I’ve seen the following tweet about git making its way around Twitter recently:
And it’s true, you can do most of your work with those commands.
What if you want to fork and contribute to an open source project (perhaps even this site!) on GitHub, GitLab or BitBucket though? You’re going to need a few more commands so that you can keep your fork up to date, namely remote, fetch and merge. Let’s see how they are used.
Fork away
When you fork a project, you make a copy of it under your own namespace. To work with the project you then clone the repository to your own machine. Let’s use a repo I have forked as an example: Twilio’s Node.js package. After forking, we clone the repo:
git clone git@github.com:philnash/twilio-node.git
We can now see the first use case for remote. Run git remote in the twilio-node directory and we see the following:
git remote
origin
OK, that’s not too useful, it just shows we have one remote repo called “origin”. Running it again with the --verbose flag shows a bit more information:
git remote --verbose
origin git@github.com:philnash/twilio-node.git (fetch)
origin git@github.com:philnash/twilio-node.git (push)
That’s better, this shows that the remote repository is the one we cloned and that we can fetch from and push to it.
Making changes
When contributing something back to the repo, it is easiest to do so on a branch. That keeps the master branch clean and makes it easy to keep up to date. Setting up your changes is covered by the six commands Cory listed in his tweet; you create a new branch with branch, checkout the branch, make the changes you want, commit as many times as necessary and finally push the branch to the origin, your fork. Then you can create your pull request and hope it gets accepted.
Sometime later down the line you find you want to make another pull request, but the original repo has moved on. You need to update your fork so that you’re working with the latest code. You could delete your fork and go through the process of forking and cloning the repo again, but that’s a lot of unnecessary work. Instead, add the upstream repository as another remote for the repo and work with it from the command line.
Adding the upstream
To do this, we use the remote command again, this time to add a new remote. Find the original repo’s URL and add it as a new remote. By convention this remote is called “upstream” though you can call it whatever you want.
git remote add upstream git@github.com:twilio/twilio-node.git
Now if we inspect the repo’s remotes again, we will see both the origin and the upstream. You can use -v as a shortcut for --verbose.
git remote -v
origin git@github.com:philnash/twilio-node.git (fetch)
origin git@github.com:philnash/twilio-node.git (push)
upstream git@github.com:twilio/twilio-node.git (fetch)
upstream git@github.com:twilio/twilio-node.git (push)
Fetching the latest
To bring the repo up to date, we can now fetch the latest from the upstream repo. It looks like this:
git fetch upstream
remote: Counting objects: 6427, done.
remote: Compressing objects: 100% (549/549), done.
remote: Total 6427 (delta 5156), reused 5105 (delta 4939), pack-reused 934
Receiving objects: 100% (6427/6427), 2.54 MiB | 1.28 MiB/s, done.
Resolving deltas: 100% (5399/5399), completed with 391 local objects.
From github.com:twilio/twilio-node
6a6733a8..73656c50 master -> upstream/master
git fetch downloads objects and refs from the repository, but doesn’t apply it to the branch we are working on. We want to apply the updates to the master branch, so make sure it is checked out.
git checkout master
Merging it all together
To bring the master branch up to date with the remote merge the remote’s master branch into our own, like so:
git merge upstream/master
If you have kept your work off the master branch this will go ahead smoothly. Finally push these updates to the fork so that it is up to date too.
git push origin master
Now everything on the master branch of the fork is up to date. If you need to update a different branch, substitute the branch name master for the branch you are working with.
Shortcuts
If you have been working with git pull then you may have already seen a potential shortcut. pull is the combination of fetch and merge, so to perform these two actions in one command you can run:
git pull upstream master
Fork happy
That’s all you need to know for keeping your fork up to date when contributing to open source. Add the upstream as a new remote repo, fetch the upstream repo and merge the branch you want to update.
For more detail on the various commands here, take a look at “Working with Remotes” from the Pro Git book.
And remember, if you get stuck with something with git, check out Oh shit, git! for all your expletive based ways to save yourself.
Happy forking!
Top comments (25)
Thanks for sharing Phil! I created an alias for rebasing my forks of upstream remotes, which is what I currently do when working on dev.to's codebase. gitaliases.iamdeveloper.com/#file-... if you're interested.
Ooh, a
pull --rebase upstream master. Pro level upstream work!I have shied away from too many aliases from my git setup, purely so that I can remember the actual commands and work with git anywhere. That's quite the list you have there and I bet I and others would learn quite a bit from reading through the ones you have aliased, which I assume you also use relatively frequently. Have you thought of writing anything up on all of these?
I probably should. I mention it to others occasionally in comments. At the moment though, I've started, in my head at least, a "Converting dev.to's search to Preact", so gotta get that one out first 😉, i.e. github.com/thepracticaldev/dev.to/...
