EMODnet-Biology-PhaseV

Code and notes for the EMODnet Biology project Phase V, in which our goals will be:

1. To make DIVAnd interpolation method available in R
2. Create gridded maps (heatmaps) of a specific species (birds? mammals?)
3. Compare the results obtained with DIVAnd with other methods available in R.

Installation and configuration

The instructions are given for a machine running under Ubuntu (22.04.3 LTS -- Jammy).

Installation of R

Documentation: https://cran.r-project.org/web/packages/JuliaCall/readme/README.html

sudo apt install r-base

Editor [optional]

If you want to work with an editor, the most widely used is Rstudio. Visual Studio Code also comes with various extensions to work with R (REditorSupport, R Debugger, ...).

Installation of Julia

We suggest to use the juliaup tool (https://github.com/JuliaLang/juliaup), which makes easier the installation, upgrade and management of different versions of Julia. On Linux or Mac:

curl -fsSL https://install.julialang.org | sh

The Julia version that will be used can be obtained with the command:

juliaup status

which gives, in our case:

 Default  Channel  Version                 Update 
--------------------------------------------------
          1.10     1.10.0+0.x64.linux.gnu         
          rc       1.10.0+0.x64.linux.gnu         
       *  release  1.10.0+0.x64.linux.gnu 

Installation of Julia in R

In a R session:

install.packages("JuliaCall")

You are asked if you want to use a personal library (type "yes"):

Warning in install.packages("JuliaCall") :
  'lib = "/usr/local/lib/R/site-library"' is not writable
Would you like to use a personal library instead? (yes/No/cancel) 

Tested R versions

• Eye Holes

version.string R version 4.3.2 (2023-10-31)
nickname       Eye Holes

• Bird Hippie

version.string R version 4.1.2 (2021-11-01)
nickname       Bird Hippie    

• _Shortstop Beagle _

version.string R version 4.2.3 (2023-03-15)
nickname       Shortstop Beagle  

Installation of netCDF

R relies on the utility nc-config to it has to be installed:

sudo apt-get install libnetcdf-dev

then the library can be installed:

  install.packages("ncdf4")

Installation of other packages [optional]

No used so far but let's keep it here for the time being.

jsonlite

When tested with Visual Studio Code, the editor required to install jsonlite

install.packages("jsonlite")

even if the installation may not be a strict requirement.

logger

The logger package is also installed for the logging purposes.

install.packages("logger")

Plotting libraries

In Julia, the plots are create with the PyPlot module, which is calling Python matplotlib functions. It might be more relevant to use only R library for the plotting tasks:

install.packages("ggplot2")
install.packages("ggmap")
install.packages("sf")
install.packages("terra")
install.packages("rnaturalearth")
install.packages("rnaturalearthdata")

Library	Description
ggplot2	creating graphics, based on The Grammar of Graphics
ggmap	spatial data and models on top of static maps
rnaturalearth	World Map Data from Natural Earth
rnaturalearthdata	World Map Data from Natural Earth Used in 'rnaturalearth'
sf	simple features, to encode spatial vector data
terra	spatial data analysis with vector and raster data

Configure Julia

You may want to specify the path to the Julia executable with the command julia_setup:

library(JuliaCall)
julia_setup(JULIA_HOME = path.expand("~/.juliaup/bin/"))

If successful, this command will give:

Juliaup configuration is locked by another process, waiting for it to unlock.
Julia version 1.10.0 at location /home/ctroupin/.julia/juliaup/julia-1.10.0+0.x64.linux.gnu/bin will be used.
Loading setup script for JuliaCall...
Finish loading setup script for JuliaCall.

so you can test if R is actually going to use the correct Julia executable.

Install Julia packages

We use the command julia_install_package_if_needed( ) for the installation.
Some errors happened with using julia_install_package( ).

Documentation: https://search.r-project.org/CRAN/refmans/JuliaCall/html/julia_package.html

julia_install_package_if_needed("Statistics")
julia_install_package_if_needed("DIVAnd")
julia_install_package_if_needed("NCDatasets")

Loading Julia packages

Before running the julia_command(" "), ensure that the Julia packages are already installed (with the Julia version specified in the variable JULIA_HOME).

julia_command("using Statistics")

Note: it is also possible to start a Julia session within R:

system("julia")

NCDatasets

It may be necessary to issue this command before starting the R session, in order to ensure the correct libcurl is used:

export LD_PRELOAD=${HOME}/.julia/juliaup/julia-1.10.0+0.x64.linux.gnu/lib/julia/libcurl.so.4.8.0

(with the obvious adaptations in the path and in the library number).
The previous command ensures that the file will be loaded before any other library.

julia_command("using NCDatasets")
julia_command("using DIVAnd")

If the commands worked, the outputs are:

Precompiling NCDatasets
  2 dependencies successfully precompiled in 5 seconds. 44 already precompiled.

and

Precompiling DIVAnd
  1 dependency successfully precompiled in 4 seconds. 192 already precompiled.

Quick test

export LD_PRELOAD=${HOME}/.julia/juliaup/julia-1.10.2+0.x64.linux.gnu/lib/julia/libcurl.so.4.8.0
Rscript DIVAnd_simple_1D.R

Other solution: install everything with Conda?

conda install conda-forge::r-base

which juliaup
/home/ctroupin/conda_root/bin/juliaup