GitMedia extension allows you to use Git with large media files without storing the media in Git itself.
Setup the attributes filter settings.
(once after install)
$ git config filter.media.clean "git-media filter-clean"
$ git config filter.media.smudge "git-media filter-smudge"
Setup the .gitattributes
file to map extensions to the filter.
(in repo - once)
$ echo "*.mov filter=media -crlf" > .gitattributes
Staging files with those extensions will automatically copy them to the media buffer area (.git/media) until you run 'git media sync' wherein they are uploaded. Checkouts that reference media you don't have yet will try to be automatically downloaded, otherwise they are downloaded when you sync.
Next you need to configure git to tell it where you want to store the large files. There are five options:
- Storing remotely in Amazon's S3
- Storing locally in a filesystem path
- Storing remotely via SCP (should work with any SSH server)
- Storing remotely in atmos
- Storing remotely via WebDav
Here are the relevant sections that should go either in ~/.gitconfig
(for global settings)
or in clone/.git/config
(for per-repo settings).
[git-media]
transport = <scp|local|s3|atmos|webdav>
autodownload = <true|false>
# settings for scp transport
scpuser = <user>
scphost = <host>
scppath = <path_on_remote_server>
# settings for local transport
localpath = <local_filesystem_path>
# settings for s3 transport
s3bucket = <name_of_bucket>
s3key = <s3 access key>
s3secret = <s3 secret key>
# settings for atmos transport
endpoint = <atmos server>
uid = <atmos_uid>
secret = <atmos secret key>
tag = <atmos object tag>
# settings for webdav transport
webdavurl = <webdav root url>
# user and password are taken from netrc if omitted
webdavuser = <user for basic auth, optional>
webdavpassword = <password for basic auth>
webdavverifyserver = <Net::Dav.verify_server setting, true by default>
webdavbinarytransfer = <Net::Dav.new :curl option value, false by default>
(in repo - repeatedly)
$ (hack, stage, commit)
$ git media sync
You can also check the status of your media files via
$ git media status
Which will show you files that are waiting to be uploaded and how much data that is. If you want to upload & delete the local cache of media files, run:
$ git media clear
If you want to replace file in git-media with changed version (for example, video file has been edited), you need to explicitly tell git that some media files has changed:
$ git update-index --really-refresh
If autodownload is set to true, required files will automatically be downloaded when checking out or pulling. Default is false
$ git config --global media.autodownload true
$ git clone [email protected]:alebedev/git-media.git
$ cd git-media
$ sudo gem install bundler
$ bundle install
$ gem build git-media.gemspec
$ sudo gem install git-media-*.gem
It is important to switch off git smart newline character support for media files.
Use -crlf
switch in .gitattributes
(for example *.mov filter=media -crlf
) or config option core.autocrlf = false
.
If installing on windows, you might run into a problem verifying certificates for S3 or something. If that happens, modify
C:\Ruby191\lib\ruby\gems\1.9.1\gems\right_http_connection-1.2.4\lib\right_http_connection.rb
And add at line 310, right before @http.start
:
@http.verify_mode = OpenSSL::SSL::VERIFY_NONE
Copyright (c) 2009 Scott Chacon. See LICENSE for details.