226

Is there a flag to pass to git when doing a clone, say don't clone the .git directory? If not, how about a flag to delete the .git directory after the clone?

2
  • 46
    This question is not a dupe of How to do a "git export" (like "svn export"). This question asks on how to clone a (remote) repository without the .git directory. The alleged dupe asks how to export an existing repository where you already have the .git directory. Commented Oct 21, 2013 at 20:18
  • 3
    Agree. You cannot git archive remote repository as the "possible duplicate" solution says. Commented Aug 5, 2014 at 8:28

10 Answers 10

310

Use

git clone --depth=1 --branch=master git://someserver/somerepo dirformynewrepo
rm -rf ./dirformynewrepo/.git
  • The depth option will make sure to copy the least bit of history possible to get that repo.
  • The branch option is optional and if not specified would get the default branch.
  • The second line will make your directory dirformynewrepo not a Git repository any more.
  • If you're doing recursive submodule clone, the depth and branch parameter don't apply to the submodules.
Sign up to request clarification or add additional context in comments.

9 Comments

For those interested, it's one of the word designators, part of the history expansion feature in Bash and zsh.
Is there a way to get the depth=1 of all branches in the repo? not just master or specified branch
you do that with the --no-single-branch option, since --depth implies --single-branch, you can un-imply it with --no-single-branch. (Taken from the git clone man page)
Using !$, while technically correct in this instance, does nothing to help people who don't recognise the syntax understand the answer to the actual question. It also relies on the second command directly following the first, and within the same shell. If anyone omits those implicit requirements, they may end up deleting something else entirely. Therefore I suggest your answer could be improved by explicitly specifying the name dirformynewrepo as an argument to the rm command as it will make the connection between both commands much clearer.
why wouldn't you use git --git-dir=/dev/null instead
|
36

since you only want the files, you don't need to treat it as a git repo.

rsync -rlp --exclude '.git' user@host:path/to/git/repo/ .

and this only works with local path and remote ssh/rsync path, it may not work if the remote server only provides git:// or https:// access.

4 Comments

that will not take .gitignore into consideration?
itub is right, this is just copying the work tree. git checkout used to do that properly, now it does not any longer.
so, what do you do if the server only provides git:// access?
there is probably a way to create/convert an ignore list/file for rsync
27

Alternatively, if you have Node.js installed, you can use the following command:

npx degit GIT_REPO

npx comes with Node, and it allows you to run binary node-based packages without installing them first (alternatively, you can first install degit globally using npm i -g degit).

Degit is a tool created by Rich Harris, the creator of Svelte and Rollup, which he uses to quickly create a new project by cloning a repository without keeping the git folder. But it can also be used to clone any repo once...

Comments

12
git clone --separate-git-dir=$(mktemp -u) --depth=1 <repo> <dir> && rm <dir>/.git

I like this solution more because I don't like rm -rfing things automatically. It just rms a .git file, which means it could never accidentally rm -rf a wrong .git directory

It has a dependency on mktemp command so it'll work *nix systems (from what I see this needs further work for the mktemp to work on MacOS, so if anyone wants to comment a working solution I'll add it)

In zsh, I made that a function so I ensure a dir value is defined:

alias np='node-project'
function node-project() {
  dir=${1:-.}
  git clone --separate-git-dir=$(mktemp -u) --depth=1 <my-node-repo> $dir && rm $dir/.git
}

Explanation

The --separate-git-dir flag lets you specify a path for the .git directory. The resulting "project" will have a .git file (not a directory) whose content will be a single line:

gitdir: <the dir you specified in the flag>

Because we used a tmp dir with the mktemp command, the actual .git directory contents will end up in a tmp dir. We also use a --depth=1 so it takes less space on tmp dirs.

1 Comment

Point of clarification - this still creates a .git equivalent folder, it's just stored in wherever mktemp chooses to put it instead of the git clone destination. When you rm <dir>/.git as per this answer, you're effectively just removing a symlink to the tmp folder created elsewhere with the normal .git files. The files are being left until the OS processes automatically cleans up the temporary directory used by --separate-git-dir. So, to properly clean up, you may still want the rm -rf, just for the git folder stored in the temp folder instead of in the clone target.
8

For those who doubt the --depth 1 solution because it still download the .git directory and you need to manually remove it afterward, maybe you need to know how git clone actually works.

When you normally clone a repo, git download all your files (spanning across commits) into the .git directory. When you clone with --depth 1, git only downloads the latest version of the files into .git. After that, git will checkout or retrieve those files from .git into the working directory (no more download).

And oftentimes, because the file objects inside .git is compressed, you will save more bandwidth by downloading the files with git clone --depth 1 rather than downloading the uncompressed files. And for some people with slow internet, that is worth the price (the need to run rm -rf).

I personally think the git archive solution is better but since it's not supported by GitHub, --depth 1 is the way to go.

4 Comments

If it's GitHub, why not just curl -L https://github.com/${owner}/${repo}/tarball/master | tar xz ?
@DesNerger Because we can easily clone from private repo with ssh
The objects are compressed, but aren't they stored as a series of diffs over their initial representations (many commits ago)? Does git fetch those diffs, or somehow just the latest version, when using --dept 1?
Subversion store diffs. Git doesn't store diffs. It stores the files and each commit stores the hash of the files.
2

git archive --remote already implements this.

5 Comments

Any combination of git archive --remote I tried from posts online failed me with Invalid command: 'git-upload-archive '<repo>'
Try: git archive --remote [email protected]:gitlab-org/gitlab.git -o gitlab_15.9.3ee.tgz v15.9.3-ee
LFS files may not be part of the archive.
Same problem here, just like @GuidoTarsia said. mr.wolle's suggestion does not help either.
It seems to be a GitHub limitation. They provide their own way to generate "download zip" URLs from the web UI. Why they do not simply support standard Git features, too, eludes me, however.
2

If the repository is on GitHub, you can simply download a ZIP file of any tag:

curl -L https://github.com/<user>/<repo-name>/archive/refs/tags/v1.2.3.zip | tar xz

This will download and uncompress the repository. The contents will be in <repo-name>-1.2.3, with no .git folder.

2 Comments

Very neat trick, but there are a lot of Git repositories that are not hosted on GitHub.
Good point. Well. They should simply move to github. I am joking. Well, I am sure it is possible to di similar thing also onnother hosts.
1
git clone --depth=1 --branch=master git://someserver/somerepo dirformynewrepo1
rd /s /q  .\dirformynewrepo1\.git

this works for windows systems

Comments

1

If you are targeting GitHub repositories, their archive URL follows a specific pattern. Based on this, you can add this function to your .zshrc:

ghdown() {
  repo="$1"
  branch="${2:-main}"
  file="$(basename "$repo")-$branch.zip"
  wget -O "$file" "https://github.com/$repo/archive/refs/heads/$branch.zip"
  unzip $file && rm $file
}

and use it like this:

ghdown llvm/llvm-project
ghdown emacs-mirror/emacs master

Comments

-4

You can always do

git clone git://repo.org/fossproject.git && rm -rf fossproject/.git

3 Comments

download everything and delete the dir later? for instance emacs has 95% of 1GB in the .git. so --depth=1 is the solution here clearly.
i still see the .git folder when i do depth=1
@AryehArmon of course you see it. Using --depth does not prevent the creation of .git, it just makes it a shallow copy, so much smaller size for .git. But if you do not want it there at all, you need to remove it afterwards.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.