Getting Started

I recommend doing tutorials before really diving in, as the way version control, and particularly git, works will be very confusing if you don't really grok it.

I would recommend not using the git docs unless you really need to as they, like most Unix tools, are not written for normal people and are therefore a bit obtuse and difficult to parse. Most people have tried to grok the concepts and put them in human terms and thus this site and other tutorials exist. The docs are good, but use them to complete your knowledge not as the basis.


To clone a repo to your system, use git clone path/to/repo. If you don't need/want the history (useful for saving space), precede the path with --depth=1.


Add files you want to commit: git add {file or folder name}. You can also add all by using git add ., but this is not recommended, as you don't want to add all unless you are POSITIVE it doesn't contain cruft or unwanted changes, which is rarely the case. An easy way to check what has changed in your files since last commit is to use the --patch flag.


Using this flag will let you track files without adding their contents to staging. This is useful especially with using the --patch flag below on new or currently untracked files. This flag can also be shortened to -N.

For example, if you were already tracking file1 and refactoring part of file1 into file2, you would normally not be able to commit only part of file2. By first using git add --intent-to-add/-N file2, the file would now be tracked but not yet staged. From here, you could use --patch and continue by adding portions at a time.[14]


Add only portions of a file as 'hunks': git add --patch/-p {file or folder name} (file or folder is optional). This will prompt you with options:

Amend Last Commit

Stage the changes you want to make and use git commit --amend to alter the previous commit. If you have already pushed your changes to a shared branch, do not do this as you will bork everyone else's history and it's a pain to fix.

Split Up Previous Commits Into Smaller Commits[15]

To split up your most recent commit, you can use git reset HEAD~. This will retain the changes you made in the last commit, but remove the actual commit of those changes from your history. From here you can use --patch and select which portions you want to commit.

To split up an older commit, you will want to use an interactive rebase. Don't rebase if you don't know what you are doing.

  1. Start by specifying how many commits you want to go back within this command: git rebase -i HEAD~N, replacing N with the number of commits you want to go back.
  2. In the interactive rebase screen, replace the word pick with edit to stop git at that commit before proceeding with the rebase. Once done, save and exit the editor. This will activate the git rebase process.
  3. Git should have stopped at your specified hash, which you can see with a git status. It should say something like You are currently editing a commit while rebasing branch 'branch_name' on 'f05d5ce'.
  4. From there, you can continue as if you were splitting your most recent commit: git reset HEAD~.
  5. Commit each change as desired and when complete, use git rebase --continue to finish the rebasing process.

Case Sensitivity

Git by default is not case sensitive when it comes to filenames and will ignore renaming of file structures that are only by case. To change this default behavior, use git config core.ignorecase false. Note that setting this option to false on a case-insensitive file system is generally a bad idea. Doing so will lead to weird errors. For example, renaming a file in a way that only changes letter case will cause git to report spurious conflicts or create duplicate files.

If it is a file or files, you can change their name as normal and then within git via this command: git mv -f filename.txt Filename.txt. This will make it show up in your commit. If it is a directory, you will need to do this with every file within the directory.


If you know that your code worked at commit aaaaaaaa and is no longer working at commit bbbbbbbb, you can use git bisect to isolate where the bad commit must be.

git bisect start
# assuming that you are at your HEAD, which is where you know the code is not working
git bisect bad
git bisect good bbbbbbbb

This will start the bisect process, where at each step you can check to see if your bug is present. If it is present, you can say git bisect bad, and if it is not present, you can say git bisect good. This will eventually leave you at the commit where your bug first appeared.

Squash and Merge[19]


before you merge a feature branch back into your main branch (often master or develop), your feature branch should be squashed down to a single buildable commit, and then rebased from the up-to-date main branch.

If you follow this process it guarantees that ALL commits in master build and pass tests. This simple fact makes debugging an issue much easier. You can use git bisect when trying to find the source of a bug. Git bisect becomes almost completely ineffective if there are broken commits on the master branch; if you jump to a commit that isn’t clean, it’s difficult or impossible to tell if it introduced the bug. [...] [A] drawback is that we lose some granularity when we squash our commits. If you really want to have multiple commits for a feature, at least squash down so that each commit builds and passes tests.


Most simple is to use the git merge --squash option[20]. If you want to merge starting-branch into target-branch:

Another more intensive way to do it that gives you more control is to rebase the new branch off of the target branch you wish to merge into. Using interactive rebase, select every commit except the oldest one and set to squash. Select that first commit and set it to reword. Once the rebase starts on exit, you will be able to change the commit message that will hold all of the squashed commits. Once that is complete, merge it into the target branch.


All preceded by git.

Action Command
Get current branch name git branch --show-current
See all branches branch
See the remote branches your local repo knows branch -a
Checkout an existing branch checkout {name}
Create and name a new branch branch {name}
Create and checkout a new branch checkout -b {name}
Create an orphaned branch checkout --orphan {branch_name}
Push the current branch push origin HEAD
Push the current branch with force push origin HEAD --force-with-lease or --force if you must but know the difference
Rename a branch. Assuming you're currently on the branch you want to rename branch -m {name}
Delete a fully pushed and merged branch branch -d {name}
Force delete a local branch branch -D {name}
Delete a merged remote branch push {remote_name} --delete {branch_name}
Merge another branch into currently checked-out branch merge {other-branch}

Orphan Branches

Commits usually have one parent, their previous iteration. The root commit (very beginning) of your repo has none, and when you merge commit a fork with another branch or the master branch.

An orphan branch is a new root commit, where:

The first commit made on this new branch will have no parents, and it will be the root of a new history, totally disconnected from all the other branches and commits. In other words, it creates a new root commit and uses it as a starting point for your new branch.

