Git, the distributed revision-control system that's become the default tool for source code control in the open source world, turns 13 on April 7. One of the more frustrating things about using Git is how much you need to know to use it effectively. This can also be one of the more awesome things about using Git, because there's nothing quite like discovering a new tip or trick that can streamline or improve your workflow.
In honor of Git's 13th birthday, here are 13 tips and tricks to make your Git experience more useful and powerful, starting with some basics you might have overlooked and scaling up to some real power-user tricks!
Editor's Note: We have updated this article to include 13 tips for 13 years of Git; we previously misstated that Git was 12 years old.
1. Your ~/.gitconfig file
The first time you tried to use the git command to commit a change to a repository, you might have been greeted with something like this:
*** Please tell me who you are.
Run
git config --global user.email "you@example.com"
git config --global user.name "Your Name"
to set your account's default identity.What you might not have realized is that those commands are modifying the contents of ~/.gitconfig, which is where Git stores global configuration options. There are a vast array of things you can do via your ~/.gitconfig file, including defining aliases, turning particular command options on (or off!) on a permanent basis, and modifying aspects of how Git works (e.g., which diff algorithm git diff uses or what type of merge strategy is used by default). You can even conditionally include other config files based on the path to a repository! See man git-config for all the details.
2. Your repo's .gitconfig file
In the previous tip, you may have wondered what that --global flag on the git config command was doing. It tells Git to update the "global" configuration, the one found in ~/.gitconfig. Of course, having a global config also implies a local configuration, and sure enough, if you omit the --global flag, git config will instead update the repository-specific configuration, which is stored in .git/config.
Options that are set in the .git/config file will override any setting in the ~/.gitconfig file. So, for example, if you need to use a different email address for a particular repository, you can run git config user.email "also_you@example.com". Then, any commits in that repository will use your other email address. This can be super useful if you work on open source projects from a work laptop and want them to show up with a personal email address while still using your work email for your main Git configuration.
Pretty much anything you can set in ~/.gitconfig, you can also set in .git/config to make it specific to the given repository. In any of the following tips, when I mention adding something to your ~/.gitconfig, just remember you could also set that option for just one repository by adding it to .git/config instead.
3. Aliases
Aliases are another thing you can put in your ~/.gitconfig. These work just like aliases in the command shell—they set up a new command name that can invoke one or more other commands, often with a particular set of options or flags. They're super useful for longer, more complicated commands you use frequently.
You can define aliases using the git config command—for example, running git config --global --add alias.st status will make running git st do the same thing as running git status—but I find when defining aliases, it's frequently easier to just edit the ~/.gitconfig file directly.
If you choose to go this route, you'll find that the ~/.gitconfig file is an INI file. INI is basically a key-value file format with particular sections. When adding an alias, you'll be changing the [alias] section. For example, to define the same git st alias as above, add this to the file:
[alias]
st = status(If there's already an [alias] section, just add the second line to that existing section.)
4. Aliases to shell commands
Aliases aren't limited to just running other Git subcommands—you can also define aliases that run other shell commands. This is a fantastic way to deal with a recurring, infrequent, and complicated task: Once you've figured out how to do it once, preserve the command under an alias. For example, I have a few repositories where I've forked an open source project and made some local modifications that don't need to be contributed back to the project. I want to keep up-to-date with ongoing development work in the project but also maintain my local changes. To accomplish this, I need to periodically merge the changes from the upstream repo into my fork—which I do by using an alias I call upstream-merge. It's defined like this:
upstream-merge = !"git fetch origin -v && git fetch upstream -v && git merge upstream/master && git push"The ! at the beginning of the alias definition tells Git to run the command via the shell. This example involves running a number of git commands, but aliases defined in this way can run any shell command.
(Note that if you want to copy my upstream-merge alias, you'll need to make sure you have a Git remote named upstream pointed at the upstream repository you've forked from. You can add this by running git remote add upstream <URL to repo>.)
5. Visualizing the commit graph
If you work on a project with a lot of branching activity, sometimes it can be difficult to get a handle on all the work that's happening and how it's all related. Various GUI tools allow you to get a picture of different branches and commits in what's called the "commit graph." For example, here's a section of one of my repositories visualized with the GitLab commit graph viewer:
opensource.com
If you're a dedicated command-line user or somebody who finds switching tools to be distracting, it's nice to get a similar view of the commit graph from the command line. That's where the --graph argument to the git log command comes in:
opensource.com
This is the same section of the same repo visualized with the following command:
git log --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit --date=relativeThe --graph option adds the graph to the left side of the log, --abbrev-commit shortens the commit SHAs, --date=relative expresses the dates in relative terms, and the --pretty bit handles all the other custom formatting. I have this aliased to git lg, and it is one of my top 10 most frequently run commands.
6. A nicer force-push
Sometimes, as hard as you try to avoid it, you'll find that you need to run git push --force to overwrite the history on a remote copy of your repository. You may have gotten some feedback that caused you to do an interactive rebase, or you may simply have messed up and want to hide the evidence.
