$ brew tap nvie/tap
$ brew install nvie/tap/git-toolbelt
If not using Homebrew, you will need to have GNU coreutils
installed, for the realpath
utility. Git for Windows users see #29.
Helper tools to make everyday life with Git much easier. Commands marked with ⭐️ are my personal favorites and are commands I use almost every day.
Everyday helpful commands:
Statistics:
Commands to help novices out:
Commands that simplify scripting. These commands typically only return exit codes and have no output.
Advanced usage:
Returns the name of the current branch, if any. Why doesn't this come with git?
$ git current-branch
master
Alias to git rev-parse --abbrev-ref HEAD
.
Returns the name of the default main branch for this repository. Historically
master
, but could also be main
if you've changed the default branch name.
Since there's no way of reliably telling what the default branch name is for
a repo, this script will probe for the existence of local branches named either
main
or master
. The first one found is used.
$ git main-branch
master
Returns the SHA value for the specified object, or the current branch head, if nothing is provided.
$ git sha <some-object>
Typical example:
$ git sha HEAD
f688d7543c5d52f5f78b3db1b0dd1616059299a4
$ git sha -s HEAD
f688d75
Shows the commit SHA for the latest commit.
Returns a list of locally modified files. In contrast to git status, it does not include any detailed file status, and never includes non-existing files.
This makes it ideal for the following use-case:
$ vim (git modified)
If you want to locally modified files that are already staged, too, use:
$ vim (git modified -i)
Like git-modified, but for printing a list of files that have been modified since master (or whatever commit specified). In contrast to git status, it does not include any detailed file status, and never includes non-existing files.
Opens all files modified on your branch (since you branched off master
).
$ vim (git modified-since)
Adds a commit with a message of only ---'s, so that it visually separates commits in the history. This is incredibly useful when doing more complex rebase operations. (They should be used as a temporary measure, and ideally taken out of the history again when done rebasing.)
Inspired by Magit's spinoff
command. Creates and checks out
a new branch starting at and tracking the current branch. That
branch in turn is reset to the last commit it shares with its
upstream. If the current branch has no upstream or no unpushed
commits, then the new branch is created anyway and the previously
current branch is not touched.
This is useful to create a feature branch after work has already began on the old branch (likely but not necessarily "master").
Pushed the current branch out to origin
, and makes sure to setup tracking of
the remote branch. Shorthand for git push -u origin <current-branch>
.
Accepts options, too, so you can use
$ git push-current -f
to force-push.
Tests if HEAD
is pointing to a branch head, or is detached.
Shows the differences made on the current branch, compared to the main branch (or the given branch).
Returns a list of local or remote branches, but contrary to Git's default commands for this, returns them machine-processable. In the case of remote branches, can be asked to return only the branches in a specific remote.
For git active-branches
, a branch is deemed "active" if its head points to
a commit authored in the last 3 weeks, by default. An arbitrary date can be
specified using either git active-branches -s <date>
or -a <date>
(mnemonic: "since" or "after"), using any date format
supported by git log
.
Tests if the given local branch, remote branch, or tag exists.
Returns a list of local branches, ordered by recency:
$ git recent-branches
foo
master
bar
qux
Print the name of the remote tracking branch of the current or given local branch name, if one exists. Errors otherwise.
Returns a list of commits that are still in your local repo, but haven't been
pushed to origin
. git has-local-commits
is the scriptable equivalent that
only returns an exit code if such commits exist.
Tests if X is merged into Y:
$ git contains X Y # does X contain Y?
$ git is-ancestor X Y # is X an ancestor of Y?
CAVEAT:
Even though they might look like opposites, X contains Y
does not mean not (X is-ancestor Y)
, since (1) X and Y can point to the same commit, or the
branches may have no common history and thus be unrelated completely.
Mimics the index / staging area to match the working tree exactly. Adds files, removes files, etc.
Alias to git add --all
.
Unstages everything. Leaves the working tree intact.
Alias to git reset HEAD
.
Ever created a merge accidentally, or decided that you didn't want to merge
after all? You can undo the last merge using git undo-merge
.
Ever committed too soon, or by accident? Or on the wrong branch? You can now undo your last commit and you won't lose any data. All the changes in the commit will be staged (like right before the commit) and the commit itself is gone.
Deletes all branches that have already been merged into the main branch. Keeps other branches lying around. Removes branches both locally and in the origin remote. Will be most conservative with deletions.
Amend all local staged changes into the last commit. Ideal for fixing typo's, when you don't want to re-edit the commit message.
$ git commit -m "Something cool."
$ vim somefile.txt # fix typo
$ git add somefile.txt
$ git fixup # merge this little change back into the last commit
Interactively lets you pick a commit to fixup with. (Uses fzf
for the
interactive picking. Use brew install fzf
to install this tool separately.)
Use -r
to trigger an interactive rebase right afterwards.
Convenience command for quickly switching to a branch <name>
. If such local
branch does not exist, but there is a remote branch named origin/<name>
, then
a local branch is created and the remote is tracked.
Say you want to rebuild your last commit, but want to keep the commit message. git delouse empties the last commit on the current branch and places all changes back into the working tree.
Since the commit remains in history, you can now rebuild the commit by "git amend"'ing or "git fixup"'ing, instead of making new commits.
Splits the last commit into N+1 commits, where N is the number of files in the last commit. The first commit is an empty commit with the original commit message and author details, and the following commits add (or delete) one file each, keeping the subject line of the original commit message.
After running git shatter-by-file
, you'll typically want to run git rebase --interactive
to start fixing up changes to files, etc. For that purpose, the
original commit message is kept in there (in the empty first commit), so make
sure to use it.
