Class: Capistrano::Deploy::SCM::Git
- Defined in:
- lib/capistrano/recipes/deploy/scm/git.rb
Overview
An SCM module for using Git as your source control tool with Capistrano 2.0. If you are using Capistrano 1.x, use this plugin instead:
http://scie.nti.st/2007/3/16/capistrano-with-git-shared-repository
Assumes you are using a shared Git repository.
Parts of this plugin borrowed from Scott Chacon’s version, which I found on the Capistrano mailing list but failed to be able to get working.
FEATURES:
* Very simple, only requiring 2 lines in your deploy.rb.
* Can deploy different branches, tags, or any SHA1 easily.
* Supports prompting for password / passphrase upon checkout.
(I am amazed at how some plugins don't do this)
* Supports :scm_command, :scm_password, :scm_passphrase Capistrano
directives.
CONFIGURATION
Use this plugin by adding the following line in your config/deploy.rb:
set :scm, :git
Set :repository
to the path of your Git repo:
set :repository, "someuser@somehost:/home/myproject"
The above two options are required to be set, the ones below are optional.
You may set :branch
, which is the reference to the branch, tag, or any SHA1 you are deploying, for example:
set :branch, "master"
Otherwise, HEAD is assumed. I strongly suggest you set this. HEAD is not always the best assumption.
You may also set :remote
, which will be used as a name for remote tracking of repositories. This option is intended for use with the :remote_cache
strategy in a distributed git environment.
For example in the projects config/deploy.rb
:
set :repository, "#{scm_user}@somehost:~/projects/project.git"
set :remote, "#{scm_user}"
Then each person with deploy priveledges can add the following to their local ~/.caprc
file:
set :scm_user, 'someuser'
Now any time a person deploys the project, their repository will be setup as a remote git repository within the cached repository.
The :scm_command
configuration variable, if specified, will be used as the full path to the git executable on the remote machine:
set :scm_command, "/opt/local/bin/git"
For compatibility with deploy scripts that may have used the 1.x version of this plugin before upgrading, :git
is still recognized as an alias for :scm_command.
Set :scm_password
to the password needed to clone your repo if you don’t have password-less (public key) entry:
set :scm_password, "my_secret'
Otherwise, you will be prompted for a password.
:scm_passphrase
is also supported.
The remote cache strategy is also supported.
set :repository_cache, "git_master"
set :deploy_via, :remote_cache
For faster clone, you can also use shallow cloning. This will set the ‘–depth’ flag using the depth specified. This cannot be used together with the :remote_cache strategy
set :git_shallow_clone, 1
For those that don’t like to leave your entire repository on your production server you can:
set :deploy_via, :export
To deploy from a local repository:
set :repository, "file://."
set :deploy_via, :copy
AUTHORS
Garry Dolley scie.nti.st Contributions by Geoffrey Grosenbach topfunky.com
Scott Chacon http://jointheconversation.org
Alex Arnell http://twologic.com
and Phillip Goldenburg
Instance Attribute Summary
Attributes inherited from Base
Instance Method Summary collapse
-
#checkout(revision, destination) ⇒ Object
Performs a clone on the remote machine, then checkout on the branch you want to deploy.
- #command ⇒ Object
-
#diff(from, to = nil) ⇒ Object
Returns a string of diffs between two revisions.
-
#export(revision, destination) ⇒ Object
An expensive export.
-
#handle_data(state, stream, text) ⇒ Object
Determines what the response should be for a particular bit of text from the SCM.
-
#head ⇒ Object
When referencing “head”, use the branch we want to deploy or, by default, Git’s reference of HEAD (the latest changeset in the default branch, usually called “master”).
-
#log(from, to = nil) ⇒ Object
Returns a log of changes between the two revisions (inclusive).
- #origin ⇒ Object
-
#query_revision(revision) ⇒ Object
Getting the actual commit id, in case we were passed a tag or partial sha or something - it will return the sha if you pass a sha, too.
-
#sync(revision, destination) ⇒ Object
Merges the changes to ‘head’ since the last fetch, for remote_cache deployment strategy.
Methods inherited from Base
default_command, #initialize, #local, #local?, #next_revision, #scm
Constructor Details
This class inherits a constructor from Capistrano::Deploy::SCM::Base
Instance Method Details
#checkout(revision, destination) ⇒ Object
Performs a clone on the remote machine, then checkout on the branch you want to deploy.
