by Ben Nadel for future Ben Nadel
The git API is so vast and so esoteric, I can barely manage to keep a fraction of one-percent of it in my head. As such, I wanted to outline a few of the commands that I commonly (and sometimes uncommonly) reach for. This way, when I inevitably get stuck and my brain fails me, I have something that I can refer back to.
Future Ben, you are welcome!
- I want to show the status of the current branch.
- I want to create a new branch that is based on the current branch.
- I want to checkout the previous branch that I was on.
- I want to list the files that have been modified in the current working tree.
- I want to view the changes that were made in a given commit.
- I want to list the files that were changed in a given commit.
- I want to view the changes that were made across multiple commits.
- I want to view the changes that were made in a given file.
- I want to view the contents of a file in a given commit.
- I want to open the contents of a file in a given commit in my editor.
- I want to copy a file from a given commit into my current working tree.
- I want to copy the last commit from another branch into my branch.
- I want to copy an earlier commit from the current branch to the
head. - I want to update the files in the current commit.
- I want to edit the current commit message.
- I want to copy
masterinto my feature branch. - I want to revert the merge of my feature branch into
master. - I want to extract changes that I accidentally made to
master. - I want to undo the changes that I've made to my branch.
- I want to remove unpublished changes from my branch.
- I want to see which branches have already been merged into
master. - I want to see which branches have not yet been merged into
master. - I want to delete my feature branch.
- I want to delete a remote branch.
- I want to update
masterbecause mypushwas rejected. - I want to remove a file from my staging area.
- I want to squash several commits into one (or more) commits.
- I want to squash several commits into one commit without using
rebase. - I want to temporarily set-aside my feature work.
- I want to keep my changes during conflict resolution.
- I want to update the upstream of a repository.
- I want to clone only one branch of a repository.
- I want to clone only the latest commit of one branch of a repository
The status command shows differences between the working tree, the index, and head commit.
git statusIn general, you want to implement new features in short-lived "feature branches" so that changes can be isolated. You can use the checkout command to create a new branch based on the current branch:
git checkout master
# Creates a new branch, `my-feature`, based on `master`.
git checkout -b my-featureIn some of the git commands, the - token refers to the "last branch" that you had checked-out. This makes it very easy to jump back-and-forth between two branches:
git checkout master
git checkout my-feature
# At this point, the `-` refers to the `master` branch.
git checkout -
# At this point, the `-` refers to the `my-feature` branch.
git checkout -The - token can also be used to merge-in the last branch that you had checked-out:
git checkout my-feature
# ... changes to the working tree (your file system).
git add .
git commit -m "Awesome updates."
git checkout master
# At this point, the `-` refers to the `my-feature` branch.
git merge -The - token can also be used to cherry-pick the most recent commit of the last branch that you had checked-out:
git checkout my-feature
# ... changes to the working tree (your file system).
git add .
git commit -m "Minor tweaks."
git checkout master
# At this point, the `-` refers to the `my-feature` branch.
git cherry-pick -By default, when you call git diff, you see all of the content that has been modified in the current working tree (and not yet staged). However, you can use the --stat modifier to simply list the files that have been modified:
git diff --statWhen show is given a branch name, it will default to head - the last or most-recent commit on the given branch:
git checkout master
# Outputs the changes made to the `head` commit of the current (`master`)
# branch.
git show
# Outputs the changes made to the `head` commit of the `my-feature` branch.
git show my-featureYou can also use the show command to target a specific commit that is not the head commit. This can be done with a specific commit hash; or, a relative commit operator like ~:
# Outputs the changes made in the commit with the given hash.
git show 19e771
# Outputs the changes made in in a previous commit of the current (`master`)
# branch.
git show head~ # Show changes in first parent.
git show head~~ # Show changes in first parent's first parent.
git show head~~~ # Show changes in first parent's first parent's first parent.