Splits the last commit into 2 or more commits. Takes one or more regex values
(which are fed to grep -Ee
), and will split the last commit by file paths
matching each of the regexes.
For example:
$ git cleave client/ server/
Will split the last commit into 2 (or 3) commits. The first one will contain
all the files containing client/
, the second will contain all the files
matching server/
. If there are files that don't match either of those, then
a 3rd commit will be made with the "remainder".
Another example:
$ git cleave '.*\.js$'
This will split off all Javascript files from a commit.
Ever been on a branch and really wanted to quickly commit a change to a different branch? Given that this is possible without merge conflicts, git commit-to will allow you to do so, without checking out the branch necessarily.
$ git branch
master
* mybranch
$ git status
M foo.txt
M bar.txt
$ git add foo.txt
$ git commit-to master -m "Add foo to master."
$ git add bar.txt
$ git commit -m "Add bar to mybranch."
Every been on a branch, just made a commit, but really want that commit available on other branches as well? You can now cherry-pick this commit to any branch, staying on the current branch. (Given the change won't lead to a merge conflict.)
$ git branch
master
* mybranch
$ git add foo.txt
$ git commit -m "Really useful thing."
$ git cherry-pick-to master HEAD
$ git branch # did not switch branches
master
* mybranch
Helper function that determines whether the current directory has a Git repo
associated to it. Scriptable equivalent of git repo
.
git root
prints the root location of the working tree.
$ cd /path/to/worktree
$ cd some/dir/in/worktree
$ pwd
/path/to/worktree/some/project/dir
$ git root
/path/to/worktree
git repo
prints the location of the Git directory, typically .git
, but
could differ based on your setup. Will return with a non-zero exit code if not
in a repo.
$ cd /path/to/my/worktree
$ git repo
.git
$ cd /tmp
$ git repo
fatal: Not a git repository (or any of the parent directories): .git
git initial-commit
prints the initial commit for the repo.
$ git initial-commit
48c94a6a29e9e52ab63ce0fab578101ddc56a04f
Helper function that determines whether there are local changes in the working tree, by returning a 0 (local changes) or 1 (no local changes) exit code.
Don't care about your local working copy's state and really want to revert back to whatever is recorded in the history? git drop-local-changes lets you do this.
This covers aborting rebases, undoing partial merges, resetting the index and removing any unknown local files from the work tree. Anything that is already committed remains safe.
??? issue a git pull, too? Typical beginners will want this.
The stash behaviour you (probably) always wanted. This actually stashes everything what's in your index, in your working tree, and even stashes away your untracked files, leaving a totally clean working tree.
Using "git stash pop" will recover all changes, including index state, locally modified files, and untracked files.
Updates all local branch heads to the remote's equivalent. This is the same as checking out all local branches one-by-one and pulling the latest upstream changes. Will only update if a pull succeeds cleanly (i.e. is a fast-forward pull).
This trio of subcommands makes it easy to inspect merge status of local branches. Use them to check whether any local branches have or haven't been merged into the target branch (defaults to master).
git-merge-status is a useful command that presents both lists in a single overview (not for machine processing).
This command, "git branches-containing [
git-branches-containing is useful to see if a branch has been merged, and, if so, which releases contain the feature/fix (if you use release branches).
Shows contribution stats for the given committer, like "most productive day", "most productive hour", "average commit size", etc.
Don't care about your local working copy's state and really want to switch to another branch? git force-checkout lets you do this.
Switching branches can be prevented by git. For good reasons, mostly. Git is designed to prevent you from losing data potentially. Examples are there are local unmerged files, or some files that would be overwritten by doing the checkout.
By using force-checkout you basically give git the finger, and check out a branch anyway. You do agree to lose data when using this command.
$ git checkout master
error: Your local changes to the following files would be overwritten by checkout:
foo/bar.txt
Please, commit your changes or stash them before you can switch branches.
Aborting
$ git force-checkout master
Switched to branch 'master'
Generates a summary for all local branches that will merge uncleanly—i.e. will lead to merge conflicts later on.
$ git branch
develop
* mybranch
master
other-branch
$ git conflicts
develop... merges cleanly
master... merges cleanly
other-branch... CONFLICTS AHEAD
Git supports marking files "skip worktree", meaning any change in the file locally will not be shown in status reports, or be added when you stage all files. This feature can be useful to toggle some switches locally, or experiment with different settings, without running the risk of accidentally committing this local data (that should remain untouched in the repo).
Notice that status reports won't show these files anymore, so it's also easily to lose track of these marked assumptions, and you probably run into weird issues if you don't remember this. (This is the reason why I put these scripts in the "advanced" category.)
Basic usage:
$ git status
M foo.txt
M bar.txt
M qux.txt
$ git skip foo.txt
$ git status
M bar.txt
M qux.txt
$ git show-skipped
foo.txt
$ git commit -am 'Commit everything.'
$ git status
nothing to commit, working directory clean
$ git is-clean && echo "clean" || echo "not clean"
not clean
$ git unskip -a
$ git status
M foo.txt
As you can see, git-is-clean
is aware of any lurking "skipped" files, and
won't report a clean working tree, as these assumed unchanged files often block
the ability to check out different branches.
Commits all local changes under a commit message of "WIP". Great for quickly
creating "savepoint" commits. If there is a mix of staged changes, and
unstaged changes, and new files, will commit each of these as a separate
commit, all titled "WIP". Effectively, running git-wip
once will potentially
lead to anywhere between 0 and 3 "WIP" commits being created.