Manage large binary files with git
A tool for managing large binary files in git repositories.
Checking large binary files into a distributed version control system is a bad idea because repository size quickly becomes unmanagable. Every operation takes longer to complete and fresh clones become something that you do start doing overnight. Using git-fat allows you to separate the storage of largefiles from the source while still having them in the working directory for your project.
- Cloning the source code remains fast because binaries are not included
- Binary files really exist in your working directory and are not soft-links
- Only depends on Python 2.7, rsync and ssh
- Download only the files you need with pattern matching
- Supports anonymous downloads of files over http
You can install git-fat using pip.
pip install git-fat
Or you can install it simply by placing it on your path.
curl https://raw.github.com/cyaninc/git-fat/master/git_fat/git_fat.py \ | sudo tee /usr/local/bin/git-fat && sudo chmod +x /usr/local/bin/git-fat
First, create a git attributes file in the root of your repository. This file determines which files get converted to git-fat files.
cat >> .gitattributes <<EOF *.deb filter=fat -crlf *.gz filter=fat -crlf *.zip filter=fat -crlf EOF
Next, create a .gitfat configuration file in the root of your repo that contains the location of the remote store for the binary files. Optionally include the ssh user and port if non-standard. Also, optionally include an http remote for anonymous clones.
[rsync] remote = storage.example.com:/path/to/store user = git port = 2222 [http] remote = http://storage.example.com/store
Commit those files so that others will be able to use them.
Initalize the repository. This adds a line to .git/config telling git what command to run for the fat filter is in the .gitattributes file.
git fat init
Now when you add a file that matches a pattern in the .gitattributes file, it will be converted to a fat placeholder file before getting commited to the repository. After you’ve added a file remember to push it to the fat store, otherwise people won’t get the binary file when they try to pull fat-files.
git fat push
After we’ve done a new clone of a repository using git-fat, to get the additional files we do a fat pull.
git fat pull
Or if you’re doing an anonymous pull, and the repository managers support it.
git fat pull-http
To list the files managed by git-fat
git fat list
To get a summary of the orphan and stale files in the repository
git fat status
Orphans are files that exist as placeholders in the working copy. Stale files are files that are in the .git/fat/objects directory, but have no working copy associated with them (e.g. old versions of files).
To find files over a certain size, use git fat find. This example finds all objects greater than 10MB in git’s database and prints them out.
git fat find 10485760
For many commands, git-fat by default only checks the current HEAD for placeholder files to clone. This can save on bandwidth for frequently changing large files and also saves on processing time for very large repositories. To force commands to search the entire history for placeholders and pull all files, call git-fat with -a. e.g.
git fat -a pull
If you add git-fat to an existing repository, the default behavior is to not convert existing binary files to git-fat. Converting a file that already exists in the history for git would not save any space. Once the file is changed or renamed, it will then be added to the fat store.
To setup an http server to accept git-fat requests, just configure a webserver to have a url serve up the git-fat directory on the server, and point the .gitfat http remote to that url.
You can retroactively import a repository to git-fat using a combination of find and index-filter used with git’s filter-branch command.
darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ git fat find 5123123 761a63bf287867da92eb420fca515363c4b02ad1 9437184 flowerpot.tar.gz 6c5d4031e03408e34ae476c5053ee497a91ac37b 10485760 whale.tar.gz darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ git fat find 5123123 | cut -d' ' -f3- > /tmp/towel darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ cat /tmp/towel flowerpot.tar.gz whale.tar.gz darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ ll total 19M drwxrwxr-x 3 darthurdent darthurdent 4.0K Dec 10 13:42 . drwxrwxrwt 6 root root 76K Dec 10 13:42 .. drwxrwxr-x 6 darthurdent darthurdent 4.0K Dec 10 13:42 .git -rw-r--r-- 1 darthurdent darthurdent 9.0M Dec 10 13:37 flowerpot.tar.gz -rw-r--r-- 1 darthurdent darthurdent 10M Dec 10 13:37 whale.tar.gz darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ git filter-branch --index-filter 'git fat index-filter /tmp/towel'\ --tag-name-filter cat -- --all Rewrite 28cfba441aac92992c3f80dae97cd1c19b3befad (2/2) Ref 'refs/heads/master' was rewritten darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ ll total 19M drwxrwxr-x 3 darthurdent darthurdent 4.0K Dec 10 13:42 . drwxrwxrwt 6 root root 76K Dec 10 13:42 .. drwxrwxr-x 6 darthurdent darthurdent 4.0K Dec 10 13:42 .git -rw-rw-r-- 1 darthurdent darthurdent 64 Dec 10 13:42 .gitattributes -rw-rw-r-- 1 darthurdent darthurdent 9.0M Dec 10 13:42 flowerpot.tar.gz -rw-rw-r-- 1 darthurdent darthurdent 10M Dec 10 13:42 whale.tar.gz darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ cat .gitattributes flowerpot.tar.gz filter=fat -text whale.tar.gz filter=fat -text darthurdent at betelgeuse in /tmp/git-fat-demo (master) $ git cat-file -p $(git hash-object whale.tar.gz) #$# git-fat 8c206a1a87599f532ce68675536f0b1546900d7a 10485760
- More friendly configuration for multiple fat remotes
- Private setting of a different remote. (e.g. local mirrors)
- Generic backend interface
- Better Documentation
- Improved Testing
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size git-fat-0.2.1.tar.gz (13.2 kB)||File type Source||Python version None||Upload date||Hashes View|