What is the most efficient mechanism (in respect to data transferred and disk space used) to get the contents of a single file from a remote git repository?
So far I've managed to come up with:
git clone --no-checkout --depth 1 git@github.com:foo/bar.git && cd bar && git show HEAD:path/to/file.txt
This still seems overkill.
What about getting multiple files from the repo?
In git version 1.7.9.5 this seems to work to export a single file from a remote
git archive --remote=ssh://host/pathto/repo.git HEAD README.md | tar xO
This will cat the contents of the file README.md
.
Following on from Jakub's answer. git archive
produces a tar or zip archive, so you need to pipe the output through tar to get the file content:
git archive --remote=git://git.foo.com/project.git HEAD:path/to/directory filename | tar -x
Will save a copy of 'filename' from the HEAD of the remote repository in the current directory.
The :path/to/directory
part is optional. If excluded, the fetched file will be saved to <current working dir>/path/to/directory/filename
In addition, if you want to enable use of git archive --remote
on Git repositories hosted by git-daemon, you need to enable the daemon.uploadarch config option. See https://kernel.org/pub/software/scm/git/docs/git-daemon.html
HEAD
with the commit ID that you want to use. HEAD
is an alias that refers to either the currently checked out commit (if applicable) or the tip of the default branch. I wrote the above answer years ago and learned this morning that GitHub doesn't support git archive
, so that makes it a lot less useful.
v
as another option to tar -x
doesn't hut. Also it may be good to note that it works also for a specific folder, not only a single file : git archive --remote=git://git.foo.com/project.git HEAD path/to/folder/ | tar -xv
If there is web interface deployed (like gitweb, cgit, Gitorious, ginatra), you can use it to download single file ('raw' or 'plain' view).
If other side enabled it, you can use git archive's '--remote=<URL>
' option (and possibly limit it to a directory given file resides in), for example:
$ git archive --remote=git@github.com:foo/bar.git --prefix=path/to/ HEAD:path/to/ | tar xvf -
git config daemon.uploadarch true
on the remote repository. By default git daemon disables remote archive with "fatal: remote error: access denied or repository not exported: ..."
git archive
approach was my first try - but then I noticed that requiring tar
on the client machine wasn't exactly convenient for Windows users. We ended up fetching from our local cgit
server. It works, but it's not as fast as I'd like it to be (and it still requires running unix2dos
or similiar on Windows machines since we store files with Unix line endings in the Git repository).
Not in general but if you are using Github:
For me wget
to the raw url turned out to be the best and easiest way to download one particular file.
Open the file in the browser and click on "Raw" button. Now refresh your browser, copy the url and do a wget
or curl
on it.
wget example:
wget 'https://github.abc.abc.com/raw/abc/folder1/master/folder2/myfile.py?token=DDDDnkl92Kw8829jhXXoxBaVJIYW-h7zks5Vy9I-wA%3D%3D' -O myfile.py
Curl example:
curl 'https://example.com/raw.txt' > savedFile.txt
curl https://example.com/raw.txt > savedFile.txt
To export a single file from a remote:
git archive --remote=ssh://host/pathto/repo.git HEAD README.md | tar -x
This will download the file README.md
to your current directory.
If you want the contents of the file exported to STDOUT:
git archive --remote=ssh://host/pathto/repo.git HEAD README.md | tar -xO
You can provide multiple paths at the end of the command.
It looks like a solution to me: http://gitready.com/intermediate/2009/02/27/get-a-file-from-a-specific-revision.html
git show HEAD~4:index.html > local_file
where 4
means four revision from now and ~
is a tilde as mentioned in the comment.
git show HEAD:./my_other_file > local_file
if the file isn't in your root dir:)
I use this
$ cat ~/.wgetrc
check_certificate = off
$ wget https://raw.github.com/jquery/jquery/master/grunt.js
HTTP request sent, awaiting response... 200 OK
Length: 11339 (11K) [text/plain]
Saving to: `grunt.js'
wget https://raw.github.com/bk322/bk_automates/master/bkubuntu/bkubuntu.bash
ERROR: Certificate verification error for raw.github.com: unable to get local issuer certificate.
To connect to raw.github.com insecurely, use '--no-check-certificate'.
curl -H 'Cache-Control: no-cache, no-store' https://raw.githubusercontent.com/org/repo/master/file > outfile
otherwise it does not download if the file has already been downloaded
A nuanced variant of some of the answers here that answers the OP's question:
git archive --remote=git@archive-accepting-git-server.com:foo/bar.git \
HEAD path/to/file.txt | tar -xO path/to/file.txt > file.txt
It seems to me the easiest way to use the following:
wget https://github.com/name/folder/file.zip?raw=true
-O your-file-name
at the end of the command above.
If you repository supports tokens (for example GitLab) then generate a token for your user then navigate to the file you will download and click on RAW output to get the URL. To download the file use:
curl --silent --request GET --header 'PRIVATE-TOKEN: replace_with_your_token' \
'http://git.example.com/foo/bar.sql' --output /tmp/bar.sql
I solved in this way:
git archive --remote=ssh://git@gitlab.com/user/mi-repo.git BranchName /path-to-file/file_name | tar -xO /path-to-file/file_name > /path-to-save-the-file/file_name
If you want, you could replace "BranchName" for "HEAD"
If no other answer worked (i.e. restrictive GitLab access), you can do a "selective-checkout" by:
git clone --no-checkout --depth=1 --no-tags URL git restore --staged DIR-OR-FILE git checkout DIR-OR-FILE
Although this solution is 100% git compliant and you can checkout a directory, it's not disk nor network optimal as doing a wget/curl on a file.
