To reproduce the results of TGS-Lite and TGS-Lite+ given in the paper, you need to follow the steps mentioned below.
- Step 1: Install R version 3.5.1 in your experimental environment (here)
- Step 2: Execute the TGS-Lite and TGS-Lite+ algorithms (here).
In this section, the following sub-steps are described:
- Installing or compiling R (here)
- Installing the ‘packrat’ dependency management package for R (here)
The experimental environment uses R version 3.5.1 (hereafter, R-3.5.1). Therefore, it is expected that any R version, same or higher than that, should suffice. For R-3.5.1 installation in Windows or macOS, please consult the following official manuals, respectively.
- 'Installing R under Windows': https://cran.r-project.org/doc/manuals/r-release/R-admin.html#Installing-R-under-Windows
- Download R version 3.5.1 for Windows: https://cran.r-project.org/bin/windows/base/old/3.5.1
- 'Installing R under macOS': https://cran.r-project.org/doc/manuals/r-release/R-admin.html#Installing-R-under-macOS
For Unix and Unix-alike OSs (like - Linux OSs), R-3.5.1 can be compiled to a destination directory of your choice from the R-3.5.1 source tarball ( https://github.com/sap01/TGS-Lite-supplem/blob/master/sourcecode/R-3.5.1.tar.gz ). Compilation is different from a system-wide installation. A R compilation inside a specific directory does not affect the system-wide installation or any other compilation inside any other directory. Thus, you may create multiple compilations in your system, where each compilation refers to a distinct version of R.
Please keep the source tarball and the ‘compileR351.sh’ shell script ( https://github.com/sap01/TGS-Lite-supplem/blob/master/sourcecode/compileR351.sh ) inside the same directory. Then open the shell script and replace the value of the variable ‘COMPILEPATH’ to a directory of your choice, where you wish to have the compiled R-3.5.1. For example:
COMPILEPATH=/home/saptarshi/R/R-3.5.1
Kindly make sure that the given path exists. The script does not create the path itself. After the compilation path is assigned, execute the shell script as shown below: (here, the '$' symbol represents the bash command prompt)
$ ./compileR351.sh
The execution may take a few minutes to complete. What to do if the execution is successful and what to do if it fails, are described in the following two paragraphs, respectively.
If ‘compileR351.sh’ Completes Execution Successfully: If 'compileR351.sh' completes execution successfully, it builds a compiled version of R-3.5.1 inside the given directory. This specific compilation can now be invoked from bash using the following command: (basically, append ‘/bin/R’ to the compilation path)
$ /home/saptarshi/R/R-3.5.1/bin/R
R version 3.5.1 (2018-07-02) -- "Feather Spray"
Copyright (C) 2018 The R Foundation for Statistical Computing
...
>
Symbol '>' represents the R command prompt.
If the compilation path is very long, writing this command can be a tedious process. Instead, go to your home directory and append the following line to your '.bashrc' file:
alias R351=/home/saptarshi/R/R-3.5.1/bin/R
Save and exit '.bashrc'. But the current bash session keeps using the older version of '.bashrc'. To mitigate this issue, source (reload) '.bashrc' as shown below:
$ source .bashrc
It creates an alias to the long command. Now, you may use the alias as follows:
$ R351
R version 3.5.1 (2018-07-02) -- "Feather Spray"
Copyright (C) 2018 The R Foundation for Statistical Computing
...
>
For executing any R script with R-3.5.1, please use the following command in bash:
$ /home/saptarshi/R-3.5.1/bin/Rscript path/to/script/script_to_run.R
If ‘compileR531.sh’ Fails: If 'compileR351.sh' fails (e.g. 'configure: error'), the reason behind the failure is that the OS does not have some of the required dependencies. Resolutions of a number of possible errors are discussed at: http://pj.freefaculty.org/blog/?p=315 ; if required, please find its offline PDF version at: https://github.com/sap01/TGS-Lite-supplem/blob/master/sourcecode/building-r-devel-PJ.pdf . Source tarballs of most of the dependency packages are stored at: http://r.research.att.com/src/ . Also the ‘Installing R under Unix-alikes’ official manual can be consulted at: https://cran.r-project.org/doc/manuals/r-release/R-admin.html#Installing-R-under-Unix_002dalikes .