One of the hazards with force pushes happens when somebody else has made changes on top of the same branch in the remote copy of the repository. When you force-push your rewritten history, those commits will be lost. This is where git push --force-with-lease comes in—it will not allow you to force-push if the remote branch has been updated, which will ensure you don't throw away someone else's work.
7. git add -N
Have you ever used git commit -a to stage and commit all your outstanding changes in a single move, only to discover after you've pushed your commit that git commit -a ignores newly added files? You can work around this by using the git add -N (think "notify") to tell Git about newly added files you'd like to be included in commits before you actually commit them for the first time.
8. git add -p
A best practice when using Git is to make sure each commit consists of only a single logical change—whether that's a fix for a bug or a new feature. Sometimes when you're working, however, you'll end up with more than one commit's worth of change in your repository. How can you manage to divide things up so that each commit contains only the appropriate changes? git add --patch to the rescue!
This flag will cause the git add command to look at all the changes in your working copy and, for each one, ask if you'd like to stage it to be committed, skip over it, or defer the decision (as well as a few other more powerful options you can see by selecting ? after running the command). git add -p is a fantastic tool for producing well-structured commits.
9. git checkout -p
Similar to git add -p, the git checkout command will take a --patch or -p option, which will cause it to present each "hunk" of change in your local working copy and allow you to discard it—basically reverting your local working copy to what was there before your change.
This is fantastic when, for example, you've introduced a bunch of debug logging statements while chasing down a bug. After the bug is fixed, you can first use git checkout -p to remove all the new debug logging, then you git add -p to add the bug fix. Nothing is more satisfying than putting together a beautiful, well-structured commit!
10. Rebase with command execution
Some projects have a rule that each commit in the repository must be in a working state—that is, at each commit, it should be possible to compile the code or the test suite should run without failure. This is not too difficult when you're working on a branch over time, but if you end up needing to rebase for whatever reason, it can be a little tedious to step through each rebased commit to make sure you haven't accidentally introduced a break.
Fortunately, git rebase has you covered with the -x or --exec option. git rebase -x <cmd> will run that command after each commit is applied in the rebase. So, for example, if you have a project where npm run tests runs your test suite, git rebase -x npm run tests would run the test suite after each commit was applied during the rebase. This allows you to see if the test suite fails at any of the rebased commits so you can confirm that the test suite is still passing at each commit.
11. Time-based revision references
Many Git subcommands take a revision argument to specify what part of the repository to work on. This can be the SHA1 of a particular commit, a branch name, or even a symbolic name like HEAD (which refers to the most recent commit on the currently checked out branch). In addition to these simple forms, you can also append a specific date or time to mean "this reference, at this time."
This becomes very useful when you're dealing with a newly introduced bug and find yourself saying, "I know this worked yesterday! What changed?" Instead of staring at the output of git log trying to figure out what commit was changed when, you can simply run git diff HEAD@{yesterday}, and see all the changes that have happened since then. This also works with longer time periods (e.g., git diff HEAD@{'2 months ago'}) as well as exact dates (e.g., git diff HEAD@{'2010-01-01 12:00:00'}).
You can also use these date-based revision arguments with any Git subcommand that takes a revision argument. Find full details about which format to use in the man page for gitrevisions.
12. The all-seeing reflog
Have you ever rebased away a commit, then discovered there was something in that commit you wanted to keep? You may have thought that information was lost forever and would need to be recreated. But if you committed it in your local working copy, it was added to the reference log (reflog), and you should still be able to access it.
Running git reflog will show you a list of all the activity for the current branch in your local working copy and also give you the SHA1 of each commit. Once you've found the commit you rebased away, you can run git checkout <SHA1> to check out that commit, copy any information you need, and run git checkout HEAD to return to the most recent commit in the branch.
13. Cleaning up after yourself
Whoops! It turns out my basic math skills aren't quite up to the same level as my Git ones. Git was originally released in 2005, which means it turns 13 this year, not 12. To make up for the mistake, here's a 13th tip to bring us up to a baker's dozen.
If you use a branching-based workflow, overtime on a long-lived project, unless you're fastidious about cleaning up as each branch is merged, eventually you will end up with a bunch of branches. This can make finding a branch of interest difficult, and you won't be able to see the forest for the... branches, if you will. Even worse, if you have a number of active branches in play, it can be really tedious to figure out whether a branch has been merged (and can be safely deleted) or if it still remains unmerged and should be left alone. Fortunately, Git has your back here: Just run git branch --merged to get a list of branches that have been merged into your current branch, or git branch --merged to find ones that have been merged into some other branch. By default, this will list branches in your local working copy, but if you include --remote or -r into the command, it will also list merged branches that only exist on the remote.
Important note: if you plan on using the output from git branch --merged to clean up those merged branches, you should be aware it will also include the current branch in the output (because, after all, the current branch is merged to the current branch!). Make sure you exclude that branch from anything destructive (or if you forgot to, see tip #12 to learn how the reflog can help you get your branch back, hopefully...)
That's all folks!
Hopefully at least one of these tips has taught you something new about Git, a 13-year-old project that's continuing to innovate and add new features. What's your favorite Git trick?
7 Comments