# Outputs the changes made in a previous commit of the `my-feature` branch.
git show my-feature~
git show my-feature~~
git show my-feature~~~Just as with git diff, you can limit the output of the git show command using the --stat modifier. This will list the files that were changed in the given commit:
# Outputs the list of files changed in the commit with the given hash.
git show 19e771 --statWhile the show command can show you changes in a given commit, you can use the diff command to show changes across multiple commits:
git checkout master
# Outputs the changes between `head~` and `head` of the current branch. If
# only one commit is provided, other commit is assumed to be `head`.
git diff head~
# Outputs the changes between the first commit and the second commit.
git diff head~~~..head~~And, since branch names are really just aliases for commits, you can use a branch name in order to show the changes between one branch and your branch:
git checkout my-feature
# At this point, the following are equivalent and output the changes between
# the `head` commit of the `master` branch and the `head` commit of the
# `my-feature` branch.
git diff master
git diff master..head
git diff master..my-featureBy default, the show command shows all of the changes in a given commit. You can limit the scope of the output by using the -- modifier and identifying a filepath:
# Outputs the changes made to the `README.md` file in the `head` commit of the
# `my-feature` branch.
git show my-feature -- README.md
# Outputs the changes made to the `README.md` file in the `19e771` commit.
git show 19e771 -- README.mdBy default, the show command shows you the changes made to a file in a given commit. However, if you want to view the entire contents of a file as defined at that time of a given commit, regardless of the changes made in that particular commit, you can use the : modifier to identify a filepath:
# Outputs the contents of the `README.md` file as defined in the `head` commit
# of the `my-feature` branch.
git show my-feature:README.md
# Outputs the contents of the `README.md` file as defined in the `19e771`
# commit.
git show 19e771:README.mdSince you're working on the command-line, the output of any git-command can be piped into another command. As such, you can use the show command to open a previous commit's file-content in your editor or viewer of choice:
# Opens the `README.md` file from the `head` commit of the `my-feature` branch
# in the Sublime Text editor.
git show my-feature:README.md | subl
# Opens the `README.md` file from the `19e771` commit in the `less` viewer.
git show 19e771:README.md | lessNormally, the checkout command will update the entire working tree to point to a given commit. However, you can use the -- modifier to copy (or checkout) a single file from the given commit into your working tree:
git checkout my-feature
# While staying on the `my-feature` branch, copy the `README.md` file from
# the `master` branch into the current working tree. This will overwrite the
# current version of `README.md` in your working tree.
git checkout master -- README.mdWhen you don't want to merge a branch into your current working tree, you can use the cherry-pick command to copy specific commit-changes into your working tree. Doing so creates a new commit on top of the current branch:
git checkout master
# Copy the `head` commit-changes of the `my-feature` branch onto the `master`
# branch. This will create a new `head` commit on `master`.
git cherry-pick my-featureSometimes, after you understand why reverted code was breaking, you want to bring the reverted code back into play and then fix it. You could use the revert command in order to "revert the revert"; but, such terminology is unattractive. As such, you can cherry-pick the reverted commit to bring it back into the head where you can then fix it and commit it:
git checkout master
# Assuming that `head~~~` and `19e771` are the same commit, the following are
# equivalent and will copy the changes in `19e771` to the `head` of the
# current branch (as a new commit).
git cherry-pick head~~~
git cherry-pick 19e771If you want to make changes to a commit after you've already committed the changes in your current working tree, you can use the --amend modifier. This will add any staged changes to the existing commit.
git commit -m "Woot, finally finished!"
# Oops, you forgot a change. Edit the file and stage it.
# ... changes to the working tree (your file system).
git add oops.txt
# Adds the currently-staged changes (oops.txt) to the current commit, giving
# you a chance to update the commit message.
git commit --amendIn addition to adding files to the current commit, the --amend modifier can also be used to change the current commit message:
git add .
git commit -m "This is greet."
# Oh noes! You misspelled "great". You can edit the current commit message:
git commit --amend -m "This is great."Note that if you omit the -m message portion of this command, you will be able to edit the commit message in your configured editor.
At first, you may be tempted to simply merge your master branch into your feature branch, but doing so will create an unattactive, non-intuitive, Frankensteinian commit tree. Instead, you should rebase your feature branch on master. This will ensure that your feature commits are cleanly colocated in the commit tree and align more closely with a human mental model:
git checkout my-feature
# This will unwind the commits specific to the `my-feature` branch, pull in
# the missing `master` commits, and then replay your `my-feature` commits.
git rebase masterOnce your my-feature branch has been rebased on master, you could then, if you wanted to, perform a --ff-only merge ("fast forward only") of your feature branch back into master:
git checkout my-feature
git rebase master
# Fast-forward merge of `my-feature` changes into `master`, which means there
# is no creation of a "merge commit" - your `my-features` changes are simply
# added to the top of `master`.
git checkout master
git merge --ff-only my-featureThat said, when you're working on a team where everyone uses a different git workflow, you will definitely want a "merge commit". This way, multi-commit merges can be easily reverted. To force a "merge commit", you can use the --no-ff modifier ("no fast forward"):
# Get the `my-feature` branch ready for merge.
git checkout my-feature
git rebase master
# Merge the `my-feature` branch into `master` creating a merge-commit.
git checkout master
git merge --no-ff my-featureNow, if the merge needs to be reverted, you can simply revert the "merge commit" and all commits associated with the merge will be reverted.