Keep up the awesome posts and looking forward to the next one. 🔥
Another reason to shy away from aliases is that you get used to the aliases and then what happens when you use another system and forget you set up aliases ? Don’t laugh! It easily happens and although it’s never caused me a problem it’s because I don’t use aliases that have any behaviour modifications that matter; I for example alias gdb to ‘gdb -q’ but that’s harmless if it’s not on another install. Other commands can be far more devastating.
They can be useful but always always always remember this : aliases are for convenience and although I wouldn’t say it’s convenience XOR security the two do fight each other. And this includes data security too! But as pointed out if you rely too much on aliases you train your brain to remember those and forget the real invocations. This can particularly be a problem with different OSes more generally whether you go to a different job or maybe even worse is you use both on a daily basis.
Btw it’s happened that some people have gone to a new job but the previous say admin had set up aliases or config files that modify behaviour that has devastating consequences. This could even be deliberate and even compromise the network even installing a back door before parting! Yes that’s beyond development but still worth remembering.
I procrastinated on my other post, so here you go 😉
My Git Aliases
ℕ𝕚𝕔𝕜 𝕋𝕒𝕪𝕝𝕠𝕣
Super useful post Phil, folks are going to love this one.
Thanks Ben! I'm mainly glad I managed to write something again, it's been too long!
Ah, you mean for the case in which we use git in a truly distributed fashion, rather than the centralised way we've evolved with services like GitHub? I could, but I trust that anyone who knows what they're doing cloning aross local networks probably also knows how to add remotes, fetch and merge as well!
Or even private repos. I have quite a few on my server that are only accessible by ssh and I have ingress filtering - and don’t allow password logins either.
You shouldn’t assume that people know all the features even if they do know how to do something else. And even if they do know they might benefit from it or think of something new! That’s the beauty of the Unix philosophy and the pipe.
Yup, but this article had quite a narrow scope. I could have explained much, much more, but I wanted to keep the message tight.
Just a suggestion for the document itself and at the least future documents :
Add a prompt to the command invocation (that is prepend) to make it clearer what you type and what the output is. Maybe it’s because I am an old timer and also quite literal but my favourite example of that is quotes and a dot inside in say a VI editor tutorial i.e. ‘dd.’ which would repeat the action.
Anyway whether it’s a ‘>’ or a ‘$’ or if you want to go old school even a ‘%’ (brings back memories heh) it would make the article a bit clearer.
Cheers.
You know, I go and back and forward over this a bit.
I understand that the prompt gives a little more context, which is useful. However, I also know that people like to copy and paste from articles online (you can argue this is bad, but that's not my choice) and sometimes will copy the prompt as well and the command won't work.
I've even seen someone implement a
$function in their terminal that just passes through the arguments to run them so that they aren't caught out by the inclusion of a$in online instructions (can't find the article about this right now, but I swear it exists).For me right now, I like to make things copy-pastable without any surprises, so I'll keep the article as it is. I appreciate the feedback though, thank you.
git pull upstream masterworks just fine in most of the scenarios.
That is correct and why I talked about using
pullright near the end of the post. It’s nice to know how to use the other commands too though, which is why I wanted to explain them too.Yes I read that too. You explained everything and it was good but in my opinion it was better to discuss
pullfirst and explaining things further. BTW it's a great article.Ah, cool, thanks. I thought it was better to build up from the basic components before unveiling the shortcuts. Thanks for reading!
Background is almost always better - but you could add a note at first that there’s a built in command that does both i.e. merges the two commands. And no I can’t help it. I love punning!
Actually if the fork is solely used to make PRs you never have to push the ~master branch to the fork. Just don't forget to pull the origin before making a change a push the feature/fix branch on your fork... Although if you're a bit obsessional it's nice to have the GH message "this fork is even with its origin"
That may be true, but as it's only one extra command then why not do it? And I am that obsessed about the text on GitHub telling me it's up to date!
And a big 🍻 for talking about
fetchandmergebefore offering thepullshortcut. It's much more didactic.Awesome! I needed this info twice last week, and now I have a handy reference for all those cases in the future! Bravo!
If only I'd written it last week then! Hopefully it will help in the future though, I know I always end up checking back on my own blog posts to remind myself. So if it helps anyone else, that's even better.
TIP
Also you can set default upstream for your push, by:
git push -u remote branch, i.e.:git push -u origin master.So every next time, you have to push in short way
git push.That is true, though when you
clonea repo the default upstream is normally set toorigin master. This is useful if you want to work on a different remote branch.When I push a branch that I want to make into a pull request, I normally do
git push -u origin new-branchwhich allows me to easily make updates to that branch.Thanks for adding the tip!
I don't think I could use git with those 6 commands anyway, because 'log' is so fundamental.