Git for Version Control

A version control system is a tool for managing changes to files over time by efficiently tracking all modifications. One of the most popular tools for version control is “git”. Another version control tool called subversion or SVN has also been widely used, but git has become a de facto standard. Git generally is already installed on most Linux-like systems.

Version control is essential for projects with multiple people editing the same collection of code. However, it can also help individual users more effectively and efficiently manage and document their scripts and programs.

Read more about version control and git here. In particular, you should read:

Read this post on writing better commit messages.

Optional reading:

Git from the command line

To begin tracking an existing project, move to the top folder in the project tree and type:

git init

To create a local copy of an existing repository:

git clone "url for repo"

For example, to clone the “Stats506_F20” repository from gihtub.com:

git clone https://github.com/jbhender/Stats506_F20.git

When cloning repositories to which you have push access, I suggest you append your username to the url like below:

git clone https://jbhender@github.com/jbhender/Stats506_F20.git

This will ensure you are granted push permissions and prompted for a password. This is not necessary when using ssh keys (see below).

Repositories

Git is a “distributed version control system” meaning that all copies are “local”.When you begin a version controlled project using git, the project folder is itself a repository. You do not need a remote repository to use git.

However, git is used most effectively with a remote repository. Remote git repository services such as github or bitbucket are effective means for both backing up and sharing the code you write. It is also a great way to manage work across multiple computers, such as a personal laptop and one or more servers.

Git for remote repositories

The following commands are useful for working with a remote repository.

To link a local and remote repository, use git remote:

git remote add origin url_to_remote_repo.git
git push -u origin master

If you haven’t made any local commits and want to bring your local copy up to date with the remote repository use git pull. To retrieve remote changes without merging into the local repository use git fetch followed by git status or git diff to see changes that have been made. You can then use git merge or git pull to finish merging.

Example

This example shows how to setup a git remote in your afs space.

Exercise

  1. Try the example on your own. Use this repository to practice version control while doing your assignments.

  2. Add, commit, and push the R and Rmarkdown templates from: https://github.com/jbhender/Stats506_20

    1. Clone the repo
    2. Copy the files to your 506 repo
    3. Add and commit them
    4. Push these files to the repo

Caution

When using a remote code repository like GitHub or Bitbucket or even AFS be careful not to submit any protected or sensitive data. When working with such data, it is especially important to use git status before git commit or git push to ensure you know exactly what will be uploaded to a remote repository.

Similarly, when working on a project with someone else, be sure to check whether there is any concern about sharing the code in a public or private repository.

Git Configuration

As with most programs, you can set preferences in git to improve your work flow. Specifically, this can be done from the command line using git config. Here are a few options you set everywhere you use git:

Setting your user name and email allows you to omit explicitly specifying them when using git push and git pull to interact with your remote repository.

git config --global user.name "jbhender"
git config --global user.email "jbhender@umich.edu"

To set your default editor to emacs (substitute your choice of editor) use:

git config --global core.editor emacs

This could also be done in the shell by exporting the EDITOR environment variable,

export EDITOR=emacs
echo $EDITOR

but doing so will affect all programs that rely on the EDITOR variable.

When making these changes be aware that they will only apply within the environments that share the same .gitconfig file. You may need to repeat these steps separately on the SCS servers, your personal computer(s), and the Great Lakes HPC environment.

Using .gitignore

At times you may find you have certain files that you want to keep in your local directory but not the remote repository for privacy or other reasons.

You can tell git to ignore such files by including a .gitignore file at the top of your repositories file tree. For example, running

cd ~/My_Repo
echo "*.csv" > .gitignore

from the command line will create a .gitignore file in My_Repo telling git to ignore all files with the .csv extension.

I’d recommend including the following in your .gitignore at a minimum:

  • .* : all hidden files
  • *~ : temporary backups created by Emacs
  • .DS_store: if you ever work on Mac
  • Various log and error file types:
    • *.log
    • *.out
    • *.Rout

To track a file usually ignored by git use git add -f.

Exercise

  1. Use your chosen editor to create a .gitignore file in your 506 repo with the extensions above.
  2. Create a new file “test.tmp” in your 506 repo.
  3. Run git status. Where does “test.tmp” appear?
  4. Add *.tmp to your .gitignore for 506
  5. Run git status. Why does “test.tmp” no longer show up?
  6. Clean up the repo by deleting “test.tmp”

Resources

Read more about configuring git here.

For routine use of git from your personal computer, you may want to connect to your remote repository using ssh. Using ssh keys you can communicate with remotes without supplying your password for each interaction. Read more here.

Those working in a Windows environment may be interested in the Git for Windows program which also includes a BASH shell.

There is also a GUI application GitHub Desktop that some may find useful.

The RStudio IDE also has support for Git. See the tab Git near the Environment tab.

Using git for homework

I encourage you to use git for your work in this class. However, please refrain from posting solutions to problem sets to public repositories created at GitHub or Bitbucket for this course prior to the due date. Instead, set up a private repo.

After the solutions are released, I suggest you do post your (possibly corrected) code to a public repo and include a concise description of what you’ve learned in a README.md written in markdown. Your repository can serve as a sort of “digital portfolio” for your computing and analysis skills that you can share with potential employers.

Course Homepage