Is it possible to shallow clone a specific commit in a repository, i.e. with depth 1? Something like
git clone http://myrepo.git 728a4d --depth 1 to get the repository state as it is at the commit with SHA 728a4d...?
The motivation is to avoid having to clone the whole repository, then check out that specific commit, when we're only interested in the state of the repository at that specific commit.
Starting with Git 2.5.0 (which needs to be available at both the client and server side) you can set uploadpack.allowReachableSHA1InWant=true on the server side to enable fetching of specific SHA1s (must be the full SHA1, not an abbreviation):
git init git remote add origin <url> git fetch --depth 1 origin <sha1> git checkout FETCH_HEAD Note that I did not find a syntax to do this with git clone directly.
Update: Starting with Git 2.49.0, there's a new --revision option for git clone "which fetches history leading up to the specified revision, regardless of whether or not there is a branch or tag pointing at it". Together with --depth 1 this should be a yet better option.
error: pathspec 'FETCH_HEAD' did not match any file(s) known to git
NOTE: My example doesn't help to clone to by a commit hash but it will help to clone a tag and have a lightweight repository.
If you have to have only one commit in your "clone" and you are going to use commit hash, short answer is NO.
I use this command construction (tested on v2.13.2.windows.1) for tags:
git clone --depth 1 [email protected]:VENDOR/REPO.git --branch 1.23.0 --single-branch Full example:
$ git clone --depth 1 [email protected]:Seldaek/monolog.git --branch 1.23.0 --single-branch Cloning into 'monolog'... remote: Counting objects: 201, done. remote: Compressing objects: 100% (188/188), done. remote: Total 201 (delta 42), reused 32 (delta 5), pack-reused 0 Receiving objects: 100% (201/201), 190.30 KiB | 0 bytes/s, done. Resolving deltas: 100% (42/42), done. Note: checking out 'fd8c787753b3a2ad11bc60c063cff1358a32a3b4'. You are in 'detached HEAD' state. You can look around, make experimental changes and commit them, and you can discard any commits you make in this state without impacting any branches by performing another checkout. If you want to create a new branch to retain commits you create, you may do so (now or later) by using -b with the checkout command again. Example: git checkout -b <new-branch-name> $ cd monolog .git dir size (267K vs 2.6M by using full clone):
$ du -h --max-depth=0 .git 267K .git I'd like to denote, --branch can take a tag/branch.
https://git-scm.com/docs/git-clone#git-clone---branchltnamegt
--branchcan also take tags and detaches the HEAD at that commit in the resulting repository.
UPD
In a nutshell, it can take "refs". You may read more here: What does the git error message “Server does not allow request for unadvertised object” mean?
Also, there is no tricks like:
git fetch --depth 1 origin <COMMIT_HASH> Thanks @BenjiWiebe for pointing me in my mistake.
--branch does not support commits, only branch names... this doesn't solve it at all. As it not true.
Try using while in bash:
git clone --depth=1 $url i=1; while ! git show $sha1; do git fetch --depth=$((i+=1)); done This is pretty slow because it fetches each commit individually; you could increase the increment (to fetch commits in batches and improve performance over a network) but it's still a brute force approach.
The immediate answer is: You can't do it using a git clone directly.
Why? A detailed explanation can be found here: Why Isn't There A Git Clone Specific Commit Option?
What else can you do?
# Create empty repository to store your content git clone <url> git reset <sha-1> --hard More info:
git clone <url> --branch <branch_name> --single-branch <folder_name>
git clone <url> --depth=1 --branch <branch_name> --single-branch <folder_name>
As @sschuberth commented out: --depth implies --single-branch.
Instead of clone use the fetch command:
# fetch a commit (or branch or tag) of interest # In this case you will have the full history of this commit git fetch origin <sha1> 
--depth implies --single-branch, so you could drop it in that case.
git fetch actually accept an SHA1 for refspec?A Github specific version would be to use aria2c or another download manager:
REPO_URL="https://github.com/nwchemgit/nwchem" COMMIT_HASH="$1" TAR_GZ_URL="${REPO_URL}/archive/${COMMIT_HASH}.tar.gz" if [ -d "$REPO_DIR" ]; then echo "Directory '$REPO_DIR' already exists. Skipping clone." else echo "Cloning repository from $REPO_URL into $REPO_DIR..." mkdir -p "$REPO_DIR" cd "$REPO_DIR" cd ../ aria2c -x 16 "$TAR_GZ_URL" -o "nwchem.tar.gz" tar -xf "nwchem.tar.gz" mv nwchem-* nwchem fi The answer from https://stackoverflow.com/a/43136160/4411491 works, but it creates a detached head. The following slight modification will avoid this, resulting in a regular single-commit checkout:
# Set those variables: repourl='https://some.test/somerepo.git' commit=4e1243bd22c66e76c2ba9eddc1f91394e57f9f83 # Paste the remaining commands unmodified into your shell. : ${repourl:?}; reponame=${repourl##*/}; reponame=${reponame%.git} git init -- "${reponame:?}" cd -- "${reponame:?}" git remote add origin "${repourl:?}" git fetch -q --depth=1 origin "${commit:?}" git reset --hard FETCH_HEAD Note that it is imperative that $commit is set to a full commit ID. An abbreviation of the commit ID does not work.