If you performed a --ff-only merge of your feature branch into master, there's no "easy" solution. You either have to reset the branch to an earlier commit (rewriting history); or, you have to revert the individual commits in the merge.
If, however, you performed a --no-ff merge ("no fast forward") that created a "merge commit", all you have to do is revert the merge commit:
git checkout master
# Merge the feature branch in, creating a "merge commit".
git merge --no-ff my-feature
# On noes! You didn't mean to merge that in. Assuming that the "merge commit"
# is now the `head` of `master`, you can revert back to the commit's fist
# parent, the `master` branch: -m 1. And, since `head` and `master` are the
# same commit, the following are equivalent:
git revert -m 1 head
git revert -m 1 masterSometimes, after you've finished working on your feature branch, you execute git checkout master, only find that you've been accidentally working on master the whole time (error: "Already on 'master'"). To fix this, you can checkout a new branch and reset your master branch:
git checkout master
# > error: Already on 'master'
# While on the `master` branch, create the `my-feature` branch as a copy of
# the `master` branch. This way, your `my-feature` branch will contain all of # your recent changes.
git checkout -b my-feature
# Now that your changes are safely isolated, get back into your `master`
# branch and `reset` it with the `--hard` modifier so that your local index
# and file system will match the remote copy.
git checkout master
git reset --hard origin/masterIf you've edited some files and then change your mind about keeping those edits, you can reset the branch using the --hard modifier. This will update the working tree - your file structure - to match the structure of the last commit on the branch (head).
Caution: You will lose data when using the --hard option.
git checkout my-feature
# ... changes to the working tree (your file system).
git add .
# Remove the file from staging AND remove the changes from the file system.
git reset --hardIf you call git reset without the --hard option, it will reset the staging to match the head of the branch, but it will leave your file system in place. As such, you will be left with "unstaged changes" that can be modified and re-committed.
If you've committed changes to the local copy of a remote (ie, published) branch, but you want to undo those changes, you can reset the local branch to match the remote branch:
git checkout my-feature
# Update the remote copy of the `my-feature` branch in order to make sure that
# you are working with the most up-to-date remote content.
git fetch origin my-feature
# Now, reset the local copy of `my-feature` to match the published copy. This
# will update your index and your local file system to match the published
# version of `my-feature`.
git reset --hard origin/my-featureFrom any branch, you can locate the merged-in branches (that can be safely deleted) by using the --merged modifier:
git checkout master
# List all of the local branches that have been merged into `master`. This
# command will be relative to the branch that you have checked-out.
git branch --mergedFrom any branch, you can locate the unmerged branches by using the --no-merged modifier:
git checkout master
# List all of the local branches that have NOT YET been merged into `master`.
# This command will be relative the branch you have checked-out.
git branch --no-mergedAfter you're merged your feature branch into master, you can delete your feature branch using the branch command:
# Merge your `my-feature` branch into `master` creating a "merge commit."
git checkout master
git merge --no-ff my-feature
# Safely delete the merged-in `my-feature` branch. The `-d` modifier will
# error-out if the given branch has not yet been merged into the current
# branch.
git branch -d my-featureIf you want to abandon a feature branch, you can use the -D modifier to force-delete it even if it has not yet been merged into master:
git checkout master
# Force-delete the `my-feature` branch even though it has not been merged
# into the `master` branch.
git branch -D my-featureWhen you delete a branch using git branch -d, it deletes your local copy; but, it doesn't delete the remote copy from your origin (ex, GitHub). To delete the remote copy, you have to push the branch using the : prefix:
git checkout master
# Safely delete the local copy of your `my-feature` branch. The `-d` modifier
# will error-out if the given branch has not been fully-merged into `master`.
git branch -d my-feature
# Delete the remote copy of the `my-feature` branch from the origin. The `:`
# prefix sends this through as a "delete" command for the given branch.
git push origin :my-featureIf you've committed changes to master but you forgot to pull recent changes from the remote master branch, your next push will be rejected with an error that looks like, "Updates were rejected because the tip of your current branch is behind its remote counterpart". To fix this, you can use the --rebase modifier:
git checkout master
git merge --no-ff my-feature
# Oh noes! You forgot to pull in the latest remote copy of `master` before you
# merged your `my-feature` commits. No problem, just `--rebase` your local
# `master` on the remote branch. This will move your local changes to the tip
# of the `master` branch.
git pull --rebase
# Now that you've pulled-in the remote changes, you should be able to push
# your updated `master` branch.
git push origin masterIf you accidentally added too many files to the staging area (in preparation for a git commit), you can use the rm --cached command to remove them from the staging area but keep them in the working tree:
git add .
