This README details how to use variables in R and bash so that you can use relative file paths i.e. relative/file/path as opposed to /path/to/my/relative/file/path. This means when sharing code it should be more reproducible by others as they will be using the same file paths as you.

Make a GitHub repository. Here the name of our GitHub repository is relative_file_paths. It doesn’t matter if the repo is public or private.

All of my projects use the same starting structure (here our project is caled relative_file_paths):

##                        levelName
## 1  relative_file_paths/         
## 2   ¦--docs/                    
## 3   ¦--manuscript/              
## 4   ¦--data/                    
## 5   ¦--analysis/                
## 6   ¦--scripts/                 
## 7   ¦--environment/             
## 8   ¦   °--environment.sh       
## 9   ¦--.Renvironment            
## 10  ¦--.Rproj.user              
## 11  ¦--relative_file_paths.Rproj
## 12  ¦--.git/                    
## 13  ¦--.gitignore               
## 14  °--README.Rmd

To make this structure I use a standard script which first creates the project directory and sub-directories. It then makes two environment files which we use to pull our global file paths into our analysis scripts with. It also makes a .gitignore file for our project and then initializes the git repository.

#!/bin/bash
cd "/path/to/001_projects"

projectname=$"relative_file_paths"


echo "Making new project called $projectname in:"
pwd

#script to set up new project on laptop and new git repo that is linked to github
cd "/path/to/001_projects"

mkdir $projectname
cd $projectname

mkdir docs
mkdir manuscript
mkdir data
mkdir analysis
mkdir scripts
mkdir environment

# make environment for sourcing file paths in bash and R
cd ../
echo "LOCAL="$PWD"" > $projectname/environment/environment.sh
echo "LOCAL="$PWD"" > $projectname/.Renviron
cd $projectname


# make .gitignore
## general list of file extensions that could contain data
(
echo *.csv
echo *.tsv
echo *.dta
echo *.txt
echo *.dat
echo *.[rR]data
echo *.[rR]data
# R file formats
echo .Rproj.user
echo .Rhistory
echo .RData
echo .Ruserdata
echo *.Rproj
echo *.Rmd
# genetic data file formats
echo *.bgen
echo *.gen
# files generated from BlueCrystal jobs
echo out*
echo error*
echo j*.sh.e*
echo j*.sh.o*
echo *.sh.e*
echo *.sh.o*
# UK Biobank data file formats
echo *.enc_ukb
echo *.enc
echo *.cwa
# common directories
echo manuscript
echo data
echo analysis
echo environment
) > .gitignore

touch README.Rmd
git init
git add .gitignore scripts/
git commit -m "first commit"
git remote add origin https://github.com/YOUR_GITHUB_USERNAME/$projectname.git
git push -u origin master

echo "Finished"

I then make a new RStudio project and assign it to this existing directory by selecting File > New Project > Existing Directory > Browse and then selecting the recently made project directory.

This project directoy is set up with two files: .Renvironment and environment/environment.sh. Both files have the exact same content a line of code that reads 001_projects='/path/to/001_projects. This allows us to source this environment variable in our scripts so we can use relative file paths.

I want to make a plot using some data stored in a text file in my data/ directory. First we need to tell R what are working directory is, but we dont want to use a full file path. So we can grab the full file path from the .Renviron file we made with our set-up script using Sys.getenv("LOCLA") where LOCAL is the variable name we assigned that file path. Then we can set our working directory as LOCAL i.e. /path/to/001_projects and make some data and save it to data/:

# set environment
LOCAL <- Sys.getenv("LOCAL")
setwd(LOCAL)

# make text file of some data and save it in data/
data <- data.frame(a = rnorm(rep(1:100, 100)),
                   b = pnorm(rep(1:100, 100)))
write.table(data, "relative_file_paths/data/data_frame.txt", 
            row.names = FALSE, col.names = TRUE, quote = FALSE, sep = "\t")

Now we can access that data in a seperate script using the same process of retrieving the variable we want to use to set our working directory:

# set environment
LOCAL <- Sys.getenv("LOCAL")
setwd(LOCAL)

# load data
data <- read.table("relative_file_paths/data/data_frame.txt", header = T, sep = "\t")

# plot data
library(ggplot2)
a <- ggplot(data = data,
       aes(a)) + 
  geom_histogram(binwidth = 0.01, alpha = 1)

b <- ggplot(data = data,
       aes(b)) + 
  geom_histogram(binwidth = 0.01, alpha = 1)

library(cowplot)
plot_grid(a,b)

Lets say we want to make a file using bash in Terminal. We can’t access the LOCAL variable in our .Renviron folder using bash. Instead we have to access it through a text file. In this case we’ve made this text file available in our environment/ directory as environment.sh.

# set environment
source environment/environment.sh

# set working directory
cd "$LOCAL/relative_file_paths"

# make folder
echo "We made this file to show how to use environment variables in bash" > "$LOCAL/relative_file_paths/bash_file.test"

# show directory files
head "$LOCAL/relative_file_paths/bash_file.test"

## We made this file to show how to use environment variables in bash

You can see that heading the file shows that it has been made. And the fact it’s been made means it is in the right location.