For single file, just use wget command.
https://i.stack.imgur.com/V05bP.png
Then, the browser will open a new page with url start with https://raw.githubusercontent.com/...
just enter the command in the terminal:
#wget https://raw.githubusercontent.com/...
A while the file will put in your folder.
If your Git repository hosted on Azure-DevOps (VSTS) you can retrieve a single file with Rest API.
The format of this API looks like this:
https://dev.azure.com/{organization}/_apis/git/repositories/{repositoryId}/items?path={pathToFile}&api-version=4.1?download=true
For example:
https://dev.azure.com/{organization}/_apis/git/repositories/278d5cd2-584d-4b63-824a-2ba458937249/items?scopePath=/MyWebSite/MyWebSite/Views/Home/_Home.cshtml&download=true&api-version=4.1
This is specific for git repos hosted on GitHub
Try the 'api' command of Github's command line app, gh
, to make an authenticated call to Github's 'get repository contents' endpoint.
The basic command is:
$gh api /repos/{owner}/{repo}/contents/<path_to_the_file>
As an added bonus, when you do this from inside a directory that contains a clone of the repo you're trying to get the file from, the {owner} and {repo} part will be automatically filled in.
https://docs.github.com/en/rest/reference/repos#get-repository-content
The response will be a JSON object. If the
To get the file contents, you can curl the value of the "download_url", or just decode the 'content' field. You can do that by piping the base64 command, like this:
$gh api /repos/{owner}/{repo}/contents/<path-to-the-file> --jq '.content' | base64 -d
I use curl, it works with public repos or those using https basic authentication via a web interface.
curl -L --retry 20 --retry-delay 2 -O https://github.com/ACCOUNT/REPO/raw/master/PATH/TO/FILE/FILE.TXT -u USER:PASSWORD
I've tested it on github and bitbucket, works on both.
Yisrael Dov's answer is the straightforward one, but it doesn't allow compression. You can use --format=zip
, but you can't directly unzip that with a pipe command like you can with tar, so you need to save it as a temporary file. Here's a script:
#!/bin/bash
BASENAME=$0
function usage {
echo "usage: $BASENAME <remote-repo> <file> ..."
exit 1
}
[ 2 -gt "$#" ] && { usage; }
REPO=$1
shift
FILES=$@
TMPFILE=`mktemp`.zip
git archive -9 --remote=$REPO HEAD $FILES -o $TMPFILE
unzip $TMPFILE
rm $TMPFILE
This works with directories too.
Github Enterprise Solution
HTTPS_DOMAIN=https://git.your-company.com
ORGANISATION=org
REPO_NAME=my-amazing-library
FILE_PATH=path/to/some/file
BRANCH=develop
GITHUB_PERSONAL_ACCESS_TOKEN=<your-access-token>
URL="${HTTPS_DOMAIN}/raw/${ORGANISATION}/${REPO_NAME}/${BRANCH}/${FILE_PATH}"
curl -H "Authorization: token ${GITHUB_PERSONAL_ACCESS_TOKEN}" ${URL} > "${FILE_PATH}"
GITHUB_PERSONAL_ACCESS_TOKEN
?
The following 2 commands worked for me:
git archive --remote={remote_repo_git_url} {branch} {file_to_download} -o {tar_out_file}
Downloads file_to_download
as tar
archive from branch
of remote repository whose url is remote_repo_git_url
and stores it in tar_out_file
tar -x -f {tar_out_file}.tar
extracts the file_to_download
from tar_out_file
If you want to get a file from a specific hash + a remote repository I've tried git-archive and it didn't work.
You would have to use git clone and once the repository is cloned you would have then to use git-archive to make it work.
I post a question about how to do it more simpler in git archive from a specific hash from remote
for bitbucket directly from browser (I used safari...) right-click on 'View Raw" and choose "Download Linked File":
https://i.stack.imgur.com/mQLZK.png
If you don't mind cloning the entire directory, this small bash/zsh function will have the end result of cloning a single file into your current directory (by cloning the repo into a temp directory and removing it afterwards).
Pro: You only get the file you want
Con: You still have to wait for the whole repo to clone
git-single-file () {
if [ $# -lt 2 ]
then
echo "Usage: $0 <repo url> <file path>"
return
fi
TEMP_DIR=$(mktemp -d)
git clone $1 $TEMP_DIR
cp $TEMP_DIR/$2 .
rm -rf $TEMP_DIR
}
If your goal is just to download the file there's a hassle-free application called gget
:
gget github.com/gohugoio/hugo 'hugo_extended_*_Linux-ARM.deb'
The above example would download single file from hugo
repository.
https://github.com/dpb587/gget
Related to @Steven Penny's answer, I also use wget. Furthermore, to decide which file to send the output to I use -O .
If you are using gitlabs another possibility for the url is:
wget "https://git.labs.your-server/your-repo/raw/master/<path-to-file>" -O <output-file>
Unless you have the certificate or you access from a trusted server for the gitlabs installation you need --no-check-certificate as @Kos said. I prefer that rather than modifying .wgetrc but it depends on your needs.
If it is a big file you might consider using -c option with wget. To be able to continue downloading the file from where you left it if the previous intent failed in the middle.
Success story sharing
git archive --remote=ssh://host/pathto/repo.git HEAD README.md | tar -x
cat README.md
tar -xO
to output to STDOUT for piping, e.g.FILE=README.md && git archive --remote=ssh://host/pathto/repo.git HEAD "$FILE" | tar -xO "$FILE"