Once R is installed, the next step is to install the ‘packrat’ package (https://CRAN.R-project.org/package=packrat). It provides a dependency management system for the algorithmic implementations. In other words, it automatically downloads and installs the required packages with the required versions when the implementations of the TGS-Lite and TGS-Lite+ algorithms are being installed. In the experimental environment, ‘packrat’ package version 0.4.9-3 is used. Therefore, it is expected that installing the same or higher version would be sufficient. In the experimental environment, a specific directory, namely 'src downloaded packages' is created to download required external packages, like - 'packrat'. Then the latest version of ‘packrat’ (version 0.4.9-3 at that time) is downloaded inside it and installed to the 'library' directory. This whole process is done through the following commands:
%% Go to the R compilation’s home directory
$ cd /home/saptarshi/R/R-3.5.1
%% Create a directory for downloading external R packages
$ mkdir src_downloaded_packages
%% Open a R session
$ R351
## Download the latest version of the ’packrat’ package
## from ’https://cloud.r-project.org/’
## to ’/home/saptarshi/R/R-3.5.1/src_downloaded_packages’
## and install it inside ’/home/saptarshi/R/R-3.5.1/library’.
> install.packages(’packrat’, ’/home/saptarshi/R/R-3.5.1/library’,
dependencies = TRUE, repos = ’https://cloud.r-project.org/’,
destdir = ’/home/saptarshi/R/R-3.5.1/src_downloaded_packages’)Finally, a directory with the name 'projects' is created where the algorithmic implementations would be installed.
$ mkdir /home/saptarshi/R/R-3.5.1/projects
The R implementations of the TGS-Lite and TGS-Lite+ algorithms are saved as a R project tarball (https://github.com/sap01/TGS-Lite-supplem/blob/master/sourcecode/TGS-Lite-2019-04-15.tar.gz). Copy this file to '/home/saptarshi/R/R-3.5.1/projects':
$ mv TGS-Lite-2019-04-15.tar.gz /home/saptarshi/R/R-3.5.1/projects
After that, unbundle the project using 'packrat':
%% Go to the project directory
$ cd /home/saptarshi/R/R-3.5.1/projects
%% Open a R prompt
$ R351
## Set your favourite CRAN repo, e.g.,
## https://cran.rstudio.com/
> options(repos=structure(c(CRAN="https://cran.rstudio.com/")))
## Attach ’packrat’ package
> library(packrat)
## Unbundle the project inside the current directory
> packrat::unbundle(’TGS-Lite-2019-04-15.tar.gz’, getwd())Once unbundled, a new project directory will be created with name 'TGS-Lite'. Go inside the project directory:
%% Go to /home/saptarshi/R/R-3.5.1/projects/TGS-Lite
$ cd TGS-Lite
Directory 'TGS-Lite' contains all required R scripts and two sub-directories: 'packrat' and 'asset'. The 'packrat' sub-directory is for internal management of 'packrat' and not to be interfered with. The 'asset' sub-directory is the place where the input and the output files are stored. Copy all the dataset files inside this sub-directory. For example, let us assume that the directory corresponding to https://github.com/aaiitg-grp/TGS/tree/master/datasets , in your local computer, is '/home/saptarshi/datasets'. Then copy all the files from that directory to 'TGS-Lite/asset':
$ scp /home/saptarshi/datasets/* asset
The driver R script is 'TGS-Lite/TGS.R'. It takes the user-defined input parameters in a JSON file format (http://www.json.org/). The JSON file must reside in the ‘asset’ sub-directory. ‘TGS.R’ can be executed with the following command:
%% Assuming ’TGS-Lite/asset/input.json’ contains the user-defined parameters.
%% Note that only ’input.json’ is used instead of ’asset/input.json’. This
%% is because ’TGS.R’ is programmed to search for the input JSON files in
%% the ’asset’ sub-directory.
%% The ’&’ symbol instructs the execution process to start in the background.
$ nohup time /home/saptarshi/R/R-3.5.1/bin/Rscript TGS.R input.json &
%% Prints the process ID
[1] 8172
Since, the execution process is performed in the background, the bash command prompt can be used for other tasks. Once the execution is complete, a ‘Done’ message is automatically displayed in bash. However, if you wish to monitor the execution of the process, you may do so with the ‘top’ command as shown below:
%% Show details of the process with ID 8172
$ top -p 8172