132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 132 def checkout(revision, destination) git = command remote = origin args = [] args << "-o #{remote}" unless remote == 'origin' if depth = variable(:git_shallow_clone) args << "--depth #{depth}" end execute = [] execute << "#{git} clone #{verbose} #{args.join(' ')} #{variable(:repository)} #{destination}" # checkout into a local branch rather than a detached HEAD execute << "cd #{destination} && #{git} checkout #{verbose} -b deploy #{revision}" if variable(:git_enable_submodules) execute << "#{git} submodule #{verbose} init" execute << "#{git} submodule #{verbose} sync" execute << "#{git} submodule #{verbose} update --recursive" end execute.join(" && ") end |
#command ⇒ Object
231 232 233 234 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 231 def command # For backwards compatibility with 1.x version of this module variable(:git) || super end |
#diff(from, to = nil) ⇒ Object
Returns a string of diffs between two revisions
201 202 203 204 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 201 def diff(from, to=nil) from << "..#{to}" if to scm :diff, from end |
#export(revision, destination) ⇒ Object
An expensive export. Performs a checkout as above, then removes the repo.
159 160 161 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 159 def export(revision, destination) checkout(revision, destination) << " && rm -Rf #{destination}/.git" end |
#handle_data(state, stream, text) ⇒ Object
Determines what the response should be for a particular bit of text from the SCM. Password prompts, connection requests, passphrases, etc. are handled here.
239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 239 def handle_data(state, stream, text) host = state[:channel][:host] logger.info "[#{host} :: #{stream}] #{text}" case text when /\bpassword.*:/i # git is prompting for a password unless pass = variable(:scm_password) pass = Capistrano::CLI.password_prompt end "#{pass}\n" when %r{\(yes/no\)} # git is asking whether or not to connect "yes\n" when /passphrase/i # git is asking for the passphrase for the user's key unless pass = variable(:scm_passphrase) pass = Capistrano::CLI.password_prompt end "#{pass}\n" when /accept \(t\)emporarily/ # git is asking whether to accept the certificate "t\n" end end |
#head ⇒ Object
When referencing “head”, use the branch we want to deploy or, by default, Git’s reference of HEAD (the latest changeset in the default branch, usually called “master”).
122 123 124 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 122 def head variable(:branch) || 'HEAD' end |
#log(from, to = nil) ⇒ Object
Returns a log of changes between the two revisions (inclusive).
207 208 209 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 207 def log(from, to=nil) scm :log, "#{from}..#{to}" end |
#origin ⇒ Object
126 127 128 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 126 def origin variable(:remote) || 'origin' end |
#query_revision(revision) ⇒ Object
Getting the actual commit id, in case we were passed a tag or partial sha or something - it will return the sha if you pass a sha, too
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 213 def query_revision(revision) raise ArgumentError, "Deploying remote branches is no longer supported. Specify the remote branch as a local branch for the git repository you're deploying from (ie: '#{revision.gsub('origin/', '')}' rather than '#{revision}')." if revision =~ /^origin\// return revision if revision =~ /^[0-9a-f]{40}$/ command = scm('ls-remote', repository, revision) result = yield(command) revdata = result.split(/[\t\n]/) newrev = nil revdata.each_slice(2) do |refs| rev, ref = *refs if ref.sub(/refs\/.*?\//, '').strip == revision.to_s newrev = rev break end end raise "Unable to resolve revision for '#{revision}' on repository '#{repository}'." unless newrev =~ /^[0-9a-f]{40}$/ return newrev end |
#sync(revision, destination) ⇒ Object
Merges the changes to ‘head’ since the last fetch, for remote_cache deployment strategy
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 |
# File 'lib/capistrano/recipes/deploy/scm/git.rb', line 165 def sync(revision, destination) git = command remote = origin execute = [] execute << "cd #{destination}" # Use git-config to setup a remote tracking branches. Could use # git-remote but it complains when a remote of the same name already # exists, git-config will just silenty overwrite the setting every # time. This could cause wierd-ness in the remote cache if the url # changes between calls, but as long as the repositories are all # based from each other it should still work fine. if remote != 'origin' execute << "#{git} config remote.#{remote}.url #{variable(:repository)}" execute << "#{git} config remote.#{remote}.fetch +refs/heads/*:refs/remotes/#{remote}/*" end # since we're in a local branch already, just reset to specified revision rather than merge execute << "#{git} fetch #{verbose} #{remote} && #{git} reset #{verbose} --hard #{revision}" if variable(:git_enable_submodules) execute << "#{git} submodule #{verbose} init" execute << "for mod in `#{git} submodule status | awk '{ print $2 }'`; do #{git} config -f .git/config submodule.${mod}.url `#{git} config -f .gitmodules --get submodule.${mod}.url` && echo Synced $mod; done" execute << "#{git} submodule #{verbose} sync" execute << "#{git} submodule #{verbose} update" end # Make sure there's nothing else lying around in the repository (for # example, a submodule that has subsequently been removed). execute << "#{git} clean #{verbose} -d -x -f" execute.join(" && ") end |