Delete All Merged Branches

To delete all branches that have been merged, start out by pulling from remote to ensure you are using the most up to date info. Then check which branches are merged by running git branch --merged. We will be using egrep to grab these and put them into a macro that will delete each merged branch that doesn't fit within the regex. To ensure a branch that contains a certain keyword does not get deleted, add it to the egrep -v "(^...)" line, with each keyword separated by an OR |. This is all then passed into xargs where it sends each merged branch as an argument to git branch -d. Note that this uses -d so it will only delete branches that have been merged into the current branch. This most likely won't include rebased branches.

git branch --merged | egrep -v "(^\*|master|main|dev|develop|staging)" | xargs git branch -d

Forking a Repo

A fork is a repo that "fork"s off of another repo to do work separate from the original repo. This can be useful if you want to contribute to a project but don't have rights to update it directly, want to make a new version of the project based on what already exists, make a parallel version that focuses on specific features while still maintaining the updates in the original repo, etc.

Github lets you click "Fork" and you're done, but what is really happening under the hood is helpful to know what you are doing. To make a fork of a repo[25]:


To see who made changes and when on a given file, you can use git blame [options] path/to/file/in/repo. This will show you all the lines and who touched them last at what commit.

The -s option will remove the committer's name from the display.

The -L x[,y] option will show line x to line y, or just line x if y is omitted. If y is preceded by a +, that will show from line x to line x+y

See Changes In Current Feature Branch

To see all the commits, use git log start...feature. To see all the changes themselves, use git diff start. start is the branch your feature branch was originally started from.

Find Changes Involving Specific String

You can scan your git history for instances where a change occurred involving a given string. If you want to test this only against a given file in your repository, follow your search string with -- path/to/file before the pipe.

git log -S 'someString' | awk -F' ' '/commit/ { print $2; }'

Cherry Pick[16]

When should I use cherry-pick? The short answer is: as rarely as possible.

For this reason, I have not spent much time learning how to get this to work, as the only experience I have had was one of my colleagues using this and causing hours of work trying to fix the side effects.

I have only ever used this when I accidentally added a commit to the wrong branch. So anyway, proceed with caution.

If you have a commit from one branch you want to apply to another, this can be done by checking out your destination branch and calling cherry-pick with the hash of the desired commit(s) you want to bring over.

git checkout target-branch
git cherry-pick d3db33f

Now commit d3db33f is on target-branch.

Using the -n flag at the end will bring in the changes found at the commit(s) without the commits themselves.


A revert allows you to go back to a previous commit, reversing those previous between commit X (HEAD) and commit Y (the selected commit). This will create a reversal commit for each commit to be undone between X and Y. In the case of a single commit to be undone, it will make a single reversal commit.

To revert one or many commits, open your git log and find the commit you want to revert back to. Save this hash. Then enter the following:

$ git revert [-m 1] <commit-hash>

The -m flag is used to revert a merge; to set where the mainline is, which parent you are wanting to revert back to. Most of the time it is 1.

For example, lets say you accidentally committed to the main branch and pushed up. You can revert this with git revert HEAD or git revert {hash}, creating a commit dialog and fixing your mistake through adding to your commit history.


Stashing allows you to temporarily save changes without committing anything.

To stash some changes, use git stash save 'message about stash', or just git stash if you want to leave out the message.

To see what you have stashed, use git stash list to see all stashed code. Then see what is in each stash by using git stash show {index}, and you can see it diffed by using the -p flag.

Stashes can be listed recalled via the indexes shown: git stash apply {index} or git stash pop {index} if you want to remove it from the stashes after being applied.

You can git stash drop {index} individual stashes or git stash clear the entire stash.

Recovering Dropped Commits or Stashes[26]

Use git show all dangling commits, blobs, and trees (in this case, I'll filter by commits with awk):

git log $( git fsck --no-reflog | awk '/dangling commit/ {print $3}' )

You can see what is in each of these commits by using git show $COMMIT_HASH.

Once you find the stash you want to recover, you can recover it by either applying the stash to the current branch, or just making a new branch:

# Apply stash
git stash apply $COMMIT_HASH
# New branch
git branch branch-name $COMMIT_HASH

Resolving Merge Conflicts in Pull Requests

To resolve a merge conflict in a pull request, you can use git merge {target-branch}, where {target-branch} is the branch you are hoping to eventually merge your code into. This will create your merge conflicts locally, where you can create a new commit to handle them and rebase if you want to clean it up.


If you receive a message like this when trying to push up to your repo:

remote: HTTP Basic: Access denied
fatal: Authentication failed for ''


remote: Invalid username or password.
fatal: Authentication failed for ''

Then you may have revoked or renewed your token and need to reset it. To do so, type this in the terminal:

$ git credential reject

And when you receive a new line, enter in url= followed by the url that showed up previously in the error message. Then use ctrl-c or ctrl-d to quit.


On trying to push again, you will get prompted for authentication and it will work.


Github deprecated standard HTTPS for SSH, so if you continue to get this error, generate a new token and use that token in place of your password.


Handle Windows newlines (CRLF => LF):


Simplify git Call

In your Bash or equivalent rc file, set an alias of g to git. Surprisingly has saved me quite a bit of mistypes and extra key typing.

Use vim as Editor

Good Commit and PR Hygiene

[C]ode review is significantly more effective if devs deliberately review the tests first, and most complex files next[.] -- mhoye[22]


  26. Recovering dropped stashes and commits
  27. How to Write Better Git Commit Messages – A Step-By-Step Guide
Incoming Links

Last modified: 202401040446