The only few Git commands that I use

Git is great but quite complicated to fully master¹. Fortunately, being familiar with a few commands is usually enough. Here’s basically all the git commands that I use.

(First of all, I need to claim that I prefer using a GUI. I either use RStudio’s Git panel or Git’s GUI.)

Clone

git clone {git-url} {local-folder}: clone a remote repository to local

Branch

git checkout {branch-name}: checkout to a branch
git checkout -b {new-branch-name}: create a new branch and checkout to it
git branch -d {branch-name}: delete a local branch (should use -D if the branch hasn’t been merged)
git merge {another-branch}: merge the content of another branch to the current branch

Repository

git remote -v: display all the remote repository
git remote add/rm {repo-alias} {git-url}: add or remove a remote repository

Local - Remote

git fetch {repo-alias}: fetch the content from the remote repository
git push {repo-alias} {branch-name}: push the local content to remote
git push -u {repo-alias} {branch-name}: push and set the remote branch as the upstream, which means you don’t have to specify the remote repo and branch when using git pull and git push later
git pull {repo-alias} {branch-name}: equals to git fetch first then git merge {repo-alias/branch-name}

Commit

git commit -a -m "your message": commit all the code changes with a message.

Clean up

git branch -d $(git branch --merged=master | grep -v master)
git fetch --prune

Advanced

git reset --hard HEAD~n: discard the last n commits hardly. --hard means reverting to the previous status. Often used when you think your last n commits are nonsense.
git reset --soft HEAD~n: discard the last n commits softly. --soft means keeping the changes. Often used when you want to squash the commits or remove the just-unintentionally-included files.
git revert HEAD: generates a new commit to revert the change of that commit. It’s similar to git reset with th difference that git reset erase commit but git revert add commit.
git merge --abort: imagine you pull from the upstream and a CONFLICT happens. You can solve the conflict but you may also consider to abort the merge. This command allows to abort the current merge.
git rebase --onto master B C: (Well, frankly speaking, I’m writing this article because of this command) Often used when you made a commit, which should have been committed on another branch. So you need to rebase the commit.
Let’s say you are working on the D branch, developping a cool new feature. You need to use an existing function but find a bug inside. The correct procedure is: git checkout master, modify the code, commit, git checkout D and git merge master to have the fix on D. Remember that you are immersed in developping the new feature (good programmers do) so there’s a chance you simply commit the fix on D and go. After a few more commits on D, you realize the previous commit should have been played on the branch master.
You have three options now. The first is to git revert --soft all the previous commits and re-commit the work except the fix. The good commit history may not be well preserved. Another option is to re-do the fix on the master branch but you may find conflicts when branch D gets merged. The last option (usually the best) is to git rebase master B C, where B and C stands for the SHA-1 strings² of last commit before the fix and of the fix.
You may find git rebase’s documentation useful (I admit I don’t really know how to interactively play with git rebase and the above command is the only one I use).
git clean -df && git checkout . to discard all local file changes. This is useful when you add / edit lots of local files and want to completely discard all your wasted efforts.

Misc

Well that’s not literally all. I may use git config (first time set-up on the new computer), git tag (tag a version) and git status (when I have to disable the Git function in RStudio for one project, due to the anti-virsus issue).

References

Florent Destremau, A simple way to clean up your git project branches

Two books I recommend on Git. The Git related chapter of Hadley’s R packages is well-written and a great place to get your hands dirty. To get a deeper understanding of Git, Pro Git book is the best choice. ↩︎
You can think SHA-1 strings are the unique names of each commit. They’re the “non-sense” strings displayed in RStudio’s Git Panel (the last column with the name SHA). You can also see them by executing git log. ↩︎