PATH=$LD_LIBRARY_PATH
conda env config vars set env LD_LIBRARY_PATH=$CONDA_PREFIX/jre/lib/amd64/server:$LD_LIBRARY_PATH
conda env config vars set env GEA_CONFIG=$CONDA_PREFIX/lib/R/library/grassGEA/extdata/config.yaml
conda env config vars set env GEA_EXTDATA=$CONDA_PREFIX/lib/R/library/grassGEA/extdata
conda env config vars set env GEA_SCRIPTS=$CONDA_PREFIX/lib/R/library/grassGEA/scripts
conda deactivate 

conda activate /usr/local/usrapps/maize/sorghum/conda/envs/r_env

#
# upon activation it throws the following warning:

# WARNING: overwriting environment variables set in the machine
# overwriting variable LD_LIBRARY_PATH

# Let's hope the change in path will be ok

Now I can install rJava from R (I did try from conda,but it did not work because of that LD_LIBRARY_PATH variable).

But first activate the conda environment

conda activate /usr/local/usrapps/maize/sorghum/conda/envs/r_env
R

Install rTASSEL

# this will install rJava as a dependency

if (!require("devtools")) install.packages("devtools")
 devtools::install_bitbucket(
    repo = "bucklerlab/rTASSEL",
     ref = "master",
    build_vignettes = FALSE
)
# RcppTOML a configr  dependency did not compile using system gcc [4.85]
# this was the whole reson I ended up using a conda environment for R
# install.packages('RcppTOML')

Install our package grassGEA

devtools::install_github("sawers-rellan-labs/grassGEA")
q()

yq installaion

Now I will install yq 4.20.1 to retrieve config values. The version of yq is critical because the syntax and options change a lot between 2, 3, and 4.

#Go home

cd 
mkdir yq; cd yq
wget https://github.com/mikefarah/yq/releases/download/v4.20.1/yq_linux_386.tar.gz
tar -zxvf yq_linux_386.tar.gz
mv yq_linux_386 $R_ENV/bin
ln -s $R_ENV/bin/yq_linux_386 $R_ENV/bin/yq
chmod a+x $R_ENV/bin/yq $R_ENV/bin/yq_linux_386
conda deactivate
conda activate /usr/local/usrapps/maize/sorghum/conda/envs/r_env

yq --version
# yq (https://github.com/mikefarah/yq/) version 4.20.1
yq .geno_dir $GEA_CONFIG

for local tests you should install yq 4 as well

Homebrew has yq version 4!

most likely in mac you won't be using the conda r_env

sawers-rellan-labs/grassGEA documentation built on May 16, 2023, 6:48 p.m.