Appendix C — Technical guidance - git
C.1 Installing Git
Download the installer from https://git-scm.com/downloads
NHS RAP Community of Practice have a Git Quick Start Guide written for the Terminal commands.
C.2 Set up using R
Following course materials developed by R Forwards or NHS-R Community Introduction to Git and GitHub using R which is also based on R Forwards slides. Another excellent resource is Jenny Bryan’s Happy Git and GitHub for the useR.
C.3 Removing sensitive and patient identifiable information
GitHub recommend using BFG Repo-Cleaner for a quick and efficient way of deleting files, their history from the commit and this can be used across all branches. BFG Repo-Cleaner is also good for removing very large files.
It requires Java installed and this may also require administrator rights to do as well as the BFG Repo-Cleaner .jar file.
Once downloaded put the bfg-1.14.0.jar in the folder where the file is that you wish to delete.
The documentation recommends taking a copy of the repository before making any changes. The code on the BFG Repo-Cleaner is for the Terminal or you can copy local folders.
Delete the file and commit that so that the latest commit is clean and doesn’t contain the undesired data.
Using the Terminal type (on Windows)
java -jar bfg-1.14.0.jar -–delete-files my_sensitive_file.rda
If the file and bfg-1.14.0.jar are in a subfolder amend the code for the bfg-1.14.0.jar part only:
java -jar example\subfolder\bfg-1.14.0.jar -–delete-files my_sensitive_file.rda
If it works then you should get a whole list of information about the deletion. However, if it doesn’t work then you will get information on the ways you can use the program.
If the sensitive data is part of something like a Quarto report, website or book then a corresponding html file will also have to be deleted.
Once the file history has been removed from the Git history type:
git push --force
Using the BFG Repo-Cleaner changes the Git history on main and may also do this for branches.
Anything that has already been cloned, forked or downloaded from GitHub will be unaffected and you may need to contact GitHub directly to ensure this information is removed from GitHub repositories.
C.4 Disaster planning
Any publishing of sensitive data will require an incident to be raised within your organisation and may be classed as a breach and this can cause stress and pressure in the people involved. In order to react quickly and efficiently it’s advisable to practice deleting Git histories that don’t involve sensitive information.
Prevention is also better than recovery and many of the teams and organisations listed in the Statement on Using Tools - git are working on preventative measures including using a comprehensive .gitignore, git hooks and so on. [TODO] Contributions on preventing accidental sharing of sensitive data for this book will be welcomed.
C.5 GitHub Personal Access Token
The Personal Access Token (PAT) should never be stored in any file that can be committed and pushed to GitHub. However, if this does occur GitHub will contact you to say that your PAT has been revoked and you need to set up a new one. This means that your code and history are untouched but you will need to set up a new PAT to reconnect your local Git to GitHub.
D GitHub
D.1 Personal Access Tokens
GitHub has two main security ways to connect: SSH and HTTPS and many people working on a public sector network may find that SSH is blocked through a firewall or proxy. NHS-R Community’s course Introduction to Git and GitHub using R has instructions on how to set up git and connecting to GitHub using the {usethis} package which translates the git commands so these can be used from the R Console rather than the Terminal.
D.2 Using more than one computer?
If you are using more than one computer to connect to GitHub the best way to ensure security is to set up two separate PATs.