# Oh noes! You didn't mean to add all of the files to the staging area. You
# can remove some of the staged files using the `--cached` modifier:
git rm --cached secrets.configIf you accidentally added an entire directory to the staging area, you can add the -r modifier to recursively apply the rm command:
git add .
# Oh noes! You didn't mean to add the ./config directory. You can recursively
# remove it with the `-r` modifier:
git rm --cached -r config/.When you rm files using --cached, they will remain in your working tree and will become "unstaged" changes.
Your commit history is a representation or your personality. It is a manifestation of your self-respect and the respect you have for your team. As such, you will often need to rewrite your feature branch's history before merging it into master. This allows you to get rid of intermediary commit messages like, "still working on it." and "Meh, missed a bug.". To do this, you can perform an "interactive rebase".
The "interactive rebase" gives you an opportunity to indicate how intermediary commits should be rearranged. Some commits can be "squashed" (combined) together. Others can omitted (remove). And others can be edited. When performing an interactive rebase, you have to tell git which commit to use as the starting point. If you're on an up-to-date feature branch, the starting point should be master.
# Create the `my-feature` branch based on `master`.
git checkout master
git checkout -b my-feature
# ... changes to the working tree (your file system).
git add .
git commit -m "Getting close."
# ... changes to the working tree (your file system).
git add .
git commit -m "Missed a bug."
# ... changes to the working tree (your file system).
git add .
git commit -m "Uggggg! Why is this so hard?"
# ... changes to the working tree (your file system).
git add .
git commit -m "Woot, finally got this working."
# At this point, your commit history is sloppy and would bring much shame on
# your family if it ended-up in `master`. As such, you need to squash the
# commits down into a single commit using an interactive rebase. Here, you're
# telling `git` to use the `master` commit as the starting point:
git rebase -i masterAs this point, git will open up an editor that outlines the various commits and asks you how you want to rearrange them. It should look something like this, with the commits listed in ascending order (oldest first):
pick 27fb3d2 Getting close.
pick e8214df Missed a bug.
pick ce5ed14 Uggggg! Why is this so hard?
pick f7ee6ab Woot, finally got this working.
# Rebase b0fced..f7ee6ab onto b0fced (4 commands)
#
# Commands:
# p, pick = use commit
# r, reword = use commit, but edit the commit message
# e, edit = use commit, but stop for amending
# s, squash = use commit, but meld into previous commit
# f, fixup = like "squash", but discard this commit's log message
# x, exec = run command (the rest of the line) using shell
# d, drop = remove commitAt this point, you can identify the later commits as needing to be squashed (s) down into the oldest commit (the one you are picking):
pick 27fb3d2 Getting close.
s e8214df Missed a bug.
s ce5ed14 Uggggg! Why is this so hard?
s f7ee6ab Woot, finally got this working.Once saved, git will prompt you to provide a cleaner commit message. And, once provided, your four shameful commits will be squashed down into a single, cohesive, meaningful commit.
In the vast majority of cases, if your git workflow is clean and true and your feature branch is short-lived, an interactive rebase should be straightforward and pain-free. However, once you make the mistake of periodically merging master into your feature branch, you are inviting a world of hurt. In such a case, you can use the merge command with the --squash modifier as an escape hatch.
When you run git merge --squash, you copy the file-changes from one branch into another branch without actually copying the commit meta-data. Instead, the changes are brought-over as staged changes on top of the current branch. At that point, you can commit all the staged changes as a single commit.
Assuming your my-feature branch needs to be "fixed", you can use the following workflow:
# Assuming that the `my-feature` branch is the branch that needs to be fixed,
# start off my renaming the branch as a backup (the `-m` modifier performs a
# rename):
git checkout my-feature
git branch -m my-feature-backup
# Now, checkout the `master` branch and use it to create a new, clean
# `my-feature` branch starting point.
git checkout master
git checkout -b my-feature
# At this point, your `my-feature` branch and your `master` branch should be
# identical. Now, you can "squash merge" your old feature branch into the new
# and clean `my-feature` branch:
git merge --squash my-feature-backup
# All the file-changes should now be in the `my-feature` branch as staged
# edits ready to be committed.
git commit -m "This feature is done and awesome."
# Delete the old backup branch as it is no longer needed. You will have to
# force delete (`-D`) it since it was never merged into `master`.
git branch -D my-feature-backupASIDE: You should almost never need to do this. If you find yourself having to do this a lot; or, you find yourself dealing with a lot of "conflict resolution", you need to reevaluate your
gitworkflow. Chances are, your feature branches are living way too long.
The life of a developer is typically "interrupt driven". As such, there is often a need to briefly set aside your current work in order to attend to more pressing matters. In such a case, it is tempting to use git stash and git stash pop to store pending changes. But, do not do this. Stashing code requires unnecessary mental overhead. Instead, simply commit the changes to your current feature branch and then perform an interactive rebase later on in order to clean up your commits:
# Oh noes! Something urgent just came up - commit your changes to your feature
# branch and then go attend to the more pressing work.
git add .
git commit -m "Saving current work - jumping to more urgent matters."
git checkout masterNow, you never have to remember where those pending changes are. This guarantees that you won't lose your work.
If you were working directly on master when urgent matters came up, you can still avoid having to use git stash. To keep your work, simply checkout a new branch and the commit your pending changes to that new branch:
# Oh noes! Something urgent just came up - checkout a new branch. This will
# move all of your current work (staged and unstaged) over to the new branch.
git checkout -b temp-work
# Commit any unstaged changes.
git add .
git commit -m "Saving current work - jumping to more urgent matters."
git checkout masterNow, your master branch should be back in a pristine state and your temp-work branch can be continued later.
If your Git workflow is healthy - meaning that you have short-lived feature branches - conflicts should be very few and far between. In fact, I would assert that getting caught-up in frequent conflicts is an indication that something more fundamental to your workflow is primed for optimization.
That said, conflicts do happen. And, if you want to resolve a conflict by selecting "your version" of a file, you can use git checkout --theirs in a merge conflict, a cherry-pick conflict, and a rebase conflict.
In a merge conflict, --theirs indicates the branch being merged into the current context:
git checkout master
git merge --no-ff my-feature
# Oh noes! There is a conflict in "code.js". To keep your version of the
# code.js file, you can check-it-out using --theirs and the file path:
git checkout --theirs code.js
git add .
git merge --continueSimilarly, in a cherry-pick conflict, --theirs indicates the branch being cherry-picked into the current context:
git checkout master
git cherry-pick --no-ff my-feature
# Oh noes! There is a conflict in "code.js". To keep your version of the
# code.js file, you can check-it-out using --theirs and the file path:
git checkout --theirs code.js
git add .
git merge --continueIn a rebase conflict, --theirs indicates the branch that is being replayed on top of the current context (See Aside):
git checkout my-feature
git rebase master
# Oh noes! There is a conflict in "code.js". To keep your version of the
# code.js file, you can check-it-out using --theirs and the file path:
git checkout --theirs code.js
git add .
git rebase --continue$ git remote set-url origin https://github.com/USERNAME/REPOSITORY.git
List your existing remotes in order to get the name of the remote you want to change. `` $ git remote -v
origin git@github.com:USERNAME/REPOSITORY.git (fetch) origin git@github.com:USERNAME/REPOSITORY.git (push) ``
ASIDE: Using
--theirsin arebasecan seem confusing because you are already in "your" feature branch. As such, it would seem logical that your version of the code would be targeted with--ours, not--theirs. However, arebaseoperates kind of like a series ofcherry-pickoperations. You can think of arebaseas doing the following:
- Check-out an earlier, common commit between your feature branch and the target branch.
- Cherry-pick your feature branch commits onto the earlier commit.
- Replace your feature branch with this temporary branch
With this mental model, "your" version - targeted using
--theirs- is the version being cherry-picked into the "current context" (the temporary branch).
git reset -- hard commit hash
git push push -f
After resetting you have to make a force pull to reset the head to that particular commit
##i-want-to-clone-only-one-branch
git clone -b master --single-branch https://gitlab.com/repo.git eposta
##i-want-to-clone-only-one-branch-and-the-latest-commit-only
git clone -b master --single-branch --depth 1 https://gitlab.com/repo.git eposta